Upload
ferlyan-huang-
View
213
Download
0
Embed Size (px)
Citation preview
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 125
Corporate governance and the cost of debt Evidence from director limited
liability and indemni1047297cation provisions
Michael Bradley a Dong Chen b
a Duke University 1 Towerview Dr Durham NC 27708 USAb University of Baltimore 1420 N Charles St Baltimore MD 21201 USA
a r t i c l e i n f o a b s t r a c t
Article history
Received 21 April 2010
Received in revised form 2 August 2010
Accepted 2 August 2010
Available online 10 August 2010
We 1047297nd that 1047297rms that provide limited liability and indemni1047297cation for their directors enjoy
higher credit ratings and lower yield spreads We argue that such provisions insulate corporate
directors from the discipline from potential litigation and allow them to pursue their own
interests by adopting low-risk self-serving operating strategies which coincidentally redound
to the bene1047297t of corporate bondholders Our evidence further suggests that the reduction in the
cost of debt may offset the costs of directorial shirking and suboptimal corporate policies
occasioned by this insulation which may explain why stockholders have little incentive to
rescind these legal protections
copy 2010 Elsevier BV All rights reserved
JEL class i 1047297cation
G30
G34
K22
Keywords
Limited liability provision
Indemni1047297
cationCorporate governance
Cost of debt
Risk-taking
Board of directors
1 Introduction
Much of the corporate governance literature centers on the alignment of the incentives of management and shareholders
(Shleifer and Vishny 1997 Becht et al 2003) It is often assumed that if left to their own devices managersdirectors will pursue
activities that bene1047297t themselves at the expense of the 1047297rms security holders Although self-serving behavior may reduce the
wealth of shareholders it is not clear that all such behavior will also have an adverse effect on the 1047297rms bondholders For example
shirking would be harmful to both bondholders and shareholders Taking on low-risk projects or engaging in conglomerate
mergers however may bene1047297t bondholders due to their concave payoff structure and the coinsurance effect associated withdiversi1047297ed acquisitions (Billett et al 2004) Although there is debate in the literature regarding the risk-taking preference of
managersdirectors (Amihud and Lev 1981 Holmstrom and Costa 1986 Hirshleifer and Thakor 1992 Adams et al 2005) the
evidence largely supports the proposition that managersdirectors prefer conservative investment strategies that are suboptimal
from shareholders perspective due to their concerns for the private bene1047297ts from control and their non-diversi1047297ed 1047297rm-speci1047297c
human capital (Bertrand and Mullainathan 2003 John et al 2008 Kempf et al 2009 Laeven and Levine 2009 Pathan 2009 )
Coincidentally low-risk operating strategies may result in a reduction of the ldquoagency costs of debtrdquo ( Jensen and Meckling
1976 Myers 1977) which ultimately would inure to shareholders bene1047297t Indeed it could well be the case that such reduction in
the cost of debt equals or exceeds the costs of suboptimal corporate conduct and shirking by managersdirectors which would
Journal of Corporate Finance 17 (2011) 83ndash107
Corresponding author Tel +1 410 837 4919 fax + 1 410 837 5722
E-mail address dchenubaltedu (D Chen)
0929-1199$ ndash see front matter copy 2010 Elsevier BV All rights reserveddoi101016jjcorp1047297n201008001
Contents lists available at ScienceDirect
Journal of Corporate Finance
j o u r n a l h o m e p a g e w w w e l s e v i e r c o m l o c a t e j c o r p f i n
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 225
explain why some 1047297rms may tolerate managerialdirectorial self-serving behaviors which are expected if the 1047297rm is weakly
governed The rationale and effects of weak governance are similar to the rationale and effects of bond covenants in that
shareholders restrict their own actions in exchange for a lower cost of debt 1
In this paper we study a particular type of governance mechanism that has received little attention in the1047297nance literature but
may have strong incentive effects on corporate managersdirectors We study the effects of limited liability provisions and the
indemni1047297cation of directors henceforth ldquoLLIrdquo as speci1047297ed in corporate charters bylaws or contracts written directly with
directors The importance of these provisions for directors is suggested by the annual survey by Tillinghast-Towers Perrin (2004
2005 2006) that approximately 20 of the directors of American corporations have been the target of at least one legal action over
the past 10 years Though still not commonplace recent experience illustrates that directors can suffer enormous losses of their
personal wealth from stockholder litigation2 As a result we expect the protection from shareholder suits as afforded by LLI
provisions may have a strong effect on directors incentives to pursue their own interests In contrast the incentives provided by
stock and stock options which received far more attention in the literature may be relatively small Yermack (2004) reports that
among the Fortune 500 companies outside directors wealth increases by only 11 cents per $1000 increase in 1047297rm value
Our sample consists of the SampP 1500 1047297rms from 2002 to 2007 that have publicly traded senior unsecured bonds outstanding
Our proxy for the strength of the LLI provisions at the 1047297rm level is an index ndash the ldquoL-indexrdquo ndash which is the sum of three indicator
variables (1) the existence of liability limitation provisions in corporate charters or bylaws (2) the indemni 1047297cation of litigation
expenses provided by corporate charters or bylaws and (3) the existence of indemni1047297cation contracts with individual directors
The data for these provisions are taken from RiskMetrics (formerly IRRC) which is the same database used by Gompers et al
(2003) (hereafter ldquoGIMrdquo) to construct their G-index
Consistent with the argument made above we show that managersdirectors of 1047297rms with a high L-index conduct low-risk
corporate policies which reduces the cost of debt signi1047297cantly Shareholders bene1047297t from such a reduction and this bene1047297t may
offset the costs of suboptimal operating strategies and shirking by managersdirectors As a result we1047297nd that 1047297rm performance is
not negatively affected by these provisions which may explain why shareholders have little incentive to rescind them ( Romano
2006)3
In this paper we identify a channel through which LLI provisions affect 1047297rm value that differs dramatically from the existing
literature Previous work 1047297nds either a non-signi1047297cant or weakly positive effect of these provisions on stock prices (Bhagat et al
1987 Netter and Poulsen 1989 Janjigian and Bolster 1990 Romano 1991 Brook and Rao 1994 Heron and Lewellen 1998 )4
Focusing exclusively on shareholders these authors assert that the empirical evidence suggests that the costs arising from the
reduced pressure from judicial intervention caused by liability limitation and director indemni1047297cation are either equal to or are
outweighed by the bene1047297ts of attracting and retaining talented directors In contrast to our argument they also conjecture that LLI
provisions encourage managersdirectors who would otherwise take overly conservative actions for fear of legal consequences to
engage in more risk-taking behaviors which would be bene1047297cial especially for 1047297rms with more growth opportunities (Bishop
1981) It is important to note here that the event-study methodology employed in these papers prevents the authors from directly
testing the implications of their theories Consequently the explanations they offer are best understood as being suggestive5 In
contrast we provide evidence that these provisions do not help to attract or retain talented directors and contrary to the
suggestions in these papers but consistent with our arguments above managersdirectors under the protection of these provisions
take on less not more risky projects especially when the 1047297rm has more growth opportunities
Our work contributes to the small strand of literature on the incentives that potential litigation has on corporate directors and
the resulting effects on the value of the 1047297rm Prior research has focused primarily on how stock and stock options board
composition and reputational concerns provide the incentives for directors to ful1047297ll their monitoring and advisory functions (eg
see the review by Hermalin and Weisbach 2003 Adams et al 2010) However very little research exists regarding how the
concerns for the potential loss of personal wealth from legal actions affect the behaviors of corporate managersdirectors Among
the current work on this topic we are the 1047297rst to include bondholders in the analysis and directly test the incentive and talent
1 The difference is also obvious While covenants are intentionally written into bond contracts to limit the actions of managersdirectors weak governance
may reduce the cost of debt as a by-product of managers low-risk operating strategy2
The infamous cases of Enron and Worldcom illustrate the point directors paid $13 million out of their own pockets in the Enron case and $2475 million inthe Worldcom suit3 It should be noted that most of these protections were enacted by the states during the height of the hostile takeover wave in the late 1980s The most
celebrated case that prompted these actions is Smith versus Van Gorkom which ultimately led to Delawares passage of Section 102(b)(7) which allows
Delaware 1047297rms to opt into a provision that eliminates any 1047297nancial liability of directors for breaches in the Duty of Care Not surprisingly a majority of Delaware
1047297rms opted into this provision (see eg Bradley and Schipani 1989) 45 other states followed Delawares lead and passed similar legislation (Romano 2006)
The 1047297rst state to enact a statute that allowed its incorporated 1047297rms to indemnify their directors for litigation expenses was New York in 1961 Today 41 other
states have similar statutes (Romano 2006) An important question that we address in this paper is why such provisions persist if their major purpose is to
insulate directors from the discipline of litigation and therefore allow them to pursue self-serving activities as some have argued (eg Bishop 1981) Historically
shareholders often oppose governance arrangement such as classi1047297ed boards and majority voting However there have been very few instances in which
shareholder activists have sought to rescind these provisions once they had been adopted ( Romano 2006)4 In contrast to our index approach these papers typically test for only one of our LLI provisions (such as director liability limitation) or the LLI provisions
separately This is mainly due to their event-study methodology5 The only paper that tests whether the director liability statute the state statute that allows 1047297rms to include the provisions for director liability limitation in
their charters is related to the attraction of directors is Heron and Lewellen (1998) who 1047297nd that 1047297rms reincorporated into a state with the statute for director
liability limitation added more independent directors subsequently But their evidence is not strong (average percent of independent directors increased from
536 to 562 in two years) Furthermore this increase could also be driven by other contemporaneous factors that were unrelated to the reincorporationdecisions which are not being controlled for
84 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 325
attraction effects of LLI provisions on managersdirectors Our results are inconsistent with the generally accepted explanation of
the effects that these provisions have on shareholder wealth We show that without considering bondholders interests the
prevailing explanations for the existence and persistence of these provisions are misleading
Our work also contributes to the general literature on corporate governance Closely related are studies by Klock et al (2005)
Ashbaugh-Skaife et al (2006 ldquoACL rdquo henceforth) and Chava et al (2009) who show that antitakeover strength of a 1047297rm as proxied
by the G-index is associated with a lower cost of debt The results they document however are subject to mixed interpretations
due to the mixed meanings of the G-index as indicating antitakeover strength on onehand and the weakness of shareholder rights
on the other In contrast we study a governance mechanism and some components of the G-index that arguably are not strong
antitakeover devices6 We show that after netting out the effect of the L-index the G-index is positively not negatively associated
with the cost of debt in stark contrast to the studies above Although not conclusive our evidence implies that the antitakeover
effect of the G-index may hurt bondholders as takeovers may bene1047297t them through the coinsurance effect (Qiu and Yu 2009)7
Finally to our knowledge little prior work has analyzed the incentives for shareholders to change a given governance
structure especially in light of the agency costs of debt8 Our results suggest an alternative channel through which governance
matters for1047297rm value weak governance may bene1047297t shareholders not because it aligns the incentives of managersdirectors with
stockholders but because adopting low-risk strategies decreases the agency costs of debt Thus our results demonstrate the
importance of including bondholders in evaluating a governance structure
The remainder of this paper is organized as follows In the next section we discuss our primary empirical models and data
Section 3 presents our empirical results Section 4 tests the talent attraction hypothesis suggested in the literatureas an alternative
explanation for our cost of debt results Section 5 conducts several robustness checks Finally Section 6 concludes
2 Empirical models and data
In this section we discuss our primary model speci1047297cations for the relation between the cost of debt and our L-index We also
discuss the major variables data sources and summary statistics The detailed de1047297nitions for all the variables in the empirical
analysis as well as their type and sources are presented in the Appendix
21 Empirical models and variables
We use two empirical methodologies to examine the relation between LLI provisions and the cost of debt First we test the
differences between the credit ratings and bond spreads of 1047297rms without the LLI provisions and 1047297rms matched by similar
characteristics that have at least one of these provisions Second we run regression models following Eq (1)
Cost of debt = f eth β 1Lminus index + β 2Gminus index + control variables0 β 3 + Time Dummies
0 β 4 + Industry Dummies
0 β 5 + β 0 + εTHORN
eth1THORN
where f is a general function of the arguments in parentheses We employ two proxies for the cost of debt the 1047297rms credit rating
as assigned by SampP and the yield spread for its publicly traded senior unsecured bonds
Following ACL we group the SampP long-term issuer credit ratings into 7 categories We then estimate an ordered-probit model
for this variable to take into account the nonlinear nature of ratings Our primary coef 1047297cient of interest is β 1 Financial control
variables are based on the existing literature (Klock et al 2005 ACL Qiu and Yu 2009) including return on assets a dummy
indicating whether ROA is negative for both the current and previous years1047297rm size book leverage ratio interest coverage ratio a
dummy indicating whether the 1047297rm has subordinated debt and equity volatility We also control for operating cash 1047298ow since a
higher cash 1047298ow implies a lower risk of 1047297nancial default ceteris paribus
We control for an extensive list of governance variables based on the literature including a dummy variable indicating whether
the 1047297rm has a classi1047297ed (staggered) board CEO option and share holdings board size and independence the percentage of
directors without any shareholdings CEO power the percentage of expert directors institutional shareholdings the number of institutional blockholders with at least 5 ownership and earnings quality (Bhojraj and Sengupta 2003 Larcker et al 2004 Klock
et al 2005 ACL Ortiz-Molina 2006 Chen 2010)9
6 The three components of our L-index are also the components of the G-index which is the sum of 24 managerial entrenchmentantitakeover indicators7 This is particularly true if bondholders insist on event-risk covenants in their indenture agreements that preclude any expropriation of their wealth caused by
a takeover8 ACL also examine why 1047297rms choose to have weak corporate governance even if it hurts their credit ratings Their explanation focuses on management not
shareholders They argue that weak governance allows managers to extract greater compensation from the 1047297rm9 Ecker et al (2006) calculate what they label as ldquo e-loadingrdquo which is a measure of a 1047297rms earnings quality The measure is based on an accrual quality (AQ)
measure which is de1047297ned as the standard deviation of the residuals of a time-series regression of the 1047297rms total current accruals on past present and future
values of the 1047297rms cash 1047298ows from operations the 1047297rms most recent change in revenues and the level of the 1047297rms property plant and equipment for each of
the preceding 1047297ve years The intuition is that if there is a meaningful and sustaining relation between a 1047297rms reported earnings and its cash 1047298ows then one can
conclude that the reported earnings are of high quality They then augment the Fama-French three-factor model and de 1047297ne the loadings on this additional AQ
factor as the e-loading As constructed this measure represents the inverse of the earnings quality In order to facilitate exposition and provide a more intuitive
interpretation of this variable we multiply the e-loading by minus1 and denote the resulting variable Earnings Quality Thus in this reformulation the higher thismeasure the higher the quality of the 1047297rms reported earnings and presumably the higher (better) the 1047297rms credit rating and the lower its bond spread
85M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 425
Since our primary regressions are at the 1047297rm level we calculate the yield spread as the weighted average spread of all
outstanding senior unsecured bonds for a 1047297rm in a given year where the weight is the ratio of the issue amount of the bond over
the total amount of all the outstanding bonds in the sample for the 1047297rm in that year The spread for an individual bond is de1047297ned as
the difference between the yield-to-maturity (YTM) of the bond and the YTM of the US Treasury Bond with the closest maturity
In the spreadregressions we also control for several issue characteristics that proxy for the liquidity andsystematic risk of a bonds
return including bond age modi1047297ed duration and convexity (Klock et al 2005 Qiu and Yu 2009) in addition to the controls
employed in the rating regressions Similar to our treatment of spreads we calculate a weighted average of these variables across
the multiple issues of each 1047297rm in our sample In our analysis of bond spreads we also entertain a 1047297rms credit rating to examine
the extent to which our independent variables are perceived similarly or differently by the ratings agencies and bond traders10 We
consider the potential simultaneous determination of credit ratings and bond spreads in the robustness checks
In all our empirical tests we include the Fama-French 48-industry and year dummy variables Standard errors are adjusted for
heteroscedasticity and clustered at the 1047297rm level (Petersen 2009)
22 Data and sample
We combine different databases in constructing the samples used in our empirical work The bond prices and issue
characteristics are taken from a proprietary database of SampP which covers all rated and publicly traded corporate bonds as of the
end of March from 2002 to 2007 The sample includes only senior unsecured seasoned bonds which is a data limitation but is
appropriate for our purposes since unsecured bondholders may be more concerned about the governance structure of the 1047297rmas
compared to secured bondholders Furthermore senior unsecured bonds are the lions share of all outstanding bonds as suggested
in the literature (Chava et al 2010)The components of the L-index and the G-index are obtained from the RiskMetrics Governance Database which covers the SampP
1500 1047297rms biennially from 2000 to 2006 We follow GIM and assume that the data are the same as those in the previous year
if that years observations are not available
Other data sources include COMPUSTAT CRSP EXECUCOMP Thomson Reuters RiskMetrics Director Database and Thomson
One Banker
When merging different data sources we take care to ensure that other variables are available to investors before the date on
which we collect data for bond spreads in order to alleviate concerns about endogeneity For example year t data for the yield
spreads are matched with year t minus1 data from COMPUSTAT
We delete 1047297rms with dual-class stocks since the special voting arrangement of these 1047297rms may imply a minimal role for other
governance structures We also delete 1047297rms whose board size is less than four as this probably re1047298ects data errors for our large
1047297rms in the sample We also winsorize the major variables at the 1st and 99th percentiles to alleviate concerns for outliers Our
1047297nal sample for the cost of debt regressions contains 1494 1047297rm-year observations from 2002 to 2007 based on 430 unique 1047297rms
23 Summary statistics
Table 1 presents the summary statistics and correlations for the variables in our primary analyses Panel A lists the summary
statistics of each of the variables We also report the summary statistics for L-dummy a dummy variable that equals one if the 1047297rm
has any of the three LLI provisions and zero otherwise The data show that the median 1047297rm in the sample has one LLI provision
and 61 ofthe1047297rms have at least one LLI provision The yield spreads in our sample are slightly larger than those reported in Klock
et al (2005) which is based on a sample earlier than ours The credit ratings of our 1047297rms are about the same as in Klock et al
(2005) and ACL The data for the G-index largely match those reported in these papers
Panel B lists the change in the L-index and the G-index over the sample period Since we assume that year t +1 data are the
same as year t data if the former are not available and our year convention is based on the bond sample we report the changes of
these two variables over three time periods from 2002 to 2003 from 2003 to 2005 and from 2005 to 2007 We only require
that we have the data for these changes (ie data may be missing for other variables) to enlarge the sample size As Panel B
indicates the L-index is remarkably stable over time For example between 2002 and 2003 only 27 of the 1206 1047297rms (22)
changed their L-index The percentage is even smaller for other time periods 11 Therefore similar to the incidences of classi1047297ed
boards as documented in the literature (Bebchuk and Cohen 2005 Masulis et al 2007 Bates et al 2008 Chen 2010 ) it appears
that most LLI provisions were also established prior to 1990 presumably at the height of hostile takeovers and changed
little afterwards Consequently similar to the arguments in these papers endogeneity (reverse causality) is not a serious concern
in this study Nevertheless we examine the potential endogeneity of our results including the omitted variable bias in the
robustness checks presented subsequently We explore the potential reason for the stability of the L-index in Section 3 In
contrast the G-index is much more variable For example between 2003 and 2005 247 of the 1047297rms changed their G-indexes
10 In personal interviews with several individuals who actually rate bonds at SampP we learned that they do not assign ratings based on a formulaic approach
Ratings are set by committees on a predetermined schedule or whenever there is a signi 1047297cant change in the situation of the 1047297rm We were assured that most
decisions were made on the basis of more qualitative situation-speci1047297c issues regarding the 1047297rms prospects for the future and the probability that the 1047297rm will
be able to honor its commitments to bondholders rather than just focusing on the 1047297rms 1047297nancials per se11
In undocumented analysis we 1047297nd that for the much longer period from 1990 to 2006 the stability of the L-index still holds for the 589 1047297rms that remain inthe SampP 1500 index from 1990 to 2006 less than 4 of them changed their L-indexes at any given year
86 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 525
Table 1
Summary statistics and correlations This table reports the summary statistics and correlations of the major variables used in the empirical analysis Panel A reports the s
L-index and the G-index over the sample period Panel C lists the correlation matrix for the major variables De1047297nitions for all the variables are in the Appendix except
logs Spread Duration Convexity Bond_age Operating cash 1047298ow ROA Leverage Intcov Volatility CEO_option CEO_share Per_inst and Earnings quality have been
Panel A Summary statistics
Variable Firm-year Observations P25 Mean Median
Spread () 1494 119 268 173
Rate 1494 4 419 4
L-index 1494 0 102 1
L-dummy 1494 0 061 1
G-index 1494 8 1021 10
Duration (years) 1494 009 015 012
Convexity (years) 1494 003 01 004
Bond_age (years) 1494 224 394 355
Operating cash 1047298ow 1494 008 011 011
ROA 1494 002 004 005
Loss 1494 0 006 0
Size ($million) 1494 272804 2856963 644443
Leverage 1494 018 028 026
Intcov 1494 454 1227 826
Subord 1494 0 014 0
Volatility 1494 03 038 035 Cboard 1494 0 066 1
CEO_option 1494 014 036 036
CEO_share 1494 0 001 0
Per_dir_noshare 1494 0 004 0
Per_ind 1494 067 076 078
Board_size 1494 9 105 10
CEO_power 1494 2 255 3
Expert director 1494 043 058 06
Per_inst 1494 065 073 075
Blockholder 1494 1 189 2
Earnings quality 1494 minus005 008 011
Panel B Change of L-index and G-index over the sample period
Number of 1047297rms Number of 1047297rms increasing index Number of 1047297rms decreasing inde
2002ndash
2003 L-index 1206 9 18 G-index 1206 420 112
2003ndash2005 L-index 1396 1 6
G-index 1396 236 109
2005ndash2007 L-index 1391 10 2
G-index 1391 161 203
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 625
Table 1 (continued)
Panel C Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Spread (1) 100
Rate (2) minus032 100
L-index (3) minus008 018 100
G-index (4) minus001 002 046 100
Operating cash 1047298ow (5) minus022 027 001 005 100
ROA (6) minus032 041 004 001 066 100
Loss (7) 025 minus033 minus005 minus007 minus032 minus049 100
Size (8) 000 026 minus004 minus012 minus019 minus006 minus004 100
Leverage (9) 018 minus026 minus005 002 004 minus024 020 011 100
Intcov (10) minus016 034 minus003 minus010 033 037 minus015 minus006 minus040 100Subord (11) 005 minus009 minus011 minus006 minus019 minus011 003 029 016 minus014 100
Volatility (12) 030 minus057 minus015 minus009 minus027 minus044 042 minus010 018 minus018 004 100
Cboard (13) minus004 minus002 minus003 046 008 002 minus005 minus012 006 minus005 minus003 minus006 100
CEO_option (14) minus006 016 004 004 000 minus005 005 minus006 minus009 010 minus013 013 minus001 100
CEO_share (15) 005 minus017 minus005 minus019 minus003 minus003 006 minus004 001 minus001 009 006 minus003 minus019 100
Per_dir_noshare (16) minus001 minus010 minus005 minus003 minus008 minus007 002 minus001 minus002 minus006 007 004 000 003 002 100
Per_ind (17) minus004 012 012 020 007 005 minus007 000 minus004 002 minus009 minus013 005 007 minus021 000
Board_size (18) minus010 041 014 008 005 009 minus011 023 minus010 010 minus001 minus031 minus003 010 minus015 000 0
CEO_power (19) 002 011 009 011 001 minus001 minus002 000 minus005 minus001 minus007 minus009 002 008 minus011 minus007 0
Expert director (20) minus007 027 015 017 001 003 000 010 minus003 004 minus008 minus013 001 019 minus021 minus007 0
Per_inst (21) minus001 minus021 minus006 014 008 010 minus007 minus014 002 minus009 000 005 009 minus002 minus016 002 0
Blockholder (22) 018 minus034 minus003 007 minus004 minus012 015 minus012 016 minus017 minus004 016 006 minus003 002 002 0
Earnings quality (23) minus020 019 minus001 005 030 037 minus027 minus008 minus001 004 004 minus037 011 minus016 006 minus002 minus0
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 225
explain why some 1047297rms may tolerate managerialdirectorial self-serving behaviors which are expected if the 1047297rm is weakly
governed The rationale and effects of weak governance are similar to the rationale and effects of bond covenants in that
shareholders restrict their own actions in exchange for a lower cost of debt 1
In this paper we study a particular type of governance mechanism that has received little attention in the1047297nance literature but
may have strong incentive effects on corporate managersdirectors We study the effects of limited liability provisions and the
indemni1047297cation of directors henceforth ldquoLLIrdquo as speci1047297ed in corporate charters bylaws or contracts written directly with
directors The importance of these provisions for directors is suggested by the annual survey by Tillinghast-Towers Perrin (2004
2005 2006) that approximately 20 of the directors of American corporations have been the target of at least one legal action over
the past 10 years Though still not commonplace recent experience illustrates that directors can suffer enormous losses of their
personal wealth from stockholder litigation2 As a result we expect the protection from shareholder suits as afforded by LLI
provisions may have a strong effect on directors incentives to pursue their own interests In contrast the incentives provided by
stock and stock options which received far more attention in the literature may be relatively small Yermack (2004) reports that
among the Fortune 500 companies outside directors wealth increases by only 11 cents per $1000 increase in 1047297rm value
Our sample consists of the SampP 1500 1047297rms from 2002 to 2007 that have publicly traded senior unsecured bonds outstanding
Our proxy for the strength of the LLI provisions at the 1047297rm level is an index ndash the ldquoL-indexrdquo ndash which is the sum of three indicator
variables (1) the existence of liability limitation provisions in corporate charters or bylaws (2) the indemni 1047297cation of litigation
expenses provided by corporate charters or bylaws and (3) the existence of indemni1047297cation contracts with individual directors
The data for these provisions are taken from RiskMetrics (formerly IRRC) which is the same database used by Gompers et al
(2003) (hereafter ldquoGIMrdquo) to construct their G-index
Consistent with the argument made above we show that managersdirectors of 1047297rms with a high L-index conduct low-risk
corporate policies which reduces the cost of debt signi1047297cantly Shareholders bene1047297t from such a reduction and this bene1047297t may
offset the costs of suboptimal operating strategies and shirking by managersdirectors As a result we1047297nd that 1047297rm performance is
not negatively affected by these provisions which may explain why shareholders have little incentive to rescind them ( Romano
2006)3
In this paper we identify a channel through which LLI provisions affect 1047297rm value that differs dramatically from the existing
literature Previous work 1047297nds either a non-signi1047297cant or weakly positive effect of these provisions on stock prices (Bhagat et al
1987 Netter and Poulsen 1989 Janjigian and Bolster 1990 Romano 1991 Brook and Rao 1994 Heron and Lewellen 1998 )4
Focusing exclusively on shareholders these authors assert that the empirical evidence suggests that the costs arising from the
reduced pressure from judicial intervention caused by liability limitation and director indemni1047297cation are either equal to or are
outweighed by the bene1047297ts of attracting and retaining talented directors In contrast to our argument they also conjecture that LLI
provisions encourage managersdirectors who would otherwise take overly conservative actions for fear of legal consequences to
engage in more risk-taking behaviors which would be bene1047297cial especially for 1047297rms with more growth opportunities (Bishop
1981) It is important to note here that the event-study methodology employed in these papers prevents the authors from directly
testing the implications of their theories Consequently the explanations they offer are best understood as being suggestive5 In
contrast we provide evidence that these provisions do not help to attract or retain talented directors and contrary to the
suggestions in these papers but consistent with our arguments above managersdirectors under the protection of these provisions
take on less not more risky projects especially when the 1047297rm has more growth opportunities
Our work contributes to the small strand of literature on the incentives that potential litigation has on corporate directors and
the resulting effects on the value of the 1047297rm Prior research has focused primarily on how stock and stock options board
composition and reputational concerns provide the incentives for directors to ful1047297ll their monitoring and advisory functions (eg
see the review by Hermalin and Weisbach 2003 Adams et al 2010) However very little research exists regarding how the
concerns for the potential loss of personal wealth from legal actions affect the behaviors of corporate managersdirectors Among
the current work on this topic we are the 1047297rst to include bondholders in the analysis and directly test the incentive and talent
1 The difference is also obvious While covenants are intentionally written into bond contracts to limit the actions of managersdirectors weak governance
may reduce the cost of debt as a by-product of managers low-risk operating strategy2
The infamous cases of Enron and Worldcom illustrate the point directors paid $13 million out of their own pockets in the Enron case and $2475 million inthe Worldcom suit3 It should be noted that most of these protections were enacted by the states during the height of the hostile takeover wave in the late 1980s The most
celebrated case that prompted these actions is Smith versus Van Gorkom which ultimately led to Delawares passage of Section 102(b)(7) which allows
Delaware 1047297rms to opt into a provision that eliminates any 1047297nancial liability of directors for breaches in the Duty of Care Not surprisingly a majority of Delaware
1047297rms opted into this provision (see eg Bradley and Schipani 1989) 45 other states followed Delawares lead and passed similar legislation (Romano 2006)
The 1047297rst state to enact a statute that allowed its incorporated 1047297rms to indemnify their directors for litigation expenses was New York in 1961 Today 41 other
states have similar statutes (Romano 2006) An important question that we address in this paper is why such provisions persist if their major purpose is to
insulate directors from the discipline of litigation and therefore allow them to pursue self-serving activities as some have argued (eg Bishop 1981) Historically
shareholders often oppose governance arrangement such as classi1047297ed boards and majority voting However there have been very few instances in which
shareholder activists have sought to rescind these provisions once they had been adopted ( Romano 2006)4 In contrast to our index approach these papers typically test for only one of our LLI provisions (such as director liability limitation) or the LLI provisions
separately This is mainly due to their event-study methodology5 The only paper that tests whether the director liability statute the state statute that allows 1047297rms to include the provisions for director liability limitation in
their charters is related to the attraction of directors is Heron and Lewellen (1998) who 1047297nd that 1047297rms reincorporated into a state with the statute for director
liability limitation added more independent directors subsequently But their evidence is not strong (average percent of independent directors increased from
536 to 562 in two years) Furthermore this increase could also be driven by other contemporaneous factors that were unrelated to the reincorporationdecisions which are not being controlled for
84 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 325
attraction effects of LLI provisions on managersdirectors Our results are inconsistent with the generally accepted explanation of
the effects that these provisions have on shareholder wealth We show that without considering bondholders interests the
prevailing explanations for the existence and persistence of these provisions are misleading
Our work also contributes to the general literature on corporate governance Closely related are studies by Klock et al (2005)
Ashbaugh-Skaife et al (2006 ldquoACL rdquo henceforth) and Chava et al (2009) who show that antitakeover strength of a 1047297rm as proxied
by the G-index is associated with a lower cost of debt The results they document however are subject to mixed interpretations
due to the mixed meanings of the G-index as indicating antitakeover strength on onehand and the weakness of shareholder rights
on the other In contrast we study a governance mechanism and some components of the G-index that arguably are not strong
antitakeover devices6 We show that after netting out the effect of the L-index the G-index is positively not negatively associated
with the cost of debt in stark contrast to the studies above Although not conclusive our evidence implies that the antitakeover
effect of the G-index may hurt bondholders as takeovers may bene1047297t them through the coinsurance effect (Qiu and Yu 2009)7
Finally to our knowledge little prior work has analyzed the incentives for shareholders to change a given governance
structure especially in light of the agency costs of debt8 Our results suggest an alternative channel through which governance
matters for1047297rm value weak governance may bene1047297t shareholders not because it aligns the incentives of managersdirectors with
stockholders but because adopting low-risk strategies decreases the agency costs of debt Thus our results demonstrate the
importance of including bondholders in evaluating a governance structure
The remainder of this paper is organized as follows In the next section we discuss our primary empirical models and data
Section 3 presents our empirical results Section 4 tests the talent attraction hypothesis suggested in the literatureas an alternative
explanation for our cost of debt results Section 5 conducts several robustness checks Finally Section 6 concludes
2 Empirical models and data
In this section we discuss our primary model speci1047297cations for the relation between the cost of debt and our L-index We also
discuss the major variables data sources and summary statistics The detailed de1047297nitions for all the variables in the empirical
analysis as well as their type and sources are presented in the Appendix
21 Empirical models and variables
We use two empirical methodologies to examine the relation between LLI provisions and the cost of debt First we test the
differences between the credit ratings and bond spreads of 1047297rms without the LLI provisions and 1047297rms matched by similar
characteristics that have at least one of these provisions Second we run regression models following Eq (1)
Cost of debt = f eth β 1Lminus index + β 2Gminus index + control variables0 β 3 + Time Dummies
0 β 4 + Industry Dummies
0 β 5 + β 0 + εTHORN
eth1THORN
where f is a general function of the arguments in parentheses We employ two proxies for the cost of debt the 1047297rms credit rating
as assigned by SampP and the yield spread for its publicly traded senior unsecured bonds
Following ACL we group the SampP long-term issuer credit ratings into 7 categories We then estimate an ordered-probit model
for this variable to take into account the nonlinear nature of ratings Our primary coef 1047297cient of interest is β 1 Financial control
variables are based on the existing literature (Klock et al 2005 ACL Qiu and Yu 2009) including return on assets a dummy
indicating whether ROA is negative for both the current and previous years1047297rm size book leverage ratio interest coverage ratio a
dummy indicating whether the 1047297rm has subordinated debt and equity volatility We also control for operating cash 1047298ow since a
higher cash 1047298ow implies a lower risk of 1047297nancial default ceteris paribus
We control for an extensive list of governance variables based on the literature including a dummy variable indicating whether
the 1047297rm has a classi1047297ed (staggered) board CEO option and share holdings board size and independence the percentage of
directors without any shareholdings CEO power the percentage of expert directors institutional shareholdings the number of institutional blockholders with at least 5 ownership and earnings quality (Bhojraj and Sengupta 2003 Larcker et al 2004 Klock
et al 2005 ACL Ortiz-Molina 2006 Chen 2010)9
6 The three components of our L-index are also the components of the G-index which is the sum of 24 managerial entrenchmentantitakeover indicators7 This is particularly true if bondholders insist on event-risk covenants in their indenture agreements that preclude any expropriation of their wealth caused by
a takeover8 ACL also examine why 1047297rms choose to have weak corporate governance even if it hurts their credit ratings Their explanation focuses on management not
shareholders They argue that weak governance allows managers to extract greater compensation from the 1047297rm9 Ecker et al (2006) calculate what they label as ldquo e-loadingrdquo which is a measure of a 1047297rms earnings quality The measure is based on an accrual quality (AQ)
measure which is de1047297ned as the standard deviation of the residuals of a time-series regression of the 1047297rms total current accruals on past present and future
values of the 1047297rms cash 1047298ows from operations the 1047297rms most recent change in revenues and the level of the 1047297rms property plant and equipment for each of
the preceding 1047297ve years The intuition is that if there is a meaningful and sustaining relation between a 1047297rms reported earnings and its cash 1047298ows then one can
conclude that the reported earnings are of high quality They then augment the Fama-French three-factor model and de 1047297ne the loadings on this additional AQ
factor as the e-loading As constructed this measure represents the inverse of the earnings quality In order to facilitate exposition and provide a more intuitive
interpretation of this variable we multiply the e-loading by minus1 and denote the resulting variable Earnings Quality Thus in this reformulation the higher thismeasure the higher the quality of the 1047297rms reported earnings and presumably the higher (better) the 1047297rms credit rating and the lower its bond spread
85M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 425
Since our primary regressions are at the 1047297rm level we calculate the yield spread as the weighted average spread of all
outstanding senior unsecured bonds for a 1047297rm in a given year where the weight is the ratio of the issue amount of the bond over
the total amount of all the outstanding bonds in the sample for the 1047297rm in that year The spread for an individual bond is de1047297ned as
the difference between the yield-to-maturity (YTM) of the bond and the YTM of the US Treasury Bond with the closest maturity
In the spreadregressions we also control for several issue characteristics that proxy for the liquidity andsystematic risk of a bonds
return including bond age modi1047297ed duration and convexity (Klock et al 2005 Qiu and Yu 2009) in addition to the controls
employed in the rating regressions Similar to our treatment of spreads we calculate a weighted average of these variables across
the multiple issues of each 1047297rm in our sample In our analysis of bond spreads we also entertain a 1047297rms credit rating to examine
the extent to which our independent variables are perceived similarly or differently by the ratings agencies and bond traders10 We
consider the potential simultaneous determination of credit ratings and bond spreads in the robustness checks
In all our empirical tests we include the Fama-French 48-industry and year dummy variables Standard errors are adjusted for
heteroscedasticity and clustered at the 1047297rm level (Petersen 2009)
22 Data and sample
We combine different databases in constructing the samples used in our empirical work The bond prices and issue
characteristics are taken from a proprietary database of SampP which covers all rated and publicly traded corporate bonds as of the
end of March from 2002 to 2007 The sample includes only senior unsecured seasoned bonds which is a data limitation but is
appropriate for our purposes since unsecured bondholders may be more concerned about the governance structure of the 1047297rmas
compared to secured bondholders Furthermore senior unsecured bonds are the lions share of all outstanding bonds as suggested
in the literature (Chava et al 2010)The components of the L-index and the G-index are obtained from the RiskMetrics Governance Database which covers the SampP
1500 1047297rms biennially from 2000 to 2006 We follow GIM and assume that the data are the same as those in the previous year
if that years observations are not available
Other data sources include COMPUSTAT CRSP EXECUCOMP Thomson Reuters RiskMetrics Director Database and Thomson
One Banker
When merging different data sources we take care to ensure that other variables are available to investors before the date on
which we collect data for bond spreads in order to alleviate concerns about endogeneity For example year t data for the yield
spreads are matched with year t minus1 data from COMPUSTAT
We delete 1047297rms with dual-class stocks since the special voting arrangement of these 1047297rms may imply a minimal role for other
governance structures We also delete 1047297rms whose board size is less than four as this probably re1047298ects data errors for our large
1047297rms in the sample We also winsorize the major variables at the 1st and 99th percentiles to alleviate concerns for outliers Our
1047297nal sample for the cost of debt regressions contains 1494 1047297rm-year observations from 2002 to 2007 based on 430 unique 1047297rms
23 Summary statistics
Table 1 presents the summary statistics and correlations for the variables in our primary analyses Panel A lists the summary
statistics of each of the variables We also report the summary statistics for L-dummy a dummy variable that equals one if the 1047297rm
has any of the three LLI provisions and zero otherwise The data show that the median 1047297rm in the sample has one LLI provision
and 61 ofthe1047297rms have at least one LLI provision The yield spreads in our sample are slightly larger than those reported in Klock
et al (2005) which is based on a sample earlier than ours The credit ratings of our 1047297rms are about the same as in Klock et al
(2005) and ACL The data for the G-index largely match those reported in these papers
Panel B lists the change in the L-index and the G-index over the sample period Since we assume that year t +1 data are the
same as year t data if the former are not available and our year convention is based on the bond sample we report the changes of
these two variables over three time periods from 2002 to 2003 from 2003 to 2005 and from 2005 to 2007 We only require
that we have the data for these changes (ie data may be missing for other variables) to enlarge the sample size As Panel B
indicates the L-index is remarkably stable over time For example between 2002 and 2003 only 27 of the 1206 1047297rms (22)
changed their L-index The percentage is even smaller for other time periods 11 Therefore similar to the incidences of classi1047297ed
boards as documented in the literature (Bebchuk and Cohen 2005 Masulis et al 2007 Bates et al 2008 Chen 2010 ) it appears
that most LLI provisions were also established prior to 1990 presumably at the height of hostile takeovers and changed
little afterwards Consequently similar to the arguments in these papers endogeneity (reverse causality) is not a serious concern
in this study Nevertheless we examine the potential endogeneity of our results including the omitted variable bias in the
robustness checks presented subsequently We explore the potential reason for the stability of the L-index in Section 3 In
contrast the G-index is much more variable For example between 2003 and 2005 247 of the 1047297rms changed their G-indexes
10 In personal interviews with several individuals who actually rate bonds at SampP we learned that they do not assign ratings based on a formulaic approach
Ratings are set by committees on a predetermined schedule or whenever there is a signi 1047297cant change in the situation of the 1047297rm We were assured that most
decisions were made on the basis of more qualitative situation-speci1047297c issues regarding the 1047297rms prospects for the future and the probability that the 1047297rm will
be able to honor its commitments to bondholders rather than just focusing on the 1047297rms 1047297nancials per se11
In undocumented analysis we 1047297nd that for the much longer period from 1990 to 2006 the stability of the L-index still holds for the 589 1047297rms that remain inthe SampP 1500 index from 1990 to 2006 less than 4 of them changed their L-indexes at any given year
86 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 525
Table 1
Summary statistics and correlations This table reports the summary statistics and correlations of the major variables used in the empirical analysis Panel A reports the s
L-index and the G-index over the sample period Panel C lists the correlation matrix for the major variables De1047297nitions for all the variables are in the Appendix except
logs Spread Duration Convexity Bond_age Operating cash 1047298ow ROA Leverage Intcov Volatility CEO_option CEO_share Per_inst and Earnings quality have been
Panel A Summary statistics
Variable Firm-year Observations P25 Mean Median
Spread () 1494 119 268 173
Rate 1494 4 419 4
L-index 1494 0 102 1
L-dummy 1494 0 061 1
G-index 1494 8 1021 10
Duration (years) 1494 009 015 012
Convexity (years) 1494 003 01 004
Bond_age (years) 1494 224 394 355
Operating cash 1047298ow 1494 008 011 011
ROA 1494 002 004 005
Loss 1494 0 006 0
Size ($million) 1494 272804 2856963 644443
Leverage 1494 018 028 026
Intcov 1494 454 1227 826
Subord 1494 0 014 0
Volatility 1494 03 038 035 Cboard 1494 0 066 1
CEO_option 1494 014 036 036
CEO_share 1494 0 001 0
Per_dir_noshare 1494 0 004 0
Per_ind 1494 067 076 078
Board_size 1494 9 105 10
CEO_power 1494 2 255 3
Expert director 1494 043 058 06
Per_inst 1494 065 073 075
Blockholder 1494 1 189 2
Earnings quality 1494 minus005 008 011
Panel B Change of L-index and G-index over the sample period
Number of 1047297rms Number of 1047297rms increasing index Number of 1047297rms decreasing inde
2002ndash
2003 L-index 1206 9 18 G-index 1206 420 112
2003ndash2005 L-index 1396 1 6
G-index 1396 236 109
2005ndash2007 L-index 1391 10 2
G-index 1391 161 203
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 625
Table 1 (continued)
Panel C Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Spread (1) 100
Rate (2) minus032 100
L-index (3) minus008 018 100
G-index (4) minus001 002 046 100
Operating cash 1047298ow (5) minus022 027 001 005 100
ROA (6) minus032 041 004 001 066 100
Loss (7) 025 minus033 minus005 minus007 minus032 minus049 100
Size (8) 000 026 minus004 minus012 minus019 minus006 minus004 100
Leverage (9) 018 minus026 minus005 002 004 minus024 020 011 100
Intcov (10) minus016 034 minus003 minus010 033 037 minus015 minus006 minus040 100Subord (11) 005 minus009 minus011 minus006 minus019 minus011 003 029 016 minus014 100
Volatility (12) 030 minus057 minus015 minus009 minus027 minus044 042 minus010 018 minus018 004 100
Cboard (13) minus004 minus002 minus003 046 008 002 minus005 minus012 006 minus005 minus003 minus006 100
CEO_option (14) minus006 016 004 004 000 minus005 005 minus006 minus009 010 minus013 013 minus001 100
CEO_share (15) 005 minus017 minus005 minus019 minus003 minus003 006 minus004 001 minus001 009 006 minus003 minus019 100
Per_dir_noshare (16) minus001 minus010 minus005 minus003 minus008 minus007 002 minus001 minus002 minus006 007 004 000 003 002 100
Per_ind (17) minus004 012 012 020 007 005 minus007 000 minus004 002 minus009 minus013 005 007 minus021 000
Board_size (18) minus010 041 014 008 005 009 minus011 023 minus010 010 minus001 minus031 minus003 010 minus015 000 0
CEO_power (19) 002 011 009 011 001 minus001 minus002 000 minus005 minus001 minus007 minus009 002 008 minus011 minus007 0
Expert director (20) minus007 027 015 017 001 003 000 010 minus003 004 minus008 minus013 001 019 minus021 minus007 0
Per_inst (21) minus001 minus021 minus006 014 008 010 minus007 minus014 002 minus009 000 005 009 minus002 minus016 002 0
Blockholder (22) 018 minus034 minus003 007 minus004 minus012 015 minus012 016 minus017 minus004 016 006 minus003 002 002 0
Earnings quality (23) minus020 019 minus001 005 030 037 minus027 minus008 minus001 004 004 minus037 011 minus016 006 minus002 minus0
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 325
attraction effects of LLI provisions on managersdirectors Our results are inconsistent with the generally accepted explanation of
the effects that these provisions have on shareholder wealth We show that without considering bondholders interests the
prevailing explanations for the existence and persistence of these provisions are misleading
Our work also contributes to the general literature on corporate governance Closely related are studies by Klock et al (2005)
Ashbaugh-Skaife et al (2006 ldquoACL rdquo henceforth) and Chava et al (2009) who show that antitakeover strength of a 1047297rm as proxied
by the G-index is associated with a lower cost of debt The results they document however are subject to mixed interpretations
due to the mixed meanings of the G-index as indicating antitakeover strength on onehand and the weakness of shareholder rights
on the other In contrast we study a governance mechanism and some components of the G-index that arguably are not strong
antitakeover devices6 We show that after netting out the effect of the L-index the G-index is positively not negatively associated
with the cost of debt in stark contrast to the studies above Although not conclusive our evidence implies that the antitakeover
effect of the G-index may hurt bondholders as takeovers may bene1047297t them through the coinsurance effect (Qiu and Yu 2009)7
Finally to our knowledge little prior work has analyzed the incentives for shareholders to change a given governance
structure especially in light of the agency costs of debt8 Our results suggest an alternative channel through which governance
matters for1047297rm value weak governance may bene1047297t shareholders not because it aligns the incentives of managersdirectors with
stockholders but because adopting low-risk strategies decreases the agency costs of debt Thus our results demonstrate the
importance of including bondholders in evaluating a governance structure
The remainder of this paper is organized as follows In the next section we discuss our primary empirical models and data
Section 3 presents our empirical results Section 4 tests the talent attraction hypothesis suggested in the literatureas an alternative
explanation for our cost of debt results Section 5 conducts several robustness checks Finally Section 6 concludes
2 Empirical models and data
In this section we discuss our primary model speci1047297cations for the relation between the cost of debt and our L-index We also
discuss the major variables data sources and summary statistics The detailed de1047297nitions for all the variables in the empirical
analysis as well as their type and sources are presented in the Appendix
21 Empirical models and variables
We use two empirical methodologies to examine the relation between LLI provisions and the cost of debt First we test the
differences between the credit ratings and bond spreads of 1047297rms without the LLI provisions and 1047297rms matched by similar
characteristics that have at least one of these provisions Second we run regression models following Eq (1)
Cost of debt = f eth β 1Lminus index + β 2Gminus index + control variables0 β 3 + Time Dummies
0 β 4 + Industry Dummies
0 β 5 + β 0 + εTHORN
eth1THORN
where f is a general function of the arguments in parentheses We employ two proxies for the cost of debt the 1047297rms credit rating
as assigned by SampP and the yield spread for its publicly traded senior unsecured bonds
Following ACL we group the SampP long-term issuer credit ratings into 7 categories We then estimate an ordered-probit model
for this variable to take into account the nonlinear nature of ratings Our primary coef 1047297cient of interest is β 1 Financial control
variables are based on the existing literature (Klock et al 2005 ACL Qiu and Yu 2009) including return on assets a dummy
indicating whether ROA is negative for both the current and previous years1047297rm size book leverage ratio interest coverage ratio a
dummy indicating whether the 1047297rm has subordinated debt and equity volatility We also control for operating cash 1047298ow since a
higher cash 1047298ow implies a lower risk of 1047297nancial default ceteris paribus
We control for an extensive list of governance variables based on the literature including a dummy variable indicating whether
the 1047297rm has a classi1047297ed (staggered) board CEO option and share holdings board size and independence the percentage of
directors without any shareholdings CEO power the percentage of expert directors institutional shareholdings the number of institutional blockholders with at least 5 ownership and earnings quality (Bhojraj and Sengupta 2003 Larcker et al 2004 Klock
et al 2005 ACL Ortiz-Molina 2006 Chen 2010)9
6 The three components of our L-index are also the components of the G-index which is the sum of 24 managerial entrenchmentantitakeover indicators7 This is particularly true if bondholders insist on event-risk covenants in their indenture agreements that preclude any expropriation of their wealth caused by
a takeover8 ACL also examine why 1047297rms choose to have weak corporate governance even if it hurts their credit ratings Their explanation focuses on management not
shareholders They argue that weak governance allows managers to extract greater compensation from the 1047297rm9 Ecker et al (2006) calculate what they label as ldquo e-loadingrdquo which is a measure of a 1047297rms earnings quality The measure is based on an accrual quality (AQ)
measure which is de1047297ned as the standard deviation of the residuals of a time-series regression of the 1047297rms total current accruals on past present and future
values of the 1047297rms cash 1047298ows from operations the 1047297rms most recent change in revenues and the level of the 1047297rms property plant and equipment for each of
the preceding 1047297ve years The intuition is that if there is a meaningful and sustaining relation between a 1047297rms reported earnings and its cash 1047298ows then one can
conclude that the reported earnings are of high quality They then augment the Fama-French three-factor model and de 1047297ne the loadings on this additional AQ
factor as the e-loading As constructed this measure represents the inverse of the earnings quality In order to facilitate exposition and provide a more intuitive
interpretation of this variable we multiply the e-loading by minus1 and denote the resulting variable Earnings Quality Thus in this reformulation the higher thismeasure the higher the quality of the 1047297rms reported earnings and presumably the higher (better) the 1047297rms credit rating and the lower its bond spread
85M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 425
Since our primary regressions are at the 1047297rm level we calculate the yield spread as the weighted average spread of all
outstanding senior unsecured bonds for a 1047297rm in a given year where the weight is the ratio of the issue amount of the bond over
the total amount of all the outstanding bonds in the sample for the 1047297rm in that year The spread for an individual bond is de1047297ned as
the difference between the yield-to-maturity (YTM) of the bond and the YTM of the US Treasury Bond with the closest maturity
In the spreadregressions we also control for several issue characteristics that proxy for the liquidity andsystematic risk of a bonds
return including bond age modi1047297ed duration and convexity (Klock et al 2005 Qiu and Yu 2009) in addition to the controls
employed in the rating regressions Similar to our treatment of spreads we calculate a weighted average of these variables across
the multiple issues of each 1047297rm in our sample In our analysis of bond spreads we also entertain a 1047297rms credit rating to examine
the extent to which our independent variables are perceived similarly or differently by the ratings agencies and bond traders10 We
consider the potential simultaneous determination of credit ratings and bond spreads in the robustness checks
In all our empirical tests we include the Fama-French 48-industry and year dummy variables Standard errors are adjusted for
heteroscedasticity and clustered at the 1047297rm level (Petersen 2009)
22 Data and sample
We combine different databases in constructing the samples used in our empirical work The bond prices and issue
characteristics are taken from a proprietary database of SampP which covers all rated and publicly traded corporate bonds as of the
end of March from 2002 to 2007 The sample includes only senior unsecured seasoned bonds which is a data limitation but is
appropriate for our purposes since unsecured bondholders may be more concerned about the governance structure of the 1047297rmas
compared to secured bondholders Furthermore senior unsecured bonds are the lions share of all outstanding bonds as suggested
in the literature (Chava et al 2010)The components of the L-index and the G-index are obtained from the RiskMetrics Governance Database which covers the SampP
1500 1047297rms biennially from 2000 to 2006 We follow GIM and assume that the data are the same as those in the previous year
if that years observations are not available
Other data sources include COMPUSTAT CRSP EXECUCOMP Thomson Reuters RiskMetrics Director Database and Thomson
One Banker
When merging different data sources we take care to ensure that other variables are available to investors before the date on
which we collect data for bond spreads in order to alleviate concerns about endogeneity For example year t data for the yield
spreads are matched with year t minus1 data from COMPUSTAT
We delete 1047297rms with dual-class stocks since the special voting arrangement of these 1047297rms may imply a minimal role for other
governance structures We also delete 1047297rms whose board size is less than four as this probably re1047298ects data errors for our large
1047297rms in the sample We also winsorize the major variables at the 1st and 99th percentiles to alleviate concerns for outliers Our
1047297nal sample for the cost of debt regressions contains 1494 1047297rm-year observations from 2002 to 2007 based on 430 unique 1047297rms
23 Summary statistics
Table 1 presents the summary statistics and correlations for the variables in our primary analyses Panel A lists the summary
statistics of each of the variables We also report the summary statistics for L-dummy a dummy variable that equals one if the 1047297rm
has any of the three LLI provisions and zero otherwise The data show that the median 1047297rm in the sample has one LLI provision
and 61 ofthe1047297rms have at least one LLI provision The yield spreads in our sample are slightly larger than those reported in Klock
et al (2005) which is based on a sample earlier than ours The credit ratings of our 1047297rms are about the same as in Klock et al
(2005) and ACL The data for the G-index largely match those reported in these papers
Panel B lists the change in the L-index and the G-index over the sample period Since we assume that year t +1 data are the
same as year t data if the former are not available and our year convention is based on the bond sample we report the changes of
these two variables over three time periods from 2002 to 2003 from 2003 to 2005 and from 2005 to 2007 We only require
that we have the data for these changes (ie data may be missing for other variables) to enlarge the sample size As Panel B
indicates the L-index is remarkably stable over time For example between 2002 and 2003 only 27 of the 1206 1047297rms (22)
changed their L-index The percentage is even smaller for other time periods 11 Therefore similar to the incidences of classi1047297ed
boards as documented in the literature (Bebchuk and Cohen 2005 Masulis et al 2007 Bates et al 2008 Chen 2010 ) it appears
that most LLI provisions were also established prior to 1990 presumably at the height of hostile takeovers and changed
little afterwards Consequently similar to the arguments in these papers endogeneity (reverse causality) is not a serious concern
in this study Nevertheless we examine the potential endogeneity of our results including the omitted variable bias in the
robustness checks presented subsequently We explore the potential reason for the stability of the L-index in Section 3 In
contrast the G-index is much more variable For example between 2003 and 2005 247 of the 1047297rms changed their G-indexes
10 In personal interviews with several individuals who actually rate bonds at SampP we learned that they do not assign ratings based on a formulaic approach
Ratings are set by committees on a predetermined schedule or whenever there is a signi 1047297cant change in the situation of the 1047297rm We were assured that most
decisions were made on the basis of more qualitative situation-speci1047297c issues regarding the 1047297rms prospects for the future and the probability that the 1047297rm will
be able to honor its commitments to bondholders rather than just focusing on the 1047297rms 1047297nancials per se11
In undocumented analysis we 1047297nd that for the much longer period from 1990 to 2006 the stability of the L-index still holds for the 589 1047297rms that remain inthe SampP 1500 index from 1990 to 2006 less than 4 of them changed their L-indexes at any given year
86 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 525
Table 1
Summary statistics and correlations This table reports the summary statistics and correlations of the major variables used in the empirical analysis Panel A reports the s
L-index and the G-index over the sample period Panel C lists the correlation matrix for the major variables De1047297nitions for all the variables are in the Appendix except
logs Spread Duration Convexity Bond_age Operating cash 1047298ow ROA Leverage Intcov Volatility CEO_option CEO_share Per_inst and Earnings quality have been
Panel A Summary statistics
Variable Firm-year Observations P25 Mean Median
Spread () 1494 119 268 173
Rate 1494 4 419 4
L-index 1494 0 102 1
L-dummy 1494 0 061 1
G-index 1494 8 1021 10
Duration (years) 1494 009 015 012
Convexity (years) 1494 003 01 004
Bond_age (years) 1494 224 394 355
Operating cash 1047298ow 1494 008 011 011
ROA 1494 002 004 005
Loss 1494 0 006 0
Size ($million) 1494 272804 2856963 644443
Leverage 1494 018 028 026
Intcov 1494 454 1227 826
Subord 1494 0 014 0
Volatility 1494 03 038 035 Cboard 1494 0 066 1
CEO_option 1494 014 036 036
CEO_share 1494 0 001 0
Per_dir_noshare 1494 0 004 0
Per_ind 1494 067 076 078
Board_size 1494 9 105 10
CEO_power 1494 2 255 3
Expert director 1494 043 058 06
Per_inst 1494 065 073 075
Blockholder 1494 1 189 2
Earnings quality 1494 minus005 008 011
Panel B Change of L-index and G-index over the sample period
Number of 1047297rms Number of 1047297rms increasing index Number of 1047297rms decreasing inde
2002ndash
2003 L-index 1206 9 18 G-index 1206 420 112
2003ndash2005 L-index 1396 1 6
G-index 1396 236 109
2005ndash2007 L-index 1391 10 2
G-index 1391 161 203
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 625
Table 1 (continued)
Panel C Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Spread (1) 100
Rate (2) minus032 100
L-index (3) minus008 018 100
G-index (4) minus001 002 046 100
Operating cash 1047298ow (5) minus022 027 001 005 100
ROA (6) minus032 041 004 001 066 100
Loss (7) 025 minus033 minus005 minus007 minus032 minus049 100
Size (8) 000 026 minus004 minus012 minus019 minus006 minus004 100
Leverage (9) 018 minus026 minus005 002 004 minus024 020 011 100
Intcov (10) minus016 034 minus003 minus010 033 037 minus015 minus006 minus040 100Subord (11) 005 minus009 minus011 minus006 minus019 minus011 003 029 016 minus014 100
Volatility (12) 030 minus057 minus015 minus009 minus027 minus044 042 minus010 018 minus018 004 100
Cboard (13) minus004 minus002 minus003 046 008 002 minus005 minus012 006 minus005 minus003 minus006 100
CEO_option (14) minus006 016 004 004 000 minus005 005 minus006 minus009 010 minus013 013 minus001 100
CEO_share (15) 005 minus017 minus005 minus019 minus003 minus003 006 minus004 001 minus001 009 006 minus003 minus019 100
Per_dir_noshare (16) minus001 minus010 minus005 minus003 minus008 minus007 002 minus001 minus002 minus006 007 004 000 003 002 100
Per_ind (17) minus004 012 012 020 007 005 minus007 000 minus004 002 minus009 minus013 005 007 minus021 000
Board_size (18) minus010 041 014 008 005 009 minus011 023 minus010 010 minus001 minus031 minus003 010 minus015 000 0
CEO_power (19) 002 011 009 011 001 minus001 minus002 000 minus005 minus001 minus007 minus009 002 008 minus011 minus007 0
Expert director (20) minus007 027 015 017 001 003 000 010 minus003 004 minus008 minus013 001 019 minus021 minus007 0
Per_inst (21) minus001 minus021 minus006 014 008 010 minus007 minus014 002 minus009 000 005 009 minus002 minus016 002 0
Blockholder (22) 018 minus034 minus003 007 minus004 minus012 015 minus012 016 minus017 minus004 016 006 minus003 002 002 0
Earnings quality (23) minus020 019 minus001 005 030 037 minus027 minus008 minus001 004 004 minus037 011 minus016 006 minus002 minus0
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 425
Since our primary regressions are at the 1047297rm level we calculate the yield spread as the weighted average spread of all
outstanding senior unsecured bonds for a 1047297rm in a given year where the weight is the ratio of the issue amount of the bond over
the total amount of all the outstanding bonds in the sample for the 1047297rm in that year The spread for an individual bond is de1047297ned as
the difference between the yield-to-maturity (YTM) of the bond and the YTM of the US Treasury Bond with the closest maturity
In the spreadregressions we also control for several issue characteristics that proxy for the liquidity andsystematic risk of a bonds
return including bond age modi1047297ed duration and convexity (Klock et al 2005 Qiu and Yu 2009) in addition to the controls
employed in the rating regressions Similar to our treatment of spreads we calculate a weighted average of these variables across
the multiple issues of each 1047297rm in our sample In our analysis of bond spreads we also entertain a 1047297rms credit rating to examine
the extent to which our independent variables are perceived similarly or differently by the ratings agencies and bond traders10 We
consider the potential simultaneous determination of credit ratings and bond spreads in the robustness checks
In all our empirical tests we include the Fama-French 48-industry and year dummy variables Standard errors are adjusted for
heteroscedasticity and clustered at the 1047297rm level (Petersen 2009)
22 Data and sample
We combine different databases in constructing the samples used in our empirical work The bond prices and issue
characteristics are taken from a proprietary database of SampP which covers all rated and publicly traded corporate bonds as of the
end of March from 2002 to 2007 The sample includes only senior unsecured seasoned bonds which is a data limitation but is
appropriate for our purposes since unsecured bondholders may be more concerned about the governance structure of the 1047297rmas
compared to secured bondholders Furthermore senior unsecured bonds are the lions share of all outstanding bonds as suggested
in the literature (Chava et al 2010)The components of the L-index and the G-index are obtained from the RiskMetrics Governance Database which covers the SampP
1500 1047297rms biennially from 2000 to 2006 We follow GIM and assume that the data are the same as those in the previous year
if that years observations are not available
Other data sources include COMPUSTAT CRSP EXECUCOMP Thomson Reuters RiskMetrics Director Database and Thomson
One Banker
When merging different data sources we take care to ensure that other variables are available to investors before the date on
which we collect data for bond spreads in order to alleviate concerns about endogeneity For example year t data for the yield
spreads are matched with year t minus1 data from COMPUSTAT
We delete 1047297rms with dual-class stocks since the special voting arrangement of these 1047297rms may imply a minimal role for other
governance structures We also delete 1047297rms whose board size is less than four as this probably re1047298ects data errors for our large
1047297rms in the sample We also winsorize the major variables at the 1st and 99th percentiles to alleviate concerns for outliers Our
1047297nal sample for the cost of debt regressions contains 1494 1047297rm-year observations from 2002 to 2007 based on 430 unique 1047297rms
23 Summary statistics
Table 1 presents the summary statistics and correlations for the variables in our primary analyses Panel A lists the summary
statistics of each of the variables We also report the summary statistics for L-dummy a dummy variable that equals one if the 1047297rm
has any of the three LLI provisions and zero otherwise The data show that the median 1047297rm in the sample has one LLI provision
and 61 ofthe1047297rms have at least one LLI provision The yield spreads in our sample are slightly larger than those reported in Klock
et al (2005) which is based on a sample earlier than ours The credit ratings of our 1047297rms are about the same as in Klock et al
(2005) and ACL The data for the G-index largely match those reported in these papers
Panel B lists the change in the L-index and the G-index over the sample period Since we assume that year t +1 data are the
same as year t data if the former are not available and our year convention is based on the bond sample we report the changes of
these two variables over three time periods from 2002 to 2003 from 2003 to 2005 and from 2005 to 2007 We only require
that we have the data for these changes (ie data may be missing for other variables) to enlarge the sample size As Panel B
indicates the L-index is remarkably stable over time For example between 2002 and 2003 only 27 of the 1206 1047297rms (22)
changed their L-index The percentage is even smaller for other time periods 11 Therefore similar to the incidences of classi1047297ed
boards as documented in the literature (Bebchuk and Cohen 2005 Masulis et al 2007 Bates et al 2008 Chen 2010 ) it appears
that most LLI provisions were also established prior to 1990 presumably at the height of hostile takeovers and changed
little afterwards Consequently similar to the arguments in these papers endogeneity (reverse causality) is not a serious concern
in this study Nevertheless we examine the potential endogeneity of our results including the omitted variable bias in the
robustness checks presented subsequently We explore the potential reason for the stability of the L-index in Section 3 In
contrast the G-index is much more variable For example between 2003 and 2005 247 of the 1047297rms changed their G-indexes
10 In personal interviews with several individuals who actually rate bonds at SampP we learned that they do not assign ratings based on a formulaic approach
Ratings are set by committees on a predetermined schedule or whenever there is a signi 1047297cant change in the situation of the 1047297rm We were assured that most
decisions were made on the basis of more qualitative situation-speci1047297c issues regarding the 1047297rms prospects for the future and the probability that the 1047297rm will
be able to honor its commitments to bondholders rather than just focusing on the 1047297rms 1047297nancials per se11
In undocumented analysis we 1047297nd that for the much longer period from 1990 to 2006 the stability of the L-index still holds for the 589 1047297rms that remain inthe SampP 1500 index from 1990 to 2006 less than 4 of them changed their L-indexes at any given year
86 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 525
Table 1
Summary statistics and correlations This table reports the summary statistics and correlations of the major variables used in the empirical analysis Panel A reports the s
L-index and the G-index over the sample period Panel C lists the correlation matrix for the major variables De1047297nitions for all the variables are in the Appendix except
logs Spread Duration Convexity Bond_age Operating cash 1047298ow ROA Leverage Intcov Volatility CEO_option CEO_share Per_inst and Earnings quality have been
Panel A Summary statistics
Variable Firm-year Observations P25 Mean Median
Spread () 1494 119 268 173
Rate 1494 4 419 4
L-index 1494 0 102 1
L-dummy 1494 0 061 1
G-index 1494 8 1021 10
Duration (years) 1494 009 015 012
Convexity (years) 1494 003 01 004
Bond_age (years) 1494 224 394 355
Operating cash 1047298ow 1494 008 011 011
ROA 1494 002 004 005
Loss 1494 0 006 0
Size ($million) 1494 272804 2856963 644443
Leverage 1494 018 028 026
Intcov 1494 454 1227 826
Subord 1494 0 014 0
Volatility 1494 03 038 035 Cboard 1494 0 066 1
CEO_option 1494 014 036 036
CEO_share 1494 0 001 0
Per_dir_noshare 1494 0 004 0
Per_ind 1494 067 076 078
Board_size 1494 9 105 10
CEO_power 1494 2 255 3
Expert director 1494 043 058 06
Per_inst 1494 065 073 075
Blockholder 1494 1 189 2
Earnings quality 1494 minus005 008 011
Panel B Change of L-index and G-index over the sample period
Number of 1047297rms Number of 1047297rms increasing index Number of 1047297rms decreasing inde
2002ndash
2003 L-index 1206 9 18 G-index 1206 420 112
2003ndash2005 L-index 1396 1 6
G-index 1396 236 109
2005ndash2007 L-index 1391 10 2
G-index 1391 161 203
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 625
Table 1 (continued)
Panel C Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Spread (1) 100
Rate (2) minus032 100
L-index (3) minus008 018 100
G-index (4) minus001 002 046 100
Operating cash 1047298ow (5) minus022 027 001 005 100
ROA (6) minus032 041 004 001 066 100
Loss (7) 025 minus033 minus005 minus007 minus032 minus049 100
Size (8) 000 026 minus004 minus012 minus019 minus006 minus004 100
Leverage (9) 018 minus026 minus005 002 004 minus024 020 011 100
Intcov (10) minus016 034 minus003 minus010 033 037 minus015 minus006 minus040 100Subord (11) 005 minus009 minus011 minus006 minus019 minus011 003 029 016 minus014 100
Volatility (12) 030 minus057 minus015 minus009 minus027 minus044 042 minus010 018 minus018 004 100
Cboard (13) minus004 minus002 minus003 046 008 002 minus005 minus012 006 minus005 minus003 minus006 100
CEO_option (14) minus006 016 004 004 000 minus005 005 minus006 minus009 010 minus013 013 minus001 100
CEO_share (15) 005 minus017 minus005 minus019 minus003 minus003 006 minus004 001 minus001 009 006 minus003 minus019 100
Per_dir_noshare (16) minus001 minus010 minus005 minus003 minus008 minus007 002 minus001 minus002 minus006 007 004 000 003 002 100
Per_ind (17) minus004 012 012 020 007 005 minus007 000 minus004 002 minus009 minus013 005 007 minus021 000
Board_size (18) minus010 041 014 008 005 009 minus011 023 minus010 010 minus001 minus031 minus003 010 minus015 000 0
CEO_power (19) 002 011 009 011 001 minus001 minus002 000 minus005 minus001 minus007 minus009 002 008 minus011 minus007 0
Expert director (20) minus007 027 015 017 001 003 000 010 minus003 004 minus008 minus013 001 019 minus021 minus007 0
Per_inst (21) minus001 minus021 minus006 014 008 010 minus007 minus014 002 minus009 000 005 009 minus002 minus016 002 0
Blockholder (22) 018 minus034 minus003 007 minus004 minus012 015 minus012 016 minus017 minus004 016 006 minus003 002 002 0
Earnings quality (23) minus020 019 minus001 005 030 037 minus027 minus008 minus001 004 004 minus037 011 minus016 006 minus002 minus0
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 525
Table 1
Summary statistics and correlations This table reports the summary statistics and correlations of the major variables used in the empirical analysis Panel A reports the s
L-index and the G-index over the sample period Panel C lists the correlation matrix for the major variables De1047297nitions for all the variables are in the Appendix except
logs Spread Duration Convexity Bond_age Operating cash 1047298ow ROA Leverage Intcov Volatility CEO_option CEO_share Per_inst and Earnings quality have been
Panel A Summary statistics
Variable Firm-year Observations P25 Mean Median
Spread () 1494 119 268 173
Rate 1494 4 419 4
L-index 1494 0 102 1
L-dummy 1494 0 061 1
G-index 1494 8 1021 10
Duration (years) 1494 009 015 012
Convexity (years) 1494 003 01 004
Bond_age (years) 1494 224 394 355
Operating cash 1047298ow 1494 008 011 011
ROA 1494 002 004 005
Loss 1494 0 006 0
Size ($million) 1494 272804 2856963 644443
Leverage 1494 018 028 026
Intcov 1494 454 1227 826
Subord 1494 0 014 0
Volatility 1494 03 038 035 Cboard 1494 0 066 1
CEO_option 1494 014 036 036
CEO_share 1494 0 001 0
Per_dir_noshare 1494 0 004 0
Per_ind 1494 067 076 078
Board_size 1494 9 105 10
CEO_power 1494 2 255 3
Expert director 1494 043 058 06
Per_inst 1494 065 073 075
Blockholder 1494 1 189 2
Earnings quality 1494 minus005 008 011
Panel B Change of L-index and G-index over the sample period
Number of 1047297rms Number of 1047297rms increasing index Number of 1047297rms decreasing inde
2002ndash
2003 L-index 1206 9 18 G-index 1206 420 112
2003ndash2005 L-index 1396 1 6
G-index 1396 236 109
2005ndash2007 L-index 1391 10 2
G-index 1391 161 203
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 625
Table 1 (continued)
Panel C Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Spread (1) 100
Rate (2) minus032 100
L-index (3) minus008 018 100
G-index (4) minus001 002 046 100
Operating cash 1047298ow (5) minus022 027 001 005 100
ROA (6) minus032 041 004 001 066 100
Loss (7) 025 minus033 minus005 minus007 minus032 minus049 100
Size (8) 000 026 minus004 minus012 minus019 minus006 minus004 100
Leverage (9) 018 minus026 minus005 002 004 minus024 020 011 100
Intcov (10) minus016 034 minus003 minus010 033 037 minus015 minus006 minus040 100Subord (11) 005 minus009 minus011 minus006 minus019 minus011 003 029 016 minus014 100
Volatility (12) 030 minus057 minus015 minus009 minus027 minus044 042 minus010 018 minus018 004 100
Cboard (13) minus004 minus002 minus003 046 008 002 minus005 minus012 006 minus005 minus003 minus006 100
CEO_option (14) minus006 016 004 004 000 minus005 005 minus006 minus009 010 minus013 013 minus001 100
CEO_share (15) 005 minus017 minus005 minus019 minus003 minus003 006 minus004 001 minus001 009 006 minus003 minus019 100
Per_dir_noshare (16) minus001 minus010 minus005 minus003 minus008 minus007 002 minus001 minus002 minus006 007 004 000 003 002 100
Per_ind (17) minus004 012 012 020 007 005 minus007 000 minus004 002 minus009 minus013 005 007 minus021 000
Board_size (18) minus010 041 014 008 005 009 minus011 023 minus010 010 minus001 minus031 minus003 010 minus015 000 0
CEO_power (19) 002 011 009 011 001 minus001 minus002 000 minus005 minus001 minus007 minus009 002 008 minus011 minus007 0
Expert director (20) minus007 027 015 017 001 003 000 010 minus003 004 minus008 minus013 001 019 minus021 minus007 0
Per_inst (21) minus001 minus021 minus006 014 008 010 minus007 minus014 002 minus009 000 005 009 minus002 minus016 002 0
Blockholder (22) 018 minus034 minus003 007 minus004 minus012 015 minus012 016 minus017 minus004 016 006 minus003 002 002 0
Earnings quality (23) minus020 019 minus001 005 030 037 minus027 minus008 minus001 004 004 minus037 011 minus016 006 minus002 minus0
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 625
Table 1 (continued)
Panel C Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Spread (1) 100
Rate (2) minus032 100
L-index (3) minus008 018 100
G-index (4) minus001 002 046 100
Operating cash 1047298ow (5) minus022 027 001 005 100
ROA (6) minus032 041 004 001 066 100
Loss (7) 025 minus033 minus005 minus007 minus032 minus049 100
Size (8) 000 026 minus004 minus012 minus019 minus006 minus004 100
Leverage (9) 018 minus026 minus005 002 004 minus024 020 011 100
Intcov (10) minus016 034 minus003 minus010 033 037 minus015 minus006 minus040 100Subord (11) 005 minus009 minus011 minus006 minus019 minus011 003 029 016 minus014 100
Volatility (12) 030 minus057 minus015 minus009 minus027 minus044 042 minus010 018 minus018 004 100
Cboard (13) minus004 minus002 minus003 046 008 002 minus005 minus012 006 minus005 minus003 minus006 100
CEO_option (14) minus006 016 004 004 000 minus005 005 minus006 minus009 010 minus013 013 minus001 100
CEO_share (15) 005 minus017 minus005 minus019 minus003 minus003 006 minus004 001 minus001 009 006 minus003 minus019 100
Per_dir_noshare (16) minus001 minus010 minus005 minus003 minus008 minus007 002 minus001 minus002 minus006 007 004 000 003 002 100
Per_ind (17) minus004 012 012 020 007 005 minus007 000 minus004 002 minus009 minus013 005 007 minus021 000
Board_size (18) minus010 041 014 008 005 009 minus011 023 minus010 010 minus001 minus031 minus003 010 minus015 000 0
CEO_power (19) 002 011 009 011 001 minus001 minus002 000 minus005 minus001 minus007 minus009 002 008 minus011 minus007 0
Expert director (20) minus007 027 015 017 001 003 000 010 minus003 004 minus008 minus013 001 019 minus021 minus007 0
Per_inst (21) minus001 minus021 minus006 014 008 010 minus007 minus014 002 minus009 000 005 009 minus002 minus016 002 0
Blockholder (22) 018 minus034 minus003 007 minus004 minus012 015 minus012 016 minus017 minus004 016 006 minus003 002 002 0
Earnings quality (23) minus020 019 minus001 005 030 037 minus027 minus008 minus001 004 004 minus037 011 minus016 006 minus002 minus0
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 725
The time-series stability of the L-index motivates our choice of ordered-probit and pooled OLS as the primary econometric
models (Zhou 2001)
Panel C reports the correlation matrix for the major variables of interest The L-index is negatively correlated with bond spreads
and positively correlated with bond ratings which is consistent with our hypothesis that LLI provisions reduce the costs of debt
The correlation between the L-index and the G-index is 046 Since the components of the L-index are also the components of the
G-index it may be better to control for a ldquoRemaining G-indexrdquo an index comprised of the components in the G-index that are not
in the L-index for a clearer interpretation of the results However for ease of comparison with the existing literature we still
employ the G-index in our primary analysis In undocumented robustness checks we repeat all the analyses with this alternative
index The results remain qualitatively similar
3 Empirical results
Previously we argued that weak governance as proxied by the number of LLI provisions may permit corporate directors to
adopt relatively low-risk corporate policies that decrease the agency costs of debt and thereby bene1047297t the 1047297rms stockholders This
bene1047297t may equal or outweigh the costs from shirking and the suboptimal operating strategies Consequently these provisions
may have a non-negative impact on 1047297rm performance To test this hypothesis we 1047297rst examine the relation between the L-index
and the cost of debt We then test the proposition that managersdirectors of 1047297rms with more LLI provisions adopt relative low-
risk projects and present evidence of shirking by the directors of these 1047297rms as manifested by a lower quality of monitoring
Finally we present evidence regarding the relation between the L-index and 1047297rm value Our purpose here is to examine whether
there is any credence to our proposition that shareholders have no incentive to rescind existing LLI provisions
31 LLI provisions and the cost of debt
We 1047297rst test the differences between the credit ratings andbond spreads of 1047297rms without any LLI provisions and1047297rms matched
by similar characteristics that have at least one LLI provision We construct our matched sample using two matching procedures
(1) industry and1047297rm size and (2) industry and credit rating The second procedure is employed since a 1047297rms credit rating may be
considered as a summary statistic of the1047297rms 1047297nancial condition Panel A of Table 2 presents the t-tests results12 Consistent with
our expectations bond spreads (credit ratings) of the 1047297rms with LLI provisions are signi1047297cantly lower (higher) than those of the
1047297rms without these provisions based on samples matched by industry and 1047297rm size and the results for bond spreads hold when
samples are matched by industry and credit rating These results suggest that compared with the average spread of the 1047297rms
without any LLI provisions the average spread of the 1047297rms in the same industry and with similar sizes and with at least one of
these provisions is reduced by 142 These results are consistent with our prediction that 1047297rms with LLI provisions enjoy a lower
cost of debtNext we run regressions following speci1047297cation (1) We 1047297rst 1047297t an ordered-probit model for credit ratings controlling for
1047297nancial and governance variables as well as year and industry 1047297xed effects The results are presented as Model 1 in Panel B of
Table 2 As shown the L-index is strongly and positively associated with ratings The coef 1047297cient on the L-index is 0174 and is
signi1047297cant at the 1 level Interestingly the coef 1047297cient on the G-index is negative and weakly signi1047297cant at the 10 level rather
than positive as reported in ACL13
The coef 1047297cients on the 1047297nancial variables all have their expected signs and are generally signi1047297cant CEO option holdings are
positively related to ratings which is inconsistent with the 1047297ndings in Ortiz-Molina (2006) who documents a positive relation
between CEO options and bond spreads We note the difference in the de1047297nition of these two option variables We de1047297ne CEO
option holdings as the BlackndashScholes value of CEO option awards as a fraction of CEO total compensation while Ortiz-Molina
(2006) de1047297nes his measure as the number of shares in unexercised options over the total shares outstanding If we use this
de1047297nition then the coef 1047297cient on CEO option holdings is indeed negative and signi1047297cant This result is interesting but puzzling We
leave for future research an explanation of this conundrum
Consistent with the 1047297ndings reported in Ortiz-Molina (2006) CEO shareholdings are negatively related to credit ratings Thecoef 1047297cients on the fraction of directors without any shares and the number of institutional blockholders are also negative and
signi1047297cant consistent with the results reported by ACL Our measures of earnings quality and board size are positively related to
credit ratings The latter relation is consistent with the results in Larcker et al (2004)
To gauge the economic signi1047297canceof theL-index on ratings we run a probit regression in Model 2 of Panel B in Table 2 where the
dependent variable is a dummy variable which is equal to one if the1047297rms rating is investment-grade (better than SampPs BB+ rating)
and zero otherwise For ease of interpretation we also use the L-dummy and control for the Remaining G-index (rather than the
G-index) in the statistical model and report the marginal effects The marginal effect on the L-dummy suggests that adding one
provision for a 1047297rm initially without any LLI provisions increases the probability that the1047297rm receives an investment-grade rating by
38 which is economically signi1047297cant for this single variable
12 Results using the Wilcoxon signed-rank tests are similar13 We note that ACL do not control for operating cash 1047298ow and volatility the two variables that we 1047297nd to be strongly related to ratings as expected
Furthermore ACL do not entertain as many governance control variables as we do Finally their sample covers only one year 2003 It should be noted that we areable to replicate their 1047297ndings if we use the same control variables and sample year as they do
89M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 825
Table 2
LLI Provisions and the cost of debt This table tests the effects of director limited liability and indemni 1047297cation provisions on the cost of debt Panel A presents the
results of thet-tests forthe difference of thecost of debt for1047297rms withoutany LLIprovisionsand 1047297rmswith at least oneLLI provision andmatched by industry and
1047297rm size or industry and credit rating between 2002 and 2007 Panel B reports the regression results of the LLI provisions on the cost of debt The sample consists
ofthe SampP 1500 1047297rmsbetween 2002 and2007 withpublic seniorunsecured bondsThe dependentvariable forModel1 is the1047297rms creditrating(Rate) de1047297ned by
grouping SampP long-term domestic issuer credit ratings (data280 in COMPUSTAT) into seven categories See the Appendix for the details The dependent variable
for Model 2 is a dummy variable that equals one if the 1047297rms credit rating is investment-grade (rateN=4) and zero otherwise Marginal effects instead of
coef 1047297cients are reported for ease of interpretation in this model Remaining G-index equals G-index ndashL-index The dependent variable for Models 3ndash5 is the bond
spread de1047297nedas theweightedaverage spreadof all thesenior unsecuredbondsof the1047297rm with the weight beingtheissue sizeof the bondas a percentageof the
total aggregate issue sizes of all outstanding bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de 1047297ned as the differencebetween the yield-to-maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of
other variables Model 1 uses ordered-probit model Model 2 uses probit model Models 3 ndash5 use pooled OLS model All models include the Fama-French
48-industry and year dummies and constant terms Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in
parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
Panel A t-tests
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and 1047297rm size (2) Difference (2)ndash(1)
Predicted sign Difference
Observations 931 931
Spread 3147 2699 ndash minus0448
Observations 1847 1847
Rate 3446 4154 + 0708
Sample with L-index= 0 (1) Sample with L-indexN0 and matched by industry and credit rating (2)
Observations 830 830
Spread 3391 2952 ndash minus0439
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
L-index 0174
(2571)
minus0370
(minus3187)
minus0335
(minus2928)
L-dummy 0038
(1996)
G-index minus0054
(minus1778)
0157
(2629)
0149
(2589)
0073
(1536)
Remaining G-index minus0009
(minus
1725)Operating cash 1047298ow 4887
(4040)
0245
(1341)
0112
(0036)
0310
(0100)
0221
(0070)
ROA 2412
(2495)
0144
(0945)
minus4645
(minus1413)
minus2927
(minus0832)
minus4836
(minus1451)
Loss minus0157
(minus0852)
minus0225
(minus4764)
1158
(1782)
0895
(1490)
1166
(1752)
Size 0566
(8572)
0054
(5131)
0172
(1273)
0225
(1582)
0163
(1186)
Leverage minus1239
(minus2373)
minus0141
(minus1890)
2266
(2096)
1764
(1735)
2196
(2007)
Intcov 0013
(3189)
0003
(1528)
0012
(2088)
0010
(1978)
0012
(2166)
Subord minus0249
(minus1646)
minus0128
(minus3923)
minus0154
(minus0505)
minus0202
(minus0646)
minus0103
(minus0328)
Volatility minus5681
(minus9651)
minus0592
(minus7932)
2715
(2450)
1398
(1217)
3014
(2671)Cboard 0035
(0264)
0019
(0849)
minus0707
(minus2827)
minus0671
(minus2800)
minus0464
(minus2010)
CEO_option 0572
(3055)
0099
(3080)
minus0920
(minus2274)
minus0631
(minus1520)
minus0941
(minus2324)
CEO_share minus3247
(minus2051)
minus0159
(minus0728)
minus0018
(minus0007)
0517
(0212)
minus0997
(minus0381)
Per_dir_noshare minus0847
(minus1956)
minus0126
(minus1980)
minus1252
(minus1396)
minus1373
(minus1522)
minus1290
(minus1408)
Per_ind 0315
(0864)
0037
(0675)
0187
(0286)
0238
(0372)
0249
(0379)
Board_size 0052
(1846)
0002
(0575)
minus0052
(minus0864)
minus0054
(minus0916)
minus0061
(minus0988)
CEO_power minus0014
(minus0357)
minus0003
(minus0451)
0091
(1121)
0080
(0993)
0088
(1074)
Expert director 0108
(0420)
0020
(0544)
minus0765
(minus1532)
minus0656
(minus1354)
minus0761
(minus1513)
90 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 925
Models 3 through 5 report the results of our spread regressions based on speci1047297cation (1) Model 3 reports the results without
controlling for ratings Consistent with the results of the ratings regressions the L-index is negative and strongly associated with
bond spreads while the G-index is positive Since the L-index is a sub-index of the G-index the marginal effect of adding one
component to the L-index should be the sum of the coef 1047297cients for the L-index and the G-index which is minus0213 This implies a123 decrease of the bond spread for the median 1047297rm in the sample and is certainly economically signi1047297cant In contrast adding
one provision of the G-index is associated with a 91 increase in the spread for the median 1047297rm Consistent with our results with
respect to credit ratings CEO option holdings and earnings quality are negatively while institutional blockholders are positively
associated with bond spreads14
In Model 4 we add the rating dummies to test whether rating agencies and bond traders view our independent variables
similarly If they do the coef 1047297cients on those variables should lose signi1047297cance after controlling for the rating dummies (Bhojraj
and Sengupta 2003) The results suggest that except for volatility and CEO option holdings all other variables appear to be viewed
differently by the rating agencies and bond traders Note that the L-index continues to be negative and highly signi1047297cant This
result implies that the effect of the LLI provisions on bondholders is not (fully) captured by a 1047297rms credit rating The same is true
for a number of other independent variables15 We focus on the speci1047297cation in Model 3 in subsequent analysis
Our results that the G-index is negatively related to ratings and positively related to spreads are in contrast with most of the
existing literature (eg Klock et al 2005 ACL Chava et al 2009)16 We note here the importance of netting out the negative effect
of the L-index in interpreting the G-index in Model 5 where we do not include the L-index as one of the explanatory variablesOther variables are the same as in Model 3 Interestingly the G-index is no longer signi 1047297cant This result suggests a different
interpretation of the L-index as compared with other components of the G-index We control for the G-index throughout this
paper
The results presented in Table 2 suggest that LLI provisions bene1047297t bondholders In the next two subsections we examine the
relations between these provisions and the risk-taking behavior of managersdirectors and the diligence with which directors
discharge their monitoring function
Table 2 (continued)
Panel B Regression results
(1) (2) (3) (4) (5)
Rate Investment Spread Spread Spread
Per_inst minus0517
(minus1011)
0063
(0804)
minus1149
(minus1046)
minus0541
(minus0467)
minus0943
(minus0853)
Blockholder minus0090
(minus2096)
minus0018
(minus2664)
0286
(3708)
0268
(3579)
0276
(3623)
Earnings quality 0304
(1822)
0041
(1649)
minus1377
(minus3091)
minus1367
(minus3061)
minus1357
(minus3004)
Duration minus14217
(minus7346)
minus13949
(minus7619)
minus14435
(minus7336)
Convexity 2608
(3766)
2587
(4057)
2696
(3878)
Bond_age 0191
(4314)
0197
(4534)
0177
(4034)
Rate2 minus10154
(minus2700)
Rate3 minus10643
(minus2845)
Rate4 minus11435
(minus3044)
Rate5 minus11723
(minus3120)
Rate6 minus10825
(minus2849)
Rate7 minus11150
(minus2932)
Observations 1494 1414 1494 1494 1494
Pseudo R 2 043 055
Adjusted R 2 035 037 034
14 The results with respect to CEO option holdings and earnings quality hold with a 1047297rst-difference (change) regression model which is presumably subject to
less endogeneity concerns15 In short bond ratings are not suf 1047297cient statistics for explaining (predicting) bond spreads16 We also use the entrenchment index (E-index) as in Bebchuk et al (2009) in place of the G-index and obtain similar results Controlling for the L-index the
E-index is negative but not signi1047297cant for credit ratings and positive and strongly signi1047297cant for bond spreads The L-index continues to be positive andsigni1047297cant for ratings and negative and signi1047297cant for spreads
91M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1025
32 LLI provisions and risk-taking behavior
The extant literature suggests that the protection of personal wealth from stockholder suits motivates managersdirectors to
take the appropriate (value-maximizing) amount of risk as otherwise they may be overly conservative for fear of potential
litigation arising from the realization of the lower payoff of a risky investment However the immunization from litigation costs
and damages may also enable managersdirectors to pursue their personal interests by conducting low-risk investment policies at
the expense of the 1047297rms shareholders due to their concerns for retaining their private bene1047297ts from control and their non-
diversi1047297ed 1047297rm-speci1047297c human capital We test these competing hypotheses by examining four variables that we believe to be
measures of risk-taking behavior RampD expenditures capital expenditures the number of diversifying acquisitions and equity
volatility In the 1047297rst three regressions we control for lagged Tobins Q ROA 1047297rm size book leverage and cash 1047298ow In the
regression for the number of diversifying acquisitions we also include a dummy variable for whether the 1047297rm acquires another
1047297rm in a given year to account for the possibility that1047297rms with a higher L-index also conduct more acquisitions in general which
may mechanically imply more diversifying acquisitions The control variables for the equity volatility regression follow Cheng
(2008) and include ROA one-year lagged ROA two-year lagged ROA 1047297rm size leverage board size independent directors CEO
power capital expenditures 1047297rm age and the number of business segments The OLS regression results are reported in Table 3
The results show that LLI provisions are associated with lower rather than higher risk-taking behavior of managersdirectors
The coef 1047297cients on capital expenditures and equity volatility are both negative Moreover1047297rms with a higher L-index make more
conglomerate mergers which also reduces 1047297rm risk In undocumented analysis we 1047297nd that the number of acquisitions is not
signi1047297cantly related to the L-index suggesting that our result for conglomerate mergers is not driven by general empire building
(Baumol 1959) but rather by risk-reducing empire building (Treynor and Black 1976) These low-risk activities should work to
Table 3
LLI provisions and corporate risk-taking These models present the results of pooled OLS regressions for the effect of director limited liability and indemni1047297cation
provisionson therisk-taking incentives of directors The sampleconsists of the SampP 15001047297rms over2002and 2007Thedependent variablefor Model 1 is the RampD
expenses as a percentage of total assets The dependent variable for Model 2 is the capital expenditure as a percentage of total assets The dependent variable for
Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in
the same industry as de1047297ned by the Fama-French 48 industries The dependent variable for Model 4 is the standard deviation of monthly stock returns over the
past 60 months See the Appendix for the de1047297nitions of other variables All models include Fama-French 48-industry and year dummies and a constant term
Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the
10 5 and 1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0001
(1009)
minus0003
(minus3023)
0027
(2330)
minus0012
(minus2726)G-index minus0000
(minus1189)
0000
(0302)
minus0005
(minus1358)
minus0004
(minus2724)
Lagged Q 0010
(8687)
0005
(7352)
0001
(0171)
ROA minus0144
(minus9865)
0013
(1793)
minus0040
(minus0823)
minus0253
(minus9311)
Size minus0005
(minus6330)
minus0002
(minus2724)
0040
(4463)
minus0017
(minus5650)
Leverage minus0017
(minus2636)
minus0002
(minus0471)
minus0038
(minus0794)
minus0011
(minus0463)
Cash1047298ow minus0001
(minus0846)
minus0001
(minus2778)
minus0003
(minus1061)
Acquirer 0716
(27495)
Board_size minus0010
(minus5914)Per_ind minus0037
(minus1697)
CEO_power minus0006
(minus2510)
ROA at t-1 minus0218
(minus5982)
ROA at t-2 minus0156
(minus3955)
Capexp 0024
(0282)
Firm_age minus0001
(minus3861)
Segment 0001
(1034)
Observations 7225 7225 7225 5665
Adjusted R 2 055 039 033 056
92 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1125
the bene1047297t of bondholders which is consistent with the results reported in Table 2 The evidence in Table 3 also suggests that the
G-index is negatively associated with equity volatility though the magnitude (minus0004) is only one-third of that of L-index
(minus0012) This evidence is consistent with the low-risk-taking of managers of 1047297rms with a high G-index ( John et al 2008)
It is notable from Model 4 of Table 3 that CEO power is negative and signi1047297cant rather than positive as reported in Adams et al
(2005) and Cheng (2008)17 The signs of the other independent variables are generally consistent with the 1047297ndings reported in
these papers in particular board size which is negative and consistent with Cheng (2008)
Though the results in Table 3 suggest thatin general the latitudeprovidedby LLI provisions does not encourage risk-taking behavior
on the part of managersdirectors it is possible that risk-taking is most acute in growth 1047297rms where such behavior is expected to be
most bene1047297cialforshareholdersWe testthishypothesis by interacting a dummy variableindicating highgrowth opportunities withthe
L-index andthe G-index respectively We use the lagged value of Tobins Q to proxy for a 1047297rms growth opportunities18 If LLI provisions
permit the managers and directors of growth 1047297rms to undertake high-risk projects this interaction term should be positive for the
regressions involving RampD and capital expenditures and equity volatility and negative for the number of diversifying acquisitions
However the results in Table 4 do not support these predictions The interaction term High lagged Q L-index is positive for capital
expenditures but is insigni1047297cant In contrast the signi1047297cant interaction terms have signs which are contrary to the predictions of the
enhanced risk-taking hypothesis the term is negative for RampD and volatility and positive for the number of diversifying acquisitions
The results reported in the table also indicate that none of the interaction terms involving the G-index is signi1047297cant
To further test the relation between LLI provisions and the risk-taking behavior of corporate managersdirectors we examine
the relation between the cost of debt and our L-index for 1047297rms that are in 1047297nancial distress It is expected that the con1047298ict of
interest between shareholders and bondholders intensi1047297es with the deterioration of a 1047297rms credit quality (Bodie and Taggart
1978 John and John 1993) Thus if LLI provisions encourage directors to take more risk it would hurt bondholders more the
closer the1047297rm is to default We use two proxies for the 1047297rms closeness to default a dummy variable indicating whether the credit
rating is speculative (worse than or equal to SampPs BB+rating) and Altmans Z-score A lower Z-score indicates that the 1047297rm is
closer to default We interact these indicators for a 1047297
rms credit quality with the L-index and the G-index and rerun our spreadregressions Models 1 and 2 in Table 5 report the results
As indicated by the statistics the results do not support the claim that LLI provisions increase the risk-taking behavior of managers
directors the interaction term Speculative L-index is negative and not signi1047297cant while Low Z-score L-index is negative and
signi1047297cant at the 5 level suggesting that when a 1047297rm is closer to default LLI provisions bene1047297t bondholders more which is consistent
with the low-risk-taking rather than the high-risk-taking behavior of managersdirectors Interestingly the interaction term Low
Z-score G-index is positive and weakly signi1047297cant which is consistent with the joint hypothesis that the G-index is a proxy for the
antitakeover strength of the1047297rm and that takeovers are bene1047297cial to bondholders perhaps due to the coinsurance effect which is most
acute for1047297nancially troubled1047297rms Surprisingly the Low Z-score dummy is negative rather than positive as expected This is probably
17 We note the difference between our de1047297nition of the CEO power and theirs they de1047297ne a CEO as being powerful if heshe is the founder of the 1047297rm the sole
insider on the board and the chairman of the board In undocumented analysis we control for the dummies indicating whether the CEO is also the chairman and
whether heshe is the sole insider on the board The results are similar although these dummies are not signi1047297cant We do not have the data for whether the CEO
is also the founder of the 1047297rm But there should be little reason to suspect any bias on our L-index18 The results using the market-to-book ratio as a proxy for the growth opportunities are similar
Table 4
LLI provisionsgrowth opportunities and corporaterisk-taking Thesemodels present the results of pooled OLS regressions for the effect of director limited liability
and indemni1047297cation provisions on the risk-taking incentives of the board conditional on a 1047297rms growth opportunities The sample consists of the SampP 1500 1047297rms
over 2002 and 2007 The dependent variable for Model 1 is the RampD expenses as a percentage of total assets The dependent variable for Model 2 is the capital
expenditure as a percentage of total assets The dependent variable for Model 3 is the number of diversifying acquisitions the 1047297rm made in a given year with a
diversifying acquisition de1047297ned as the 1047297rm acquiring another 1047297rm not in the same industry as de1047297ned by the Fama-French 48 industries The dependent variable
forModel4 is thestandard deviationof monthly stock returns over thepast 60 months High laggedQ is a dummy variable that equalsone if thelagged TobinsQ is
above the sample median and zero otherwise See the Appendix for the de1047297nitions of other variables All models include the control variables as in the
corresponding models in Table 3 the Fama-French 48-industry and year dummies and a constant term Thesevariables are omitted to save space Standard errors
are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and1 levels respectively
(1) (2) (3) (4)
RampD Capexp Numdaqr Volatility
L-index 0003
(3208)
minus0004
(minus3124)
0010
(0727)
minus0006
(minus1226)
High lagged Q L-index minus0004
(minus1814)
0002
(0922)
0037
(1805)
minus0015
(minus2215)
High lagged Q 0009
(1625)
0012
(2468)
0077
(1287)
0024
(1028)
G-index minus0000
(minus1589)
0000
(0863)
minus0003
(minus0693)
minus0003
(minus1872)
High lagged Q G-index 0000
(0477)
minus0000
(minus0618)
minus0003
(minus0483)
minus0001
(minus0242)
Observations 7225 7225 7225 5103Adjusted R 2 056 040 033 055
93M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1225
the result of multicollinearity the interaction term Low Z-score G-index and G-index are highly correlated Low Z-score dummy and
some of the 1047297nancial controls including ROA leverage ratio and interest-coverage ratio are also highly correlated19
33 LLI provisions and directorial shirking
The results in Subsection 32 are consistent with the hypothesis that directors protected by LLI provisions have the
incentives and ability to pursue their self interests and conduct low-risk corporate policies which bene1047297t corporate
bondholders However self-serving behaviors that involve directorial malfeasance or shirking do not necessarily aid
bondholders If the major function of corporate directors is to monitor management shirking may result in a lower
monitoring quality which would be harmful to all security holders 20 We test this hypothesis by examining two proxies
Table 5
LLI provisions directorial incentives and the cost of debt These models present the results of pooled OLS regressions for the effect of director limited liability and
indemni1047297cation provisions on bond spreads conditional on a 1047297rms closeness to 1047297nancial default and agency costs The sample consists of the SampP 1500 1047297rms
between 2002and 2007with public seniorunsecured bondsThe dependentvariable forall themodels is the bond spread de1047297nedas theweightedaverage spread
of allthe seniorunsecured bonds of the1047297rmwith theweight being theissue size of thebond as a percentage of thetotal aggregateissuesizes of all theoutstanding
bonds in the sample of the issuing 1047297rm in a given year The spread for an individual bond is de1047297ned as the difference between the yield-to-maturity (YTM) of the
bond and the YTM of the Treasury Bond matched by the closest maturity Speculative is a dummy variable that equals one if the credit rating is speculative-grade
(Rateb4) and zero otherwise Low Z-score is dummy variable that equals one if the 1047297rms Altmans Z-score is at or below the sample median and zero otherwise
High free cash 1047298ow is a dummy variable that equals one if the 1047297rms free cash 1047298ow is above the sample median and zero otherwise where the calculation of free
cash 1047298ow follows Lehn and Poulsen (1989) No blockholder is a dummy that equals one if the 1047297rm does not have an institutional blockholder with at least 5ownership and zero otherwise See the Appendix for the de1047297nitions of other variables All models use pooled OLS regressions and include the control variables as
inModel3 of Table 2 theFama-French48-industryand year dummies anda constant term These variables areomitted to save spaceStandarderrors areadjusted
for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels
respectively
(1) (2) (3) (4) (5)
Spread Spread Spread Spread Spread
L-index minus0316
(minus2377)
0013
(0099)
minus0379
(minus3265)
minus0580
(minus3121)
minus0499
(minus3759)
Speculative L-index minus0137
(minus0476)
Speculative 0845
(0770)
G-index 0150
(2564)
minus0034
(minus0501)
0146
(2494)
0184
(2218)
0171
(2655)Speculative G-index 0019
(0152)
Low Z-score L-index minus0408
(minus2021)
Low Z-score minus2217
(minus1848)
Low Z-score G-index 0245
(2193)
PPS minus1259
(minus2971)
minus1279
(minus3141)
minus1280
(minus3086)
High free cash 1047298ow L-index 0406
(1955)
High free cash 1047298ow 0483
(0630)
High free cash 1047298ow G-index minus0084
(minus1041)No blockholder L-index 0688
(3091)
No blockholder 0168
(0211)
No blockholder G-index minus0050
(minus0628)
Observations 1494 823 1487 1484 1487
Adjusted R 2 035 043 035 035 035
19 In undocumented analysis we 1047297nd that the correlation between the Low Z-score dummy and bond spread is 027 and the correlation between the interaction
term Low Z-score G-index and G-index is 097 suggesting that the unexpected sign on Low Z-score dummy may be due to multicollinearity We also run the
spread regression without the interaction term Low Z-score G-index ROA leverage interest coverage ratio Low Z-score dummy is positive and strongly
signi1047297cant at the 5 level consistent with expectations The interaction term Low Z-score L-index is still negative but no longer signi1047297cant20 A lower monitoring quality may also enable managers to pursue corporate strategies for their own bene1047297ts As argued in this paper such strategies may
include taking low-risk corporate policies Thus shirking by itself is not necessarily harmful to bondholders However here we are examining the consequences
of shirking by directors in deciding to retain or dismiss a 1047297rms CEO and setting her compensation Shirking in this context may be harmful to the 1047297rmsbondholders
94 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1325
for the monitoring quality of corporate directors CEO turnover-performance sensitivity (TPS) and pay-performance
sensitivity (PPS) It is expected that if directors monitor more diligently both TPS and PPS should be higher The results of
these tests are presented in Table 6
Model 1 reports the results of a probit regression on TPS The control variables follow the literature ( Weisbach 1988 Denis etal 1997 Huson et al 2001 Dahya et al 2002) We delete the observations if the 1047297rm was a takeovertarget in a given year as the
Table 6
LLI provisions and directorial shirking These models present the results of panel regressions for the effect of director limited liability and indemni1047297cation
provisions on CEO turnover-performance sensitivity (TPS)and pay-performancesensitivity(PPS) The sample consists of the SampP1500 1047297rmsand coversyears from
2002 to 2006 for the TPS regression and from 2002 to 2007 for the PPS regression The dependent variable for Model 1 is the CEO turnover de1047297ned as a dummy
variable that equals one if the current CEO is no longer the CEO next year and zero otherwise Firms which are the takeover targets are deleted from thesample in
Model 1 The dependent variable for Model 2 is the change of CEO total compensation over previous year in $100000s Firms whose CEOs have experienced
turnover in current or previous year or whose CEOs tenure is less than two years are deleted from the sample High L-index (G-index) is a dummy variable
that equals one if the 1047297rms L-index (G-index) is above the sample median and zero otherwise Δ denotes the change of a variable over previous year See the
Appendix for the de1047297nitions of other variables Both models include the Fama-French 48-industry and year dummies and a constant term Model 1 uses probit
model Model 2 uses pooled OLS model Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
CEO turnover ΔCEO compensation
L-index minus0003
(minus0080)
0141
(0159)
Return High L-index minus0034
(minus0164)
G-index minus0008
(minus0530)
0277
(0996)
Return High G-index minus0169
(minus0861)
Return minus0348
(minus2807)
ΔSW 0003(3116)
ΔSW High L-index minus0002
(minus2190)
ΔSW High G-index 0003
(2859)
CEO_share minus3184
(minus2903)
CEO_power 0081
(2835)
CEO_tenure minus0005
(minus0927)
CEO_age 0017
(3315)
Per_ind minus0155
(minus0642)
Board_size 0059(3263)
Per_inst 0439
(1991)
Size minus0037
(minus1355)
Volatility 0206
(0943)
ΔPer_ind 7917
(0821)
ΔBoard_size minus0226
(minus0252)
ΔPer_inst 4753
(0605)
ΔSize 12523
(2543)
ΔVolatility 4289(0338)
ΔROA minus4087
(minus0516)
ΔMb 0431
(0358)
Observations 4237 4202
Pseudo R 2 006
Adjusted R 2 004
95M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1425
CEOs of these 1047297rms may experience some abnormal turnover The interaction term between the High L-index (High G-index) and
the stock return is not signi1047297cant where the High L-index (High G-index) is a dummy indicating whether the 1047297rms L-index
(G-index) is higher than the median value in the sample These results suggest that neither index is related to CEO turnover
We follow Jensen and Murphy (1990) and run an OLS regression on the change in total CEO compensation (ΔCEO
compensation) on the change in market capitalization (ΔSW) and interact the High L-index (High G-index) dummy withΔSW to
test the effect of the L-index (G-index) on PPS We also control for the change in independent directors board size institutional
shareholdings1047297rm size equity volatility ROA and the market-to-book ratio Following Coles et al (2007) we delete the 1047297rm-year
observations that experienced a CEO turnover in the current or previous year and require that the CEOs tenure to be at least two
years This ensures that we do not include the change in CEO compensation for fractional years Interestingly the interaction term
High L-index ΔSW is negative and strongly signi1047297cant at the 1 level Moving from a low L-index 1047297rm to a high L-index 1047297rm
decreases the PPS from 0003 to 0001 In contrast moving from a low G-index 1047297rm to a high G-index 1047297rm doubles the PPS from
0003 to 0006
Though the results in Table 6 show that more LLI provisions are associated with less diligent directors (directorial
shirking) it is not clear whether a lower PPS is bad for bondholders as it is for shareholders since a lower PPS may
suggest that the CEO compensation is more aligned with the claim structure of bondholders21 We test this by controlling
for PPS in the spread regression where the PPS is de1047297ned as ΔCEO compensation
ΔSW The results are shown in Model 3 of
Table 5 Indeed PPS is negative and strongly associated with bond spreads suggesting that a lower PPS is also harmful for
bondholders
To further test the shirking hypothesis we examine whether bondholders bene1047297t less from a higher L-index if the agency costs
of the 1047297rm are high The hypothesis is that diligent directorial monitoring should be more valuable for 1047297rms with higher agency
costs and hence the lower monitoring quality associated with the LLI provisions would hurt security holders more for those1047297
rmsincluding bondholders We use two proxies for a 1047297rms agency costs free cash 1047298ow ( Jensen 1986) and whether the 1047297rm has an
institutional blockholder (Bertrand and Mullainathan 2001) It is expected that 1047297rms with higher free cash 1047298ows and 1047297rms
without an institutional blockholder incur higher agency costs We follow Lehn and Poulsen (1989) to calculate the free cash1047298ow
and interact a dummy indicating whether the free cash 1047298ow is above the sample median with the L-index (G-index) in the spread
regression Model 4 in Table 5 shows the results Consistent with the shirking hypothesis the interaction term High free cash
1047298ow L-index is positive and weakly signi1047297cant at the 10 level which suggests that 1047297rms with higher agency costs bene1047297t less
from a higher L-index In Model 5 we interact the dummy for whether the 1047297rm has an institutional blockholder with the L-index
(G-index) The interaction term (No blockholder L-index) is positive and strongly signi1047297cant at the 1 level In fact the
coef 1047297cients on this interaction term (0688) and L-index (minus0499) indicate that if the 1047297rm does not have an institutional
blockholder a higher L-index hurts bondholders It is also noteworthy that none of the interaction terms involving the G-index is
signi1047297cant
34 The incentive of shareholders to rescind the LLI provisions
In this subsection we examine the incentives of shareholders to rescind LLI provisions Since most of these provisions were
established in the mid- to late 1980s during the heyday of hostile takeovers ( Bradley et al 1988) we focus on rescissions or the
lack thereof rather than initiations of LLI provisions Our previous results suggest that LLI provisions reduce the cost of debt which
presumably bene1047297ts shareholders The costs of these provisions however are suboptimal low-risk operating strategies and
directorial shirking The relative magnitudes of the bene1047297t and costs of these provisions determine their net effect on 1047297rm value
We therefore examine how LLI provisions affect 1047297rm performance to gauge the incentive of shareholders to rescind these
provisions
We 1047297rst examine whether there is any evidence that the number of LLI provisions is positively associated with a 1047297rms litigation
risk as an indication for whether 1047297rms implement these provisions optimally An optimal adoption of these provisions implies a
lack of incentives of shareholders to change them We follow the existing literature and practitioners advice in choosing the 1047297rm-
level variables that may be related to litigation risk It is expected that directors of larger1047297
rms1047297
nancially distressed1047297
rms growth1047297rms and1047297rms with more shareholders may bene1047297t more from the LLI provisions as these 1047297rms are more likely to be sued(Core
1997 Gische 2000) We use Altmans Z-score to proxy for the extent of a 1047297rms 1047297nancial distress Model 1 in Table 7 presents the
OLS regression results with the L-index as the dependent variable As expected 1047297rm size and the number of shareholders are
positively and the Z-score is negatively related to the L-index The coef 1047297cient on market-to-book is positive but not signi1047297cant
Overall the results support the claim that 1047297rms useLLI provisions rationally in that 1047297rms with higherlitigation risk also grant their
directors more provisions22
21 Indeed in unreported analysis we 1047297nd that the L-index (G-index) is associated with higher (lower) CEO cash compensation22 These results are best interpreted as the average effect of these litigation risk factors on the cross-sectional variation of the LLI provisions since these factors
1047298uctuate much more than the provisions Our intent here is to examine that even though 1047297rms seem not have changed these provisions often once they were
established in the 1980s the establishment and strength of them may be driven by the litigation risk The results suggest that the cross-sectional variations for
the ranking of these litigation factors may be much smaller than their time-series 1047298uctuations hence even using a sample as recent as ours the results still
suggest that litigation risk is positively correlated with the L-index But there is a caveat to interpreting these results the regression coef 1047297cients are understoodas cross-sectional correlations rather than causal effects
96 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1525
We now examine how the L-index affects 1047297rm performance We use Tobins Q as a proxy for performance following the
convention in the literature Control variables include 1047297rm size ROA intangible assets leverage RampD and capital expenditures
1047297rm age number of business segments and a dummy variable indicating whether the 1047297rm is in the SampP 500 index (SP500) The
costs of suboptimal operating strategies and directorial shirking on shareholders are well established in the literature There is
Table 7
LLIprovisions andlitigation risk and1047297rm performance These models present theregression results forlitigation risk on thenumber of director limited liability and
indemni1047297cation provisions the effect of the cost of debt on 1047297rm performance and the effect of the LLI provisions on 1047297rm performance The samples for Models 1
and 4 consist of the SampP 1500 1047297rms over 2002 and 2007 The sample for Model 2 consists of the SampP 1500 1047297rms between 2002 and 2006 with public senior
unsecured bonds The sample for Model 3 consists of the SampP 1500 1047297rms between 2002 and 2007 with public senior unsecured bonds The dependent variable for
Model 1 is the number of director limited liability and indemni1047297cation provisions at the 1047297rm level The dependent variable for Model 2 is the change of Tobins Q
over previous year The dependent variable for Models 3 and 4 is Tobins Q ldquoPredicted reduction of the cost of debt due to L-index rdquo is the predicted component of
the bond spread due to the L-index based on Model 3 of Table 2 (without the negative sign) scaled by total bond spread Δ denotes the change of a variable over
previous year See the Appendix for the de1047297nitions of other variables All models include a constant term Models 1 3 and 4 further include the Fama-French 48-
industry and year dummies The coef 1047297cient for Z-score has been multiplied by 1000 Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rmlevel T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
L-index ΔQ Q Q
Size 0128
(6537)
minus0168
(minus3598)
minus0195
(minus6097)
Z-score minus0003
(minus3369)
MB 0019
(0973)
Shareholder 0094
(4282)
ΔSpread minus0004
(minus2806)
ΔRate 0062(2055)
Predicted reduction of the cost
of debt due to L-index
0015
(2140)
L-index 0012
(0377)
G-index minus0015
(minus2016)
minus0021
(minus2373)
ΔROA 1159
(3857)
ΔSize minus0619
(minus4753)
ΔIntangible minus0348
(minus0852)
ΔLeverage 0359
(1210)
ΔRampD minus2269(minus0821)
ΔCapexp 2002
(3631)
ΔSegment 0001
(0236)
ROA 5004
(5711)
2386
(7616)
Intangible 0313
(1818)
0897
(5594)
Leverage 0817
(2384)
0107
(0541)
RampD 10989
(6030)
5604
(8594)
Capexp 2405
(3319)
4923
(8481)
Firm_age minus0000(minus0379)
minus0002(minus0971)
Segment 0006
(0817)
minus0007
(minus0938)
SP500 0339
(4684)
0837
(8483)
Observations 5397 1628 2208 7297
Adjusted R 2 014 014 053 036
97M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1625
Table 8
LLI provisions and attraction and retention of talented directors These models present the results of pooled OLS regressions for the effect of director limited
liability and indemni1047297cation provisions on the attraction and retention of talented directors during the sample period The sample for Model 1 consists of the SampP
1500 1047297rms between 2002 and 2006 The samples for Models 2ndash5 consist of the SampP 15001047297rms between 2002 and 2007 The dependent variable for Model 1 is the
numberof departingdirectors as a fraction of board sizein a given year Thedependent variable forModel2 is thenumber of newly added directorsas a fraction of
board size Thedependent variable forModel3 is theaverage numberof outside directorships fornewlyaddeddirectorsThe dependentvariable forModel 4 is the
average number of outside directorships for the whole board The dependent variable for Model 5 is the number of independent directors who are active
executives of other 1047297rms as a fraction of board size See the Appendix for the de1047297nitions of other variables Δ denotes the change of a variable over previous year
Firms which are the takeover targets are deleted from the samples for Models 1 and 2 All models include the Fama-French 48-industry and year dummies and a
constant term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
L-index 0001
(0285)
0002
(0748)
minus0046
(minus1535)
minus0007
(minus0445)
minus0003
(minus1065)
G-index 0000
(0056)
minus0000
(minus0031)
0013
(1176)
0011
(2010)
0003
(2892)
Return minus0012
(minus2627)
0008
(1848)
minus0013
(minus0191)
minus0027
(minus1377)
Return at t-1 minus0010
(minus2482)
minus0003
(minus0652)
minus0088
(minus1508)
minus0030
(minus1665)
Return at t-2 minus0003
(minus0911)
minus0002
(minus0474)
minus0006
(minus0110)
minus0023
(minus1722)
ROA minus0020
(minus0987)
minus0011
(minus0807)
minus0008
(minus0041)
minus0277
(minus3315)
Size 0001
(0984)
0004
(2884)
0140
(5878)
0172
(15616)
0012
(6193)
Dir_age 0002
(3253)
minus0001
(minus1467)
0018
(2072)
0017
(3787)
Dir_tenure minus0001
(minus3069)
minus0004
(minus2141)
minus0008
(minus1713)
minus0015
(minus2906)
Dirship 0006
(1677)
minus0002
(minus0607)
Dir_share 0001
(0042)
0008
(0456)
0164
(0683)
0107
(0536)
CEO out 0013
(3787)
0011
(3568)
ΔPer_inst 0033
(1661)
0041
(2501)
Mb 0051
(1750)
0036
(2960)
0004
(2154)
Board_size minus0032
(minus2327)
minus0026
(minus4373)
Interlock minus0189
(minus1849)
0105
(1917)
Num_oe 0197
(5086)
0271
(14920)
Per_noe 0351
(1048)
0877
(7223)
Blk_idir 0221
(1671)
minus0082
(minus1686)
Blockholder 0025
(1143)
0015
(2020)
Idir_share minus0042
(minus1164)Firm_age 0001
(5112)
Segment minus0001
(minus0842)
RampD 0050
(0935)
Leverage minus0024
(minus1941)
CEO_share minus0005
(minus0152)
CEO_tenure minus0002
(minus5799)
CEO_age minus0000
(minus0262)
Lagged board_size minus0000
(minus0356)
98 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1725
little empirical evidence on the bene1047297t of a reduced cost of debt on 1047297rm performance however23 We therefore 1047297rst examine
whether bond spreads (credit ratings) are negatively (positively) associated with Tobins Q To reduce the concerns of
endogeneity we adopt a 1047297rst-difference (change) regression In addition to the change in bond spreads and credit ratings we also
include changes of each of the control variables as mentioned above except for 1047297rm age The change of SP500 is not included in
this regression due to the lack ofvariation of this variable within the sample period Model 2 in Table 7 presents the results Indeed
the change of bond spread (credit rating) is negatively (positively) associated with the change of Tobins Q suggesting that a
reduced cost of debt is associated with an improved 1047297rm performance as expected
In Model 3 we further examine this by testing whether the predicted reduction of the cost of debt due to our L-index is
associated with higher performance similar to the methodology in Core et al (1999) We use the empirical model in Model 3 of
Table 2 to calculate the predicted reduction of the cost of debt due to the L-index scaled by bond spread (=0370 L-index
Spread) We use an OLS model rather than a change regression since the stability of the L-index implies that the variability of this
predicted reduction of the cost of debt is mainly driven by its denominator Therefore a change regression on this variable is
similar to Model 2 and is not meaningful We also control for the G-index in the OLS model The results are presented in Model 3
The predicted reduction of the cost of debt due to the L-index is indeed positively associated with Tobins Q Furthermore
consistent with GIM the G-index is associated with lower 1047297rm performance
The results in Models 2 and 3 are consistent with our expectation that a reduced cost of debt associated with LLI provisions
bene1047297ts shareholders which results in improved1047297rm performance However this bene1047297t needs to be balanced against the costs of
these provisions as documented earlier Therefore in Model 4 we examine the net effect of the L-index on 1047297rm performance The
results show that the L-index is positive but not signi1047297cantly associated with1047297rm performance which suggest that the bene1047297t o f a
lower debt 1047297nancing cost may offset the costs of suboptimal operating strategies and directorial shirking induced by LLI
provisions24 Therefore shareholders may have little incentive to alter these provisions This is consistent with the remarkable
stability of the L-index as we observed in Table 1
The results in this section call for a fresh look at the widely-studied G-index Our evidence suggests that 1047297rms with a higher
G-index have a higher PPS which should bene1047297t both shareholders and bondholders We also provide weak evidence that the
G-index is associated with lower 1047297rm risk All these results suggest that bondholders should bene1047297t from the provisions in the
G-index However evidence presented in Table 2 shows the opposite effect Furthermore the relation between performance
and the G-index is negative These results are hard to reconcile with the interpretation of the G-index as proxying for weak
shareholder rights However our evidence is compatible with the interpretation of the G-index as a measure of the antitakeover
strength of the 1047297rm and that takeovers are bene1047297cial to bondholders (Billett et al 2004 Qiu and Yu 2009)
4 Test of the ldquotalent attractionrdquo hypothesis
The literature discussed previously offers an alternative explanation for the bene1047297cial effects of LLI provisions they may help
the 1047297rm to attract and retain talented directors It could well be the case that this effect may drive the negative relation between
the L-index and the cost of debt In this section we examine thisalternative hypothesis by testing whether ourL-index is associated
with a lower director departure rate andor a higher director addition rate and whether the L-index is associated with higher
ability of the 1047297rms board members25
Director turnover is de1047297ned as the percentage of the board members who experienced a turnover in a given year Addition
variables are de1047297ned similarly In addition to the L-index and the G-index other control variables included in the analysis follow
Farrell and Whidbee (2000) and Yermack (2004) We run OLS regressions for both departure and addition models The results are
presented as Models 1 and 2 in Table 8
Table 8 (continued)
(1) (2) (3) (4) (5)
Dpt_dir Add_dir Add_dirship Dirship Per_ie
MampA at t-1 0000
(0077)
ROA at t-1 0009
(0819)
Observations 3766 4020 1917 4771 3329Adjusted R 2 002 009 008 043 017
23 We thank an anonymous referee for pointing this out24 We also use accounting measures of performance to further examine this result As it turns out L-index is positive and signi1047297cant (at the 1 level) for ROA
and positive but insigni1047297cant for ROE25 We also examine whether LLI provisions are related to the attraction and retention of talented independent directors as well as active executives from other
1047297rms since our sample period is characterized by an increased demand and a decreased supply of these types of directors ( Chen 2007 Linck et al 2009) LLI
provisions could be most valuable for 1047297rms during this time period in attracting and retaining these directors However the results are not signi1047297cant similar tothe results with respect to the whole board
99M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1825
It is notable that this sample period favors 1047297nding the talent attraction effect of LLI provisions since it is characterized by
heightened litigation risk for corporate directors after the conspicuous corporate scandals as evidenced by Enron and Worldcom
Yet our results in Models 1 and 2 do not support this hypothesis the L-index is not signi1047297cant in either model We note that the G-
index is also insigni1047297cant in these regressions These results suggest that 1047297rms with a high L-index or G-index are no different in
attracting and retaining directors as other 1047297rms26
Even though LLI provisions may not affect the departure or addition of directors they may still help 1047297rms to recruit talented
directors We test this hypothesis by examining whether the L-index is associated with the addition and retention of talented
directors We rely on two proxies for the ability (talent) of a board the number of outside directorships held by the board
members and the percentage of the board members who are independent and are active executives from other 1047297rms (Fich 2005
Chen 2007) We run three models with the following dependent variables (1) the average number of outside directorships for
newly added directors (2) the average number of outside directorships for the entire board and (3) and the percentage of the
board members who are active executives of other1047297rms The control variables are based on Booth and Deli (1996) for the1047297rst two
regressions and Boone et al (2007) for the third regression The results are presented in Models 3 ndash5 of Table 8
Again the L-index is not signi1047297cant in any of these models Furthermore the coef 1047297cient on the L-index is negative rather than
positiveas predicted Incontrastthe G-index is positiveand signi1047297cant for theaveragedirectorshipsof thewhole board andthe percent
of independent active executives suggesting that the G-index rather than L-index may help to attract more talented directors27
Overall the results in this section do not support the generally held view that LLI provisions help to attract and retain talented
directors28
5 Robustness checks
In this section we perform a number of robustness checks to further substantiate our 1047297ndings
51 DampO insurance and omitted variable bias
Since DampO insurance also protects directors from potential litigation it may incentivize directors similar to the LLI provisions
However data on the type and amount of DampO insurance held by the directors of public corporations are con1047297dential and virtually
impossible to obtain29 This data de1047297ciency may give rise to a potential omitted variable bias We discuss the robustness of our
empirical results in light of this bias in this subsection
First as argued in Bhagat et al (1987 p726) DampO insurance may not have as strong of an incentive effect as LLI provisions The
presence of DampO insurance may create a stronger incentive for shareholders to sue and lawsuits often damage the reputations of
managersdirectors This potential loss constrains to a certain extent managerial malfeasance In contrast director limited liability
provisions preclude shareholders from suing the directors on the basis of a breach of duty of care Moreover indemni1047297cation does
not induce stockholders to sue more frequently since there is no external reservoir from which to pay the damages Boyer (2005)even argues that the purpose of DampO insurance is not to protect corporate directors per se but shareholders themselves since
directors are already protected by LLI provisions
Second Core (2000) and Chalmers et al (2002) argue that the purchase of DampO insurance represents opportunistic behavior on
the part of managers and as such is often closely related to the quality of corporate governance If this is the case our extensive
control of governance variables in the cost of debt regressions may signi1047297cantly alleviate this omitted variable bias
Finally even if our governance controls do not fully capture the effects of DampO insurance the direction of the potential omitted
variable bias depends on whether DampO insurance and LLI provisions are substitutes or complements Prior literature is not very
helpful in answering this question But since our sample period is characterized by dramatic developments in the DampO insurance
market (Tillinghast-Towers Perrin (2004 2005 2006)) the stability of the LLI provisions imply that these provisions are
26 Since SOX and the new exchange listing standards impose higher independence standards on public 1047297rms these results may be driven by the forced
compliance by the regulations We exclude all 1047297
rms that were not in compliance with the regulations in 2002 by having less than a majority of independentdirectors on their boards The results are similar27 We are aware of the possibility that the number of directorships may not proxy for the effectiveness of directors even though the directors holding more
directorships may be abler Fich and Shivdasani (2006) document that if the majority of outside directors hold at least two outside directorships and hence are
considered ldquo busyrdquo the board will not be effective To further gauge the effect of the L-index (and the G-index) on the ability and effectiveness of the board we
rerun the regressions in Table 8 and include a dummy variable indicating whether the directors of the 1047297rm are busy which is de1047297ned as having at least one
director holding at least two outside directorships in the regression of the average directorship of newly added directors and the majority of the board members
holding at least two outside directorships in the regression of the average directorship of the whole board The results are similar In particular the G-index does
not appear to attract ldquobusyrdquo directors28 We realize ability is inherently hard to measure To further test the talent attraction hypothesis we examine whether the LLI provisions bene1047297t bondholders
more when the 1047297rms demand for abler directors is high Our 1047297rst variable to proxy for this demand is a principal factor from the standardized values of 1047297rm size
the number of business segments and leverage following Coles et al (2008) We also follow Brook and Rao (1994) and use a variable which indicates whether
the 1047297rm is underperforming We interact the dummies for higher advising demand based on these two variables with the L-index in the spread regressions The
coef 1047297cients on the two interaction terms are both insigni1047297cant inconsistent with the talent attraction hypothesis29 Linck et al (2009) searched all proxy and 1047297nancial statements of New York incorporated 1047297rms (New York requires 1047297rms to disclose information about DampO
insurance) and SampP 500 1047297rms in 1998 2001 and 2004 but were only able to 1047297nd information for DampO insurance for 27 New York 1047297rms and 12 SampP 500 1047297rms Our
sample only has 73 New York incorporated 1047297rms Therefore it would be futile to search the DampO insurance information for these 1047297rms and run regressions tocheck the robustness of our results
100 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 1925
substitutes rather than complements to DampO insurance30 Hence assuming DampO insurance has similar incentive effects as LLI
provisions our regression results will be biased downward This suggests that the signi1047297cant results we document would be even
stronger without the bias and hence we only need to be careful in interpreting the insigni1047297cant results in particular the results in
our tests of the talent attraction hypothesis Based on the results reported in Table 8 the results for the attraction of new directors
(Model 2) may be mostly affected by this bias since the signs of the coef 1047297cients are positive as expected but statistically
insigni1047297cant However even if adding DampO insurance data would render the results signi1047297cant the results in Model 3 of Table 8
suggest that the added directors for 1047297rms with a higher L-index are no abler than other 1047297rms (or may even be less abler
considering the bias) which is still inconsistent with the talent attraction hypothesis
52 Endogeneity
The evidence in this paper is consistent with the observation that most LLI provisions were established before 1990 and
changed little afterwards This observation should alleviate much of the concerns regarding endogeneity especially the issue of
reverse causality Nevertheless we employ change regressions in this subsection to further check the robustness of the results for
the effect of LLI provisions on the cost of debt The change regressions are immune to the bias induced by an omitted but time-
invariant variable that is correlated with both the L-index and the cost of debt The stability of the L-index as evidenced in Table 1
however suggests that most of the observations on the change of the L-index (ΔL-index) are zeros which will dilute the power of
the tests dramatically This issue is especially acute since during our sample period there were only 18 1047297rms in our sample that
changed their L-index31 To reach a balance between sample size and the power of the tests we match these 18 1047297rms with 1047297rms
that did not alter their L-indexes based on industry and similar characteristics We then run change regressions for bond spreads
based on samples that are composed of these 181047297rms and their matched 1047297rms In addition toΔL-index we also include the change
of the G-index (ΔG-index) in the regressions since these two changes may be highly correlated We use four factors for the second
matching variable (in addition to industry) 1047297rm size leverage ratio equity volatility and credit rating Leverage and equity
volatility are used as the matching criteria because they are highly signi1047297cant for the cost of debt based on the results in Table 2
The results of the change regressions are reported in Table 9
32
Regardless of the matching criteria the results in this table show that an increase in the L-index (G-index) is associated with a
decrease (an increase) in the bond spread consistent with our results reported in Table 2 Therefore the results using the change
regressions though with small sample sizes nonetheless are consistent with our earlier 1047297ndings
We further check the robustness of our results by rerunning the ratings and spread regressions at 200633 and include only the
1047297rms that kept their L-indexes constant between 2002 and 2006 The results are also similar qualitatively
30 This is also consistent with the historical origin of the director liability statutes in the mid-1980s which was a result of states responding to skyrocketing
costs of DampO insurance (Romano 2006) due primarily to the ruling in Smith v Van Gorkom See discussion in footnote 331 The difference between this number and the number we reported in Panel B of Table 1 is due to the additional requirement here that we have the data
available for the change of bond spreads32 Model 3 only has 34 observations because for one 1047297rm that changed its L-index during the sample period there is not a matched 1047297rm based on industry and
equity volatility33 The regression is run at year 2006 instead of year 2007 because CEO option data has a large number of missing observations in 2007
Table 9
First-difference regression LLI provisions and the cost of debt These models present the change regression results of the director limited liability and
indemni1047297cation provisions on bondspreads Thesamples areobtainedby matching1047297rms thatchangedtheir L-index with1047297rms thatkept their L-index constant
between 2002 and 2007 by industry (according to Fama-French 48 industries) and either 1047297rm size or book leverage ratio or equity volatility or credit rating
Thedependent variable foreachmodel is thechange of thebondspread over previousyear wherebond spread is de1047297nedas theweightedaveragespread of all
thesenior unsecured bonds of the1047297rmwiththe weightbeing theissue size ofthe bond as a percentageof thetotal aggregateissue sizesof alloutstanding bonds
in thesample of theissuing1047297rmin a givenyearThe spreadforan individual bond is de1047297ned as the difference between the yield-to-maturity(YTM) of the bond
and the YTM of the Treasury Bond matched by the closest maturity See the Appendix for the de1047297nitions of other variables All models also include a constant
term Standard errors are adjusted for heteroscedasticity and clustered at the 1047297rm level T-statistics are reported in parentheses and indicate
signi1047297cance at the 10 5 and 1 levels respectively
(1) (2) (3) (4)
ΔSpread ΔSpread ΔSpread ΔSpread
Sample Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and 1047297rm size
Firms with ΔL-indexne0
matched with ΔL-index=0
by industry and leverage
Firms with ΔL-indexne0
matched with ΔL-index= 0 by
industry and equity volatility
Firms with ΔL-indexne0 matched
with ΔL-index=0 by industry and
credit rating
ΔL-index minus1615
(minus2341)
minus1904
(minus3433)
minus1744
(minus2502)
minus1359
(minus2178)
ΔG-index 1229
(2390)
1472
(3512)
1322
(2601)
1086
(2273)
Observations 36 36 34 36
Adjusted R 2 013 038 011 015
101M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2025
Table 10
2SLS regression credit ratings andbond spreads This table reports theresults of thesimultaneousequationsestimation for credit ratingsand bondspreads Thesample
consistsof theSampP15001047297rmsbetween2002 and2007with publicsenior unsecuredbondsThe dependentvariableforModel1is the1047297rmscredit rating(Rate)de1047297nedby
groupingSampP long-termdomesticissuer credit ratings (data280 in COMPUSTAT)into sevencategories Seethe Appendix for the details The dependentvariablefor Model
2 isthebondspreadde1047297nedas theweightedaverage spreadof allthe seniorunsecured bonds of the1047297rmwith theweight being theissuesizeof thebondas a percentage
of the total aggregate issue sizes of all outstanding bonds in the sample of the issuing1047297rm in a given year Thespread for an individual bond is de1047297ned as the difference
between theyield-to-maturity(YTM) of thebond andthe YTMof theTreasury Bond matched by closest maturitySee theAppendix for thede1047297nitionsof othervariables
Both models include the Fama-French 48-industry and year dummies and a constant term Standard errorsare adjusted for heteroscedasticity and clusteredat the1047297rm
level T-statistics are reported in parentheses and indicate signi1047297cance at the 10 5 and 1 levels respectively
(1) (2)
Rate Spread
Rate minus0041
(minus0196)
L-index 0020
(1898)
minus0363
(minus3115)
G-index minus0005
(minus0981)
0146
(2476)
Spread minus0021
(minus3186)
Lagged rate 0813
(47747)
Duration minus14131
(minus7413)
Convexity 2564(3821)
Bond_age 0185
(4308)
Operating cash 1047298ow 0619
(2022)
minus3788
(minus1927)
ROA 0995
(2890)
minus0141
(minus0068)
Loss 0057
(1076)
1585
(2644)
Size 0046
(4131)
0157
(1107)
Leverage minus0144
(minus1481)
2270
(2126)
Intcov 0001
(1591)
0012
(2123)
Subord 0007(0270)
minus0082(minus0280)
Volatility minus0313
(minus2649)
2700
(2376)
Cboard 0006
(0299)
minus0593
(minus2501)
CEO_option 0067
(1637)
minus0888
(minus2193)
CEO_share minus0499
(minus2241)
minus0611
(minus0257)
Per_dir_noshare minus0108
(minus1174)
minus1167
(minus1368)
Per_ind minus0033
(minus0476)
0187
(0298)
Board_size minus0001
(minus0286)
minus0039
(minus0672)
CEO_power minus0008(minus1065)
0097(1221)
Expert director minus0028
(minus0664)
minus0693
(minus1426)
Per_inst 0158
(1397)
minus1407
(minus1330)
Blockholder minus0022
(minus2404)
0299
(4001)
Earnings quality 0064
(1376)
minus1498
(minus3432)
Observations 1473 1473
Adjusted R 2 090 034
102 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2125
53 Simultaneous determination of credit ratings and bond spreads
In the spread regressions we have controlled for rating dummies Credit ratings and bond spreads however are likely to be
determined simultaneously We address this simultaneity in this subsection We employ 2SLS models to address this issue34
Following Qiu and Yu (2009) we use the lagged credit rating as the instrumental variable for credit rating and the three variables
for issue characteristics modi1047297ed duration convexity and bond age as the instrumental variables for bond spread Table 10
presents the results35 Similar to the results in Table 2 the L-index continues to be positive for credit ratings and negative for bond
spreads
54 Sample selection bias
Since our cost of debt sample contains only publicly traded bonds this raises concerns regarding a sample selection bias We
use Heckmans selection model to address this bias We follow Ertugrul and Hegde (2008) and use three instruments for access to
the public bond market 1047297rm age inclusion in the SampP 500 index and listing on the NYSE Other control variables in the 1047297rst stage
include 1047297rm size market-to-book ratio ROA equity volatility leverage and the Fama-French 48-industry and year dummies We
calculate the inverse Mills ratio and include it as an additional variable in the spread regression The results are very similar to
those obtained by OLS models The coef 1047297cient on the L-index isminus0362 and the coef 1047297cient on the G-index is 0187 which are very
close to the results we obtained in Model 3 of Table 2 Both are also highly signi1047297cant at the 1 level
55 LLI provisions and state laws
Since 1047297rm-level provisions to limit the liability of directors or indemnify their litigation expenses are governed by state
statutes and state laws may cluster our 1047297ndings of the signi1047297cant effects of LLI provisions on the cost of debt may suggest the
signi1047297cance of other state laws which may be correlated with states liability or indemni1047297cation statutes36 We thus include
controls for the relevant state laws as suggested by the literature to check the robustness of our1047297ndings (Romano 2006 Mansi et
al 2009 Francis et al 2010) We 1047297rst consider the statutes to relax the 1047297duciary standards as well as to cap the damages for
liability litigation payable by directors37 Presumably these statutes may have similar effects as the LLI provisions However
controlling for these statutes does not alter any of our results Thetwo dummy variables that indicate incorporations in a state with
a relaxed culpability standard or damage cap are associated with lower costs of debt but not signi1047297cantly so38 We next consider
state payout restriction laws Mansi et al (2009) document that 1047297rms incorporated in states with more restrictive payout statutes
enjoy a lower cost of debt Including a variable indicating the minimum asset-to-liability ratio for payout however does not
change any of our results39 Finally we consider state antitakeover laws Francis et al (2010) 1047297nd that a larger number of these
laws are associated with lower bond spreads We create an index variable from the RiskMetrics database indicating the number of
antitakeover laws for a 1047297rm incorporated in that state40 None of our results are affected
56 L-index versus L-dummy
Throughout the paper we have used the L-index the sum of the three indicators for the LLI provisions as our proxy for the
strength of the protection at the 1047297rm level to insulate directors from potential litigation This index approach presumes that the
three LLI provisions have similar incentive effects on corporate directors GIM make a similar assumption in creating their 24-
component G-index However this may not be the case empirically In undocumented analysis we 1047297nd that the three provisions
are positively and signi1047297cantly correlated with each other The three provisions are all positively related to ratings and negatively
related to spreads individually Although only the relation between director limited liability provisions and ratings is statistically
signi1047297cant all three provisions are signi1047297cant for spreads We replace our L-index with the L-dummy and rerun all the regressions
The results are similar The summary statistics presented in Table 1 also suggest that 31 of the 1047297rms do not have any LLI
provisions If we delete these 1047297rms from the sample then the effect of the L-index on bond spreads is still negative but no longer
signi1047297cant at the conventional levels41 These results suggest that the initiation of an LLI provision has a larger impact on
directorial incentives and the cost of debt than an addition to preexisting LLI provisions
34 Results using 3SLS models are similar35 In undocumented analysis we also show that the results in Table 2 remain if we use OLS models to run the regressions for credit ratings or control for credit
ratings (instead of the rating dummies) in the bond spreads regressions36 For example most states follow Delawares model for their statutes on director liability limitation These statutes leave the decision to individual 1047297rms
shareholders to authorize the inclusion of the provisions to limit directors liability Therefore 1047297rm-level LLI provisions are governed by the state statutes37 We are indebted to Roberta Romano for providing the coding of these state statutes38 Perhaps this provides one rationale for why these statutes were historically far less popular than the director liability limitation and indemni 1047297cation statutes
Only 6 states have the statutes for relaxed culpability standards while only one has the damage cap statute (Romano 2006)39 We thank John Wald for posting the data for this variable in his website40 These include business combination laws fair price laws control share acquisition laws constituency statutes anti-greenmail (recapture of pro1047297ts) laws and
cash-out laws41 It is signi1047297cant at the 15 level however
103M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2225
57 Additional robustness checks
Finally we conduct several robustness checks to further buttress our major 1047297ndings Ertugrul and Hegde (2008) show that
the amount of stock and options granted to directors is negatively associated with the 1047297rms bond spread We calculate the
annual stock and option awards by dividing the number of annual restricted stocks and options by the number of shares
outstanding and include these as additional controls in the spread regression Including these independent variables
dramatically reduces the sample size to 962 Nevertheless the results are qualitatively similar However director option grant is
positively instead of negatively associated with bond spreads in contrast with Ertugrul and Hegde (2008) while the amount of
stock held by directors is not signi1047297cantly related to spreads The signi1047297cance of director options however disappears when we
run change regressions
We exclude all 1047297nance and utility 1047297rms from the sample (one-digit SIC code ldquo6rdquo and two-digit SIC code ldquo49rdquo) and rerun all the
regressions Theresults are similar We also run thespread regressions with only thelargestissue of each1047297rmin a given year The results
are also qualitatively similar Finally we run thebondspread regressions using allof thebonds in our sample rather than aggregating the
different issues into a 1047297rm-level bond spread We cluster the standard errors at the 1047297rm level The results are also similar
6 Conclusion
The basic thesis of this paper is that managerial self-serving behavior may not be detrimental to bondholders as undisciplined
managers may pursue risk-reducing activities that bene1047297t bondholders Since the reduced cost of debt inures to shareholders
bene1047297t it is also not clear whether weak governance is uniformly bad for shareholders
We suggest that the provision of limited liability and indemni1047297cation (LLI) for corporate directors represents a weak form of
corporate governance in the sense that these provisions allow managersdirectors more latitude to pursue self-interested
operating strategies Because of their concerns for the private bene1047297ts from control and their non-diversi1047297ed human capital
investment managersdirectors have incentives to undertake low-risk operating strategies that result in a lower cost of debt
which ultimately inures to the bene1047297t of the 1047297rms stockholders We provide evidence of a strong negative relation between a
1047297rms cost of debt and the provision of LLI for the 1047297rms directors This bene1047297t to shareholders may offset the costs of suboptimal
operating strategies and shirking by directors Indeed we provide evidence that 1047297rm performance is not signi1047297cantly impacted by
the provision of LLI which may explain why shareholders have little incentive to rescind these provisions even though most of
them were implemented in the 1980s during the boom of hostile takeovers In most cases the provisions were designed to protect
directors from stockholder suits alleging that they breached their 1047297duciary duty by rejecting a premium hostile takeover bid Yet
they persist today perhaps because of the effects we have documented in this paper We also provide evidence that is inconsistent
with the widely held belief that the effect of these provisions is to attract and retain talented directors
The results reported in this paper call for a new interpretation of the widely-studied G-index Contrary to the existing literature
our results indicate that the G-index is negatively related to credit ratings and positively related to bond spreads Our evidence
suggests that it may be inappropriate to interpret the G-index as a proxy for weak shareholder rights Rather our results are
consistent with the view that the G-index measures the 1047297rms antitakeover strength and that takeovers are bene1047297cial to
bondholders presumably because of the coinsurance effect
Acknowledgements
The authors thank the Duke Global Capital Markets Center for 1047297nancial support We also thank the analysts at Standard amp Poors
for providing insights into the ratings process and advice regarding our research design In particular we would like to recognize
Cliff Griep Laurence Hazell and Dan Konigsburg We especially thank George Dallas who has been an invaluable resource
throughout this project We also appreciate the helpful comments of an anonymous referee
Appendix A Variable de1047297nition type and source
Variable De1047297nitions and calculations Typea Data source
Director limited liability and indemni1047297cation
L-index Board liability limitation and indemni1047297cation (LLI) index sum of the three 1047297rm-level charter
bylaw and contract provisions to limit the director liability to the extent allowed by state laws
andor to indemnify the directors for the expenses involved in a liability lawsuits should the
director follow due diligence (dirliab+ dirind+dirindc)
C RiskMetrics
L-dummy Equal s one if t he 1047297rm has any LLI provisions and zero otherwise D RiskMetrics
Cost of debt
Rate Collapsed seven categories out of the SampP long-term domestic issuer credit ratings (data280)
Speci1047297cally Rate=1 if Ratingb=CCC+ Rate=2 if CCC+bRatingb=B+ Rate=3 if
B +bRatingb=BB+ Rate=4 if BB+bRatingb=BBB+ Rate=5 if BBB+bRatingb=A+
Rate=6 if A +bRatingb=AA+ Rate=7 if RatingNAA+
C COMPUSTAT
104 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2325
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Cost of debt
Spread Weighted average spread of all the senior unsecured bonds of the 1047297rm with the
weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the outstanding bonds in the sample of the issuing 1047297rm in a given year
The spread for an individual bond is de1047297ned as the difference between the yield-to-
maturity (YTM) of the bond and the YTM of the Treasury Bond matched by the
closest maturity
C SampP Snapshot
Issue characteristics
Duration Weighted average modi1047297ed duration of all the senior unsecured bonds of the 1047297rm with
the weight being the issue size of the bond as a percentage of the total aggregate issue
sizes of all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Convexity Weighted average convexity of all the senior unsecured bonds of the 1047297rm with the
weightbeing theissue size of thebond as a percentage of the total aggregateissue sizes of
all the bonds in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
B ond_ age Weighted aver age age of all thesenior u nsecu red bonds of the 1047297rmwith theweight being
the issue sizeof the bondas a percentageof the total aggregate issuesizes of all the bonds
in the sample of the issuing 1047297rm in a given year
C SampP Snapshot
Financial variables
Operating cash 1047298ow Operating cash 1047298ow (=(dat a13-(dat a1 6-(data3 5-l agdat a35 )))data6) C COMPUSTAT
ROA Return on asset (=data18data6) C COMPUSTAT
ROA at t-1 Lagged value of return on asset C COMPUSTAT
ROA at t-2 Two-year lagged value of return on asset C COMPUSTAT
Loss Equ als one if ROA is negative in cu rr ent and p reviou s 1047297scal year and zero otherwise D COMPUSTAT
Size Firm size log of total assets (log(data6)) C COMPUSTAT
Leverage Book leverage (=(data9 + data34)data6) C COMPUSTAT
Intcov Inter est- cover age (=d ata13(data15 or data339 ifd ata15=0or d ata15 isnot available)) C C OMPUSTAT
Subord Equals one if the 1047297rm has subor dinated d ebt and zer o otherwise (data80) D C OMPUSTAT
Return The annual dividend reinvested total return to shareholders (trs1yr) C EXECUCOMP
Return at t-1 The one-year lagged return C EXECUCOMP
Return at t-2 The two-year lagged return C EXECUCOMP
ΔSW The change of market capitalization over the p reviou s year (Δdata199data25) C COMPUSTAT
Acquirer Equals one if the 1047297rm is an acquirer in a given year and zer o other wise D Thomson One Banker
Mb Market-to-book ratio ((data6-data60 +data199data25)data6) C COMPUSTAT
Q Tobins Q ((data6-data60-data74 +data199data25)data6) C COMPUSTAT
Z-score 12(data4-data5)data6 + 06data199data25(data9 + data34)) C COMPUSTAT
Cash1047298ow Sum ofincome beforeextraordinary items and depreciationovernet PPE atthebeginning
of the year ((data18+data14)lagged data8)
C COMPUSTAT
Intangible Intangible assets as a percentage of total assets (1-data8data6) C COMPUSTATGovernance variables
G-index Sum of 24 antitakeoverweak shareholder rights indicators from Gompers et al (2003) C RiskMetrics
Cboard Equals one if the 1047297rm has classi1047297ed board and zero otherwise D RiskMetrics
CEO compensation CEO total compensation TDC1100 C EXECUCOMP
CEO_option BlackndashScholes value of CEO unexercised options as a percentage of CEO total compensation C EXECUCOMP
CEO_share CEO share holding as a percentage of total shares outstanding excluding options C EXECUCOMP
Per_dir_noshare Percentage of directors with no shareholdings C RiskMetrics
Per_ind Percentage of independent directors to total board size C RiskMetrics
Board_size Number of board members C RiskMetrics
CEO_power Composite score representing the power of the CEO where a 1047297rm receives one point
if the CEO is the chairman of the board and one point for each major committee
(compensation nomination audit) that the CEO sits on
C RiskMetrics
Expert d ir ector Percentage of d irector s sitting on at least one other corp orate boards C RiskMetrics
Per_inst Percentage of institutional shareholdings C Thomson Reuters
Blockholder Number of institutional block holders with at least 5 ownership C Thomson Reuters
Earnings quality The negative of the loadings on the accrual quality (AQ) factor as augmented from theFama-French three-factor model where AQ is de1047297ned as the standard deviation of the
residuals from the regressions of the change in working capital on past current and
future cash 1047298ow from operations (Ecker et al (2006))
C Frank Ecker of Duke University
Dir_share (Idir_share) The aggregate percentage shareholdings of all (independent) directors C RiskMetrics
Interlock Dummy for whether the 1047297rm has any director interlocked with any other 1047297rms directors D RiskMetrics
Blk_idir Dummy for whether the 1047297rm has any independent director with at least 5 ownership D RiskMetrics
PPS Pay- per formance sensitivity of CEO compensation de1047297ned as the change of CEO
compensation over the change of the market capitalization
C COMPUSTATEXECUCOMP
Turnoveraddition variables
Dpt_dir Departing directors as a fraction of board size in a given year C RiskMetrics
Add_dir Newly added directors as a fraction of board size in a given year C RiskMetrics
CEO turnover Equals one if the incumbent CEO is no longer the CEO next year and zero otherwise D RiskMetrics
Ability variables
Dirship The average number of outside directorships of all directors C RiskMetrics
Add _dirship The average number of ou tsid e directorships of all newly added directors C RiskMetrics
(continued on next page)
105M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2425
References
Adams R Almeida H Ferreira D 2005 Powerful CEOs and their impact on corporate performance Rev Finan Stud 18 1403ndash1432Adams R Hermalin B Weisbach M 2010 The role of boards of directors in corporate governance a conceptual framework and survey Journal of Economic
Literature 48 (1) 58ndash107
Amihud Y Lev B 1981 Risk reduction as a managerial motive for conglomerate mergers Bell J Econ 12 605ndash
617Ashbaugh-Skaife H Collins D LaFond R 2006 The effects of corporate governance on 1047297rms credit ratings J Acc Econ 42 203ndash243BatesTWBecher DA Lemmon ML 2008 Board classi1047297cationand managerial entrenchment evidence from themarket forcorporate control J FinanEcon87
656ndash677Baumol W 1959 Business Behavior Value and Growth Macmillan New YorkBebchuk L Cohen A 2005 The costs of entrenched boards J Finan Econ 78 409 ndash433Bebchuk L Cohen A Farrell A 2009 What matters in corporate governance Rev Finan Stud 22 783ndash827Becht M Bolton P Roell A 2003 Corporate governance and control In Constantinides G Harris M Stulz R (Eds) The Handbook of Economics of FinanceBertrand M Mullainathan S 2001 Are CEOs rewarded for luck The ones without principals are Quart J Econ 116 (3) 901 ndash932Bertrand M Mullainathan S 2003 Enjoying the quiet life corporate governance and managerial preferences J Polit Economy 111 1043ndash1075Bhagat S Brickley J Coles J 1987 Managerial indemni1047297cation and liability insurance the effect on shareholder wealth J Risk Insur 54 721 ndash736Bhojraj S Sengupta P 2003 Effects of corporate governance on bond ratings and yields the role of institutional investors and outside directors J Bus 76
455ndash476Billett M King T Mauer D 2004 Bondholder wealth effects in mergers and acquisitions new evidence from the 1980s and 1990s J Finance 59 107 ndash135Bishop J 1981 Law of Corporate Of 1047297cers and Directors Indemni1047297cation and Insurance CBC Clark Boardman Callaghan Deer1047297eld IllBodie Z Taggart RA 1978 Future investment opportunities and value of call provision on a bond J Finance 33 1187 ndash1200Boone A Field L Karpoff J Raheja C 2007 The determinants of corporate board size and composition an empirical analysis J Finan Econ 85 66 ndash101
Booth J Deli D 1996 Factors affecting the number of outside directorships held by CEOs J Finan Econ 40 81ndash
104Boyer M 2005 Directors and Of 1047297cers Insurance and Shareholder Protection Working paper HEC MontrealBradley M Schipani C 1989 The relevance of duty of care standard in corporate governance Iowa Law Rev 75 1ndash74Bradley M Desai A Kim H 1988 Synergistic gains from corporate acquisitions and their division between target and acquiring 1047297rms J Finan Econ 21 3ndash40Brook Y Rao R 1994 Shareholder wealth effects of directors liability limitation provisions J Finan Quant Anal 29 481 ndash497Chalmers J Dann L Harford J 2002 Managerial opportunism evidence from directors and of 1047297cers insurance purchases J Finance 57 609ndash636Chava S Livdan D Purnanandam A 2009 Do shareholder rights affect the cost of bank loans Rev Finan Stud 22 2973 ndash3004Chava S Kumar P Warga A 2010 Managerial agency and bond covenants Rev Finan Stud 23 1120 ndash1148Chen D 2007 The Monitoring and Advisory Functions of Corporate Boards Theory and Evidence University of Baltimore Working paperChen D 2010 Classi1047297ed Boards the Cost of Debt and Firm Performance University of Baltimore Working paperCheng S 2008 Board size and the variability of corporate performance J Finan Econ 87 157 ndash176Coles J Daniel N Naveen L 2007 Co-opted Boards Costs Bene1047297ts Causes and Consequences Arizona State University Working paperColes J Daniel N Naveen L 2008 Boards does one size 1047297t all J Finan Econ 87 329ndash356Core JE 1997 On the corporate demand for directors and of 1047297cers insurance J Risk Insur 64 63ndash87Core JE 2000 The directors and of 1047297cers insurance premium an outside assessment of the quality of corporate governance J Law Econ Organ 16 449 ndash477Core JE Holthausen RW Larcker DF 1999 Corporate governance chief executive of 1047297cer compensation and 1047297rm performance J Finan Econ 51 371ndash406Dahya J McConnell JJ Travlos NG 2002 The Cadbury committee corporate performance and top management turnover J Finance 57 461 ndash483
Denis DJ Denis DK Sarin A 1997 Ownership structure and top executive turnover J Finan Econ 45 193ndash
221Ecker F Francis J Kim I Olsson P Schipper K 2006 A returns-based representation of earnings quality Acc Rev 81 749 ndash780
Appendix A (continued)
Variable De1047297nitions and calculations Typea Data source
Per_ie The number of independent directors who are active executives of other 1047297rms as a
percentage of total board size
C RiskMetrics
Risk-taking variable
RampD RampD expenses as a percentage of 1047297rm assets (data46data6) missing value coded as zero C COMPUSTAT
Capex p Capit al expendi ture as a percentage of 1047297rm assets (data128data6) missing value coded
as zero
C COMPUSTAT
Numdaqr Number of diversifying acquisitions the 1047297rm made in a given year A 1047297rm made a
diversifying acquisition if the 1047297rm acquires another 1047297rm not in the same industry
according to the Fama-French 48 industries
C Thomson One Banker
Volatility Standard deviation of the monthly stock returns over the past 60 months C EXECUCOMP
Other CEOdirector characteristics
Dir_age Average age of all directors C RiskMetrics
CEO_age CEO age C RiskMetrics
Dir_tenure The average tenure of directors in years C RiskMetrics
CEO_tenure The tenure for CEO in years C RiskMetrics
CEO out Equals one if the CEOs tenure is less than the average tenure of directors indicating that
the CEO that appointed the average director was no longer with the 1047297rm
D RiskMetrics
Num_oe The number of outside directors (independent plus grey directors) who are active
executives of other 1047297rms
C RiskMetrics
Per_noe The percentage of outside directors (independent plus grey directors) who are not active
executives of other 1047297rms to board size
C RiskMetrics
Other 1047297rm characteristics
Firm_age The number of years since the 1047297rm was listed in CRSP C CRSP
Segment The number of business segments of the 1047297rm C COMPUSTATSEGMENTS
MampA at t-1 Equals one if the 1047297rm hasundergone anyMampA transactions in the previous year and zero
otherwise
D Thomson One Banker
Shareholder The number of shareholders (data1001000) in millions C COMPUSTAT
SP500 Equals one if the 1047297rm is included in the SampP 5 00 index and zero otherwise D COMPUSTAT
a C denotes non-binary variable D denotes binary variable
106 M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107
7232019 Corp Governance-cost of Debt
httpslidepdfcomreaderfullcorp-governance-cost-of-debt 2525
Ertugrul M Hegde S 2008 Board compensation practices and agency costs of debt J Corp Finance 14 512ndash531Farrell K Whidbee D 2000 The consequences of forced CEO succession for outside directors J Bus 73 597ndash627Fich E 2005 Are some outside directors better than others evidence from director appointments by fortune 1000 1047297rms J Bus 78 1943ndash1971Fich E Shivdasani A 2006 Are busy boards effective monitors J Finance 61 689 ndash724Francis B Iftekhar H Kose J Maya W 2010 The effect of state antitakeover laws on the 1047297rms bondholders J Finan Econ 96 127ndash154Gische D 2000 Directors and Of 1047297cers Liability Insurance Available online at httplibrary1047297ndlawcom2000Jan1241472htmlGompers P Ishii J Metrick A 2003 Corporate governance and equity prices Quart J Econ 118 107ndash115Hermalin B Weisbach M 2003 Boards of Directors as an Endogenously Determined Institution a Survey of the Economic Literature FRBNY Economic Policy
Review April pp 7ndash26Heron R Lewellen W 1998 An empirical analysis of reincorporation decision J Finan Quant Anal 33 549 ndash568
Hirshleifer D Thakor AV 1992 Managerial conservatism project choice and debt Rev Finan Stud 5 437 ndash470Holmstrom B Costa JRI 1986 Managerial incentives and capital management Quart J Econ 101 835 ndash860Huson MR Parrino R Starks LT 2001 Internal monitoring mechanisms and CEO turnover a long-term perspective J Finance 56 2265ndash2297
Janjigian V Bolster P 1990 The elimination of director liability and stockholder returns an empirical investigation J Finan Res 13 53ndash60 Jensen M 1986 Agency costs of free cash 1047298ow corporate 1047297nance and takeovers Am Econ Rev 76 323ndash329 Jensen M Meckling W 1976 Theory of the 1047297rm managerial behavior agency costs and ownership structure J Finan Econ 3 305ndash360 Jensen M Murphy K 1990 Performance pay and top-management i ncentives J Polit Economy 98 225ndash264 John TA John K 1993 Top-management compensation and capital structure J Finance 48 949ndash974 John K Litov L Yeung B 2008 Corporate governance and risk-taking J Finance 63 1679ndash1728Kempf A Ruenzi S Thiele T 2009 Employment risk compensation incentives and managerial risk taking evidence from the mutual fund industry J Finan
Econ 92 92ndash108Klock M Mansi S Maxwell W 2005 Does corporate governance matter to bondholders J Finan Quant Anal 40 693 ndash719Laeven L Levine R 2009 Bank governance regulation and risk taking J Finan Econ 93 259 ndash275Larcker D Richardson S Tuna I 2004 Does Corporate Governance Really Matter University of Pennsylvania Working paperLehn K Poulsen A 1989 Free cash 1047298ow and stockholder gains in going private transactions J Finance 44 771 ndash787Linck JNetter J Yang T 2009 Theeffects andunintended consequences of theSarbanesndashOxley Act on thesupply anddemand fordirectors Rev FinanStud22
3287ndash3328
Mansi S Maxwell W Wald J 2009 Credit protection laws and the cost of debt J Law Econ 52 701ndash726Masulis RW Wang C Xie F 2007 Corporate governance and acquirer returns J Finance 62 1851ndash1889Myers S 1977 Determinants of corporate borrowing J Finan Econ 5 147ndash175Netter J Poulsen A 1989 State corporation laws and shareholders the rent experience Finan Manage 18 29ndash40Ortiz-Molina H 2006 Top management incentives and the pricing of corporate public debt J Finan Quant Anal 41 317 ndash340Pathan S 2009 Strong boards CEO power and bank risk-taking J Bank Finance 33 1340 ndash1350Petersen M 2009 Estimating standard errors in 1047297nance panel data sets comparing approaches Rev Finan Stud 22 435ndash480Qiu J Yu F 2009 The market for corporate control and the cost of debt J Finan Econ 93 505ndash524Romano R 1991 The shareholder suit litigation without foundation J Law Econ Organ 7 55 ndash87Romano R 2006 The states as a laboratory legal innovation and state competition for corporate charters Yale J Regul 23 209 ndash247Shleifer A Vishny R 1997 A survey of corporate governance J Finance 52 737ndash783Tillinghast-Towers Perrin 2004 2005 2006 Directors and of 1047297cers liability surveyTreynor J Black F 1976 Corporate investment decisions In Myers S (Ed) Modern Developments in Financial Management New York Praeger pp 310 ndash327Weisbach MS 1988 Outside directors and CEO turnover J Finan Econ 20 (1-2) 431 ndash460Yermack D 2004 Remuneration retention and reputation incentives for outside directors J Finance 59 2281ndash2308Zhou X 2001 Understanding the determinants of managerial ownership and the link between ownership and performance comment J Finan Econ 62
559ndash571
107M Bradley D Chen Journal of Corporate Finance 17 (2011) 83ndash107