Journal of Banking & Finance 34 (2010) 2481–2499

Contents lists available at ScienceDirect

Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf

Control/ownership structure, creditor rights protection, and the cost of debt

financing: International evidence
Narjess Boubakri a,b,*, Hatem Ghouma c
a
American University of Sharjah, United Arab Emirates
b
HEC Montreal, Canada
c
Al-Akhawayn University, Morocco

a r t i c l e i n f o a b s t r a c t

Article history: We explore the effect of governance on bond yield-spreads and ratings in a multinational sample of firms.
Received 23 May 2008 We find strong evidence that ultimate ownership (i.e., the voting/cash-flow rights wedge) and family con-
Accepted 9 April 2010 trol have a positive and significant effect on bond yield-spreads, and a negative and significant effect on
Available online 13 April 2010
bond ratings. Control in the hands of widely held financial firms has a positive effect on bond ratings only,
while State control has no effect on either bond yield-spreads or ratings. We also find that a higher pro-
JEL classification: tection of debtholders’ rights generally reduces bond yield-spreads and increases bond ratings. Our
G34
results additionally show that, for both bondholders and rating agencies, the enforcement of debt laws
G38
F34
is crucially important. Finally, we document a negative effect of debt covenants on debt costs when there
is a high expropriation risk and poor creditor rights protection.
Keywords: Ó 2010 Elsevier B.V. All rights reserved.
Bond yields and ratings
Ultimate ownership
Investor protection
Debt enforcement

1. Introduction
Shleifer and Vishny (1997) define corporate governance as ‘‘the
ways through which suppliers of capital to corporations assure them-
selves of getting a return on their investment.”1 A large body of empirical
research links corporate governance (particularly ownership struc-
ture) to the managers–shareholders agency problems (Jensen, 1986;
Morck et al., 1989; Walsh and Seward, 1990; Fluck, 1999; among
others). However, little attention has been devoted to the sharehold-
ers–bondholders agency conflict. Indeed, as Jensen and Meckling
(1976) show, shareholders may expropriate wealth from bondholders
by undertaking risky new projects that will allow them to reap most of
the gains, while bondholders bear most of the cost (Klock et al., 2005).
In this paper, we focus on the link between the ultimate owner-
ship of the firm and its bond yield and rating. We look at ultimate
ownership along two dimensions: (1) the discrepancy between
voting and cash-flow rights of major shareholders, and (2) ultimate
shareholder identity. The voting/cash-flow rights wedge measures
the extent and the likelihood of expropriation by majority share-
holders (Claessens et al., 2000; Faccio and Lang, 2002). Although
bondholders may also face the risk that managers behave opportu-
nistically, we choose to focus on the conflict between controlling
shareholders and bondholders based on the following rationale:
the literature to date shows that, outside the US, ownership struc-
ture is highly concentrated (Denis and McConnell, 2003).2 This evi-
dence makes expropriation by major shareholders more likely in an
international setting such as ours. Moreover, managers’ opportunis-
tic behaviour is tackled to a large extent by the fact that controlling
shareholders usually appoint managers among their relatives (La
Porta et al., 1999; Claessens et al., 2000; Faccio and Lang, 2002), thus
aligning both parties’ interests. This situation in fact increases the
likelihood that expropriation by controlling shareholders occurs.
Assuming that bondholders anticipate such behaviour, they will re-
quire a higher premium, resulting in a higher debt cost.
Previous studies also suggest that the identity of the firms’
(ultimate) owners is important to both bondholders and rating
agencies. For example, families usually appoint managers among
their relatives (see for example Faccio et al., 2001). This collusion

2
* Corresponding author at: American University of Sharjah, United Arab Emirates. Non-US studies that analyze ownership structure include, among others: Prowse
Tel.: +971 6 515 2587; fax: +971 6 558 5065. (1992) in Japan, Franks and Mayer (2001) in Germany, Xu and Wang (1997) in China,
E-mail addresses: nboubakri@aus.edu (N. Boubakri), H.Ghouma@aui.ma (H. Valadares and Leal (2000) in Brazil, Faccio and Lang (2002) in West Europe, and
Ghouma). Claessens et al. (2000) in East Asia. Denis and McConnell (2003) offer an excellent
1
Shleifer and Vishny (1997, p. 737). literature review of this evidence.

0378-4266/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2010.04.006
2482 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
may be perceived by bondholders as an increased risk of expropria-
tion. Their rational response will be to require higher yields. In the
same vein, rating agencies are likely to award lower ratings to family
firms. However, one could also argue that families are more likely to
adopt value-maximizing strategies to ensure the firm’s survival gi-
ven that they intend to pass it on to subsequent generations (Ander-
son et al., 2003). This behaviour may benefit bondholders and other
stakeholders, and may result in lower bond yield-spreads and high-
er ratings. Which of these two effects will dominate remains an open
empirical question that only few studies tried to address (see for
example Anderson et al. (2003) on controlling families and Bhojraj
and Sengupta (2003) on the impact of institutional ownership, both
set in the US). The common ground of these studies is that they use
ownership-based measures (generally the direct ownership stake)
to assess the power of the main shareholder. In this paper, we use
control-based rather than ownership-based measures. The data on
the identity of ultimate owners allows us to determine who is per-
ceived by bondholders and rating agencies as ultimately represent-
ing a potential risk of expropriation.
Using a multinational sample of debt issuing firms from devel-
oped and developing countries, we assess how the quality of the
institutional environment conditions the agency cost of debt across
institutionally diverse environments. Our study thus contributes to
the scarce academic literature on the link between governance
mechanisms and the cost of debt financing, and contributes to
our understanding of the functioning of fixed income securities’
markets around the world. As Shleifer and Vishny (1997) note,
empirical research on creditor governance is indeed an under re-
searched area in the corporate governance literature.
While some recent studies analyze the relationship between
governance mechanisms and debt costs, (Sengupta, 1998; Bhojraj
and Sengupta, 2003; Anderson et al., 2003; Mansi et al., 2004;
Ashbaugh et al., 2006), their evidence is drawn from the US, and
thus cannot be generalized to other countries with less favourable
legal environments. The lack of evidence on this issue is puzzling
since debt constitutes an important external source of financing
for publicly traded firms around the globe.
This paper adds to the international corporate governance liter-
ature on other grounds as well: for instance, the available literature
focuses primarily on direct ownership. Our analysis relies instead
on ultimate ownership and allows us to control for the extent
and likelihood of expropriation by controlling shareholders (i.e.,
extent of agency conflicts within the firm). Few existing studies
look at the potential impact of investor protection and overall qual-
ity of institutions in the country on the firms’ cost of debt financ-
ing. Miller and Puthenpurackal (2002) offer the first empirical
evidence on the importance of investor protection in the debt mar-
ket. The authors analyze the cost of debt for a sample of 260 Yan-
kee bonds issued by non-US firms and report that investors charge
higher bond costs to firms that are located in countries with poor
investors’ rights protection and those that ‘‘do not have a prior his-
tory of ongoing disclosure.” A more recent study by Ellul et al.
(2005) provides some evidence on the impact of legal institutions
on debt costs by analyzing US firms and foreign firms that issue
ADRs in the US. However, such an approach is likely to suffer from
a selection bias problem since ADR firms have to comply with
(internal) corporate governance standards that are generally im-
posed by the American legislator and the Securities Exchange Com-
mission (SEC). As a consequence, the firms used in Ellul et al.
(2005)’s study are more likely to exhibit a better governance than
their local counterparts that do not issue ADRs. Finally, our frame-
work provides us with a valuable opportunity to identify the set of
institutions that the legislator needs to adjust to foster the devel-
opment and well functioning of financial debt markets.
Based on a sample of corporate bond issues in 19 countries from
East Asia and Western Europe, we find that the wedge between
ownership and control (i.e., our proxy for expropriation) affects
significantly both the bond yield and rating. With respect to the
controlling shareholder’s identity, we find that family control is
perceived as a potential risk of expropriation by both bondholders
and rating agencies as it loads a positive statistical effect on bond
spreads, and a negative statistical effect on bond ratings. Thus, con-
trary to Anderson et al. (2003) who show that US family firms are
seen as a protector of bondholders’ rights, our finding suggests that
this type of owner is more likely to harm bondholders in other
markets. Through their controlling position, families are able to ex-
tract private benefits that are costly to all stakeholders, including
bondholders. Furthermore, families often avoid ownership dilution
in order to keep a tight control over the firm, which leads them to
prefer debt to equity financing, hence the higher leverage of such
firms. Finally, we find that control in the hands of widely held
financial firms affects positively bond ratings (only), while State
control affects neither bond yields nor ratings.
Next, we analyze the effect of the institutional environment on
corporate bond yields and ratings. We consider a large set of na-
tional governance mechanisms that encompasses regulatory insti-
tutions that previous studies (e.g., La Porta et al., 1998; Dyck and
Zingales, 2004; Djankov et al., 2007, 2008) have shown to play a
significant role in preserving investors’ rights. Our results show
that higher investors’ (and essentially debt-holders’) protection
generally reduces bond spreads, and increases corporate bond rat-
ings. However, we document that the creditor rights index (i.e.
restrictions that directly protect their rights) does not matter for
bond spreads and ratings, while most debt enforcement measures
load statistically and economically significant coefficients. This re-
sult suggests that both debtholders and rating agencies value debt
enforcement rather than the mere existence of debt laws. Thus,
ceteris paribus, authorities who are seeking to develop bond mar-
kets should put more emphasis on the enforcement of laws pro-
tecting creditors (for example, by creating credit registries)
rather than seek to create and enact new laws for the book.
Finally, the framework of this study offers the opportunity to
test whether debt covenants affect the relation between bond
yield-spreads and corporate governance. Cremers et al. (2007) find
that the impact of a controlling shareholder on bond yield-spreads
is smaller for bonds with covenants. The authors conclude that
bond covenants could help ‘‘in the convergence of shareholders
and bondholders interests”, and hence reduce the agency cost of
debt. We test this hypothesis in an international framework, and
we find evidence that in general, protective covenants alleviate
the potential risk of expropriation. In fact, the presence of cove-
nants reduces bond yield-spreads by 12.4 bps for firms with higher
voting/cash-flow rights wedge. However, covenants seem to be
ineffective in reducing spreads if the ultimate owner is a family. Fi-
nally, debt convenants mitigate, but do not totally eliminate, the
spreads in poor creditor rights environments.
We organize the remainder of this paper as follows. Section 2
presents the theoretical framework by describing the relation
between corporate governance and the bond yields and ratings. In
Section 3 we describe our models, the variable measurements, the
sample and data sources, and we provide descriptive statistics.
Section 4 discusses our empirical evidence and Section 5 concludes.
2. Corporate governance and the agency cost of debt:
Hypothesis development
2.1. Theoretical framework: the agency cost of debt
Recent studies show that lenders do not only rely on the firm’s
past profitability and on the issue characteristics in order to infer
the expected cash flows (and default probability). In fact, investors
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
also price the firm’s corporate governance structure. Specifically,
debtholders facing the managers’ opportunistic behavior, will
rationally require larger yields from firms with less disciplined
managers. Similarly, for firms where they anticipate potential
wealth expropriation by shareholders (e.g., through overinvest-
ment), debtholders will require a higher premium (i.e., charge a
higher cost of debt financing to these firms).
We discuss in the next two sections, the empirical literature
that relates corporate governance to bond yields and ratings in or-
der to derive our main hypotheses.
2.2. Corporate governance, ultimate ownership, bond yield-spreads
and credit rating
Few recent studies investigate the relation between internal
corporate governance mechanisms and bond yield-spreads and
ratings. All available evidence is based on US markets, and the vari-
ables related to corporate governance differ across studies, which
makes general inferences relatively difficult. For example, Seng-
upta (1998) analyzes the association between the firm’s corporate
disclosure quality and the cost of its debt, and finds it to be nega-
tively related to disclosure quality, especially for those firms exhib-
iting higher variances in stock returns (as a proxy for market
uncertainty, and hence the firm’s default risk).
Anderson et al. (2003) observe that ownership concentration in
the hands of the founding family is negatively associated to the
agency cost of debt. In fact, family controlled firms enjoy a lower
cost of debt financing of about 32 basis points than non-family
firms. With their undiversified holdings, families seem to be more
concerned with the firm’s survival, in order to pass it onto subse-
quent generations. As a result, they tend to avoid risky projects
which alleviates the possibilities to expropriate bondholders,
hence the lower observed bond yield-spreads.3
Bhojraj and Sengupta (2003) document lower bond yields and
higher ratings for firms with higher institutional ownership and a
large proportion of outside directors. More recently, Ashbaugh
et al. (2006) report that bond ratings are positively affected by
the quality of financial transparency and by board independence,
ownership and expertise. Moreover, ratings are negatively related
to shareholder rights, to the CEO being also Chief of the board,
and to ownership concentration.
Other recent studies that examine the impact of shareholder
control on bondholders are Cremers et al. (2007) and Klock et al.
(2005). Cremers et al. (2007) document that shareholders’ control
is associated with lower yields for well protected firms (against
takeovers) while Klock et al. (2005) extend Gompers et al. (2003)
to analyze the impact of anti-takeover provisions on bond yield-
spreads. Their analysis shows that anti-takeover governance provi-
sions, while (not) beneficial to (shareholders) managers, are priced
by debtholders. They find that firms with strong anti-takeover pro-
visions are associated with a lower cost of debt financing of about
34 basis points relative to firms with weak anti-takeover
provisions.4
This literature review on the effect of corporate governance on
bond yields and ratings is, however, country specific since all stud-
ies analyze US data. Furthermore, these studies restrict their anal-
ysis to a variable number of corporate governance mechanisms.
Some of them (e.g. Billett et al. (2004)) use an event-study frame-
3 5
For a study on diversification in the banking industry and its impact on the cost of
debt, please refer to Deng et al. (2007).
4
Some other studies analyze the impact on bondholders’ wealth around some
major corporate events that directly affect the firm’s governance structure such as
Leverage Buyouts (e.g. Lehn and Poulsen, 1988; Marais et al., 1989; Warga and Welch,
1993), equity offering (Elliott et al., 2009), CEO turnover (Adams and Mansi, 2009),
and Mergers & Acquisitions (e.g., Dennis and McConnell, 1986; Maquieira et al., 1998;
Billett et al., 2004).
2483
work to compute abnormal returns (yields) bearing the risk that
the main conclusions may be contaminated by other events. All
these factors prevent us from generalizing the available evidence
to countries outside the US.
In this study, we use a multinational sample to explore this is-
sue and proxy for the quality of corporate governance by the like-
lihood of expropriation by different types of ultimate controlling
shareholders. We thus use control-based rather than ownership-
based measures (unlike Anderson et al. (2003) and Bhojraj and
Sengupta (2003)). These arguments lead to the formulation of the
following hypothesis:
H1: The extent of expropriation is positively (negatively) related to
bond yield spreads (ratings)
2.3. Creditor rights protection, bond yield-spreads and ratings
Debtholders’ rights are remarkably heterogeneous around the
world, particularly between developing and developed countries.5
The first formal work that examines investors’ protection in an inter-
national framework is by La Porta et al. (1998). In addition to share-
holders’ rights, the authors explore the extent to which debtholders
(and more generally creditors) are protected in 49 developed and
emerging countries, and observe that Common-law countries offer
a stronger legal protection for debtholders compared to French-ci-
vil-law countries. More recently, Djankov et al. (2007) document a
significant relation between the development of private credit
(claims on the private sector by banks) and legal creditor rights
(measured by either the creditor rights index or by the existence
of public and private registries). Well protected debt claimants
would normally require lower interest rates.
Miller and Puthenpurackal (2002) examine the impact of legal
protection and the information disclosure on the cost of public
yankee bonds issued by non-US firms and find that better investor
protection and more detailed information disclosure reduce the
costs of these bonds. The authors use the creditor rights and the
rule of law indices developed by La Porta et al. (1998) to proxy
for the level of investors’ protection. In our study, however, and
in addition to controlling for the expropriation risk (internal gover-
nance), we rely on a large set of institutional indicators that are
shown to improve investors’ rights protection (La Porta et al.,
1998; Djankov et al., 2008; Dyck and Zingales, 2004).
Finally, a recent study by Ellul et al. (2005) reports that family
firms benefit from lower bond yield-spreads in countries where
investor protection is better. This evidence is based on a sample
of international firms that issued ADRs in the US market, which
could eventually lead, as discussed previously, to a selection bias
since these firms must normally comply with corporate gover-
nance standards imposed by the SEC.
Based on this literature, we expect that cross-country differ-
ences in the level of debtholders’ protection will lead (1) debthold-
ers to require higher (risk) premia (i.e. higher yields), and (2) rating
agencies to assign lower bond ratings to firms from countries
where debtholders are the least protected. Thus, we can formulate
our second testable hypothesis as follows:
H2: A better creditor rights protection is negatively (positively)
related to bond yield-spreads (ratings)
In some countries (such as Greece), laws that prevent automatic liquidation and
encourage the reorganization of the defaulting company actually protect non-senior
debt claimants against senior debt claimants, while in some other countries, senior
debt claimants have the prerogative to hold up the collateral even in the case of
reorganization. Further, in some countries (e.g. France), laws favour reorganization
plans in order to preserve jobs, while in others, like the UK for instance, laws focus
primarily on the enforcement of financial contracts between management and
creditors, and allow these latter to choose between liquidation and reorganization.
2484 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

To establish the relation between investor protection and debt Table 1

Data distribution. This table provides a description of the frequency distribution of
costs, we consider two aspects: laws on the books and the extent issued bonds. Panel A presents distribution by years. Panel B presents distribution by
of its enforcement. Miller and Puthenpurackal (2002) show that country and region. Panel C presents distribution by the type of the bond, i.e. whether
bondholders value the existence of strong laws protecting creditor it is a Yankee Bond or not. Panel D presents distribution by the currency of
rights as well as the effective enforcement of these laws. La Porta denomination. Panel E presents distribution by the number of issues per firm.
et al. (2006) find that the existence of laws that facilitate private Number Percentage Cumulative %
contracting is a stronger determinant of financial markets develop- Panel A: Issues by year
ment than the public regulatory enforcement. Bhattacharya and 1994 4 1.30 1.30
Daouk (2002) document that the initial enactment of laws prohib- 1995 9 2.93 4.23
iting insider trading does not affect the cost of equity, while the 1996 14 4.56 8.79
1997 25 8.14 16.94
first time these laws are enforced reduces it by about 5%. Another
1998 26 8.47 25.41
study, more related to ours, by Esty and Megginson (2003) ana- 1999 35 11.40 36.81
lyzes the impact of creditor rights and legal enforcement on the 2000 43 14.01 50.81
debt ownership structure for a sample of global syndicated loans. 2001 80 26.06 76.87
Their results suggest that banks prefer larger and more diffuse syn- 2002 71 23.13 100

dicates’ structures to deter voluntary default in countries with poor Total 307 100
legal enforcement mechanisms. Based on this evidence, we control Panel B: Issues by country and region
for variables that reflect both the quality of debt laws and the ex- Austria 4 1.3 1.3
tent of their enforcement in the country. Finland 1 0.33 1.63
France 58 18.89 20.52
Germany 32 10.42 30.94
Indonesia 1 0.33 31.27
3. Methodology and descriptive statistics Ireland 3 0.98 32.25
Italy 3 0.98 33.22
3.1. Sample and data sources Japan 7 2.28 35.5
Malaysia 7 2.28 37.79
Norway 15 4.89 42.67
We first start by merging the samples with ultimate ownership The Philippines 16 5.21 47.88
compiled by Faccio and Lang (2002) for Western European corpo- Portugal 1 0.33 48.21
rations, and Claessens et al. (2000) for East Asian corporations. Singapore 6 1.95 50.16
Our analysis is cross-sectional in nature since ultimate ownership South Korea 20 6.51 56.68
Sweden 16 5.21 61.89
for European firms is for either one year between 1996 and 1999,
Thailand 7 2.28 64.17
and as of December 1996 for Asian firms. We consider all firms that UK 110 35.83 100
have complete information on voting and cash-flow divergence,
Europe 243 79.15 79.15
and on the identity of the ultimate owner. We obtain an initial Asia 64 20.85 100
database of more than 8000 firms from 22 countries. We then
Total 307 100
match this database with bond data from The Fixed Investment
Panel C: Yankee bonds
Securities Database. We keep only issues between 1994 and 2002
Yes 218 71.01 97.07
that offer fixed interest rates. Since ultimate ownership does not No 80 26.06 26.06
change substantially over time (Faccio and Lang, 2002; Claessens Missing information 9 2.93 100
et al., 2000), we believe that it remained relatively steady around Total 307 100
the years they were collected at. Assuming that bond issues be-
Panel D: Issues by currency denomination
tween 1994 and 2002 exhibit the ownership data collected by Fac- US dollar denomination 200 65.15 65.15
cio and Lang (2002) and Claessens et al. (2000), only 568 issues Non-US dollar denomination 91 29.64 94.79
survive the matching between these samples and the Fixed Invest- Missing information 16 5.21 100
ment Securities Database. Total 307 100
Data on issuer characteristics (performance, risk, size, and Panel E: Issue frequency
leverage) are from Worldscope. Regulatory institutions are from Firms with 1 issue 88 60.69 60.69
La Porta et al. (1998) (Creditor rights and Corruption), Djankov Firms with 2 issues 21 14.48 75.17
Firms with 3 issues 17 11.72 86.90
et al. (2008) (Estimated cost of the insolvency proceeding, Effi-
Firms with more than 3 issues 19 13.11 100
ciency of the bankruptcy process), Djankov et al. (2007) (Public
Total 145 100
registries and Contract enforcement days), and Dyck and Zingales
(2004)) (Newspaper circulation). Hand-matching the data on
bonds and ultimate ownership with the data on issuer characteris-
tics and regulatory institutions results in a final sample of 307 is-
firms (61%) issued only one bond between 1994 and 2002. However,
sues. Unfortunately, we are not able to compute the bond
around 14% of the firms issued two bonds, 12% issued three bonds
spreads for 51 issues which reduces the number of observations
and 13% issued more than three bonds.
for the SPREAD model to 256. Table 1 reports the distribution
and the description of our sample. Around 21% of our sample con-
sists of bonds issued by Asian firms. Yankee bonds, i.e. bonds is- 3.2. Specifications
sued by foreign firms in the US markets, represent 71% of the
sample (218 bonds).6 Our sample includes no Eurobonds. An addi- To test the relation between governance mechanisms and bond
tional 3% of the total sample is dropped as we lack information on yields and ratings, we use the two following general specifications:
the bond type. Finally, Panel E of the same table shows that the
307 bonds were issued by 145 different firms. The majority of these (1) Bond spreads = f (ultimate ownership variables, creditor pro-
tection variables, control variables).
6
For a complete review of the Yankee bond market, see Miller and Puthenpurackal (2) Bond ratings = f ((ultimate ownership variables, creditor
(2002). protection variables, control variables).
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
The Bond spread model is estimated using OLS while the Rating
model is estimated using an ordered probit model, given that the
dependent variable is ordinal (S&P ratings are classified in seven
ordering categories) (see Appendix B).
Our models include three major potential groups of determi-
nants of bond yields and ratings (i.e., ultimate ownership variables,
creditor rights protection variables, and control variables).7
3.3. Variables
We measure corporate bond spreads (SPREAD) by subtracting
the yield to maturity on a US treasury bond from the yield to matu-
rity on the corporate bond issue. The US treasury and the corporate
bonds should (ideally) have similar maturities and coupon.8
As credit rating measures, we use the S&P credit ratings (RAT-
ING) that assess the creditworthiness of the obligor with respect
to its debt obligations. There are 22 ratings ranging from highest
(AAA) to lowest (D). We follow Ashbaugh et al. (2006) and trans-
form these ratings into seven ordering numerical categories as pre-
sented in Appendix A.
3.3.1. Ultimate ownership and the likelihood of expropriation
Conflicts between minority and controlling shareholders (and
affiliated managers) could lead to expropriation practices by these
latter, such as not paying dividends, pursuing non-profit-maximiz-
ing strategies, transferring cash to other firms in which they have
interests. As in Claessens et al. (2000), we first measure the likeli-
hood (and extent) of expropriation by the wedge between the vot-
ing and cash-flow rights of the largest shareholders (C_O).
Previous studies suggest that the identity of the firms’ (ulti-
mate) owners is of interest for both bondholders and rating agen-
cies, as control concentration in the hands of few shareholders
could enhance (or reduce) expropriation. To test whether the con-
trolling shareholder’s identity affects bonds yield-spreads and rat-
ings, we introduce the dummy variables FAMILY, STATE, and
WHELDFIN, to capture the type of the ultimate owner (respectively
for family, State, and widely held financial firms). Finally, we add
another dummy variable, MANAGER, to determine whether the
appointment of a manager among family relatives has an impact
on the bondholders/rating agencies perception of risk factors.
3.3.2. Creditor rights protection
The role of regulatory institutions in preserving investor rights
depends on two factors: the existence of the rights per se and
the quality of their legal enforcement. According to Esty and Megg-
inson (2003, p. 41): ‘‘In addition to ensuring they have legal rights,
creditors must also ensure their rights are enforceable in the host
7
In order to tackle the presence of multicollinearity between our variables, we
proceed as follows. First, we test for multicollinearity in each regression we run. To do
so, we use the regression collinearity diagnostic procedures of Belsley et al. (1980)
that examine the ‘‘conditioning” of the matrix of explanatory variables. This
procedure consists in the computation of the condition number (the largest singular
value of the matrix). Belsley et al. (1980) suggest that a value of 30 (or higher) implies
collinearity problems. Second, and for each regression with a condition number of 30
or more, we use the Gram–Schmidt orthogonalization technique, which produces a
set of new orthogonal variables from the original ones. Each new variable is created in
such a way that the effects of the other variables are removed. By using these
orthogonal variables in our regression, we ensure that we are measuring the ‘‘right
effect” of each variable.
8
An alternative measure could be the use of a domestic treasury bond for each
country instead of US Treasury Bonds. We prefer using the US Treasury Bonds for
essentially two reasons. Firstly, the computation of spreads and risk premium is
usually done with reference to a risk free security. Since the US government is
unlikely to go bankrupt, we use its bonds as a risk free security. Secondly and most
importantly, the use of the US treasury bonds makes the comparison across countries
easier since it provides a common basis (or reference) to compare to. In spite of this,
we conduct our analyses using the available data on domestic treasury bonds for each
country. As discussed later in the text, our main results remain overall unchanged.
2485
country.” Following La Porta et al. (1998), Dyck and Zingales
(2004), Djankov et al. (2007, 2008), we choose the following indi-
cators to assess the quality of investor (particularly creditors) pro-
tection in a given country:
(i) Creditor rights (CREDRIGHTS): This index assesses the extent
of creditor rights in the country. It ranges from 0 (poor cred-
itor protection) to 4 (strong creditor protection). We expect
this index to be negatively (positively) associated with bond
yields (ratings).
(ii) Public registry (PUBREGIS): Public credit registries are dat-
abases managed by governments (e.g. through Central Banks
or any other public agency). Their main function is to pro-
vide lenders with information (that they have already col-
lected) on borrowers. The existence of a public credit
registry should negatively (positively) affect bond yield-
spreads (ratings).
(iii) Estimated cost of insolvency proceedings (COST_INSLV): It con-
sists on all kinds of costs borne by all parties as a percentage
of the value of the insolvency estate (e.g., court and attorney
fees, bankruptcy administrator fees, accountant fees, and
publication fees). The cost of insolvency should have a posi-
tive effect on bond yield-spreads and a negative effect on
bond ratings.
(iv) Efficiency of the bankruptcy process (EFFDBTENFORC)9: It is
measured by the present value of the terminal value of the
firm after bankruptcy costs. It reflects the value preserved in
debt enforcement proceedings: higher values indicate higher
efficiency of debt enforcement. Thus, we expect this variable
to be negatively (positively) related to debt yields (ratings).
(v) Contract enforcement days (ENFORCDYS): It is the number of
days needed to resolve a payment dispute (through courts)
that emerges from a simple debt contract, as a measure of
the quality of law enforcement. This variable should nor-
mally be positively (negatively) related to bond yield-
spreads (ratings).
(vi) Newspaper circulation (NEWS): It is the ratio of daily newspa-
pers divided by population. It reflects the public pressure on
dominant shareholders. Since it is expected to reduce expro-
priation, it should also be negatively (positively) related to
bond yield-spreads (ratings).
(vii) Corruption (CORRUPTION): It is an assessment of corruption
in the government. Since lower scores are for higher levels
of corruption, we expect this variable to have a negative
(positive) effect on bond yield-spreads (ratings).
Note that the first variable, creditor rights index, indicates the
quality of the laws that exist in the country, while the other vari-
ables (specifically, Public registry, Estimated cost of the insolvency
proceeding, Efficiency of the bankruptcy process, and Contract
enforcement days) measure the quality of debt laws’ enforcement.
3.3.3. Control variables
Typically, bond yield-spreads and ratings can be explained by
three factors: (i) the issuer characteristics that allow lenders and
underwriters to perceive the likelihood of default of the firm
(e.g., leverage, firm size, firm profitability, industry, market risk);
(ii) the issue characteristics that include the maturity of the debt,
the size of the issue, and some other special features such as the
debt seniority, the existence of call provisions and the existence
of a sinking fund; and (iii) the country macroeconomic conditions
such as the business cycle of the economy.
9
For more details on the computation of this variable, see Djankov et al. (2007).
2486 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
We include the following control variables: At the firm level, we
control for the issue characteristics (maturity, and issue size),10 and
for the issuer characteristics (firm size, leverage, performance, indus-
try, and risk). At the country level, we control essentially for the
country inflation rate, and the level of economic and financial devel-
opment (debt market size, and GDP growth). A detailed description
of the variables appears in Appendix A.
The Spread model can thus be written as follows:
SPREADi ¼ a0 þ a1  LMAT i þ a2  LISIZEi þ a3  ASSET i þ a4  STDINC i
þ a5  ROIi þ a6  LEVERAGEi þ a7  INFLATIONi
þ a8  GDPGROTHi þ a9  DEBTMKTSIZEi þ b1  C Oi
þ b2  FAMILY i þ b3  MANAGERi þ b4  STATEi
þ b5  WHELDFINi þ d1  CREDRIGHTSi þ d2  PUBREGISi
þ d3  COST INSLV i þ d4  EFFDBTENFORC i
þ d5  ENFORCDYSi þ d6  NEWSi þ d7  CORRUPTION i
þ Country Dummies þ Industry Dummies
þ Year Dummies þ ei :
And the Rating model as follows:
PrðRATINGi ¼ rÞ ¼ Uða1  LMAT i þ a2  LISIZEi þ a3  ASSET i
þ a4  STDINC i þ a5  ROIi þ a6  LEVERAGEi
þ a7  INFLATIONi þ a8  GDPGROTHi
þ a9  DEBTMKTSIZEi þ b1  C Oi þ b2  FAMILY i
þ b3  MANAGERi þ b4  STATEi þ b5  WHELDFIN i
þ d1  CREDRIGHTSi þ d2  PUBREGISi
þ d3  COST INSLV i þ d4  EFFDBTENFORC i
þ d5  ENFORCDYSi þ d6  NEWSi
þ d7  CORRUPTIONi þ Country Dummies
þ Industry Dummies þ Year Dummies þ ei Þ
where r 2 f1; 2; 3; 4; 5; 6; 7g:
3.4. Descriptive statistics
Table 2 reports descriptive statistics of the variables used in the
analysis. In our sample, the average spread is about 150.4 bps,
while the average rating is 4.63, which falls between A- and
BBB+ in our transformation scale. Moreover, the mean for the var-
iable C_O is 3.6%. Around 28% of the firms are controlled by fami-
lies. This proportion is about 19% and 20% for State and widely held
financial firms, respectively. Table 2 also presents Pearson correla-
tions between our two key variables (SPREAD and RATING), and all
the other potential explanatory variables. In general (except for
some variables), SPREAD is positively and significantly correlated
while RATING is negatively and significantly correlated with ulti-
mate ownership variables. Furthermore, both variables are signifi-
cantly correlated with most country variables, and generally in the
expected directions. We can note at this stage the high and signif-
icant correlations between RATING and most other variables. Since
these variables will be introduced in our SPREAD regressions, we
choose (as in Sengupta (1998) and Bhojraj and Sengupta (2003))
to exclude the RATING variable from those models.11
10
In our sample, the absolute majority (more than 98%) of issues are not convertible
and do not have any other special characteristic that could affect their yields or
ratings.
11
Including RATING in the SPREAD models does not affect our results as we will
show later on.
4. Empirical results
4.1. The impact of ultimate ownership on bond spreads and ratings
We first examine the effect of ultimate ownership structure on
bond yield-spreads., and report the results in Table 3. The statisti-
cal significances of the reported coefficients are calculated using
robust standard errors (Huber–White–Sandwich estimator of
variance).
In column (2) of Table 3, we report the basic model in which we
do not control for corporate governance. Most variables related to
issuer and issue characteristics are significant (at 9% or better) with
the expected signs. As expected, bond yield-spreads are negatively
affected by the size of the debt market (DEBTMKTSIZE) and the le-
vel of economic growth (GDPGROWTH). This finding suggests that
investors are sensitive to the size of secondary markets (where
they can liquidate their assets), and to the economic situation of
the country as a whole since it reflects, to some extent, the future
potential of their investments. The existence of a call provision in-
creases bond spreads while the seniority and the existence of a
sinking fund have no statistical effects (they exhibit the expected
negative signs though). Finally, the coefficient of PRUS, (i.e. the fact
the firm has accessed the US market) has the expected positive sign
although it is not significant. In columns (2)–(6) of the same table,
we add the ultimate ownership measures, each one in a separate
model. These regressions show that family control increases
spreads. On the contrary, State control and widely held financial
firms’ control have the opposite effect. The level of expropriation,
despite its positive expected sign, is not statistically significant.
In model (7), we include all the governance variables as well as
an interaction term between C_O and the identity of the ultimate
owner (FAMILY, STATE, or WHELDFIN). Results show that the
expropriation measure becomes significant and positive at less
than 6% level. That is, an increase of 1% in the level of expropriation
leads to an increase in the bond yield-spreads by approximately
12.3 basis points beyond the Treasury bond spread. Among ulti-
mate owners, only FAMILY is viewed as a potential risk of expropri-
ation by bondholders. This result goes against evidence in
Anderson et al. (2003) who find that US family firms have lower
spreads than non-family US firms. This could be due to the fact
that, in their study, the authors use an ownership-based measure
rather than a control-based measure of family holdings. Finally,
as shown by the coefficients of the interaction terms, when the
ultimate owner is a family or a widely held financial firm whose
control stake exceeds the ownership stake, the effect on bond
spreads is more significant. Furthermore, bondholders do not seem
to consider State control as a risk factor, even if the State has voting
rights in excess of its cash-flow rights (the coefficient of STA-
TExC_O being insignificant).
Faccio et al. (2001) argue that dividends can be used by control-
ling shareholders to expropriate outside shareholders’ wealth. The
authors suggest that if investors are aware of the risk of expropri-
ation, they will use dividends to protect themselves. Using West
European countries as a benchmark, the authors show that inves-
tors in East Asia are less alert to expropriation. Investors in West
European countries, in contrast to their counterparts in East Asia,
anticipate more strongly the risk of expropriation and, to offset
their concerns, require higher dividends in firms where expropria-
tion is more likely. If this is indeed the case, we should expect
bondholders in Western Europe to require higher interest rates
(i.e., bond spreads), compared to bondholders in East Asia, when-
ever there is a significant separation between the ownership and
control stakes held by controlling shareholders. Put differently,
we expect the (positive) effect of the cash-flow/voting rights diver-
gence on spreads to be more prevalent in the West European sam-
ple. To test this hypothesis, we estimate model (7) for both West
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499 2487

Table 2
Summary statistics. This table reports summary statistics. SPREAD: Yield to maturity on the bond issues minus the yield to maturity on a US treasury bond of similar maturity and
coupon (in basis points). RATING: Ordinal variable taking on value from 1 to 7 that represent the S&P bond ratings. For more details on the transformation procedure, see
Appendix B. LMAT: The logarithm of the years to maturity. LISIZE: The logarithm of the size (offering amount) of the issue (in US $1000). ASSET: The annual total assets for the
year preceding the bond issue, or last available (in US $1000). STDNINC: The operational risk as measured by the standard deviation of the net annual incomes for the five years
before the bond issue, or last available. ROI: The Return On Investments as of the year before the bond issue, or last available. LEVERAGE: The ratio of total debts to total assets for
the year preceding the bond issue, or last available. CALL: A dummy variable equals to 1 if there is a call provision. SENIOR: A dummy variable equals to 1 if there is a seniority
provision. SINK: A dummy variable equals to 1 if the issue contains a sinking fund provision. PRUS: A dummy variable equals to 1 if the firm has accessed public US security
market in the past (cross-listing or debt issuance). INFLATION: Annual percentage changes of the country Consumer Price Index. DEBTMKTSIZE: It is the ratio of the sum of bank
debt of private sector and outstanding non-financial bonds to GNP. GDPGROWTH: Average annual growth of per capita GDP. C_O: A measure of the likelihood of expropriation by
excess control. It is the difference between voting and cash-flow rights of the largest shareholders at a 10% level. FAMILY: A dummy variable equals to 1 if the controlling
shareholder is a family. MANAGER: A dummy variable equals to 1 if a member of the controlling family is a CEO, Honorary Chairman, Chairman, or Vice-Chairman. STATE: A
dummy variable equals to 1 if the controlling shareholder is a State. WHELDFIN: A dummy variable equals to 1 if the controlling shareholder is a widely held financial firm.
CREDRIGHTS: An index reflecting creditor rights that ranges from 0 to 4 (La Porta et al., 1998). PUBREGIS: A dummy variable equals 1 if a public credit registry operates in the
country, 0 otherwise. COST_INSLV: The estimated cost of the bankruptcy proceeding for a firm. It is the ratio of all kinds of costs (authority costs, accountant fees, inspector fees,
bankruptcy administrator fees, . . .) to the value of the insolvency estate. EFFDBTENFORC: It is the present value of the terminal value of the firm after bankruptcy costs. It reflects
the value preserved in debt enforcement proceedings. ENFORCDYS: The number of days to resolve a payment dispute through courts. NEWS: Circulation of daily newspapers
divided by population. CORRUPTION: An index (ranges from 0 to 10) that indicates the level of corruption in the government. Low ratings indicate higher corruption levels. The
significance levels of correlations with SPREAD and RATING are given into parentheses.

Variable Mean Std. dev. Min Max Pearson correlation with SPREAD Pearson correlation with RATING
SPREAD 150.403 125.848 0 816 – 0.527
(0.000)
RATING 4.637 1.351 1 7 0.527 –
(0.000)
LMAT 9.192 7.814 0.961 99.997 0.065 0.130
(0.299) (0.022)
CALL 0.248 0.432 0 1 0.513 0.410
(0.000) (0.000)
SENIOR 0.957 0.201 0 1 0.118 0.052
(0.095) (0.260)
SINK 0.352 0.059 0 1 0.082 0.079
(0.189) (0.085)
PRUS 0.257 0.444 0 1 0.155 0.169
(0.013) (0.000)
LISIZE 12.458 1.154 6.910 18.420 0.208 0.171
(0.000) (0.002)
ASSET (in US $ millions) 108 186 0.058 754 0.15 0.271
(0.0167) (0.000)
STDNINC (in US $ millions) 0.419 0.930 0.000 5.843 0.007 0.027
(0.903) (0.633)
ROI 7.49 20.576 43.89 252.45 0.082 0.006
(0.187) (0.906)
LEVERAGE 42.788 23.941 0 206.38 0.213 0.016
(0.000) (0.773)
C_O 3.601 8.493 0 54.36 0.139 0.107
(0.025) (0.061)
FAMILY 0.278 0.448 0 1 0.231 0.435
(0.000) (0.000)
MANAGER 0.028 0.165 0 1 0.084 0.157
(0.180) (0.005)
STATE 0.188 0.391 0 1 0.137 0.162
(0.005) (0.004)
WHELDFIN 0.197 0.398 0 1 0.072 0.253
(0.247) (0.000)
INFLATION 2.226 1.446 1.583 8.844 0.15 0 0.372
(0.016) (0.000)
GDPGROWTH 3.132 2.211 0.3 11.56 0.204 0.010
(0.001) (0.849)
DEBTMKTSIZE 0.947 0.259 0.1 1.22 0.211 0.235
(0.000) (0.000)
CREDRIGHTS 2.625 1.388 0 4 0.097 0.059
(0.122) (0.302)
PUBREGIS 0.412 0.492 0 1 0.159 0.277
(0.010) (0.000)
COST_INSLV 9.221 8.167 1 38 0.271 0.245
(0.000) (0.000)
EFFDBTENFORC 72.918 20.219 17.5 96.1 0.157 0.070
(0.011) (0.218)
ENFORCDYS 224.482 156.504 60 1.390 0.111 0.159
(0.075) (0.005)
NEWS 3.099 1.284 0 8 0.166 0.223
(0.007) (0.000)
CORRUPTION 8.5 1.627 2 10 0.210 0.310
(0.000) (0.000)

European and East Asian sub-samples. The last two columns of Ta- Asian and European countries anticipate expropriation by control-
ble 3 display the results that show that bondholders in both East ling shareholders (C_O is significant). Moreover, it seems that they
2488 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

Table 3
Ultimate ownership structure and bond yield-spreads. This table reports the OLS regression results for the bond yield-spreads on the ultimate ownership structure and firm- and
issue- control variables. SPREAD: Yield to maturity on the bond issues minus the yield to maturity on a US treasury bond of similar maturity and coupon (in basis points). LMAT:
The logarithm of the years to maturity. LISIZE: The logarithm of the size (offering amount) of the issue (in US $1000). ASSET: The annual total assets for the year preceding the
bond issue, or last available (in US $1000). STDNINC: The operational risk as measured by the standard deviation of the net annual incomes for the five years before the bond issue,
or last available. ROI: The Return On Investments as of the year before the bond issue, or last available. LEVERAGE: The ratio of total debts to total assets for the year preceding the
bond issue, or last available. CALL: A dummy variable equals to 1 if there is a call provision. SENIOR: A dummy variable equals to 1 if there is a seniority provision. SINK: A dummy
variable equals to 1 if the issue contains a sinking fund provision. PRUS: A dummy variable equals to 1 if the firm has accessed public US security market in the past (cross-listing
or debt issuance). INFLATION: Annual percentage changes of the country Consumer Price Index. DEBTMKTSIZE: It is the ratio of the sum of bank debt of private sector and
outstanding non-financial bonds to GNP. GDPGROWTH: Average annual growth of per capita GDP. C_O: A measure of the likelihood of expropriation by excess control. It is the
difference between voting and cash-flow rights of the largest shareholders at a 10% level. FAMILY: A dummy variable equals to 1 if the controlling shareholder is a family.
MANAGER: A dummy variable equals to 1 if a member of the controlling family is a CEO, Honorary Chairman, Chairman, or Vice-Chairman. STATE: A dummy variable equals to 1 if
the controlling shareholder is a State. WHELDFIN: A dummy variable equals to 1 if the controlling shareholder is a Widely held financial firm.

Dependent variable: SPREAD (1) (2) (3) (4) (5) (6) (7) East Asian West European
Constant 166.350 159.463 147.708 163.544 176.941 173.687 162.953 168.405 159.203
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
LMAT 13.841 15.895 15.440 14.338 15.885 11.545 19.295 14.814 23.401
(0.021)** (0.010)*** (0.005)*** (0.015)** (0.007)*** (0.050)** (0.001)*** (0.120) (0.002)***
LISIZE 20.850 19.952 16.609 19.670 15.990 24.265 18.556 21.950 17.981
(0.001)*** (0.002)*** (0.007)*** (0.002)*** (0.017)** (0.000)*** (0.003)*** (0.086)* (0.007)***
ASSET 15.465 14.751 10.771 14.950 14.944 12.908 18.951 14.149 23.053
(0.001)*** (0.002)*** (0.018)** (0.001)*** (0.000)*** (0.011)** (0.000)*** (0.090)* (0.000)***
STDINC 16.691 18.702 18.304 17.177 19.636 20.082 17.737 3.378 20.136
(0.008)*** (0.004)*** (0.003)*** (0.006)*** (0.005)*** (0.001)*** (0.002)*** (0.786) (0.000)***
ROI 12.821 11.138 14.296 12.408 14.108 13.110 12.487 19.493 14.941
(0.007)*** (0.020)** (0.002)*** (0.011)** (0.003)*** (0.005)*** (0.011)** (0.381) (0.031)**
CALL 56.995 55.412 51.464 56.585 52.713 55.636 57.424 51.561 64.338
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
SENIOR 6.327 5.223 4.171 5.891 3.739 5.428 6.287 46.373 a
(0.217) (0.323) (0.355) (0.252) (0.424) (0.407) (0.259) (0.001)*** –a
SINK 4.698 5.042 4.689 4.918 6.028 4.160 4.703 6.814 9.687
(0.298) (0.261) (0.268) (0.284) (0.201) (0.365) (0.330) (0.039)** (0.000)***
PRUS 1.322 2.230 2.713 3.197 1.083 1.223 5.020 16.456 0.588
(0.831) (0.735) (0.663) (0.589) (0.863) (0.844) (0.425) (0.011)** (0.936)
LEVERAGE 18.218 17.194 15.284 18.352 19.777 15.838 14.444 14.788 14.092
(0.007)*** (0.013)** (0.015)** (0.006)*** (0.003)*** (0.020)** (0.024)** (0.208) (0.050)*
INFLATION 11.901 13.675 13.569 12.957 12.946 15.677 14.569 21.036 1.859
(0.037)** (0.023)** (0.012)** (0.022)** (0.022)** (0.009)*** (0.010)*** (0.042)** (0.804)
GDPGROWTH 19.984 19.220 19.813 19.308 13.797 25.560 19.387 40.079 20.775
(0.003)*** (0.004)*** (0.001)*** (0.004)*** (0.058)* (0.000)*** (0.002)*** (0.000)*** (0.000)***
DEBTMKTSIZE 10.151 8.541 13.729 10.339 14.714 10.734 11.449 22.423 1.801
(0.088)* (0.172) (0.022)** (0.083)* (0.022)** (0.067)* (0.049)** (0.019)** (0.744)
C_O 1.699 12.284 10.102 19.104
(0.127) (0.054)* (0.049)** (0.005)***
FAMILY 50.887 25.475 26.292 21.812
(0.001)*** (0.000)*** (0.005)*** (0.000)***
MANAGER 44.247 2.057 16.926 0.298
(0.147) (0.760) (0.003)*** (0.966)
STATE 46.590 4.834 4.730 1.245
(0.014)** (0.349) (0.738) (0.800)
WHELDFIN 53.584 8.223 4.193 18.884
(0.005)*** (0.164) (0.757) (0.003)***
FAMILYxC_O 14.042 10.258 11.700
(0.012)** (0.411) (0.060)*
STATExC_O 2.722 a 11.134
(0.584) a (0.019)**
WHELDFINxC_O 9.052 a 15.055
(0.048)** a (0.008)***
Country dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes
N 256 256 256 256 256 256 256 58 198
Adj. R2 48.56 48.41 49.05 51.59 48.78 49.44 49.66 89.00 53.91
F 6.71 6.67 6.69 7.29 6.63 6.78 6.83 13.24 6.49
Sig. (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
a
Dropped because of lack of variability.
*
Significance at the 10% level.
**
Significance at the 5% level.
***
Significance at the 1% level.

fear family control, since the coefficient of FAMILY is positive and
highly significant in both models. However, only European inves-
tors seem to view control by widely held financial firms as a poten-
tial risk of expropriation. Overall, these results show, to some
extent, the type of investor that Asian and European bondholders
have most confidence in (since the presence of that particular type
of investor reduces bonds yield-spreads).
Table 4 reports the results of ordered probit models for the ef-
fect of expropriation measures on bond ratings. We keep all the
control variables previously discussed. The unique difference is
that instead of introducing dummies for years, countries and
industries, we choose to include dummies to control for three par-
ticular industries more likely to affect ratings (Finance, Utility and
High-Tech), for three periods (before, during and after the 1997
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
Asian financial crisis), and for two regions (Asia or Europe). We do
so because cross-tabulating the ordered dependent variable (RAT-
ING) with many dummy variables creates too many empty cells
that prevent an ordered probit estimation. We tackle this issue
by reducing the dummies instead of merging some categories of
the dependent variable.
Column (1) presents the baseline model without including any
proxies for governance. In columns (2)–(6), we test separately for
the effect of each measure of ultimate ownership on bond ratings.
At 10% level (or better), all the governance measures affect signif-
icantly bond ratings. The level of the cash-flow and voting rights
discrepancy as well as the existence of family control in the firm
and the appointment of managers among the owner’s family mem-
bers lead rating agencies to assign lower ratings for these firms (the
coefficients for C_O, FAMILY and MANAGER are negative). In con-
trast, State control and control by widely held financial firms do
not seem to be considered by rating agencies as additional poten-
tial risk factors of expropriation (because their presence as control-
ling shareholders increases bond ratings as shown by the
coefficients of STATE and WHELDFIN). When we consider all the
governance proxies in the same model (column (7)), we still obtain
the same results, except that STATE becomes insignificant while
still keeping its positive sign. The interaction coefficients between
C_O and both FAMILY and WHELDFIN are negative and highly sig-
nificant, suggesting that rating agencies have a ‘‘negative” percep-
tion of the cash-flow/voting rights discrepancy. Overall, these
findings shed some light on the effect of ultimate owner identity
on bond ratings. Previous studies have argued that the concentra-
tion of ownership in the hands of institutions (Bhojraj and Seng-
upta, 2003) or in the hands of blockholders (Ashbaugh et al.,
2006) could enhance the private benefits hypothesis, leading rating
agencies to downgrade the scores for the firms that exhibit these
features. Evidence from Table 4 goes beyond this evidence and
shows that controlling owners do not all have the same impact
on bond ratings.
At this stage, we can compare the perceptions of bondholders
and rating agencies regarding the firms’ ownership features. By
taking a glance at column (7) of Tables 3 and 4, we can easily note
that bondholders as well as rating agencies are able to assess the
potential risk of expropriation (the cash flow and voting rights dis-
crepancy). Both seem to consider the potential costs of family con-
trol (both FAMILY and C_OxFAMILY being economically and
statistically significant in both models). This risk can materialize
in two ways: First, families can take full advantage from their con-
trolling position, and seek to extract (direct) private benefits that
harm all stakeholders (especially minority shareholders and debt-
holders). Second, families often want to keep control by avoiding
the dilution of their ownership. Thus, family firms will normally
prefer debt to equity financing, which translates in relatively high-
er leverage ratios.
4.2. The impact of regulatory institutions on bond costs
We first assess the effect of regulatory institutions on bond
spreads. Models (1)–(7) in Table 5, Panel A, illustrate the effect of
each of our proxies on bond spreads. Among our set of institutions,
only creditor rights and the existence of public registry are insig-
nificant at the 10% level. All the other institutions are significant
with their expected signs. In the last column of Table 5, we regress
bond spreads on all regulatory factors. The conclusions remain the
same except that the existence of credit public registry becomes
significant once we take into account the effect of all other institu-
tions. Thus, the existence of a credit public registry, the efficiency
of the bankruptcy procedure, and the extent of newspapers’ circu-
lation affect, as expected, negatively and significantly, corporate
bond spreads. In contrast, higher insolvency costs and longer con-
2489
tract enforcement periods lead to significantly higher bond
spreads. Moreover, higher corruption (i.e. lower values of CORRUP-
TION) increases bond spreads. These results lead us to a first con-
clusion: the legal environment is important to bondholders. This
conclusion corroborates, to some extent, the finding of Ellul et al.
(2005) that the presence of a founding family in less protective le-
gal systems exacerbates the cost of debt (and vice versa). Their re-
sult indirectly points to the role of the legal environment in
encouraging (or forcing) founding families to preserve minority
interests. Our findings suggest that, regardless of who is controlling
the firm, legal regimes could protect debtholders and guarantee
their rights.
We also note from Table 5 that, contrary to the evidence on debt
laws’ enforcement measures discussed above, the existence of
restrictions aimed at protecting creditor rights has no impact on
bond yield-spreads. In fact, while variables related to the quality
of debt laws’ enforcement (i.e., PUBREGIS, COST_INSLV, EFFDBTEN-
FORC, and ENFORCDYS) are significant, the creditor rights index
(CREDRIGHTS) that captures the existence of these laws, is not sig-
nificant at any conventional level. That is, to bondholders, enforce-
ment of debt laws is more important than the mere existence of
these laws.
Panel B of Table 5 reports evidence on the effects of regulatory
institutions on bond ratings. As previously, we start by testing the
impact of each isolated factor, and then we test the effect of all fac-
tors taken together. From columns (1) to (7), we can see that the
existence of a credit public registry as well as an extensive news-
papers’ circulation and a lower corruption level in the country lead
to higher bond ratings. When we run our regression with all the
institutional governance factors in the same model, we find that
CREDRIGHTS, COST_INSLV, and CORRUPTION have no statistical ef-
fect on bond ratings. PUBREGIS, EFFDBTENFORC, ENFORCDYS, and
NEWS are, in contrast, highly significant (at less than 1% level) with
the expected signs. We can thus conclude that, to rating agencies,
the enforcement of debt laws is more important than the existence
of laws (CREDRIGHTS being insignificant). The final model leads to
a pseudo-R2 of 25.35%.
4.3. The determinants of debt costs
In order to construct our final models for both bond yield-
spreads and ratings, we propose to introduce both ultimate owner-
ship and creditor rights measures in the same regression. Results
are reported in Table 6.
Panel A of Table 6 reports the results for bond spreads. As we
can notice in column (1) of the same panel, we get mainly the same
results as in Table 3 (column (7)) regarding the ultimate ownership
measures. Spreads are positively affected by both the discrepancy
between cash flow and voting rights (C_O) and the existence of a
controlling family (FAMILY). The interaction between C_O and
FAMILY and WHELDFIN exacerbate that positive effect. Finally,
State control is not significant, and seems to have no effect on bond
spreads. As for regulatory institutions, we also obtain similar re-
sults as in the last column of Table 5, Panel A. Particularly, the cred-
itor rights index is not as important to debtholders as debt
enforcement (PUBREGIS, COST_INSLV, and EFFDBTENFORC) (these
measures being all statistically and economically significant while
CREDRIGHTS is not). Moreover, the public pressure on controlling
shareholders (as measured by NEWS) plays a positive role in
reducing bond spreads. The overall model explains more that
54% of corporate bond spreads.
Up to now, we assumed that our set of explanatory variables
(especially the ownership measures) is exogenous. Many previous
studies suggest that this assumption may not hold (see for exam-
ple Hermalin and Weisbach (2003) for the board of directors). End-
ogeneity may be caused by (i) omitted variables that are correlated
2490 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

Table 4
Ultimate ownership structure and bond ratings. This table presents the Ordered Probit regression results of bond ratings on ultimate ownership structure, as well as firm- and
issue- control variables. We introduced dummies to control for the year, country and industry effects in all models. RATING: Ordinal variable taking on value from 1 to 7 that
represent the S&P bond ratings. For more details on the transformation procedure, see Appendix B. LMAT: The logarithm of the years to maturity. LISIZE: The logarithm of the size
(offering amount) of the issue (in US $1000). ASSET: The annual total assets for the year preceding the bond issue, or last available (in US $1000). STDNINC: The operational risk as
measured by the standard deviation of the net annual incomes for the five years before the bond issue, or last available. ROI: The Return On Investments as of the year before the
bond issue, or last available. LEVERAGE: The ratio of total debts to total assets for the year preceding the bond issue, or last available. CALL: A dummy variable equals to 1 if there
is a call provision. SENIOR: A dummy variable equals to 1 if there is a seniority provision. SINK: A dummy variable equals to 1 if the issue contains a sinking fund provision. PRUS:
A dummy variable equals to 1 if the firm has accessed public US security market in the past (cross-listing or debt issuance). INFLATION: Annual percentage changes of the country
Consumer Price Index. DEBTMKTSIZE: It is the ratio of the sum of bank debt of private sector and outstanding non-financial bonds to GNP. GDPGROWTH: Average annual growth
of per capita GDP. C_O: A measure of the likelihood of expropriation by excess control. It is the difference between voting and cash-flow rights of the largest shareholders at a 10%
level. FAMILY: A dummy variable equals to 1 if the controlling shareholder is a family. MANAGER: A dummy variable equals to 1 if a member of the controlling family is a CEO,
Honorary Chairman, Chairman, or Vice-Chairman. STATE: A dummy variable equals to 1 if the controlling shareholder is a State. WHELDFIN: A dummy variable equals to 1 if the
controlling shareholder is a widely held financial firm.

Dependent variable: RATING (1) (2) (3) (4) (5) (6) (7)
LMAT 21.364 21.303 21.508 20.144 21.381 21.926 20.839
(0.000)*** (0.000)*** (0.000)*** (0.001)*** (0.000)*** (0.000)*** (0.001)***
LISIZE 23.065 23.207 24.504 24.594 23.799 25.095 29.012
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
ASSET 32.385 32.553 35.035 33.838 32.659 34.362 38.853
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
STDINC 14.209 14.158 14.994 15.010 14.286 15.149 16.157
(0.008)*** (0.010)*** (0.005)*** (0.006)*** (0.016)** (0.000)*** (0.000)***
ROI 1.092 1.010 0.397 0.958 1.192 0.713 0.093
(0.826) (0.839) (0.929) (0.841) (0.804) (0.890) (0.985)
CALL 35.180 35.330 35.622 34.914 35.390 35.313 37.096
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
SENIOR 11.246 11.220 12.831 11.964 11.192 12.350 14.326
(0.105) (0.110) (0.054)* (0.093)* (0.083)* (0.169) (0.081)*
SINK 5.843 5.882 6.579 6.189 5.845 6.197 7.292
(0.032)** (0.032)** (0.004)*** (0.046)** (0.046)** (0.027)** (0.060)*
PRUS 5.013 5.212 4.882 5.627 4.556 5.569 6.855
(0.343) (0.330) (0.385) (0.287) (0.393) (0.307) (0.243)
LEVERAGE 3.244 3.570 4.070 4.051 3.394 1.630 4.440
(0.612) (0.577) (0.515) (0.525) (0.600) (0.786) (0.450)
INFLATION 42.693 42.583 44.848 45.381 43.536 44.585 -48.857
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
GDPGROWTH 1.453 1.834 3.577 2.324 0.745 4.755 8.232
(0.771) (0.716) (0.490) (0.654) (0.883) (0.372) (0.157)
DEBTMKTSIZE 7.619 7.327 9.596 8.606 7.827 10.072 12.201
(0.207) (0.232) (0.111) (0.130) (0.193) (0.107) (0.063)*
FINANCE 28.025 28.972 25.009 33.684 30.263 24.083 26.857
(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
UTILITY 7.138 7.927 2.556 10.037 0.494 4.987 0.899
(0.223) (0.179) (0.682) (0.084)* (0.937) (0.397) (0.890)
HIGHTEC 28.854 27.771 25.979 30.833 26.309 29.529 24.764
(0.000)*** (0.000)*** (0.001)*** (0.000)*** (0.000)*** (0.000)*** (0.001)***
YRS9798 5.621 4.801 7.648 6.326 5.093 9.660 6.479
(0.384) (0.457) (0.256) (0.314) (0.432) (0.136) (0.335)
POSTCRISIS 7.528 7.720 7.721 9.260 5.741 9.017 5.890
(0.286) (0.277) (0.254) (0.183) (0.416) (0.230) (0.410)
EUROPE 12.700 12.649 12.333 14.535 13.054 15.869 16.469
(0.048)** (0.052)* (0.070)* (0.031)** (0.043)** (0.025)** (0.041)**
C_O 10.811 15.754
(0.099)* (0.025)**
FAMILY 48.508 23.909 53.274
(0.000)*** (0.000)*** (0.000)***
MANAGER 14.363
(0.028)**
STATE 15.663 3.657
(0.010)** (0.558)
WHELDFIN 37.990 32.660
(0.000)*** (0.000)***
FAMILYxC_O 29.299
(0.000)***
STATExC_O 6.224
(0.477)
WHELDFINxC_O 11.100
(0.002)***
N 307 307 307 307 307 307 307
Pseudo-R2 15.82 16.6 20.01 17.84 16.13 19.28 24.87
Model v2 210.84 206.55 207.42 199.47 234.98 168.19 183.99
Sig. (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
*
Significance at the 10% level. (Coefficients are multiplied by 100).
**
Significance at the 5% level. (Coefficients are multiplied by 100).
***
Significance at the 1% level. (Coefficients are multiplied by 100).
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499 2491

Table 5
Regulatory institutions and bond spreads and ratings. This table reports the OLS and Probit regression results of bond spreads and ratings on regulatory institutions, as well as on
firm- and issue- control variables (only the results for the regulatory institutions are reported). We introduced dummies to control for the year, country and industry effects in all
models. Control variables’ descriptions are presented in Table 1. SPREAD: Yield to maturity on the bond issues minus the yield to maturity on a US treasury bond of similar
maturity and coupon (in basis points). RATING: Ordinal variable taking on value from 1 to 7 that represent the S&P bond ratings. For more details on the transformation
procedure, see Appendix B. LMAT: The logarithm of the years to maturity. CREDRIGHTS: An index reflecting creditor rights that ranges from 0 to 4 (La Porta et al., 1998).
PUBREGIS: A dummy variable equals 1 if a public credit registry operates in the country, 0 otherwise. COST_INSLV: The estimated cost of the bankruptcy proceeding for a firm. It is
the ratio of all kinds of costs (authority costs, accountant fees, inspector fees, bankruptcy administrator fees, . . .) to the value of the insolvency estate. EFFDBTENFORC: It is the
present value of the terminal value of the firm after bankruptcy costs. It reflects the value preserved in debt enforcement proceedings. ENFORCDYS: The number of days to resolve
a payment dispute through courts. NEWS: Circulation of daily newspapers divided by population. CORRUPTION: An index (ranges from 0 to 10) that indicates the level of
corruption in the government. Low ratings indicate higher corruption levels.

Dependent variable: SPREADS (1) (2) (3) (4) (5) (6) (7) (8)
Panel A: Bonds spreads
CREDRIGHTS 6.955 7.257
(0.245) (0.237)
PUBREGIS 8.028 15.197
(0.152) (0.001)***
COST_INSLV 20.335 21.320
(0.002)*** (0.004)***
EFFDBTENFORC 10.117 12.877
(0.099)* (0.002)***
ENFORCDYS 7.004 5.966
(0.078)* (0.033)**
NEWS 24.893 22.198
(0.000)*** (0.000)***
CORRUPTION 31.835 10.165
(0.000)*** (0.032)**
N 256 256 256 256 256 254 256 254
Adj. R2 48.30 48.36 48.26 48.35 48.38 48.47 48.36 48.37
F 6.52 6.53 6.51 6.53 6.54 6.64 6.66 5.92
Sig. (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
Dependent variable: RATING (1) (2) (3) (4) (5) (6) (7) (8)
Panel B: Bonds ratings
CREDRIGHTS 8.725 6.293
(0.106) (0.267)
PUBREGIS 17.139 41.114
(0.006)*** (0.000)***
COST_INSLV 10.072 12.535
(0.216) (0.131)
EFFDBTENFORC 1.611 28.940
(0.797) (0.000)***
ENFORCDYS 6.937 32.345
(0.202) (0.000)***
NEWS 22.942 56.413
(0.000)*** (0.000)***
CORRUPTION 19.230 5.540
(0.002)*** (0.371)
N 307 307 307 307 307 304 307 304
Pseudo-R2 15.97 16.28 15.93 15.84 16.15 17.26 16.23 25.35
Model v2 249.111 206.28 230.00 222.70 215.52 255.80 262.65 337.23
Sig. (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
*
Significance at the 10% level. (Coefficients are multiplied by 100 for the bond rating model).
**
Significance at the 5% level. (Coefficients are multiplied by 100 for the bond rating model).
***
Significance at the 1% level. (Coefficients are multiplied by 100 for the bond rating model).

with some explanatory variables, (ii) by measurement error when
measuring variables, or (iii) by simultaneity between the depen-
dent variable and at least one independent variable. In our case,
the cash flow/voting rights discrepancy (C_O) could imperfectly
measure expropriation or eventually be correlated with an omitted
variable. Furthermore, FAMILY, STATE, and WHELDFIN, reflect
some features of the firms’ ownership structure that could be
simultaneously determined with bond costs (Bhojraj and Sengupta,
2003),12 or correlated with some omitted variables. To control for
this potential endogeneity, we use the two-stage least square
(2SLS) technique that relies on instrumental variables which allows
12
Bhojraj and Sengupta (2003) document that bond yields and institutional
ownership are simultaneously determined. That is, the monitoring of the institutions
reduces debt costs, and firms with lower debt costs are more attractive to these
investors.
to solve endogeneity even if the exact cause is unknown as in our
case. As instrumental variables, we propose to use the following
variables from La Porta et al. (1998); the proxy by mail possibility,
shareholders’ preemptive rights to buy new stock issues, antidirector
rights, and the legal origin. One can easily relate the proxy by mail,
antidirector rights, and the preemptive rights with ownership struc-
ture patterns. The decision for controlling shareholders to hold (or
not) stocks of firms is closely related to the existence of such provi-
sions in the company charter. For example, we expect that large
shareholders, who seek to keep the control, will avoid investing in
firms where minority shareholders have the valuable right of mail-
ing their proxy vote. As for the legal origin, La Porta et al. (1998) sug-
gest that ownership concentration could be viewed as a substitute
for poor investor protection. That is, the concentration of ownership
in the hands of a controlling shareholder (and even his choice to get
voting rights in excess of his capital stake) are more prevalent in
 2492
Table 6
The determinants of bond spreads and ratings. This table reports the OLS and Probit regression results of bond spreads and ratings, respectively, on regulatory institutions, as well as on firm- and issue- control variables (only the results
for the governance variables are reported). We introduced dummies to control for the year, country and industry effects in all models. Control variables’ descriptions are presented in Table 1. SPREAD: Yield to maturity on the bond issues
minus the yield to maturity on a US treasury bond of similar maturity and coupon (in basis points). RATING: Ordinal variable taking on value from 1 to 7 that represent the S&P bond ratings. For more details on the transformation
procedure, see Appendix B. LMAT: The logarithm of the years to maturity. CREDRIGHTS: An index reflecting creditor rights that ranges from 0 to 4 (La Porta et al., 1998). PUBREGIS: A dummy variable equals 1 if a public credit registry
operates in the country, 0 otherwise. COST_INSLV: The estimated cost of the bankruptcy proceeding for a firm. It is the ratio of all kinds of costs (authority costs, accountant fees, inspector fees, bankruptcy administrator fees, . . .) to the
value of the insolvency estate. EFFDBTENFORC: It is the present value of the terminal value of the firm after bankruptcy costs. It reflects the value preserved in debt enforcement proceedings. ENFORCDYS: The number of days to resolve a
payment dispute through courts. NEWS: Circulation of daily newspapers divided by population. CORRUPTION: An index (ranges from 0 to 10) that indicates the level of corruption in the government. Low ratings indicate higher
corruption levels.

Panel A Panel B
OLS estimates: dependent variable bonds spreads Ordered probit estimates: dependent variable bonds ratings
Expected sign (1) (2) (3) (4) (5) Heckman Model Expected sign (1) (2) Heckman Model
C_O + 17.344 14.574 5.753 17.739 16.274 – 13.454 0.119
(0.007)*** (0.028)** (0.286) (0.031)** (0.002)*** (0.019)** (0.007)***

N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

FAMILY ? 25.01 26.584 13.742 32.322 25.737 ? 50.780 0.336
(0.000)*** (0.000)*** (0.029)** (0.000)*** (0.000)*** (0.000)*** (0.000)***
MANAGER ? 1.689 11.482 14.078 4.855 1.895 ? 27.623 0.172
(0.785) (0.097)* (0.079)* (0.581) (0.719) (0.000)*** (0.000)***
STATE ? 5.222 2.857 3.117 4.397 5.390 ? 2.051 0.019
(0.316) (0.501) (0.622) (0.529) (0.227) (0.775) (0.659)
WHELDFIN ? 7.441 10.010 9.656 26.744 8.390 ? 40.613 0.255
(0.198) (0.074)* (0.091)* (0.003)*** (0.139) (0.000)*** (0.000)***
FAMILYxC_O ? 11.889 3.600 12.492 1.649 12.451 ? 22.346 0.198
(0.033)** (0.574) (0.012)** (0.833) (0.014)** (0.000)*** (0.000)***
STATExC_O ? 6.922 1.416 0.380 10.661 6.033 ? 17.338 0.098
(0.199) (0.795) (0.960) (0.068)* (0.179) (0.003)*** (0.015)**
WHELDFINxC_O ? 11.485 4.061 4.184 17.015 10.574 ? 9.327 0.055
(0.012)** (0.343) (0.516) (0.008)*** (0.007)*** (0.024)** (0.158)
CREDRIGHTS  5.379 0.887 5.576 10.169 4.962 + 6.090 0.022
(0.361) (0.865) (0.327) (0.162) (0.299) (0.394) (0.597)
PUBREGIS  20.104 18.962 10.371 22.886 19.603 + 16.859 0.071
(0.000)*** (0.000)*** (0.014)** (0.000)*** (0.000)*** (0.032)** (0.150)
COST_INSLV + 20.096 18.345 9.897 20.197 19.635  11.092 0.074
(0.003)*** (0.006)*** (0.070)* (0.013)** (0.001)*** (0.227) (0.145)
EFFDBTENFORC  10.294 10.944 17.903 17.930 10.980 + 22.559 0.112
(0.031)** (0.006)*** (0.027)** (0.018)** (0.005)*** (0.002)*** (0.024)**
ENFORCDYS + 4.797 10.052 2.126 0.529 4.831  52.511 0.301
(0.177) (0.019)** (0.611) (0.919) (0.122) (0.000)*** (0.000)***
NEWS  21.771 17.197 3.834 34.672 21.915 + 39.143 0.227
(0.000)*** (0.000)*** (0.561) (0.000)*** (0.000)*** (0.000)*** (0.000)***
CORRUPTION  7.350 9.487 8.273 8.184 7.818 + 17.424 0.119
(0.146) (0.069)* (0.153) (0.221) (0.063)* (0.030)** (0.009)***
N 254 254 120 218 q = .054 304 q = 0.068
v2 = 0.34a v2 = 0.05a
Sig. = 0.561b Sig. = 0.817b
Adj. R2 54.22 53.12 24.75 51.73 –
Pseudo-R2 – – – – 33.02
F 6.33 6.10 2.29 4.62 –
Model v2 – – – – 251.64
Sig. (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***
a
This is the v2 statistic to test the null that q = 0.
b
The significance level corresponding to the null that q = 0.
*
Significance at the 10% level. (Coefficients are multiplied by 100 for the debt rating model).
**
Significance at the 5% level. (Coefficients are multiplied by 100 for the debt rating model).
***
Significance at the 1% level. (Coefficients are multiplied by 100 for the debt rating model).
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
civil-law countries where investor rights are less protected. Finally,
these candidate instruments are less likely to affect corporate bond
yields (thus the model error terms).
Column (2) of Table 6 (Panel A) reports the results of the 2SLS
regression. As we can see, we obtain qualitatively the same results
as previously. All variables keep their significance levels and their
expected signs, except for MANAGER, WHELDFIN, ENFORCDYS, and
CORRUPTION that become slightly significant.
Finally, and as shown in Table 1, 91 bonds (30%) are in non-US
Dollar denomination, and 16 bonds (5%) do not have information
on currency denomination. For these bonds, using US treasury
bond yield to compute bond yield spread may be inappropriate.13
To check whether our results are robust, we run our final model
using the 218 bonds denominated in US Dollars. The results are re-
ported in Table 6 Column (4) (Panel A). The overall results are sim-
ilar to the ones reported earlier (model (1) of Table 6). The only
notable difference is that WHELDFIN and the interaction between
C_O and STATE become significant, while the interaction between
C_O and FAMILY loses its significance. In short, these results confirm
our previous evidence on the particular importance of debt
enforcement.
We next run our final model for bond ratings using both the
ultimate ownership structure and the regulatory institutions. As
reported in column (1) of Table 6, Panel B, conclusions drawn from
the previous analyses on the effect of ultimate ownership structure
and regulatory institutions on corporate bond ratings remain basi-
cally the same. Our expropriation proxy (C_O) is highly significant
and negatively related to bond ratings. Family control and the
appointment of the manager among the family relatives appear
to reduce rating scores, while control in the hands of widely held
financial firms seems to increase it. As for debt enforcement mea-
sures, they are still highly influential and keep the same signs as
previously documented, namely the existence of public registries,
the efficiency of the bankruptcy process, and the contract enforce-
ment days. The coefficient of the creditor rights index remains
insignificant at any conventional level which comforts our previ-
ous conclusion that rating agencies price debt laws enforcement
rather than the existence of laws. The overall model generates a
pseudo-R2 of about 33.02%.
4.4. Debt costs and the role of covenants14
Debt covenants are agreements between the firm and its cred-
itors that are designed to protect these latter from being expropri-
ated. Common covenants include restrictions on paying dividends,
requirements for a specified leverage ratio or working capital, and
limitations on asset sales. Failure to respect these covenants may
result in debt becoming immediately due.
Previous studies have revealed the importance of debt cove-
nants in corporate governance (see for example Asquith and Wiz-
man, 1990; DeAngelo et al., 2002; Cremers et al., 2007 among
others). Cremers et al. (2007) for example document that the exis-
tence of protective covenants reduces bondholder concerns about
shareholders misbehaviour. These studies point out that the pres-
ence of a blockholder has a significantly smaller effect on bond
spreads if bonds include covenant provisions. We propose to ex-
plore whether the presence of debt covenants affects our previous
findings. Following Cremers et al. (2007),15 we re-run our final
models while including interaction terms between governance
13
We thank the reviewer for pointing this out.
14
We thank the reviewer for raising this issue.
15
Using other methods, such as mean comparison tests or running separate
regression models for issues with and without covenants leads to the same
conclusions.
2493
measures and a new variable, COV, that indicates whether the issue
has protective covenants or not.
Panel A of Table 7 presents the results for bond spreads. We
show that bonds without covenants bear a higher cost of about
14.4 bp (the coefficient of C_O). However, if the issue is protected
by covenants, the increase in the cost is negligible and can be at
most 2 pb (14.4 – 12.4, where 12.4 is the coefficient of C_OxCOV).
Thus, in the presence of debt covenant, the voting/cash-flow rights
wedge seems to impose little expropriation risk for bondholders, as
it does not notably increase the bond yield spread. We also note the
effect of debt covenants on the impact of poor creditor rights pro-
tection on bond spreads, which we assess with the coefficient of
ENFORCDYS and ENFORCDYS * COV (4.563  12.295 = 7.732).
This result suggests that poor creditor rights protection does not
increase (or even reduce) the spread of bonds with debt covenants,
although it does increase the spread for bonds without debt
covenants.16
Finally, Panel B of the same table reports generally the same
conclusions when we use bond ratings. Overall, these results sug-
gest that having protective covenants could reduce the risk of
expropriation and hence mitigate the agency cost of debt.
4.5. Robustness checks
4.5.1. Potential sample selection bias
Around 70% of the bonds in our sample are yankee bonds. This
may lead to a selection bias problem as yankee bonds are most
likely to be issued by larger corporations with better corporate
governance. Indeed, yankee bond issuers are required by US
authorities to comply with higher financial and accounting stan-
dards compared to their home countries. To deal with this poten-
tial problem, we first compare our final sample with the sample
from our initial database. Appendix C reports mean comparison
tests (both parametric and non-parametric) in terms of size, per-
formance, corporate governance features, and whether the firm
has previously accessed the US public market. Results show that,
in general, both samples have similar performance, control/owner-
ship discrepancy, and percentage of family control. However, as ex-
pected, firms in our final sample tend to have a larger size, and are
more likely to have accessed the US public market in the past. To
make sure that these differences between the original and the final
sample do not bias our results, we re-run our final models using
the Heckman sample selection model (Heckman, 1979). Model
(5) (Panel A for the Spread regression) and model (2) (Panel B for
the Rating regression) of Table 6 report the results. We use Sigel-
man and Zeng (1999) method to get the marginal effect (evaluated
at the means).17
16
We thank the reviewer for suggesting this explanation.
17
To run Heckman, we use all observations available in the initial sample. In the
selection equation (first stage equation), the regressor vector contains candidates of
relevant variables that may explain the issuer being included in our sample (i.e.,
yankee bonds that are more likely to be covered by FISD), namely: the issuer size, the
issuer performance, whether the issuer has accessed public US security market in the
past (cross-listing or debt issuance), the identity of the ultimate owner and other
issue characteristics. In addition to these issue and issuer features, we add some
country level variables as they may impact the issue being made in the US, and hence
their inclusion in FISD (Barry et al., 2009). For example, issues by firms from a
European country (such as UK) are more likely to be followed by the FISD analysts
(and hence included in our sample) than African or Asian issues. Moreover, this
selectivity (sample inclusion) could be biased toward more developed countries
possibly because of higher level of transparency and the availability of better
institutions. Thus, we also include, in our selection equation, variables such as the
GDP growth, the country creditor rights score, and other debt enforcement proxies.
The intuition behind this idea is that firms in countries with high creditor rights
protection will have less difficulty issuing bonds in the US since they are more
familiar (and used) with (relatively) high disclosure requirements in their home
countries.
2494 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

Table 7
Bond costs and covenants. This table reports the OLS and Probit regression results of bond spreads and ratings, on regulatory institutions, as well as on firm- and issue- control
variables (only the results for the governance variables are reported). We introduced dummies to control for the year, country and industry effects in all models. Control variables’
descriptions are presented in Table 1. Governance variables are interacted with COV which is a dummy variable that equals 1 if the issue contains a covenant provision. SPREAD:
Yield to maturity on the bond issues minus the yield to maturity on a US treasury bond of similar maturity and coupon (in basis points). RATING: Ordinal variable taking on value
from 1 to 7 that represent the S&P bond ratings. For more details on the transformation procedure, see Appendix B. LMAT: The logarithm of the years to maturity. CREDRIGHTS:
An index reflecting creditor rights that ranges from 0 to 4 (La Porta et al., 1998). PUBREGIS: A dummy variable equals 1 if a public credit registry operates in the country, 0
otherwise. COST_INSLV: The estimated cost of the bankruptcy proceeding for a firm. It is the ratio of all kinds of costs (authority costs, accountant fees, inspector fees, bankruptcy
administrator fees, . . .) to the value of the insolvency estate. EFFDBTENFORC: It is the present value of the terminal value of the firm after bankruptcy costs. It reflects the value
preserved in debt enforcement proceedings. ENFORCDYS: The number of days to resolve a payment dispute through courts. NEWS: Circulation of daily newspapers divided by
population. CORRUPTION: An index (ranges from 0 to 10) that indicates the level of corruption in the government. Low ratings indicate higher corruption levels.

Panel A Panel B
OLS estimates: dependent variable bonds spreads Ordered probit estimates: dependent variable bonds ratings
Expected sign (1) Expected sign (1)
C_O + 14.396 – –18.003
(0.036)** (0.007)***
FAMILY ? 25.994 ? 58.306
(0.000)*** (0.000)***
MANAGER ? 1.782 ? 29.511
(0.784) (0.000)***
STATE ? 5.485 ? 0.804
(0.315) (0.915)
WHELDFIN ? 12.539 ? 38.237
(0.035)** (0.000)***
FAMILYxC_O ? 7.616 ? 27.459
(0.221) (0.000)***
STATExC_O ? 8.569 ? 17.078
(0.124) (0.006)***
WHELDFINxC_O ? 13.123 ? 7.974
(0.005)*** (0.061)*
CREDRIGHTS  6.624 + 4.874
(0.274) (0.511)
PUBREGIS  19.741 + 15.334
(0.000)*** (0.070)*
COST_INSLV + 19.658  11.211
(0.005)*** (0.245)
EFFDBTENFORC  10.824 + 22.717
(0.039)** (0.006)***
ENFORCDYS + 4.563  57.125
(0.192) (0.000)***
NEWS  24.247 + 42.117
(0.000)*** (0.000)***
CORRUPTION  11.664 + 23.623
(0.034)** (0.006)***
C_OxCOV ? 12.418 ? 13.929
(0.083)* (0.064)*
FAMILYxCOV ? 3.127 ? 4.789
(0.667) (0.507)
WHELDFINxCOV ? 8.158 ? 19.726
(0.088)* (0.004)***
STATExCOV ? 3.709 ? 6.707
(0.410) (0.240)
CREDRIGHTSxCOV  3.434 + 9.066
(0.509) (0.178)
PUBREGISxCOV  4.403 + 17.696
(0.443) (0.006)***
COST_INSLVxCOV ? 12.716 ? 26.199
(0.150) (0.003)***
EFFDBTENFORCxCOV  5.311 + 19.537
(0.383) (0.012)**
ENFORCDYSxCOV  12.295 + 11.307
(0.010)*** (0.200)
NEWSxCOV  11.213 + 15.341
(0.055)* (0.021)**
N 254 304
Adj. R2 55.20 –
Pseudo-R2 – 37.30
F 5.70 –
Model v2 – 257.55
Sig. (0.000)*** (0.000)***
***
Significance at the 10% level. (Coefficients are multiplied by 100 for the bond rating model).
**
Significance at the 5% level. (Coefficients are multiplied by 100 for the bond rating model).
*
Significance at the 1% level. (Coefficients are multiplied by 100 for the bond rating model).
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499
For the SPREAD model, the correlation between errors (residu-
als from the selection and outcome models) as given by rho (q)
is very low (5.4%) and not statistically different from 0. This
suggests that the simple OLS approach is all we need to estimate
SPREAD.18
For the RATING model instead, we have to tackle a technical is-
sue related to the use of the Heckman (1979)’s procedure due to
the fact that the outcome variable RATING is an ordinal discrete
variable, while the original Heckman’s model requires the depen-
dent variable in the outcome regression to be continuous. We deal
with this problem in two ways. First, we estimate our outcome
model using OLS, and hence treat RATING as a continuous variable.
Alternatively, we transform RATING into a binary variable and use
a Probit model with sample selection, which some statistical soft-
ware can estimate.19 We acknowledge that both methods are not
ideal since in the first one we change the nature of the dependent
variable, while in the second we lose a lot of information when we
reduce the scale of RATING. Nevertheless, both methods lead to qual-
itatively the same results. Column (2) of Table 6 reports the results
for the OLS option.
4.5.2. Alternative measure of spreads
When computing bond spreads, we use US treasury bonds that
share similar characteristics as corporate bond issues (especially in
terms of maturity).20 Among other reasons discussed previously, we
do so due to the difficulty to obtain data on each country treasury
bond issues. We are aware that, by using US treasury bonds, we dis-
regard some characteristics of the domestic country treasury bonds.
For example, the inflation level, which is expected to have an impact
on the country treasury bond yields, is not (fully) captured by US
treasury bond yields. To overcome this potential shortcoming, we
have already included in our model macroeconomic variables such
as the country inflation rate, GDP growth and debt market size. How-
ever, and to assess the robustness of our previous results, we try to
collect additional data from all possible available sources (FISD, SDC
Platinum Thomson Financial, central banks’ websites). Overall, we are
able to obtain data on bond issues for the following countries: Aus-
tria, France, Germany, Norway, South Korea, Sweden, and UK. The fi-
nal sample falls to 120 observations. The Pearson correlation
between the Spreads using US treasury bonds and country treasury
bonds is higher than 71% (significantly different from 0 at less than
1% level). Column 3 of Table 6 (Panel A) reports the results of using
domestic country treasury bonds. As we can see, our inferences re-
main the same: we still obtain an insignificant coefficient for the
creditor rights index, while most measures of debt laws’ enforce-
ment yield economically and statistically significant coefficients, as
in our original model. With regard to the ownership features, family
control has the same effect as in our base model. There are however
small differences with respect to the expropriation measure (C_O
still has the expected sign but becomes insignificant) and the effect
of WHELDFIN and MANAGER which become significant when we use
domestic countries’ treasury bonds. This may be due to the relatively
small size of the new sample, and probably to the fact that the larg-
est portion of this sample comes from European countries. Overall,
the use of US treasury bonds has no considerable effect on the qual-
ity of our results.
18
We also compare the coefficients and significance levels with our initial model
(comparison between model (1) and (4) in the same table) and we reach the same
conclusions for all variables except for INFLATION which becomes significant in the
selection model.
19
One transformation consists in transforming the rating into a binary variable that
indicates whether the issue has an investment or a speculative grade. We are
indebted to Willam Greene for his helpful suggestions regarding sample selection
models.
20
The reader is referred to Van Landschoot (2008) for evidence on the dominance of
US interest rates in the corporate bonds markets.
2495
4.5.3. Relaxing the hypothesis of independent residuals
In our regressions, we assume that residuals are independent.
However, our sample being multinational, bond yield-spreads (rat-
ings) within each country may not be independent, which could re-
sult in residuals not being independent. In an unreported
regression, we re-ran our analyses while relaxing this assumption
(thus allowing for possible within-country dependence). Results
are mainly similar to those reported in the final model for both
bond spreads and ratings.21 Thus, the assumption of independent
residuals does not materially affect our conclusions.
4.5.4. Excluding bond issues by financial institutions
In our initial sample, we did not eliminate bond issues by banks
and insurance companies. Because of the differences that could ex-
ist between financial and non-financial firms regarding debt
financing and governance characteristics, our results could be
influenced by the inclusion of the financial sector in our sample.
After excluding financial institutions from the initial sample, we
are left with 209 and 231 observations for the bond yield-spreads
and the ratings model, respectively. When we re-run our spreads
and ratings models with these observations, we obtain almost sim-
ilar results (unreported but available upon request) as in the final
model with all bond issuers.
4.5.5. Excluding the UK sample
Issues from the UK account for close to 35% of our sample. To
test whether such weight of UK issues affects our results, we re-
run our regressions excluding observations from the UK In unre-
ported results, we document similar findings regarding all the legal
variables except for debt enforcement days (ENFORCDYS) which
loses its significance in the spread model, and for the cost of insol-
vency proceeding (COST_INSLV) which becomes significant and
negative (as expected) in the ratings model. Particularly, for both
the spreads and ratings models, the creditor rights index remains
insignificant while the debt enforcement measures still keep their
significance and signs. However, for both bond spreads and ratings
models, the expropriation measure (C_O) is not significant,
although it keeps its expected sign. A possible explanation is pro-
vided by our previous result that only Western European debthold-
ers anticipate expropriation, and are sensible to the voting/cash-
flow rights divergence. Excluding around 35% of the European
observations from our sample increases the likelihood that Asian
observations become more influential (C_O is more likely to be
insignificant as in the Asian sample).
4.5.6. Excluding the 1997 data
During the financial crisis of 1997, financial markets faced trou-
bled times, especially in Asian countries. The data collected during
that period could then be affected. To tackle this concern, we ex-
clude all 1997 issues and we re-run our final models.22 Our unre-
ported results are qualitatively similar to our original inferences.
4.5.7. Endogeneity issues
In our SPREAD regressions, we did not include the rating as a
potential explanatory variable. Some previous studies, however,
argue that ratings may be a determinant of bond yield-spreads
(see for example Chen et al., 2007 and Yu, 2005, among others).
In this case, the results of the SPREAD model may suffer from an
endogeneity problem caused by the omission of the rating variable.
We control for this situation in two ways: first, we re-run the final
SPREAD model (model (1), Table 6) by simply adding the rating as
an extra control variable. Second, as in Anderson et al. (2003), we
21
Results are available upon request.
22
In unreported results, we find similar findings when we exclude both 1997 and
1998 data from our sample.
2496 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

re-run the SPREAD model and we add the residual values of RAT- debtholders and rating agencies value the quality of debt enforce-
ING where the predicted values are generated by regressing RAT- ment, which is important for authorities who want to develop local
ING on its control variables as in Table 6.23 This second debt markets: more efforts should be devoted to the enforcement of
alternative is motivated by the fact that our RATING and SPREAD laws, such as creating credit registries and/or taking the appropri-
models have basically the same explanatory variables. Thus, simply ate measures to reduce the cost of insolvency proceedings, and
adding the raw rating variable in the SPREAD model, could bias re- the number of days it takes to resolve a payment dispute through
sults. Albeit the difference in these two approaches, the regression courts, rather than the enactment of new laws.
results (unreported but available upon request) are very similar, Finally, we explore whether the presence of debt covenants af-
and broadly confirm the conclusions drawn from the main model fects the relation between bond spreads (ratings) and corporate
of Table 6. governance, and we find that, in general, protective covenants alle-
viate the potential risk of expropriation by reducing bond spreads
5. Concluding remarks by 12.4 bps for firms with a higher voting/cash-flow rights wedge.
However, covenants seem to be ineffective in reducing spreads if
Our main goal in this study is to explore the combined effect of the ultimate owner is a family. As regards the creditor rights pro-
the ultimate ownership structure (hence the potential risk of tection, protective bond covenants are able to reduce bond yield-
expropriation by the controlling shareholders) and the quality of spreads in poor creditor rights protection environments.
creditors’ protection on the costs of corporate bonds from a set Our results are robust to several sensitivity checks such as the
of developed and developing countries. Using data on the ultimate use of the Heckman selection model to control for any potential
ownership of firms around the world, we proxy for the likelihood selection bias, the use of alternative benchmarks, the exclusion of
of expropriation by the controlling shareholders with the voting issues by financial firms, those by British firms, and those during
and cash-flow rights divergence. Our hypothesis is that ultimate the Asian financial crisis of 1997.
large shareholders with voting rights in excess of their cash-flow
rights could threaten the interests of minority stockholders as well Acknowledgments
as those of bondholders, essentially by undertaking less (or not)
profitable projects that increase the likelihood of bankruptcy. We We would like to thank Maria Boutchkova, Jean-Claude Cosset,
find that expropriation by controlling shareholders affects indeed Georges Dionne, Donald Fraser, Nabil Ghalleb, Omrane Guedhami,
bond yield-spreads and ratings. Using data on the identity of the Douglas Hodgson, Amrita Nain and Chen Zhian for their valuable
ultimate owner, we find strong evidence that family control has comments and suggestions. We also thank seminar participants
a positive and significant effect on bond spreads and a negative at the 2006 CIRPEE meetings in Val-Morin (Canada), the 2007
and significant effect on bond ratings, which goes against evidence IFC4 in Tunisia and the 2007 EFMA Symposium on Corporate Gov-
on U.S firms as documented by Anderson et al. (2003). We rational- ernance and Shareholder Activism in Bocconi University (Italy) for
ize our results as follows: First, when they are in a controlling po- their input. Valuable research assistance from Mohamed Jabir is
sition, families are more likely to extract private benefits that harm greatly appreciated. Financial support is acknowledged from SSRC
debtholders’ interests. Second, these families avoid capital dilution and the Center of Research in E-Finance (HEC Montréal). Hatem
to preserve their control, and are more likely to use debt financing Ghouma acknowledges financial support from the Tunisian Minis-
as opposed to equity financing, hence higher leverage ratios. Con- try of Higher Education, Scientific Research and Technology.
trol in the hands of widely held financial firms has a positive effect
on bond ratings only, while State control affects neither bond Appendix A. Variables description and data sources
spreads nor bond ratings.
We also test whether debtholders in East Asian countries are
Variable Description Source
less alert to expropriation by controlling shareholders than those
in West European countries (as originally hypothesized by Faccio SPREAD Yield to maturity on Fixed
et al., 2001). High fear of expropriation should result in higher pre- the bond issues Investment
miums required by debtholders. Results show that bondholders in minus the yield to Securities
both East Asian and European countries anticipate expropriation maturity on a US Database
by controlling shareholders and fear family control. However, only treasury bond of
European investors seem to consider control by widely held finan- similar maturity (in
cial firms as a potential risk of expropriation. basis points)
Our analysis considers a large number of regulatory institutions RATING Ordinal variable Fixed
to proxy for the quality of investor protection in the country, and taking on value from Investment
examines their impact on the costs of debt financing. We find that 1 to 7 that represent Securities
a better debtholders’ protection generally reduces bond yields and the S&P bond ratings. Database
increases corporate bond ratings. More importantly, we find that, to For more details on
both bondholders and rating agencies, enforcement of debt laws is the transformation
more important than the existence of laws on the book. Indeed, our procedure, see
results show that the creditor rights index has no significant effect Appendix B
on bond spreads and ratings, while most measures of debt enforce- C_O A measure of the Claessens et al.
ment (the existence of a public credit registry, the estimated cost of likelihood of (2000), Faccio
insolvency proceedings, the efficiency of the bankruptcy process, expropriation by and Lang
and the number of contract enforcement days) are generally statis- excess control. It is (2002)
tically and economically significant. This finding suggests that both the difference
between voting and
23 cash-flow rights of
The unique difference compared to the regression reported in Table 6 is that the
estimation is done using OLS technique instead of an ordered probit since we require the largest
numerical values for ratings (rather than probabilities) to subtract from the real
values and determine the residuals.
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499 2497

Appendix A (continued) Appendix A (continued)

Variable Description Source Variable Description Source
shareholders at a 10% A public registry is
level defined as a database
FAMILY A dummy variable Claessens et al. owned by public
equals to 1 if the (2000), Faccio authorities (usually
controlling and Lang the Central Bank or
shareholder is a (2002) Banking Supervisory
family Authority), that
MANAGER A dummy variable Claessens et al. collects information
equals to 1 if a (2000), Faccio on the standing of
member of the and Lang borrowers in the
controlling family is a (2002) financial system and
CEO, Honorary makes it available to
Chairman, Chairman, financial institutions.
or Vice-Chairman The variable is
STATE A dummy variable Claessens et al. constructed as at
equals to 1 if the (2000), Faccio January for every year
controlling and Lang from 1978 to 2003
shareholder is a State (2002) Cost of the The estimated cost of Djankov et al.
WHELDFIN A dummy variable Claessens et al. Insolvency the bankruptcy (2008)
equals to 1 if the (2000), Faccio (COST_INSLV) proceeding for a firm.
controlling and Lang It is the ratio of all
shareholder is a (2002) kinds of costs
Widely held financial (authority costs,
firm accountant fees,
Creditor rights An index reflecting La Porta et al. inspector fees,
(CREDRIGHTS) creditor rights. It is (1998) bankruptcy
formed by adding 1 administrator
when (1) the country fees, . . .) to the value
imposes restrictions, of the insolvency
such as creditors estate
consent or minimum Efficiency of the It is the present value Djankov et al.
dividends to file for bankruptcy of the terminal value (2008)
reorganization; (2) process of the firm after
secured creditors are (EFFDBTENFORC) bankruptcy costs. It
able to gain reflects the value
possession of their preserved in debt
security once the enforcement
reorganization proceedings. Higher
petition has been values indicate higher
approved (no efficiency of the debt
automatic stay); (3) enforcement
secured creditors are Contract The number of days Djankov et al.
ranked first in the enforcement to resolve a payment (2008)
distribution of the days dispute through
proceeds that result (ENFORCDYS) courts. It is the
from the disposition number of calendar
of the assets of a days to enforce a
bankrupt firm; and contract of unpaid
(4) the debtor does debt worth 50% of the
not retain the country’s GDP per
administration of its capita. The variable is
property pending the constructed as at
resolution of the January 2003
reorganization. The Newspaper Circulation of daily Dyck and
index ranges from 0 circulation/ newspapers divided Zingales (2004)
to 4 population by population
Public registry A dummy variable Djankov et al. (NEWS)
(PUBREGIS) equals 1 if a public (2008) Corruption An index (ranges from International
credit registry (CORRUPTION) 0 to 10) that indicates Country Risk
operates in the the level of corruption Guide (ICR)
country, 0 otherwise. in the government.

(continued on next page)

2498 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499

Appendix A (continued) Appendix A (continued)

Variable Description Source Variable Description Source
Low ratings indicate (FINANCE) A dummy variable Fixed
higher corruption that equals 1 if the Investment
levels firm operates in the Securities
Inflation Rate Annual percentage International financial sector; 0 Database
(INFLATION) changes of the Financial otherwise
country Consumer Statistics (IFM) (UTILITY) A dummy variable Fixed
Price Index that equals 1 if the Investment
Debt Market Size It is the ratio of the International firm operates in the Securities
(DEBTMKTSIZE) sum of bank debt of Financial utility sector; 0 Database
private sector and Statistics otherwise
outstanding non- (HIGHTEC) A dummy variable American
financial bonds to that equals 1 if the Electronics
GNP firm operates in the Association
Economic Average annual International high-technology (AeA) website
Development growth of per capita Financial industry; 0 otherwise
(GDPGROWTH) GDP Statistics (BFRCRISIS) A dummy variable Fixed
Maturity (LMAT) The logarithm of the Fixed that equals 1 if the Investment
years to maturity Investment bond was issued Securities
Securities before 1997; 0 Database
Database otherwise
Issue size (LISIZE) The logarithm of the Fixed (YRS9798) A dummy variable Fixed
size (offering amount) Investment that equals 1 if the Investment
of the issue (in US $ Securities bond was issued Securities
1000) Database between 1997 and Database
CALL A dummy variable 1998; 0 otherwise
equals to 1 if there is (POSTCRISIS) A dummy variable Fixed
a call provision that equals 1 if the Investment
SENIOR A dummy variable bond was issued after Securities
equals to 1 if there is 1998; 0 otherwise Database
a seniority provision (ASIA) A dummy variable Claessens et al.
SINK A dummy variable that equals 1 if the (2000), Faccio
equals to 1 if the issue firm is from Asia; 0 and Lang
contains a sinking otherwise (2002)
fund provision (EUROPE) A dummy variable Claessens et al.
PRUS A dummy variable that equals 1 if the (2000), Faccio
equals to 1 if the firm firm is from Europe; 0 and Lang
has accessed public otherwise (2002)
US security market in
the past (cross-listing
or debt issuance)
Firm size (ASSET) The annual total Worldscope
assets for the year Appendix B. S&P credit rating transformations
preceding the bond
issue, or last available
(in US $ 1000) S&P initial ratings Transformation
Risk (STDNINC) The operational risk Worldscope AAA 7
as measured by the AA+ 6
standard deviation of AA 6
the net annual AA 6
incomes for the five A+ 5
years before the bond A 5
issue, or last available A 5
Performance (ROI) The Return On Worldscope BBB+ 4
Investments as of the BBB 4
year before the bond BBB 4
issue, or last available BB+ 3
Leverage The ratio of total Worldscope BB 3
(LEVERAGE) debts to total assets BB 3
for the year preceding B+ 2
the bond issue, or last B 2
available B 2
 N. Boubakri, H. Ghouma / Journal of Banking & Finance 34 (2010) 2481–2499 2499

Appendix B (continued) Denis, D.K., McConnell, J.J., 2003. International corporate governance. Journal of
Financial and Quantitative Analysis 38, 1–36.
S&P initial ratings Transformation Dennis, D., McConnell, J.J., 1986. Corporate mergers and security returns. Journal of
Financial Economics 16, 143–187.
CCC+ 1 Djankov, S., Hart, O., McLiesh, C., Shleifer, A., 2008. Debt Enforcement Around the
CCC 1 World. NBER Working Paper N°12807.
Djankov, S., McLiesh, C., Shleifer, A., 2007. Private credit in 129 countries. Journal of
CCC 1 Financial Economics 84, 299–329.
CC 1 Dyck, A., Zingales, L., 2004. Private benefits of control: An international comparison.
C 1 Journal of Finance 59, 537–600.
Elliott, W., Prevost, A.K., Rao, R.P., 2009. The announcement impact of seasoned
D 1 equity offerings on bondholder wealth. Journal of Banking and Finance 33,
1472–1480.
Ellul, A., Guntay, L., Lel, U., 2005. External Governance and Debt Agency Costs of
Family Firms. Working Paper, Indiana University. Available at SSRN: <http://
Appendix C. Comparison between the initial and the final ssrn.com/abstract=687371>.
Esty, B.C., Megginson, W.L., 2003. Creditor rights, enforcement, and debt ownership
samples structure: Evidence from the global syndicated loan market. Journal of Financial
and Quantitative Analysis 38, 37–59.
Faccio, M., Lang, H.P.L., Young, L., 2001. Dividends and expropriation. The American
Variable Initial Final T-test Mann– Economic Review 91, 54–78.
sample sample Whitney Faccio, M., Lang, H.P.L., 2002. The ultimate ownership of western European
corporations. Journal of Financial Economics 65, 365–395.
(N = 8212)* (N = 307) test Fluck, Z., 1999. The dynamics of the management-shareholder conflict. The Review
Size (assets) 70.290 108.083 (0.052)* (0.000)*** of Financial Studies 12, 379–404.
Franks, J.R., Mayer, C., 2001. Ownership and control of German corporations. The
Performance 9.066 7.490 (0.783) (0.831) Review of Financial Studies 14, 943–977.
(ROI) Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices.
C_O (Control/ 3.662 3.601 (0.866) (0.215) Quarterly Journal of Economics 118, 107–155.
Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47,
Ownership 153–161.
wedge) Hermalin, B.E., Weisbach, M.S., 2003. Boards of directors as an endogenously
Family 0.294 0.278 (0.427) (0.425) determined institution: A survey of the economic literature. Economic Policy
Review 9, 7–26.
control Jensen, M., 1986. Agency costs of free cash flow, corporate finance, and takeovers.
State control 0.056 0.188 (0.000)*** (0.000)*** The American Economic Review 76, 323–329.
Widely Held 0.154 0.197 (0.068)* (0.068)* Jensen, M., Meckling, W., 1976. Theory of the firm: Managerial behavior, agency
costs and ownership structure. Journal of Financial Economics 3, 305–360.
Fin.
Klock, M., Mansi, S., Maxwell, W., 2005. Does corporate governance matter to
Institution bondholders. Journal of Financial and Quantitative Analysis 40, 693–720.
control La Porta, R., Lopez-De-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and finance.
Journal of Political Economy 106, 1113–1154.
Previous US 0.065 0.257 (0.000)*** (0.000)***
La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 1999. Corporate ownership around the
Issuer world. The Journal of Finance 54, 471–517.
* La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 2006. What works in securities laws.
The initial sample consists of ownership data on 2981 East Asian firms from The Journal of Finance 61, 1–32.
Claessens et al. (2000) and 5231 Western European firms from Faccio and Lang Lehn, K., Poulsen, A., 1988. Leveraged buyouts: Wealth created or wealth
(2002). redistributed? In: Weidenbaum, M., Chilton, K. (Eds.), Public Policy Toward
***
Significance at the 10% level. Corporate Takeovers. Transaction Publishers, New Jersey.
Mansi, S.A., Maxwell, W.F., Miller, D.P., 2004. Does auditor quality and tenure
matter to investors? Evidence from the bond market. The Journal of Accounting
Research 42, 753–755.
References Maquieira, C.P., Megginson, W.L., Nail, L., 1998. Wealth creation versus wealth
redistributions in pure stock-for-stock mergers. Journal of Financial Economics
Adams, J.C., Mansi, S.A., 2009. CEO turnover and bondholder wealth. Journal of 48, 3–33.
Banking and Finance 33, 522–533. Marais, L., Schipper, K., Smith, A., 1989. Wealth effects of going private for senior
Anderson, C.R., Mansi, A.S., Reeb, D.M., 2003. Founding family ownership and the securities. Journal of Financial Economics 23, 155–191.
agency cost of debt. Journal of Financial Economics 68, 263–285. Miller, D.P., Puthenpurackal, J.J., 2002. The costs, wealth effects and determinants of
Asquith, P., Wizman, T.A., 1990. Event risk, covenants, and bondholder returns in public yankee bond offerings. Journal of Financial Intermediation 11, 455–485.
leveraged buyouts. Journal of Financial Economics 27, 195–213. Morck, R., Shleifer, A., Vishny, R., 1989. Alternative mechanisms for corporate
Ashbaugh, H., Collins, D.W., LaFond, R., 2006. The effects of corporate governance on control. The American Economic Review 79, 842–852.
firms’ credit ratings. Journal of Accounting and Economics 42, 203–243. Prowse, S.D., 1992. The structure of corporate ownership in Japan. The Journal of
Barry, C., Mann, S., Mihov, V., Rodriguez, M., 2009. Interest rate changes and the Finance 47, 1121–1141.
timing of debt issues. Journal of Banking and Finance 33, 600–608. Sengupta, P., 1998. Corporate disclosure quality and the cost of debt. The
Belsley, D.A., Kuh, E., Welsch, R.E., 1980. Regression Diagnostics: Identifying Accounting Review 73, 459–474.
Influential Data and Sources of Collinearity. John Wiley, New York. Shleifer, A., Vishny, R., 1997. A survey of corporate governance. The Journal of
Bhattacharya, U., Daouk, H., 2002. The world price of insider trading. The Journal of Finance 52, 737–783.
Finance 57, 75–108. Sigelman, L., Zeng, L., 1999. Analyzing censored and sample-selected data with
Bhojraj, S., Sengupta, P., 2003. Effect of corporate governance on bond ratings and Tobit and Heckman models. Political Analysis 8, 167–182.
yields: The role of institutional investors and the outside directors. The Journal Valadares, S.M., Leal, R.P.C., 2000. Ownership and Control Structure of Brazilian
of Business 76, 455–475. Companies. Working Paper, Universidade Federal de Rio de Janeiro.
Billett, M., King, T., Mauer, D., 2004. Bondholder wealth effects in mergers and Van Landschoot, A., 2008. Determinants of yield spreads dynamics: Euro versus US
acquisitions: New evidence from the 1980’s and 1990’s. The Journal of Finance dollar corporate bonds. Journal of Banking and Finance 32, 2597–2605.
59, 107–135. Walsh, J., Seward, J., 1990. On the efficiency of internal and external corporate
Chen, L., Lesmond, D., Wei, J., 2007. Corporate yield spreads and bond liquidity. The control mechanisms. The Academy of Management Review 15, 421–458.
Journal of Finance 62, 119–149. Warga, A., Welch, I., 1993. Bondholder losses in leveraged buyouts. The Review of
Claessens, S., Djankov, S., Lang, H.P.L., 2000. The separation of ownership and Financial Studies 6, 959–982.
control in East Asian corporations. Journal of Financial Economics 58, 81–112. Xu, X., Wang, Y., 1997. Ownership structure, corporate governance, and firms’
Cremers, K.J., Nair, V., Wei, C., 2007. Governance mechanisms and bond prices. The performance: The case of Chinese stock companies. World Bank Policy Research
Review of Financial Studies 20, 1359–1388. Working Paper No. 1794. Available at SSRN: <http://ssrn.com/abstract=45303>.
DeAngelo, H., DeAngelo, L., Wruck, K., 2002. Asset liquidity, debt covenants, and Yu, F., 2005. Accounting transparency and the term structure of credit spreads.
managerial discretion in financial distress: The collapse of L.A. Gear. Journal of Journal of Financial Economics 75, 53–84.
Financial Economics 64, 3–34.
Deng, S., Elyasiani, E., Mao, C.X., 2007. Diversification and the cost of debt of bank
holding companies. Journal of Banking and Finance 31, 2453–2473.