 

a‘

‘3

g.)

 

LIBRARY
Michigan State

Universiiy

-.—~— 'v—.—._-.~—_‘

 

This is to certify that the
dissertation entitled

ESSAYS ON CORPORATE FINANCE

presented by

NOOLEE KIM

has been accepted towards fulﬁllment
of the requirements for the

 

 

 

 

Ph.D. * degree in Finance
Major Professdv’s Signature
{71? /2. 09 7?
Date

MSU is an Aﬁirmative Action/Equal Opportunity Employer

 

ou----—---------c--o--o-----o--n--u-u-.-n-o---o--I-----o--n-0-a----l—-.—n—n--u-o-o--u-o-u---o---

I PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/CIRCIDateDue.indd

 

 

ESSAYS ON CORPORATE FINANCE
By

Noolee Kim

A DISSERTATION

Submitted to
Michigan State University
In partial fulﬁllment of the requirements
For the degree of

DOCTOR OF PHILOSOPHY
Finance

2009

ABSTRACT
ESSAYS ON CORPORTAE FINANCE
By

Noolee Kim
This dissertation explores the effects of bank credit agreements on the value of the
borrowing ﬁrm and its various claimholders’ wealth. Using bond return data to
distinguish among alternative explanations regarding the valuation effect of bank loan
announcements, we Show that bondholders earn significantly positive abnormal returns
surrounding announcements of bank credit agreements. Such returns are considerably
higher for ﬁrms that are riskier and ﬁrms that perform worse, particularly for ﬁrms that
perform worse but have higher growth opportunities. The results are remarkably similar
when we use stock returns to measure abnormal returns. We also ﬁnd no evidence of a
wealth transfer from bondholders to shareholders; in fact, we ﬁnd that bond
announcement returns are signiﬁcantly positively related to stock announcement returns.
These results suggest that the positive stock market reaction to bank loan announcements
documented in prior studies does not reﬂect anticipated wealth transfers from creditors to
shareholders or anticipated tax beneﬁts to shareholders arising from higher leverage, but
instead reﬂects the anticipated value increase resulting from monitoring services that

bank loans provide to the client ﬁrms.

To my parents and sister

iii

ACKNOWLEDGEMENTS

I would like to express my deepest gratitude to my dissertation chair,
Professor Jun-Koo Kang for his constant guidance and unwavering support. My
dissertation would not have been possible without his serious commitment as an
advisor. Furthermore, I am deeply indebted to him, who as a teacher, has provided
invaluable knowledge, experience, and wisdom which have played a deﬁning role in
my dissertation and career goals.

I wish to thank the other members of my dissertation committee, Professor
Charles J. Hadlock, Professor Ted Fee, and Professor Stephen Dimmock for their
helpful cements and discussions throughout my dissertation works. I am especially
grateful to Professor Charles J. Hadlock for all the encouragement and support he has
provided ‘me as the director of the doctoral program. I am greatly indebted to
Professor Long Chen for his valuable help, guidance and» suggestions on my
dissertation.

I would like to express my appreciation to the Department Chair, Professor G.
Geoffrey Booth, for providing me with the opportunity to study in the doctoral
program and for his support throughout my time at Michigan State University. Also, I
would like to thank Professor Elizabeth Booth for her thoughtful care and help
throughout my doctoral studies. I also thank the other professors of the ﬁnance
department for their teaching and advice in the process of my development as a

scholar.

iv

I thank all the graduate students who I have studied with at Michigan State
University: Ranadeb Chaudhuri, Hoontaek Seo, William Gerken, Neslihan Yilmaz,
Kyoung-Min Kwon, Min Jung Kang, and other fellow doctoral students for their
friendship and support. Also, I thank all my friends for all their help and support.

Finally, my deepest gratitude goes to my parents and sister back in Korea. I
cannot thank them enough for their continuous support, patience and love. Without

them, this would not have been possible.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................ vii

LIST OF FIGURES ............................................................................................. viii

1 Introduction .................................................................................................. l

2 Sample Selection and Data .......................................................................... 9

3 Event Study Methodology ......................................................................... 12

4 Empirical Results ....................................................................................... 15
4.1 Univariate Results for Abnormal Stock, Bond,

and Firm Returns ............................................................................ 15

4.2 Cross-Sectional Determinants of Bond and Stock CARS .............. 20

4.3 Relation between Abnormal Bond and Stock Returns .................. 26

5 Summary and Conclusion .......................................................................... 30

APPENDIX ............................................................................................................ 33

BIBLIOGRAPHY .................................................................................................. 60

vi

LIST OF TABLES

Table 1 Predictions of Monitoring, Wealth Transfer, and Leverage

Hypotheses for Key Variables ....................................................... 34
Table 2 Distribution of Bank Loan Announcements by Year

and Borrower Industry ................................................................... 35
Table 3 Summary Statistics of Borrowers, New Bank Loans,

and Bonds Outstanding .................................................................. 36
Table 4 Mean and Median Cumulative Abnormal Returns (CAR)

around the Announcement Date (AD) ........................................... 38
Table 5 Three-Day Cumulative Abnormal Returns (CARS)

for Borrowing Firms Categorized by Firm

and Loan Seniority Characteristics ................................................ 40
Table 6 Three-Day Cumulative Abnormal Returns (CARS)

for Borrowing Firms by Borrower Performance

and Grth Opportunities .............................................................. 43
Table 7 OLS Regression of the Three-Day Cumulative Abnormal

Bond Returns on Explanatory Variables ........................................ 46
Table 8 OLS Regression of the Three-Day Cumulative Abnormal

Stock Returns on Explanatory Variables ....................................... 49
Table 9 OLS Regression of the Three-Day Cumulative Abnormal

Bond and Stock Returns on Wealth Transfer Variables ................ 52

Table 10 OLS Regression of the Three-Day Cumulative
Abnormal Bond Returns on Three-Day Cumulative
Abnormal Stock Returns ................................................................ 55

vii

LIST OF FIGURES

Figure 1 Cumulative abnormal returns from day -10 to day +10
around the loan announcement ...................................................... 58

viii

1. Introduction

It is well documented that, unlike the case for shareholders of ﬁrms that use other
types of ﬁnancing, shareholders of borrowing ﬁrms earn signiﬁcantly positive excess
stock returns surrounding the announcements of bank credit agreements (Mikkelson and
Partch (1986), James (1987), Lummer and McConnell (1989), Best and Zhang (1993),
Preece and Mullineaux (1994), Billet, Flannery, and Garﬁnkel (1995)).1 For example,
James (1987) shows an average abnormal return of 1.93% over the two-day period
around the bank loan announcement date. The positive announcement effect is stronger
for favorable loan renewals and revisions (Lummer and McConnell (1989)) and for loans
from less risky banks (Billet, Flannery, and Garfmkel (1995)).

Prior literature interprets these results as evidence that banks play a special role as
information transmitters in the capital markets and perform an important value-enhancing
intermediary function for their client ﬁrms (hereafter, the “monitoring hypothesis”). This
interpretation is controversial, however, because the positive excess stock returns
associated with announcements of bank credit agreements are subject to other alternative
explanations. For example, Peek and Rosengren (2005) argue that to avoid classifying
existing loans to troubled client ﬁrms with poor prospects as nonperforming, banks in
Japan have strong incentives to extend credit to ﬁnancially distressed ﬁrms. Supporting
this argument, Kang and Liu (2007) show that the announcement returns for Japanese
borrowing ﬁrms are signiﬁcantly positive and those for lending banks are sometimes

signiﬁcantly negative. Furthermore, the announcement returns for borrowing ﬁrms are

 

1 The existing literature shows that announcements of new equity and convertible bonds issues are
associated with a decrease in the announcing ﬁrm’s stock price. See Smith (1986) and Masulis (1988) for a
review of the evidence on the announcement effect associated with new security issues.

negatively related to those for lending banks, especially when poorly performing ﬁrms
borrow ﬁ'om ﬁnancially healthy banks.

It is also possible that, by arranging bank loans that frequently are more senior to or
have shorter maturity than current outstanding bonds, managers of borrowing ﬁrms can
reorder the priority of debt claims and have opporttmities to expropriate current
bondholders through certain corporate activities such as risky investments and share
repurchases. In these circumstances, bank loan announcements can lead to wealth transfer
from either lending banks or bondholders to shareholders, resulting in positive
announcement stock returns for the borrowing ﬁrms (hereaﬁer, the “wealth transfer
hypothesis”).

Alternatively, new bank loans increase the borrower’s leverage ratio. To the extent
that higher debt in a ﬁrm’s capital structure generates larger tax savings (Modigliani and
Miller (1963)), positive excess stock returns surrounding bank loan announcements
documented in prior studies are also consistent with the view that the higher leverage
resulting ﬁom new bank loan ﬁnancing is beneﬁcial to shareholders (hereaﬁer, the
“leverage hypothesis”).

In this paper we use daily bond return data to distinguish among these alternative
explanations for the well-documented positive abnormal stock returns surrounding bank
loan announcements. Using a sample of 254 new bank loan announcements during the
period 1995-2003, we provide one of the ﬁrst analyses that simultaneously documents the
responses of borrowers’ stock and bond prices to announcements of bank credit
agreements. Previous studies on bank loan announcement effects have exclusively used

stock prices. Although stock prices are useful in examining the effect of bank loan

announcements on shareholder wealth, they have only limited ability to distinguish
among the alterative hypotheses discussed above. In particular, these alterative
hypotheses suggest that we cannot say much about the nature of the association between
bank loan announcements and stock returns unless we examine the market responses to
bank loan announcements from other classes of securities such as corporate bonds. For
example, the positive excess stock returns surrounding announcements of bank credit
agreements documented in prior studies may not reﬂect a value increase due to bank
monitoring, but rather a wealth transfer from creditors to shareholders, or the tax beneﬁts
that shareholders enjoy from increased tax shields. Bond return data allow us to assess
the relative importance and merit of these alternative explanations that stock return data
alone cannot.

Speciﬁcally, we compare both the magnitudes and the signs of abnormal stock and
bond returns and examine their correlations. We also investigate whether abnormal stock
and bond returns differ depending on ﬁrm characteristics (e.g. risk, past performance, and
growth opportunity) and new bank loans’ relative seniority (e.g., security and maturity)
and size compared to current corporate bonds outstanding.

According to the monitoring hypothesis, banks’ loan commitments certify the quality
of borrowing ﬁrms and signal that the lending banks are willing to provide various
monitoring services. These monitoring services alleviate borrowing ﬁrms’ managerial
incentive problems, reduce information asymmetry problems between managers and
outside investors, and ease their potential ﬁnancial constraints (Campbell and Kracaw
(1980), Diamond (1984), and Fama (1985)). Since both shareholders and debtholders

beneﬁt from these valuable services, the monitoring hypothesis suggests that bank loan

announcements have a positive effect not only on the value of stocks but also on the
value of bonds, and that bond announcement returns are positively related to stock
announcement returns. To the extent that ﬁrms that are riskier, ﬁrms that perform worse,
and ﬁrms that have higher growth opportunities generally suffer more from information
asymmetries, these ﬁrms are expected to realize stronger positive excess stock and bond
returns surrounding the announcements of bank credit agreements. However, the
monitoring hypothesis does not provide any direct implications for the effect of relative
seniority and size between new bank loans and outstanding bonds on stock and bond
prices. The value-enhancing monitoring services provided by the banks should create
efﬁciency gains for both shareholders and bondholders, irrespective of the nature of
claims represented by new bank loans and outstanding bonds.

In contrast to the monitoring hypothesis, the wealth transfer hypothesis posits that,
assuming there is no wealth transfer from the lending banks to the stockholders (as shown
below), the positive announcement effects for stockholders represent a wealth transfer
ﬁom the bondholders in the same ﬁrm, suggesting positive (negative) excess stock
(bond) returns surrounding the announcements of bank credit agreements and a negative
relation between stock and bond announcement returns. In particular, the wealth transfer
hypothesis suggests that if new bank loans have relatively higher seniority (due to higher
security or shorter maturity) and larger size than bonds already issued, a wealth transfer
from bondholders to shareholders is more likely to occur. Thus, stock (bond) prices are
expected to respond more positively (negatively) in these cases. However, the wealth
transfer hypothesis'does not have any clear implications for the effect of a ﬁrm’s risk,

past performance, and growth opportunities on stock and bond prices.

Finally, assuming that borrowing ﬁrms have low risk and low leverage, the leverage
hypothesis predicts that the increase in leverage resulting from new bank loan ﬁnancing
creates large tax shields and thus is beneﬁcial to stockholders. In contrast, the impact of
such an increase in leverage on bondholder wealth is expected to be either negligible if
the ﬁrm is healthy or negative if the high leverage increases the probability of a possible
future bankruptcy and thus imposes signiﬁcant ﬁnancial distress risk on bonds
outstanding. Thus, the leverage hypothesis suggests positive (non-positive) excess stock
(bond) returns surrounding announcements of bank credit agreements and a non-positive
relation between excess stock and bond returns. To the extent that risky ﬁrms, poorly
performing ﬁrms, and ﬁrms with high growth opportunities have a high probability of
facing ﬁnancial distress, the leverage hypothesis further suggests that for these ﬁrms, the
shareholders’ beneﬁts (bondholders’ losses) from using more debt are likely to be smaller
(larger), predicting less positive (more negative) equity (bond) announcement returns
around bank loan announcements. The effect of higher leverage that results from new
bank loan ﬁnancing on bondholder wealth also depends on bank loans’ relative seniority
and size. To the extent that the adverse effect of leverage on bondholder wealth is greater
when new bank loans have relatively higher security, shorter maturity, or larger size than
extent bond issues, the leverage hypothesis predicts that bond announcement returns are
lower in these cases. However, the leverage hypothesis does not provide any direct
implications for the effect of bank loans’ relative seniority and Size on excess stock
returns. Table 1 summarizes the predictions of the monitoring, wealth transfer, and

leverage hypotheses and presents our empirical results.

We ﬁnd that the positive stock market reaction to bank loan announcements
documented in prior studies does not reﬂect either anticipated wealth transfers from
creditors to shareholders or an increase in tax beneﬁts arising from high leverage, but
instead reﬂects an anticipated value increase resulting from monitoring services that bank
loans provide to the client ﬁrms. Indeed, we ﬁnd that both bondholders and shareholders
of borrowing ﬁrms earn signiﬁcantly positive abnormal returns surrounding
announcements of bank credit agreements: the average three-day cumulative abnormal
return (CAR) for stocks is 1.48% with a p—value of 0.02, the average three-day CAR for
bonds is 1.01% with a p-value of 0.04, and the average three-day CAR for the entire ﬁrm
is 1.15% with a p-value of 0.02. These results are consistent with the monitoring
hypothesis and suggest that the market views a bank loan announcement as a value-
enhancing event for the whole ﬁrm.

We also ﬁnd that the positive announcement returns for stocks and bonds are
considerably higher for ﬁrms that are riskier (i.e., higher equity beta, higher stock return
volatility, and higher leverage), ﬁrms that perform worse (i.e., lower operating
performance and lower stock return performance), and poorly performing ﬁrms that have
higher growth opportunities (i.e., lower stock return performance, but higher Tobin’s q).
To the extent that these ﬁrms face greater information asymmetry and beneﬁt more from
a bank’s monitoring services, the results suggest that new bank loans provide value-
enhancing monitoring services to both shareholders and bondholders, rather than to
shareholders only. These results are also consistent with the monitoring hypothesis.

Turning to the effect of new bank loans’ relative seniority and size on bond prices, in

contrast to the predictions of the wealth transfer and leverage hypotheses, we ﬁnd that the

abnormal bond returns for the cases in which 1) loans are secured but bonds carry lower
security, 2) loans have longer maturity than bonds, and 3) loan size is larger than bond
size, are not Signiﬁcantly different from those for other types of bank credit agreements.
Furthermore, equity announcement returns are higher when bonds carry higher security
and lower when loans are secured but bonds carry lower security. In untabulated tests, we
also ﬁnd that the stock prices of lending banks do not respond negatively to new loan
announcements and that stock (bond) announcement returns for borrowing ﬁrms are not
signiﬁcantly related to those for lending banks. Taken together, these results suggest that
the leverage effect or the wealth transfer from either the lending banks or the bondholders
to the shareholders is unlikely to be the driving determinant of the observed valuation
effects for stocks and bonds.

Finally, we ﬁnd that abnormal bond returns are signiﬁcantly, positively related to
abnormal equity returns. This positive relation is more pronounced if the borrower’s past
operating (stock) performance is below the sample median, particularly if its Tobin’s q is
above the sample median. These results further support the monitoring hypothesis.

Overall, our analysis for the effects of bank loan announcements on bond and stock
prices provides strong support to the monitoring hypothesis. We ﬁnd no evidence
suggesting wealth transfer from either the lending banks or bondholders to shareholders.
We also ﬁnd no evidence supporting the leverage hypothesis.

Our paper is closely related to the recent study by Ongena, Roscovan, and Werker
(2007), who also examine responses from both corporate bonds and stocks to new bank
loan announcements. However, unlike our study that examines bond returns, Ongena,

Roscovan, and Werker (2007) focus on bond yield changes. Yield changes do not directly

consider maturities (durations) and thus the results using yield changes could be quite
different from those using bond returns. Furthermore, while Ongena, Roscovan, and
Werker (2007) focus more on the expected default loss on bonds, to distinguish among
alternative explanations of positive excess stock returns surrounding bank loan
announcements documented in prior studies, we utilize extensive information about
borrowing frrms’ risk, past performance, and growth opportunities, and about new bank
loans’ relative seniority and size compared to outstanding bonds. Combined, both studies
provide important new evidence, from different angles, on the bond market reaction to
bank loan announcements.

Our paper is also related to several studies that examine the announcement effect of
certain corporate events on bondholder wealth. For example, Maxwell and Stephens
(2003) examine abnormal bond returns around repurchase announcements during the
period 1980-1997 and ﬁnd that bondholders experience negative abnormal returns.
Maxwell and Rao (2003) investigate the bond price responses to spin-off announcements
during the period 1974-1997. They ﬁnd that bondholders suffer a signiﬁcant negative
abnormal return during the month of the spin-off announcement. Finally, Billett, King,
and Mauer (2004) analyze the wealth effects of mergers and acquisitions on target and
acquiring ﬁrm bondholders, using a sample of 940 deals during the period 1979-1997.
They ﬁnd signiﬁcantly positive excess returns to target bonds and signiﬁcantly negative

excess returns to acquirer bonds. Our paper adds to the literature by showing that

bondholders earn signiﬁcantly positive abnormal returns surrounding announcements of
bank credit agreements.2

This paper proceeds as follows. In Section 2, we discuss the data and sample
characteristics. Section 3 describes the methodology used to estimate stock and bond
abnormal returns surrounding bank loan announcements. In Section 4, we present the

results from our empirical analysis. Section 5 summarizes and concludes the paper.

2. Sample Selection and Data

We compile our sample by searching Factiva for loan announcements over the period
1995 to 2003. Our initial sample consists of ﬁrms that have daily bond price data in
DATASTREAM, which publishes daily stock and bond prices that are widely used by
academics and practitioners. We start the sample period in 1995 because daily bond
prices become more regularly available only after 1995. We exclude ﬁom our initial
sample ﬁrms that do not have daily stock return data in CRSP, those that do not have
ﬁnancial data in COMPUSTAT, and those that do not have bond characteristic data, such
as rating and bond security, in the Fixed Income Security Database (F ISD).

To obtain our sample of loan announcements, we search F activa by using the

99 ‘6 9’ 6‘

following key words: “line of credit , credit line”, “credit facility , credit agreement”,

99 ‘6 99 66

“credit extension , new loan , loan agreement”, “loan renewal”, “loan revision”, “loan

99 ‘6

extension”, “ﬁnance company loan”, “term loan , commercial loan”, and “bank loan”.

 

2 Unlike these studies that use monthly bond prices in Lehman Brothers Bond Database (LBBD) to
estimate abnormal bond returns, our paper uses daily bond prices in DAT ASTREAM. The LBBD contains
only monthly data and does not provide daily bond prices. One advantage of using daily bond returns in
estimating announcement effects is that it allows us to more accurately measure the responses of bond
prices to announcements of corporate events. The LBBD is available only up to 1998.

For each announcement, we also collect information on loan amount, maturity, and
security.

This procedure yields a sample of 369 loan announcements from 258 ﬁrms. Of these,
we delete 44 observations in which bond prices do not move at all during the event
window (days --170 to +10 relative to the bank loan announcement). We also delete 34
loans made to ﬁnancial and utility companies. We further exclude 12 cases in which the
ﬁrm releases other important information with the bank loan announcement. Finally, to
avoid having results confounded by multiple events that cluster during a short time
period, we eliminate 25 events that occur within 90 days after the ﬁrst loan
announcement events.3 Thus, our ﬁnal sample consists of 254 announcements of bank
credit agreements.

Table 2 summarizes the frequency of bank loan announcements by year and industry.
The years 2001 and 2002 are the most active bank credit agreement years, with 76 and 55
cases, respectively, during which time the US. stock market dropped signiﬁcantly and
ﬁrms had difﬁculty accessing external direct ﬁnancing. A breakdown of the loan
announcements by industry shows that most of the borrowing ﬁrms are classiﬁed as
manufacturing (115 cases). F orty-nine borrowing ﬁrms are in the transportation and
communication industries and 45 in services and others.

Panel A of Table 3 presents summary statistics for our sample borrowers. The average
borrowing ﬁrm’s book value of assets (market value of equity) equals $10.95 billion
($10.1 billion), with mean sales of $8.27 billion. Leverage, measured as the book value of

total debt divided by the sum of the book value of total debt and the market value of

10

equity, averages 43.2%, and the mean ratio of operating income before depreciation to
total assets is 11.4%. The average Tobin’s q (sum of the book value of debt and the
market value of equity / total assets) and the average previous six-month excess return are
1.22 and 3%, respectively. Equity beta averages 0.93 and the mean standard deviation of
borrower past excess stock returns is 0.59.

Panel B of Table 3 provides summary statistics for deal-speciﬁc characteristics of the
announced loans. The average loan size is $1.2 billion, which accounts for 11% of the
borrowing ﬁrms’ average total assets. About 62% of the announced loans are secured and
the maturity of announced loans averages 3.51 years.4

Panel C of Table 3 reports summary statistics for bonds outstanding prior to bank
credit agreements. The borrowing ﬁrms’ bonds outstanding averages $1 billion. The
mean bond security and rating are 4.65 and 4.88, respectively. 5 The average bond
maturity of 8 years is more than two times longer than the average maturity of announced

loans.6

 

3 Since we use the market-model approach to estimate abnormal returns for the events and a signiﬁcant
portion of the estimation periods (day —170 to day —21) overlaps for these 25 events, it is likely that t-
statistics in the analyses of the abnormal returns are biased upwards if we include them in our sample.

4 F activa does not provide enough information on loan security for our sample ﬁrms. Therefore, we obtain
detailed loan security information by thoroughly searching a fn'm’s 10K report and supplementing these
data with information from Dealscan. Of the 254 sample loan announcements, we are able to obtain loan
security information for 192 announcements.

5 The bond security score ranges from one to six, with a higher score indicating higher security.
Speciﬁcally, the score takes the value of one if the bond security level is junior subordinate, two if it is
junior or subordinate, three if it has no special security features, four it is senior subordinate, ﬁve if it is
senior, and six if it is senior secured. The bond rating score ranges from one to eight, with a higher score
indicating lower rating. Speciﬁcally, the score takes the value of one if the bond rating is AAA or Aaa, two
if it is AA or Aa, three if it is single A, four if it is triple BBB or Baa, ﬁve if it is BB or Ba, six if it is B,
seven if it is triple CCC or Caa, and eight if it is lower than the 7th category.

6 When a borrowing ﬁrm has multiple bonds outstanding, we estimate the bond security, rating, and
maturity as the weighted average of the bonds’ securities, ratings, and maturities, respectively, where the
weights are assigned according to the face values of the bonds.

ll

3. Event Study Methodology
We use standard event study methodology to assess the valuation effect of bank loan
announcements on borrower stock prices. We implement the test procedure by computing

daily abnormal stock returns (AR) as
ARit = Rn -(ai + ﬂiRmt)’ (1)

where Kit and Rm, are the daily stock return of borrower i at time t and the daily market
index return at time t, respectively. The CRSP value-weighted return is used as the

market return. The coefﬁcients “1' and ,3,- are ordinary least squares estimates of the

intercept and slope, respectively, of the market model regression. To compute the

abnormal returns, we estimate the borrower-speciﬁc parameters di and ,3; with an
ordinary least squares regression, using 150 daily returns beginning with day t = -170 and
ending with t = -21 relative to the announcement date.

Similarly, the daily abnormal bond return (ABR) is calculated as
ABR, = BR” — ((35,. + ,BA,1 x Chg _Spread, + 61.2 x Chg_Slope,), (2)

where BR” is the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in Baa- over Aaa-rated

bond yields, and Chg_Slope, is the change in the slope of the yield curve at time t, which

12

is deﬁned as the change in the difference between 10-year Treasury over l-year Treasury
yields. This speciﬁcation, following Fama and French (1993) and Chen, Lesmond, and
Wei (2007), assumes that the expected corporate bond return is driven by an aggregate
corporate bond factor and an interest rate factor. Both the credit spread and yield curve
data are obtained from the Federal Reserve. We estimate this bond “market model” using
150 daily bond returns beginning with day t = -170 and ending with t = -21 relative to the
announcement date. When a borrowing ﬁrm has multiple bonds outstanding, we ﬁrst
estimate the ABR for each bond, then calculate the weighted average ABR for all bonds
outstanding, where the weights are assigned according to the face value of each bond.7
The daily abnormal stock and bond returns are accumulated to obtain the stock and
bond cumulative abnormal returns (CARS), respectively, between any two dates T1 and

T2as

T2
StockCARl. (T1, T2) = 2 AR,

t=Tl

(3)

 

7 The problem of inﬁ’equently traded bonds can cause error in measuring excess returns. For example, as
noted by Scholes and Williams (1977) and Dirnson (1979), the return of infrequently traded bonds is not
likely to move along the bond market factors synchronously and thus there might be lagged effect ﬁ'om the
previous period’s return. This can lead to underestimation of betas and overestimation of excess returns. To
mitigate this problem, in untabulated tests, we also estimate an alternative bond market model:

5 A A
ABR“ = BR" — d, + 2 13:1,,- x Chg_Spread,_j + ,6in x Chg_Slope,_J
j=0

That is, we include ﬁve lags for each market factor to estimate excess bond returns. We repeat all analyses
below using excess bond returns estimated from this alternative method and obtain results that are
qualitatively similar to those reported in the paper.

13

T2
BondCAR, (T1, T2) = Z ABR, ,

I=Tl

(4)

and we compute the sample cross-sectional average cumulative abnormal stock and bond

returns, that is, stock and bond ACAR, as

N
StockACAR(T1, T2) = %Z StockCAR, (T1, T2)

i=1

(5)

N
BondACAR(T1, T2) = 'lez BondCAR, (T1, T2) .

i=1

(6)

We use the t-statistic to test the hypothesis that the stock (bond) ACARs over any
given interval are equal to zero. We also perform two additional tests to provide detailed
information on the distribution of the estimated CARS and to ensure the robustness of our
results. First, in addition to the mean CARS, we estimate the median CARS and test the
hypothesis that the stock (bond) CARS over any given interval are distributed
symmetrically around zero. Second, in untabulated tests, we winsorize the CARs at the

99% and 1% cutoff points, and ﬁnd that all of our main conclusions remain the same.

14

4. Empirical Results
4.1. Univariate Results for Abnormal Stock, Bond, and Firm Returns

Panel A of Table 4 reports stock CARS for the borrowing ﬁrms. On average,
shareholders of the borrowing ﬁrms earn statistically signiﬁcant positive gains. The
average AR (0), CAR (-1, 0), and CAR (-1, 1) are 1.47%, 1.28%, and 1.48%,
respectively, all of which are signiﬁcantly different from zero at the 5% level or lower.
These results are generally consistent with those in previous studies, which document
positive excess stock returns sm'rounding announcements of bank credit agreements
(Mikkelson and Partch (1986), James (1987), Lummer and McConnell (1989), Best and
Zhang (1993), Preece and Mullineaux (1994), Billet, Garﬁnkal, and Flannery (1995)).

Panel B of Table 4 shows bond CARs. The mean CAR (-1, 1) is 1.01%, which is
signiﬁcant at the 5% level. Thus, announcements of bank loan ﬁnancing create
statistically signiﬁcant wealth gains for both shareholders and bondholders of the
borrowing ﬁrms.

Figure 1 shows stock and bond CARs for the borrowing ﬁrms from day -10 to day
+10 around the event date. Both stock and bond CARs for the borrowers increase
signiﬁcantly from day -1 to day +1. However, there are different patterns of CAR
movements for stocks and bonds aﬁer day +1. The stock CARs peak on day +1 , and then
start falling so that the stock CAR is almost zero by the end of day +10. However, the
bond CARs continue to rise even after day +1, leveling off at day +8 and starting to fall
slightly below 1% by the end of day +10.

Panel C of Table 4 presents the CARs for the entire ﬁrm. We measure the ﬁrm CAR

as a value-weighted average of the stock CAR and the bond CAR, with the weights being

15

the market value of equity and the book value of debt measured at the ﬁscal year-end that
immediately precedes an announcement date. The average AR (0), CAR (-1, 0), and CAR
(-1, 1) are 0.94%, 0.91%, and 1.15%, respectively, all of which are statistically
signiﬁcant at the 5% level.

Overall, these results do not support the wealth transfer hypothesis or leverage
hypothesis, but are consistent with the monitoring hypothesis, which predicts that banks
perform an important value-enhancing intermediary function for both shareholders and
bondholders of the borrowing ﬁrms.8

To ﬁrrther distinguish among these three hypotheses, we partition the sample
according to some key characteristics of borrowing ﬁrms and new bank loans, and
compare bond and stock CARS (-1, 1) across these characteristics. The results are
reported in Table 5. In Panels A through D, we separate borrowing ﬁrms according to
their riskiness, as proxied by bond rating, equity beta, stock return volatility, and
leverage, respectively. We ﬁnd that the positive and signiﬁcant average stock CAR (-1 ,
1) is only evident in the subsamples of borrowers with non-investment grade rating,
borrowers with higher beta, borrowers with higher stock return volatility, and borrowers
with higher leverage. We also ﬁnd that the average stock CARs (-1, l) for these
subsamples are signiﬁcantly higher than those for borrowers with investment grade
bonds, borrowers with lower betas, borrowers with lower stock return volatility, and

borrowers with lower leverage. The results are similar for bond CARs (-1, 1), although

 

8 In untabulated tests, we ﬁnd that the stock prices of lending banks do not respond negatively to new loan
announcements and that borrower stock (bond) announcement returns are not signiﬁcantly related to lender
announcement returns. These results suggest that the wealth transfer from lending banks to shareholders
(bondholders) is unlikely to be the driving factor for the positive valuation effect of stocks (bonds).

l6

we do not ﬁnd signiﬁcant differences in bond CARs (-l, 1) across higher and lower beta
borrowers.

To examine whether abnormal returns vary by past borrower performance, in Panels
E and F we separate the sample according to past operating and stock return performance
of the borrowers. We ﬁnd that for both bond and stock CARs (-1, 1), a positive and
signiﬁcant mean CAR (-1, 1) is only evident in the subsamples of borrowers with lower
past operating performance and of those with lower past stock return performance.
Furthermore, the CARs (-1, 1) for these subsamples are signiﬁcantly higher than those
for borrowers with higher past operating performance and for those with higher past
stock return performance, respectively.

In Panel G, we divide the sample according to the growth opportunity of the
borrower. The mean bond CAR for a subsample of ﬁrms with higher Tobin’s q is positive
and signiﬁcant, and is higher than that for a subsample of ﬁrms with lower Tobin’s q.
Although the equality of the mean CARs for these two subsamples cannot be rejected at
conventional levels, the median bond CAR (-1, 1) for a subsample of ﬁrms with higher
Tobin’s q is signiﬁcantly higher than that for a subsample of ﬁrms with lower Tobin’s q.
On the other hand, the mean and median stock CARs for a subsample of ﬁrms with
higher Tobin’s q are not signiﬁcantly different from those for a subsample of ﬁrms with
lower Tobin’s q, respectively.

Overall, these results suggest that bondholders of the borrowing ﬁrms realize
signiﬁcant positive excess returns around bank loan announcements, particularly when
their ﬁrms are risky, when their ﬁrms perform poorly, or when their ﬁrms have good

future growth opportunities. Stockholders of the borrowing ﬁrms also realize higher

17

returns when their ﬁrms are risky or when their ﬁrms perform poorly, strongly supporting
the monitoring hypothesis.

Panels H through J show the effect of new bank loans’ security, maturity, and size,
relative to those of bonds outstanding, on abnormal bond and stock returns. As discussed
in the introduction, the wealth transfer hypothesis predicts that a relative advantage of
new bank loans in these aspects facilitates the potential wealth transfer ﬁom bondholders
to stockholders in the borrowing ﬁrm -—abnormal bond (stock) returns are expected to be
lower (higher) if new bank loans carry higher security, shorter maturity, and larger size
than current bonds outstanding. In addition, the leverage hypothesis predicts that a
relative advantage of new bank loans in these respects will push the current bonds into a
more inferior position, due to higher leverage and/or lower seniority, suggesting lower
bond CARs. On the other hand, the monitoring hypothesis predicts no clear link between
abnormal returns and these relative measures of seniority and size.

We ﬁnd that the average bond and stock CARs are small and insigniﬁcant for a
subsample of borrowers whose new bank loans are secured but bonds outstanding have
low security and for a subsample of borrowers whose new bank loans are larger in size
than bonds outstanding. Furthermore, for a subsample of borrowers whose new bank
loans carry shorter maturity than bonds outstanding, the average bond CAR is not
signiﬁcant, but the average stock CAR is positive and signiﬁcant. These results do not
support either the wealth transfer hypothesis or leverage hypothesis, but are rather
consistent with the monitoring hypothesis.

Next, given the importance of ﬁrm performance and growth opportunities in

explaining bond and stock abnormal returns around bank loan announcements, we

18

examine whether the interaction of these two variables has any further incremental power
to distinguish among the three alternative hypotheses above. Because the beneﬁts of bank
monitoring services are likely to be more valuable for ﬁrms that perform poorly but have
good growth opportunities, the monitoring hypothesis predicts that excess bond and stock
returns around bank loan announcements are more pronounced for these ﬁrms. In
contrast, the leverage hypothesis suggests that these ﬁrms face higher expected
bankruptcy costs and thus the beneﬁts ﬁom additional debt ﬁnancing should be smaller
for them. Consequently, the leverage hypothesis predicts that equity (bond)
announcement returns are less positive (more negative) for poorly performing ﬁrms with
high growth opportunities than for other ﬁrms. The wealth transfer hypothesis, however,
does not provide any direct implications for the interaction effect of ﬁrm performance
and growth opportunities on abnormal bond and stock returns.

In Table 6, we address this issue by dividing our sample events according to whether
the borrower’s past performance and grth opportunities (Tobin’s q) are above or below
their respective sample medians. Panels A and B report bond and stock CARs (-1, 1),
respectively, using the ratio of operating income before depreciation to total assets as a
measure of past performance. The main ﬁndings are that bond and stock CARs are
highest when poorly performing ﬁrms with good future growth opportunities borrow
funds from the banks. The mean bond and stock CARS for the subsample of ﬁrms with
poor past performance and high growth opportunities are 4.78% and 3.74%, respectively,
both of which are signiﬁcant at the 10% level. The corresponding mean and median

returns for other types of loans are small and insigniﬁcant except for the mean stock CAR

19

for a subsample of ﬁrms with poor past performance and low growth opportunities. These
results support the monitoring hypothesis.

Panels C and D of Table 6 provide bond and stock CARS (-l, 1), respectively, using
past cumulative excess stock returns as a measure of past performance. We again ﬁnd
that the mean bond and stock CARs are highest for a subsample of ﬁrms with poor past
stock performance and high growth opportunities, albeit the mean stock CAR is not

signiﬁcant.

4.2. Cross-Sectional Determinants of Bond and Stock CARs

To better understand the cross-sectional variation in bond and stock CARs, we report
the estimates from multivariate regressions. Because the univariate results above show
that equity beta, stock return volatility, leverage, past performance, and Tobin’s q are
important determinants of bond and stock CARS, we include these variables as key
explanatory variables. We also include ﬁrm size (log of total assets) as a control variable.

Table 7 reports the regression estimates using bond CARS (-1, 1) as the dependent
variable. The ﬁrst three regressions in Table 7 examine the relation between the bond
CAR and risk variables. We ﬁnd that the coefficients on equity beta and leverage in
model (1) and the coefficient on stock return volatility in model (2) are positive and
signiﬁcant. However, when we include all three of these variables in model (3), only the
coefﬁcient on stock return volatility is positive and signiﬁcant while the signiﬁcance of

the coefﬁcients on other risk variables disappears.9

 

9 In untabulated tests, we include as an alternative measure of ﬁrm risk a dummy for bond ratings that
takes the value of one if an issue has a non-investment grade and zero otherwise. We ﬁnd that the
coefﬁcient on this dummy variable is positive, but not signiﬁcant.

20

In models (4) through (6),“ we use past performance and growth opportunities as
explanatory variables. We ﬁnd that in model (4), the coefﬁcient on the ratio of operating
income before depreciation to total assets is negative and signiﬁcant at the 1% level and
the coefficient on Tobin’s q is positive and signiﬁcant at the 5% level. Thus, bondholders
of the borrowing ﬁrms realize higher returns when their ﬁrms perform poorly or when
they have good future prospects. In model (5), we replace the ratio of operating income
before depreciation to total assets with past cumulative excess stock returns and ﬁnd that
the coefficient on this variable is negative and signiﬁcant at the 1% level. However, the
coefﬁcient on Tobin’s q becomes insigniﬁcant. In model (6), we include all three of these
variables and ﬁnd that the coefficients on both the ratio of operating income before
depreciation to total assets and past cumulative excess stock returns are negative and
signiﬁcant. The coefﬁcient on ﬁrm size is negative and signiﬁcant, suggesting that the
beneﬁts of bank ﬁnancing to borrower bondholders are more pronounced for smaller
ﬁrms.

In model (7), we include all of the explanatory variables used in the previous
regressions. We ﬁnd that the coefﬁcients on the ratio of operating income before
depreciation to total assets, past cumulative excess stock returns, and ﬁrm size are
negative and signiﬁcant, and the coefficient on Tobin’s q is positive and signiﬁcant.
However, other risk variables turn out to be statistically insigniﬁcant. These results
suggest that ﬁrm performance and growth opportunities play a more important role in
explaining abnormal bond returns around bank loan announcements than ﬁrm risk does.

Finally, in models (8) and (9), we examine the interaction effect of past performance

and future growth opportunity on bond returns. We construct a dummy variable that takes

21

the value of one if the ratio of operating income before depreciation to total assets (past
cumulative excess stock return) is below the sample median and Tobin’s q is above the
sample median, and zero otherwise. The coefficient on this dummy variable is 0.03
(0.035) and signiﬁcant at the 5 % (1%) level. These results suggest that the average bond
CAR of ﬁrms that have poor past operating (stock return) performance and good growth
opportunities is higher than that of other ﬁrms by 3% (3.5%). Therefore, the interaction
effect of ﬁrm performance and growth opporttmities on bondholder wealth seems to be
both statistically and economically signiﬁcant.

In summary, both univariate and multivariate analyses suggest that bondholders of
borrowing ﬁrms earn higher returns if their ﬁrms perform poorly, if they have good
future growth opportunity, and if they are small. Abnormal bond returns are particularly
higher when ﬁrms perform poorly but have good future growth opportunities. These
results do not support the wealth transfer hypothesis or leverage hypothesis, but strongly
support the monitoring hypothesis.

Table 8 reports the regression estimates using stock CARs (-1, 1) as the dependent
variable. We use as explanatory variables those used in Table 7. In the ﬁrst regression,
we ﬁnd that the coefﬁcients on the equity beta and leverage variables are positive and
signiﬁcant. Thus, abnormal stock returns around bank loan announcements are greater for
risky ﬁrms than for safe ﬁrms. In the second regression we replace equity beta with stock
return volatility. The estimated coefficient on the stock return volatility is again positive
with a p-value of 0.01. In the third regression, we include all of the risk variables as
explanatory variables. We ﬁnd that the coefficients on stock return volatility and leverage

are positive and signiﬁcant, but the coefﬁcient on equity beta is positive and

22

insigniﬁcant. Thus, stock returns are higher when risky ﬁrms borrow funds from banks.
These ﬁndings are generally consistent with those in Table 7, which uses bond returns as
the dependent variable.

In models (4) through (6), we use past performance and growth opportunities as the
main determinants of abnormal stock returns. In all three models, the estimated
coefficients on past operating and stock return performance are negative and signiﬁcant
at the 1% level, suggesting that shareholders of borrowing ﬁrms realize higher returns
from bank ﬁnancing when their ﬁrms perform poorly in the past. These results are again
consistent with those using abnormal bond retmns as the dependent variable.

In model (7), we include all explanatory variables used in the previous regressions.
We ﬁnd that stock CARs are negatively and signiﬁcantly related to past operating and
stock return performance. The coefﬁcients on other variables, however, are not
signiﬁcant.

In the last two regressions, we examine the interaction effect of past performance and
future grth opportunity on stock returns. In model (8), the coefﬁcient on the dummy
variable for lower past operating performance and higher Tobin’s q is an insigniﬁcant
0.013 while in model (9), the coefficient on the dummy variable for lower past stock
return performance and higher Tobin’s q is a signiﬁcant 0.033.

Overall, the results in Table 8 are similar to those in Table 7. We ﬁnd that both
bondholders and shareholders of the borrowing ﬁrms beneﬁt more from bank ﬁnancing
when their ﬁrms face higher risk. When borrowing ﬁrms perform poorly, particularly

when they have good future growth opportunities, valuable monitoring services that bank

23

loans bring to the borrowing ﬁrms also beneﬁt both their shareholders and bondholders.
Taken together, these results strongly support the monitoring hypothesis.

In further analysis, we explicitly consider the direct implication of the wealth transfer
hypothesis by exploring the variables that measure the security of new bank loans and
bonds outstanding. We also consider the variables that measure new bank loans’ relative
security and size compared to bonds outstanding. The results are presented in Table 9.
The dependent variable in the ﬁrst ﬁve columns of Table 9 is the bond CAR (-1, 1). In
models (1) through (3), we use, respectively, a dummy variable that equals one if the new
bank loans are not secured and zero otherwise, a dtnnmy variable that equals one if the
bonds outstanding have higher security (i.e., the bond security level is either “senior” or
“senior secured”) and zero otherwise, and a dummy variable that equals one if the new
bank loans are secured but the bonds outstanding have low security (i.e., the bond

9’ C"

junior or subordinate,

,9 ‘6

security level is either “junior subordinate, senior
subordinate,” or “not speciﬁed”) and zero otherwise. As we discussed earlier, both the
wealth transfer and the leverage hypotheses predict that the impact of new bank loans to
current bondholders will be less detrimental in the ﬁrst two cases but more detrimental in
the last case. In other words, if the wealth transfer and the leverage hypotheses hold, we
should observe more negative coefﬁcients on the third dummy variable but less so on the
ﬁrst two dummy variables. However, we ﬁnd that the coefﬁcients on all three dummy
variables are not signiﬁcant.

To investigate the effect of relative maturity between new bank loans and existing

bonds on bondholder wealth, in model (4) we include a dummy variable that equals one if

loan maturity is longer than bond maturity and zero otherwise. Under both the wealth

24

transfer and the leverage hypotheses, the longer the maturity of loans relative to bonds,
the less likely that the current bondholders will suffer since the ﬁrm can use the proceeds
from loans to pay off the bonds as needed. Consequently, controlling for other effects, we
expect the coefficient on this dummy variable to be positive. The result, however, shows
that it is negative and insigniﬁcant.

Finally, in model (5) we replace the relative maturity dummy variable with a relative
size dummy variable that equals one if loan size is larger than bond size and zero
otherwise. The wealth transfer hypothesis predicts that, the larger the relative loan size,
the more the room for the borrowing ﬁrm to engage in wealth transfer from the current
bondholders to stockholders (through, for example, risky investments and stock
repurchases). In addition, the leverage hypothesis predicts that the larger the relative loan
size, the more detrimental it is to current bondholders. So both hypotheses predict a
negative relation between relative size disparity and bondholder returns. However, we
ﬁnd that the coefﬁcient on a relative size dummy variable is not signiﬁcant.

In the next ﬁve columns of Table 9, we use the stock CAR (-1, 1) as the dependent
variable. The coefﬁcients on an unsecured loan dummy variable in model (6), a relative
maturity dummy variable in (9), and a relative size dummy variable in (10) are not
signiﬁcant. On the other hand, the coefficient on a high bond security dummy in model
(7) is positive and signiﬁcant and the coefﬁcient on a relative security dummy in model
(8) is negative and signiﬁcant. Since the wealth transfer hypothesis predicts that the
coefﬁcient on the dummy variable in model (7) is negative and those in models (8) and

(10) are positive, these results are not consistent with the wealth transfer hypothesis.

25

Overall, we ﬁnd no evidence to support the view that the positive abnormal stock
returns surrounding announcements of bank credit agreements are due to either a wealth

transfer from the current bondholder to the stockholders or a leverage effect.

4. 3. Relation between Abnormal Bond and Stock Returns

The three hypotheses regarding the valuation effect of bank loan announcements also
suggest that abnormal bond returns and abnormal stock returns for the borrowing ﬁrms
are related, possibly due to bank monitoring, wealth transfer, or leverage effects. In this
section, we examine the relation between abnormal bond and stock returns and show how
the sensitivity of abnormal bond returns to abnormal stock returns differs across ﬁrm
performance and future growth opportunities. Table 10 reports the regression estimates.
We use the bond CARs (-1, l) for the borrowing ﬁrms as the dependent variable. In
model (1), we include the stock CARs (-l, 1) for the borrowing ﬁrms and ﬁrm size as the
explanatory variables. We ﬁnd that the coefﬁcient on the stock CAR (-1, 1) is 0.268,
which is signiﬁcant at the 1% level. Evaluating the estimated coefﬁcient at the mean
indicates that all else being constant, a 10% increase in the stock CARs is associated with
an approximately 2.68% increase in the bond CARS. Therefore, the positive relation
between the bond CARS and the stock CARs seems to be both statistically and
economically signiﬁcant.

The positive relation between the stock and the bond CARs supports the monitoring
hypothesis, which predicts that bank ﬁnancing brings beneﬁts to both shareholders and

bondholders. However, this result is not consistent with the wealth transfer hypothesis,

26

which predicts a negative relation between abnormal stock and bond returns, nor is it
consistent with the leverage hypothesis, which predicts a non-positive relation.

To examine whether the sensitivity of abnormal bond returns is different depending
on past borrower performance, in model (2) we include as explanatory variables a
dummy variable for good past operating performance and an interaction term between
this dummy and the stock CAR (-1, 1). The dummy variable for good past operating
performance takes the value of one if the ratio of operating income before depreciation to
total assets is above the sample median and zero otherwise. Here, therefore, the stock
CAR (-1, 1) measures the association between a poorly performing ﬁrm’s stock CAR (-1,
1) and its bond CAR (-1, 1). We ﬁnd that the coefﬁcients on the stock CAR (-1, 1) and its
interaction with the dummy for good past operating performance are signiﬁcantly
positive (correlation coefﬁcient = 0.393) and negative (correlation coefficient = -0.325),
respectively. Thus, relative to poorly performing borrowers, the extent of the positive
relation between the stock CAR (-1, l) and the bond CAR (-1, 1) is less for the good
performing borrowers. These results are consistent with the view that the market’s
positive valuation of bank monitoring is more pronounced when the borrowing ﬁrm
performs poorly, but are not consistent with the wealth transfer hypothesis, which
predicts a negative relation between abnormal stock and bond returns.

In model (3), we replace the dummy variable for good past operating performance in
model (2) with a dummy variable for good past stock performance. The dummy variable
for geod past stock performance takes the value of one if the previous one-year excess
returns for the borrowing ﬁrm are above the sample median and zero otherwise.

Consistent with the results in model (2), we ﬁnd that the coefficient on the stock CAR (-

27

1, l) is signiﬁcantly positive (correlation coefﬁcient = 0.291) but the coefficient on its
interaction with the dummy for good past stock performance is signiﬁcantly negative
(correlation coefﬁcient = -0.270). Thus, the positive relation between the stock CAR (-1,
1) and the bond CAR (-1 , l) is stronger for borrowers whose stocks have performed
poorly in the past. This result is again inconsistent with the wealth transfer hypothesis.

In model (4), we examine whether the link between the stock CAR (-1, 1) and the
bond CAR (-1, l) is different between borrowing ﬁrms with good future growth
opportunities and those with poor future growth opportunities. To make this comparison,
we use a dummy for high Tobin’s q and an interaction term between this dummy and the
stock CAR (-1, 1). A dummy variable for high Tobin’s q takes the value of one if the
borrower’s Tobin’s q is above the sample median and zero otherwise. The coefﬁcient on
the stock CAR (-1, 1) is signiﬁcantly positive. The coefficient on the interaction term
between the stock CAR (-1, 1) and the dummy for high Tobin’s q, however, is
insigniﬁcantly positive. Thus, the positive relation between the stock CAR (-1 , 1) and the
bond CAR (-1, 1) exists irrespective of borrower Tobin’s q.

In model (5) we repeat the analysis of the previous regressions but using a dummy
variable for poor past operating performance and high Tobin’s q and an interaction term
between this dummy and the stock CAR (-1, 1). The dummy variable for poor past
operating performance and high Tobin’s q takes the value of one if the ratio of operating
income before depreciation to total assets is below the sample median and Tobin’s q is
above the sample median and zero otherwise. The monitoring hypothesis predicts that the
signaling effect of the lending bank’s commitment to provide monitoring services is

larger when borrowers perform poorly, but have good future growth opportunities.

28

Consistent with this View, we ﬁnd that the coefficient on the interaction term between the
stock CAR (-1, l) and the dummy. for poor past operating performance and high Tobin’s
q is positive and signiﬁcant.

In model (6), we repeat the same analysis using a dummy variable for poor past stock
performance and high Tobin’s q. The dummy variable for poor past stock performance
and high Tobin’s q takes the value of one if the previous one-year excess returns for the
borrowing ﬁrm are below the sample median and Tobin’s q is above the sample median
and zero otherwise. We again ﬁnd that the coefficient on the interaction term between the
stock CAR (-1, l) and the dummy for poor past stock performance and high Tobin’s q is
positive and signiﬁcant.

In the last regression, we use a dummy variable for secured loans and low security
bonds and an interaction term between this dummy and the stock CAR (- l , 1). The
dummy variable for secured loans and low security bonds takes the value of one if new
bank loans are secured and bonds outstanding have lower security. To the extent that the
wealth transfer from borrower bondholders to either lending banks or borrower
shareholders is more likely to occur when new bank loans are secured but bonds
outstanding have lower security, the wealth transfer hypothesis predicts that the
coefficient on the interaction term is negative. Since the adverse effect of an increase in
leverage on bondholder wealth is more severe in this case, the leverage hypothesis also
predicts a negative coefficient on this interaction term. However, we ﬁnd that the

coefficient on the interaction term is positive and insigniﬁcant.

29

In untabulated tests, we include variables used in Table 7 as explanatory variables and
ﬁnd that the results are qualitatively similar to those reported in Table 10, albeit the
coefficients on interaction terms in models (3) and (4) are marginally insigniﬁcant.

Overall, the results in Table 10 indicate that the positive relation between abnormal
bond and stock returns is most pronounced for borrowing ﬁrms that perform poorly,
particularly those that have high Tobin’s q. These results suggest that the beneﬁts of bank
monitoring are greater for ﬁrms that are ﬁnancially weaker, but have good future

prospects, supporting the monitoring hypothesis.

5. Summary and Conclusion

Using daily bond and stock return data, we investigate the valuation effect of
borrowing ﬁrms surrounding announcements of bank credit agreements during the 1995
to 2003 period. We consider three competing hypotheses regarding the valuation effect of
bank loan announcements. According to the monitoring hypothesis, banks play a special
role as information transmitters in the capital markets and perform an important value-
enhancing intermediary function for both the shareholders and the bondholders of the
borrowing ﬁrms. The wealth transfer hypothesis posits that bank credit agreements lead
to wealth transfer ﬁom either lending banks or bondholders to shareholders of borrowing
ﬁrms. The leverage hypothesis suggests that, assuming borrowing ﬁrms are not risky or
current leverage is low, the additional tax beneﬁts resulting from the new bank loan
ﬁnancing increase shareholder wealth. However, the impact of an increase in leverage
due to new bank loan ﬁnancing on bondholder wealth is expected to be either negligible

if the ﬁrm is healthy or negative if the higher leverage imposes signiﬁcant ﬁnancial

30

distress risk on bondholders, implying a non-positive effect of loan announcements on
bondholder wealth.

Our results strongly support the monitoring hypothesis. We ﬁnd that bank loan
agreements are associated with positive announcement returns for both shareholders and
bondholders of borrowing ﬁrms. For example, the average three-day cumulative
abnormal return surrounding announcements of bank credit agreements is 1.48% for
shareholders, 1.01% for bondholders, and 1.15% for the entire ﬁrm, all of which are
signiﬁcant at the 5% level.

We also ﬁnd that both abnormal bond and stock returns are considerably higher for
ﬁrms that are riskier and ﬁrms that Show worse performance. In particular, the bond
(stock) market’s positive valuation of bank loan ﬁnancing is more pronounced when the
borrowing ﬁrm performs poorly but has good future prospects. For example, all else
constant, the bondholders (shareholders) of the borrowing ﬁrms with poor past stock
performance and high Tobin’s q realize a 3.5% (3.3%) higher abnormal announcement
return, signiﬁcant at the 5% level or lower, than those of other types of borrowing ﬁrms.

These results are inconsistent with either the wealth transfer hypothesis or the
leverage hypothesis. Furthermore, we ﬁnd no evidence supporting these two hypotheses
along other dimensions. In particular, we ﬁnd that the abnormal bond returns are not
statistically different between the two subgroups in each of the following cases: (1) loans
that are relatively more secured than bonds versus loans that are relatively less secured
than bonds; (2) loans that have relatively longer maturity than bonds versus loans that
have relatively shorter maturity than bonds; and (3) loans that have relative larger size

than bonds versus loans that have relatively smaller size than bonds. Abnormal stock

31

returns show similar patterns except that secured loans with lower bond security have
lower abnormal stock returns than other types of loans.

Finally, we ﬁnd that bond announcement returns are positively and signiﬁcantly
related to stock announcement returns and that this positive relation is more pronounced
when borrowing ﬁrms perform poorly, particularly when they have good future
prospects. These results further support the monitoring hypothesis, which posits that a
bank’s decision to lend to borrowers signals the lending bank’s commitment to provide
monitoring services — services that have a positive effect on both shareholder and
bondholder wealth of borrowing ﬁrms.

Overall, examining both stock and bond returns data, this paper provides new
evidence on the valuation effect of bank loan announcements and its cross-sectional
determinants. The results strongly support the view that banks perform a valuable
monitoring function for both shareholders and bondholders of borrowing ﬁrms, and reject
the wealth transfer and the leverage hypotheses as a potential explanation for the positive

excess stock returns surrounding bank loan announcements.

32

APPENDIX

33

Table 1
Predictions of Monitoring, Wealth Transfer, and Leverage Hypotheses for Key
Variables

This table summarizes the predictions of the monitoring, wealth transfer, and leverage
hypotheses for key variables. The monitoring hypothesis posits that banks play a
special role as information transmitters in the capital markets and perform an important
value-enhancing intermediary function for both shareholders and bondholders of the
borrowing ﬁrms. The wealth transfer hypothesis posits that bank credit agreements lead
to wealth transfer from either lending banks or bondholders to shareholders of the
borrowing ﬁrms. The leverage hypothesis suggests that, assuming borrowing ﬁrms are
not risky or current leverage is low, the higher leverage resulting ﬁom the new bank
loan ﬁnancing is beneﬁcial to shareholders. However, the impact of such an increase in
leverage on bondholder wealth is likely to be either negligible if the ﬁrm is healthy or
negative if the higher leverage imposes signiﬁcant ﬁnancial distress risk on
bondholders, implying a non-positive effect of loan announcements on bondholder

wealth.

 

 

Monitoring Wealth Leverage Empirical results
hypothesis transfer hypothesis
hypothesis
Abnormal stock + + + +
returns
Abnormal bond returns + - Non-positive +
Effect of high risk, poor Positive Not clear Negative Positive effect on
past performance, and effect on effect on abnormal stock returns
high grth on abnormal abnormal
abnormal stock and stock returns stock returns
bond returns
Positive Negative Positive effect on
effect on effect on abnormal
abnormal abnormal bond returns
bond returns Bond returns
Effect of higher Not clear Positive effect Not clear Negative (no) effect of
seniority (higher on abnormal higher loan security
security and shorter stock returns (shorter maturity and
maturity) and larger larger size) on
size of new bank loans abnormal stock returns
than bonds on
abnormal stock and Negative Negative No effect on abnormal
bond returns effect on effect on bond returns
abnormal abnormal
bond returns Bond returns
Correlation between + - Non-positive +
abnormal stock and
bond returns

 

34

Table 2
Distribution of Bank Loan Announcements by Year and Borrower Industry

The sample consists of 254 bank loan announcements ﬁom 1995 to 2003. The sample
is compiled by searching Factiva for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded.

 

 

Agriculture,
Year Fiﬁiﬁiﬁb d Construction Manufacturing 21681132333232; “lhgifjﬁle faint? Total
Mining
1995 0 O 4 1 l 0 6
1996 2 1 5 3 2 7 20
1997 2 0 15 1 1 6 25
1998 0 O 7 1 0 2 10
1999 1 1 5 3 l 0 11
2000 2 0 4 3 2 3 14
2001 l 1 37 18 9 10 76
2002 2 2 20 13 8 10 55
2003 2 O 18 6 4 7 37
Total 12 5 1 15 49 28 45 254

 

35

Table 3
Summary Statistics of Borrowers, New Bank Loans, and Bonds Outstanding

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching F activa for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days —l7O to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. When a borrowing ﬁrm has multiple bonds
outstanding, bond security, rating, and maturity are estimated as the weighted average
of the bonds’ securities, ratings, and maturities, respectively, where the weights are
assigned according to the face values of the bonds. The bond security score ranges from
one to six, with a higher score indicating higher security. Speciﬁcally, the score takes
the value of one if the bond security level is junior subordinate, two if it is junior or
subordinate, three if it has no special security features, four it is senior subordinate, ﬁve
if it is senior, and six if it is senior secured. The bond rating score ranges from one to
eight, with a higher score indicating lower rating. Speciﬁcally, the score takes the value
of one if the bond rating is AAA or Aaa, two if it is AA or Aa, three if it is single A,
four if it is triple BBB or Baa, ﬁve if it is BB or Ba, six if it is B, seven if it is triple
CCC or Caa, and eight if it is lower than the 7th category.

36

Table 3

 

Sample Mean Median $31.1de
srze devratron

Panel A: Characteristics of borrowing ﬁrms

 

 

 

 

 

Total assets (in billion dollars) 254 10.95 3.40 22.76
Sales (in billion dollars) 254 8.27 2.30 18.28
Market value of equity (in billion dollars) 252 10.10 1.79 27.06
Leverage (T ptal debt / (Total debt + Market 252 0.432 0.388 0245
value of equrty))
Sspgsatmg income before deprecratron / Total 253 0.114 0.116 0.108
Tobin’s q ((Market value of equity + Book
value of debt) / Total assets) 252 1'219 ””6 0'857
Past one-year cumulative excess stock returns 254 0.030 0.085 0.878
Equity beta 254 0.933 0.842 0.600
Stock return volatility (Standard deviation of 254 0.588 0.461 0.386
prevrous one-year darly excess stock returns)
Panel B: Characteristics of new bank loans

Size of loans (in billion dollars) 240 1.20 0.50 2.82
Dummy for secured loans 192 0.620 - -
Maturity of loans (in years) 155 3.507 3.000 1.837

Panel C: Characteristics of bonds outstanding
Size of bonds outstanding (in billion dollars) 254 1.02 0.34 1.90
Bond security 254 4.654 5.000 0.544
Bond rating 253 4.878 5.000 1.413
Maturity of bonds (in years) 254 7.950 6.325 5.992

 

37

Table 4
Mean and Median Cumulative Abnormal Returns (CAR) around the
Announcement Date (AD)

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching F activa for loan announcements. To be included in the
sample, borrowing frrrns must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal stock returns are computed using the
market model. The market model is estimated by using 150 trading days of stock return
data ending 20 days before the loan announcement. The CRSP value-weighted return is
used as the benchmark. Abnormal bond returns (ABR) are computed using the
following model: ABRit = BRit — (ai + ,8] x Chg_Spreadt + ,8 2 x Chg_Slopet), where
BRitis the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in the Baa- over Aaa-
rated bond yields, and Chg_SIopet is the change in the slope of the yield curve at time t,

which is deﬁned as the change in the difference between lO-year Treasury over l-year
Treasury yields. This model is estimated using 150 trading days of bond return data
ending 20 days before the loan announcement. When a borrowing ﬁrm has multiple
bonds outstanding, the ABR is estimated as the weighted average of the bonds’
abnormal returns, where the weights are assigned according to the face value of each
bond. The daily abnormal stock (bond) returns are accumulated to obtain the stock
(bond) cumulative abnormal return (CAR) between any two dates. AD refers to the
announcement date. Numbers in parentheses are p-values for the test that the
mean/median is equal to zero. The ﬁrm CAR is estimated as a value-weighted average
of the stock CAR and the bond CAR with the weights being the market value of equity
and the book value of debt measured at the ﬁscal year-end that immediately precedes an
announcement date.

38

Table 4

 

 

 

 

 

 

 

Event windows 83:16 Mean Median
Panel A: Stock CARs (%)
(A.-.) 2323.33 '83:;
In; 2:35
(AD+1) 254 30.14368) 36.17393
(W... 15:31: 3.33.:
(AD-1, AD+1) 254 1637092? frills;
Panel B: Bond CARs(%)
(ADJ) 254 (393197) (3.215;
(A1» 254 is???) 23:31?
(A...) 3.15:. 3.93:;
(AD'I’ AD) 254 I 33.51525; (3.228))
(arm... 25. (3.103;: 369:;
Panel C: Firm CARs (%)
w» 23:22.: is:
in: at;
(AD+1) 252 €6.23) 3097187)
«mm “£83" is,
(AD-1, AD+1) 252 16(1):)823‘ 35.11847)

 

39

Table 5
Three-Day Cumulative Abnormal Returns (CARs) for Borrowing Firms
Categorized by Firm and Loan Seniority Characteristics

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching F activa for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal stock returns are computed using the
market model. The market model is estimated by using 150 trading days of stock return
data ending 20 days before the loan announcement. The CRSP value-weighted return is
used as the benchmark. Abnormal bond returns (ABR) are computed using the
following model: ABR” = BRit — (ai + ,8] x Chg_Spreadt + ,8 2 x Chg_Slopet), where
Brit is the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in the Baa- over Aaa-
rated bond yields, and Chg_SIopet is the change in the Slope of the yield curve at time t,

which is deﬁned as the change in the difference between 10-year Treasury over l-year
Treasury yields. This model is estimated using 150 trading days of bond return data
ending 20 days before the loan announcement. When a borrowing ﬁrm has multiple
bonds outstanding, the ABR is estimated as the weighted average of the bonds’
abnormal retruns, where the weights are assigned according to the face value of each
bond. The daily abnormal stock (bond) returns are accumulated to obtain three-day
cumulative abnormal return (CAR) from day -1 to day +1 relative to the announcement
date. Low security bonds are those for which the security level is either “junior
subordinate,” “jtmior or subordinate,” “senior subordinate,” or “not speciﬁed.”

Numbers in parentheses are p-values for the test that the mean/median is equal to zero.

40

Table 5

 

Bond CARS (-l, 1) Stock CARs (-l, 1)

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean Median Mean Median
Panel A: By bond rating
_ 0.177 0.032 0.322 0.031
Investment grade bonds (N — 107) (0.15) (0.36) (0.39) (0.30)
. = 1.629“ -0.043 2.323" 0.644
Non-investment grade bonds (N 146) (0.06) (0.99) (0.02) (0.33)
Tes t-o f. di fference 1.452“ 0.075 2.001 * 0.613
(0.09) (0.4@ (0.07) (0.99)
Panel B: By equity beta
. _ 0.435 0.000 0.356 0.031
Below sample median (N - 127) (0.31) (0.71) (0.66) (0.92)
. _ 1.588" 0.022 2603*" 0660*
Ab” samPl" med‘a“ (N ’ ‘27) (0.08) (0.72) (0.01) (0.06)
. 1.153 0.022 2.247 * 0.629
Tes‘ff'd‘ffe’ence (0.24) (0.85) (0.07) 40.15)
Panel C: By volatility (Standard deviation of previous one-year daily excess stock returns)
. _ 0.006 0.000 -0.331 0.018
Below sample median (N - 127) (0.96) (0.63) (0.25) (0.96)
. ___ 2.018" 0.032 3290*“ 1.181"
Above sample median (N 127) (0.04) (0.28) (0.01) (0.04)
1.65th -di fference 2.012" 0.032 3621*" 1.163"
(0.04) (0.38) ($00) (0.09)
Panel D: By leverage (Total debt / (Total debt + Market value of equity))
. _ 0.020 0.015 0.103 0.029
Below sample median (N - 126) (0.92) (0.90) (0.77) (0.45)
. = 2.022" 0.003 2.859" 0.653
Ab” saml’le med‘a“ (N ‘26) (0.04) (0.56) (0.02) (0.22)
Test-of-difference 2.002“ 0.012 2.756" 0.624
(0.04) (0.83) (0.03) (0.67)
Panel B: By operating performance (Operating income before depreciation / Total assets)
. _ 2.033" -0.065 2849*" 1.224“
381°“ “mm" mm“ (N ‘ ‘26) (0.04) (0.95) (0.00) (0.03)
. _ 0.016 0.027 0.164 -0.131
Abm’e “mm“ mm“ (N " 127) (0.92) (0.29) (0.83) (0.71)
Test-of- difference 2.017" 0.092 2.685" 1.355"
(0.04) (0.39) (0.03) (0.05)
Panel F: By stock return performance (Previous one-year cumulative daily excess returns)
. _ 1.976" 0.020 2.913M 1.248"
Below sample median (N — 127) (0.04) (0.42) (0.01) (0.02)
. = 0.048 -0.009 0.046 -0.380
Ab” 53mm" ”‘1‘“ (N ‘27) (0.73) (0.83) (0.92) (0.53)
. 1.928" 0.029 2.867" 1.628"
Tes"°f’d‘ffe’°”°e (0.05) (0.47) (0.02) (0.03)

 

41

Table 5 (cont’d)

 

Panel G: By Tobin’s q (Market value of equity + Book value of total debt) / Total assets))

 

 

 

 

 

 

 

. 0.434 -0093 1.537“ 0.369
Below sample medran (N = 125) (0.36) (0.13) (0.03) (0.37)
. _ 1599* 0.097" 1.426 0.411
Above sample medran (N — 127) (0.07) (0.03) (0.16) (0.38)
Temﬁdiﬁmme 1.165 0.190“ 0.1 11 0.042
(0.24) (0.01) (0.93) (0.95)
Panel H: By relative security between loans and bonds
Loan is secured and bond has lower security (N = 44) 3625099; 300587) ((0278) 2332:;
= 1.781" 0.001 3.048m 1139*"
Others (N ‘48) (0.03) (0.86) (0.00) (0.01)
. 1572* 0.036 3819*" 1.226"
Tesmmffe’ence (0.09) (0.74) (0.00) (0.05)
Panel 1: By relative maturity between loans and bonds
. . 0.630 0.01 1 1.673" 0715*
“’3“ mam” <= b°“d mat‘m‘y (N = 13 1) (0.24) (0.65) (0.02) (0.09)
. . = 0.253 -0.093 1.844 1.656
L03“ mam” > b‘m“ mam” (N 24) (0.31) (0.87) (0.31) (0.45)
Temﬁdiﬁmm 0.377 0.104 0.171 0.941
(0.52) (0.96) (0.93) (0.61)
Panel J: By relative size between loans and bonds
. _ . _ 1291* -0.028 1.547 0.307
”a” 5‘23 <‘ b‘md 5‘28 (N ‘ 129) (0.09) (0.81) (0.14) (0.84)
. . = 0.145 0.024 1129* 0.585
”a” 317‘" > b‘md 517:6 (N 11 1) (0.24) (0.56) (0.05) (0.13)
. 1.146 0.052 0.418 0.278
Tes‘fff‘fferenc" (0.13) (0.79) (0.73) (0.35)

 

42

Table 6
Three-Day Cumulative Abnormal Returns (CARs) for Borrowing Firms by
Borrower Performance and Growth Opportunities

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching F activa for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded ffom the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal stock returns are computed using the
market model. The market model is estimated by using 150 trading days of stock return
data ending 20 days before the loan announcement. The CRSP value-weighted return is
used as the benchmark. Abnormal bond returns (ABR) are computed using the
following model: ABRit = BRit — (ai + ,B 1 x Chg__Spreadt + )6 2 x Chg_Slopet), where
Brit is the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in the Baa- over Aaa-
rated bond yields, and Chg_Slopet is the change in the slope of the yield curve at time t,

which is deﬁned as the change in the difference between 10-year Treasury over l-year
Treasury yields. This model is estimated using 150 trading days of bond return data
ending 20 days before the loan announcement. When a borrowing ﬁrm has multiple
bonds outstanding, the ABR is estimated as the weighted average of the bonds’
abnormal returns, where the weights are assigned according to the face value of each
bond. The daily abnormal stock (bond) returns are accumulated to obtain three-day
cumulative abnormal return (CAR) from day -1 to day +1 relative to the announcement
date. Growth opportunity is measured by Tobin’s q ((market value of equity + book
value of total debt) / total assets). Numbers in parentheses are p-values for the test that
the mean/median is equal to zero.

43

Table 6

 

Panel A: Bond CARS (-1, 1) across past operating performance and growth opportunities

 

 

 

Growth opportunities (Tobin’s q)
High Low Test-of-difference
N Mean Median N Mean Median Mean Median
Operating Low 41 4.784" 0.466“ 84 0.714 -0.159 4.070 0.625”
rgctome (0.08) (0.08) (0.29) (0.13) (0.14) (0.04)
e ore
depreciation High 86 0.081 0.03 1 40 -0.1 15 0.009 0.196 0.022
[Total assets (0.58) (0.13) (0.75) (0.83) (0.62) (0.40)
Test-of- 4.703 "‘ 0.435 0.829 0.168
difference (0.08) (0.19) (0.28) (0.3 1)

 

Panel B: Stock CARS (-1, 1) across past operating performance and growth opportunities

 

Growth opportunities (Tobin’s q)

 

 

High Low Test-of-difference
N Mean Median N Mean Median Mean Median
Operating Low 41 3.743" 1.351 84 2.431" 0.925 1.312 0.426
1::ng (0.08) (0.13) (0.02) (0.14) (0.57) (0.60)
ore
depreciation High 86 0.321 -0.087 40 -0.200 -0.256 0.521 0.169
fTotal assets (0.78) (0.87) (0.70) (0.56) (0.67) (0.70)
Test-of- 3 .422 1.438" 2.631“ "' 1.181
difference (0.12) (0.09) (0.02) (0.20)

 

Table 6 (cont’d)

 

Panel C: Bond CARs (-1, 1) across past stock return performance and growth opportunities

 

Growth opportunities (Tobin’s q)

 

 

High Low Test-of-difference
N Mean Median N Mean Median Mean Median
Previous Low 64 2.949" 0.132“ 61 1.025 -0.1 12 1.924 0.244
one-year (0.09) (0.09) (0.28) (0.66) (0.33) (0.16)
cumulative
excess stock High 63 0.228 0.032 64 -0. 129 -0.073"' 0.357 0.105"
returns (0.27) (0.16) (0.47) (0.08) (0. 19) (0.03)
Test-of— 2.721 0.100 1.154 0.039
difference (0.12) (0.61) (0.23) (0.59)

 

Panel D: Stock CARs (-1, 1)

across past stock return performance and growth opportunities

 

Growth opportunities (Tobin’s q)

 

 

High Low Test-of-difference
N Mean Median N Mean Median
Previous Low 64 3.117 1.338 61 2.752" 1.125 0.365 0.213
0113'”?! (0.11) (0.11) (0.03) (0.18) (0.87) (0.76)
cumulative
excess stock High 63 -0.292 -0.048 64 0.379 -0.402 0.671 0.354
retmns (0.59) (0.51) (0.58) (0.83) (0.44) (0.81)
Test-of- 3.409* 1.386“ 2.373“ 1.527
difference (0.09) (0.07) (0.09) (0.20)

 

45

Table 7
OLS Regression of the Three-Day Cumulative Abnormal Bond Returns on
Explanatory Variables

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching F activa for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded ﬁ'om the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal bond returns (ABR) are computed using
the following model: ABRit = BRit — (ai + ,8] x Chg_Spreadt + [32 x Chg_SIopet),
where BRitis the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in the Baa- over Aaa-
rated bond yields, and Chg_Slopet is the change in the slope of the yield curve at time t,

which is deﬁned as the change in the difference between 10-year Treasury over l-year
Treasury yields. This model is estimated using 150 trading days of bond return data
ending 20 days before the loan announcement. When a borrowing ﬁrm has multiple
bonds outstanding, the ABR is estimated as the weighted average of the bonds’
abnormal returns, where the weights are assigned according to the face value of each
bond. The daily abnormal bond returns are accumulated to obtain three-day cumulative
abnormal return (CAR) from day -1 to day +1 relative to the announcement date.
Numbers in parentheses are p-values. ***, **, and * denote signiﬁcance of the
parameter estimates at the 1%, 5%, and 10% levels, respectively.

46

 

 

am 8m 5 8m N8. NR 268836 8 .oz
8.8 8.: NE: 88 :8 85 6%:
88.8 82.8 28.8 88... 98.8 88.8 a: -888...
38:53 5:66 06 gone mm a "2.538. \ 5:38 05
323 mm 532 x88 8088 038383 808-25 3355
3883; 58:02: 2: 30% m_ a 3.38. \ 5.32: o5
323 m_ Somme .38 \ 2258.58 283 2:88 @3825
8.8 28.8 8.8 8.8 88 A28
588.... 3.88.8 2.88.8 3.8.8 88.? .88 was .998 won
88 5.8 8.8 888 86.8 \ 88
88.8 8.8.8 2.85.8 86:6 28> 88 + £8... 8 2:9 880.828 8 8.83
8.8 8.8
«3348.? 1.... $86- 3.5:: :0on 8088 028385 808.80 323.5
8.8 8.8
:82 mo- 588.8 was .88. \ 88888 288 2:85 wéaoao
88 E8 88
888 88.: £888 2888 8 28> Ea: + 58 :88 \ 38 868 omega:
8.8 8.8 9832 0.8.2 888
2.88.8 3.88.8 8% 8688 32.68 .6 88:8 .0858 £82.;
5 .8 88
8:8 3.88.8 88 88m
88 8.8 8.8 8.8 c ”.8 5.8
3.88... .........88.8 t. 88.8 88.8 88.8 88.8 88825
8 5 E 6 A8 5 $389 868885

 

h 033—.

47

 

 

_mm _mm .3 808.2030 .80 .02
was VmN— «Wm. 02?...—
NE... 8...... «SN... N883?
898 988:3 5:008 05 0250 8 .0 9:38. \ 5:008
2138.: 05 30.0.. m. 8808 0.08... 800.3 03338.6 808-080 30305
388 $8833 8“:68 05 0.68. m. .0 $8308 \ 5:008 08
1.886 322. 8 308.0 .88 \ 80380080.. 0880.. 0888 8880.5
8.8 .88 8.8
.. 386- 3086- 33.86- 308.. .88 .8 we:
8:8 3088 .83. \
.386 A508 .88.? 0...? «com + £5.00 80 02.3 .8523 0 3:30...
88 8.8
3.386. 2:, _ mmcd. 8.80. 0.008 8083 088383 808-080 8.030:—
88 8.8
3.2.35.9. 3.3.536. $08.“ .88. \ 80880080.. 0.5.809 08008 @8808
.88 .88 8.8
$8... :88... 88... .333 8 2...; .282 + 5% Eat \ 5% .598 omega.
8 . .8 :88 8.28 8832 0.8:. 38.6
88... a8... 88... 8.8 38-25 325:. 8 88.3.. .8838 8.3.5
.88 3.8 .88
2.86 N. 36 $86 809 83m
.88 .88 8.8
236 586 $856 809.08.
A3 :8 AS 330.88 80:82.08:

 

8.808 5 02a...

48

' Table 8
OLS Regression of the Three-Day Cumulative Abnormal Stock Returns on
Explanatory Variables

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching Factiva for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal stock returns are computed using the
market model. The market model is estimated by using 150 trading days of stock return
data ending 20 days before the loan announcement. The CRSP value-weighted return is
used as the benchmark. The daily abnormal stock returns are cumulated to obtain three-
day cmnulative abnormal return (CAR) from day -1 to day +1 relative to the
announcement date. Numbers in parentheses are p—values. ***, **, and * denote
significance of the parameter estimates at the 1%, 5%, and 10% levels, respectively.

49

 

 

EN NR .8 N8 N8 N2 8235on co 52
_2 85 v3. 3:. 2w 2:. can;
we 3 $25 838 8.8.8 88.8 $3.: mm 6333.
9:583 5:58 05 96% m_ a FEB—oh \ 5:65 05
323 mm E38 x83 $85 033383 33-28 30305
35:53 5:68 2: 03% m_ a 9:33. \ 5:58 2:
323 mm Somme. _SS \ :ouﬁoeqoc 283 2:85 wag—30
5.8 898 3.8 8&8 $8 $8
38.? ~88? 2 8.? ”83 3.85 :8... some 33% won
398 5.8 $8 33.3 _58. \ 2%
2.8.? :38? 88.8 39.6 83> :25 + be? he 2.3 3328 a ranch
898 88.8
signed. stunned- 2:32 goo; 3088 028383 anemone 98305
888 898
tact? 2.3.28? 333 EA; \ 550238 882. 0885 3338
93.8 a. .8 888
.83... 2.3.8 528... 28:3 co 8.9, Es: + 5% Eat \ 3% Bob ”@223
698 :98 @632 xosm mass
.5 8... 5835 $8 auras Saga .8 8338 nausea £35
5.8 A88
2 8.: 2:23 92. beam
398 6.8 E8 A88 2: .8 S .8
.385 38... 88... .33.? .98.? 88.? @225
A8 6 E 5 S E 8.35., 2.88385

 

w 035,—.

50

 

 

_mm _mm _mm 805338. mo .02
NS 3.6 63 can;
885 an; $2.8 N: -86.?
398 38:83 5:68 65 655 m_ a @898.
19.8.: \ 5:68 65 30—62 mm 8.86.. 6.6on 3688 635.883 56.380 8036.:—
818 38883 5:68 65 265 mm a 98:8. \
336 5:68 65 323 m_ 385 .88 \ 55623:. 282. 6888 8:825
58 5.8 5.8
88.8 22:. 885 sows 53% won
888 A385
:86 as: \ 2% 83% 83> 6.6 + 8:3 .6 2:9 $828 a 98.8.
88.8 88.8
*uahwmod. univamoqon men—50.— v—ooum mmOQXD 033—3830 avenue—Ho 9530.5
5.8 95.8
1.326. 1.32 .o- $85 :38. \ 86566868 2&3 6888 88830
5.8 5.8 c .8
r. _ 65 5&de :85 28:3 6o 8.9, $52 + 3% Bob \ 3% :88 6823
9:8 828 5.8 @532
£85 _2 o... ~25 62m 386 8% .3625 32.55% 883% 5858 88:5
5.8 2 m8 38
ES... 8 :3 N83. as 886
868 8.8 5.8
:3? 33.? RS? 8885
A8 a: E 833.8, .5838:—

 

8888 a «38.

51

Table 9
OLS Regression of the Three-Day Cumulative Abnormal Bond and Stock Returns
on Wealth Transfer Variables

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching Factiva for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days --170 to +10 relative to the bank loan
announcement), and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal stock returns are computed using the
market model. The market model is estimated by using 150 trading days of stock return
data ending 20 days before the loan announcement. The CRSP value-weighted return is
used as the benchmark. Abnormal bond returns (ABR) are computed using the
following model: ABR” = BRit — (ai + ,8] x Chg_Spreadt + ,82 x Chg_Slopet), where
Brit is the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in the Baa- over Aaa-
rated bond yields, and Chg_Slopet is the change in the slope of the yield curve at time t,

which is deﬁned as the change in the difference between 10-year Treasury over l-year
Treasury yields. This model is estimated using 150 trading days of bond return data
ending 20 days before the loan announcement. When a borrowing ﬁrm has multiple
bonds outstanding, the ABR is estimated as the weighted average of the bonds’
abnormal returns, where the weights are assigned according to the face value of each
bond. The daily abnormal stock (bond) returns are accumulated to obtain three-day
cumulative abnormal return (CAR) from day -1 to day +1 relative to the announcement
date. High security bonds are those for which the bond security level is either “senior”
or “senior secured.” Low security bonds are those for which the security level is either
“junior subordinate,” “junior or subordinate,” “senior subordinate,” or “not speciﬁed.”
Numbers in parentheses are p-values. ***, **, and * denote signiﬁcance of the
parameter estimates at the 1%, 5%, and 10% levels, respectively.

52

 

 

RN 3: 66. .8 a: 26662866 .6 .62
66.6 :6 Ed 6...: 66.2 6665
.6666 666.6 662.6 6666.6 2 66.6 am. -6233.
86.8
686.6. 36668 6% 6666 66.6666. a on: 5.3
M96”. 86668 6566 6.66 55 66.6. a £666 663
“$66“. A 6 86668 €39. 32 a; 6.66 65 6288 a 83
:6
6:66 25:68 6:68... 666 86 666m
2 6.8
E66 @6668 626886 a .63
$6.8 86.8 26.8 86.8 26.8
«N956. 359$. 63.329? 6516—9? *eeNQod. 38mm 33:0 wou—
868 5.8 c _.8 2 _.8 § .8 33.0.6 666.1666
«666.6 266.6 668.6 626.6 6666.6 66.6 36> 6666 + 6666 6 86> 68.628 6 6.66.;
86.8 86.8 86.8 86.8 86.8 .
5686.6. 2.6866. 3.66666. 56666.6. 2.68.6.6. mesa 636 686 636668 663.6 96626
66.8 :68 66.8 86.8 86.8
.566. 6666.6 :826. .3668 .6. 2.826. 6366 6E \ 6662666 6666 6686 6668660
86.8 36.8 5 .8 c 6.8 :68 26666
626.6 626.6 666.6 866.6 .2666 6 36> .852 + 666 66.8 \ 666 668 66293
$6.8 $6.8 86.8 86.8 c 6.8 6836.. 696 686
6666.6 5 6.6 66 _ 6.6 6666.6 28.6 666 666.666 6.6666 6 6666666 666658 £6625
36.8 66.8 96.8 86.8 66.8
366.6 6666.6 3666.6 6666.6 2.6266 866 5666
§ .8 5.8 2 N8 $6.8 66.8
636.6 668.6 366.6 .6686 56.6 68.35
5 A8 5 A8 A 8 66666.; 66656665

 

2 .3 «<0 666m

 

a 036.“.

53

 

 

666 66. 66. .66 66. 26.626666 6 .62
6.6 66.6 66.6 .66 66.6 66.66-".
6666.6 6666.6 6666.6 666.6 6. .66 6m 6266......
.668
6. .66 .6666... 66.6 6.66 .6... 66.6. 6. 66.6 56..
mwsﬁ .6666... 6.566 6.66 .6... 66.6. 6.. b.6666 666..
:.6.%6......6- .6666... 66.686 66. 66.. 6.66 .66 66.686 6. 666..
. .68
2.66666 .6666... 66.686. 66... 66.. 6.6..
.668
6666.6. .6568 66.6865. 6. .66..
.668 .668 .668 .668 .668
6666.6 6666.6 .6666. 6666.6. .6666 66666 66.6 66..
.668 .668 .668 .668 .668 .6666 .666 \ .666
6666.6 6.66.6. 66.6.6 6.66.6 66.6.6 66.6 66.66 6.66.. + 6.666 6 66.66 6.6628 6 6.6.666
.668 .668 .668 .668 .668
1.66666. 2.66666. 2.66666. 56666.6. 3.66666. 6.56. 66.... 668.6 63.6668 666.666 6.662..
.668 .668 .668 .668 .668
. .6. ..6. 6666.6. 1.6666 2666.6. .6. 6. .6 6.6666 .666 6 66.666666 6666 668... 66.6660
.6. .8 .668 ..6.8 .6. .8 .668 .3666
6666.6 6.66.6 .3666. .6 6666.6 :666. .6 6 86> 66.66,. + 666 .66... \ .666 .666. 66826..
.668 .668 .668 .668 .668 .6666. 6.66.... 668.6
.6666 5666.6 6666.6 66. 6.6 .66666 6.66 666.666 666.6... 6 66.6.66 6665.6. 6.6.6.;
.668 .668 6.8 .668 6.8
6666.6 .6666. 66.6.6 .6666 66.6.6 666 6.666.
26.8 .668 .668 .668 .668
6666.6. 6666.6. 6.66.6. 6666.6 6. . .6. 6866.
.6: .6. .6. .6. .8 6666...? 66666666...

 

.. .3 6.5 666.6

 

.6268 6 6.66...

54

Table 10
OLS Regression of the Three-Day Cumulative Abnormal Bond Returns on Three-
Day Cumulative Abnormal Stock Returns

The sample consists of 254 bank loan announcements ﬁ'om 1995 to 2003. The sample
is compiled by searching F activa for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database.
Financial and utility companies are excluded from the sample. Firms for which bond
prices do not move at all during the event window (days -170 to +10 relative to the
bank loan announcement), and ﬁrms that release other important information with the
bank loan announcement are also excluded. Abnormal stock returns are computed
using the market model. The market model is estimated by using 150 trading days of
stock return data ending 20 days before the loan announcement. The CRSP value-
weighted return is used as the benchmark. Abnormal bond returns (ABR) are
computed using the following model: ABR” = BRit — (ai + '31 x Chg_Spreadt + ,6 2 x
Chg_Slopet), where Brit is the daily bond return of borrower i at time t, Chg_Spreadt is

the change in aggregate credit spread at time t, which is deﬁned as the change in the
Baa- over Aaa-rated bond yields, and Chg_Slopet is the change in the slope of the yield

curve at time t, which is deﬁned as the change in the difference between lO-year
Treasury over l-year Treasury yields. This model is estimated using 150 trading days
of bond return data ending 20 days before the loan announcement. When a borrowing
ﬁrm has multiple bonds outstanding, the ABR is estimated as the weighted average of
the bonds’ abnormal returns, where the weights are assigned according to the face
value of each bond. The daily abnormal stock (bond) returns are accumulated to obtain
three-day cumulative abnormal return (CAR) from day -1 to day +1 relative to the
announcement date. Low security bonds are those for which the security level is either
“junior subordinate,” “junior or subordinate,” “senior subordinate,” or “not speciﬁed.”
Numbers in parentheses are p-values. ***, **, and * denote signiﬁcance of the

parameter estimates at the 1%, 5%, and 10% levels, respectively.

55

Table 10

 

 

Independent variables (1) (2) (3) (4)
Three-day cumulative abnormal stock returns: 0.2678**"' 0.3926*** 0.2911*** 0.1749"
(a) (0.00) (0.00) (0.00) (0.04)
Operating income before depreciation / Total -0.0081
assets is above the median (dummy): (b) (0.39)
(a) x (b) -0.3252***
(0.00)
Previous one-year cumulative excess stock -0.0133
return is above the median (dummy): (c) (0.16)
(a) x (C) -o.2703*
(0.06)
Tobin’s q ((Market value of equity + Book 0.0156
value of total debt) / Total assets) is above the (0.11)
median (dummy): (d)
(a) x ((1) 0.1343
(0.19)
Operating income before depreciation / Total
assets is below the median / Tobin’s q is
above median (dummy): (e)
(a) X (6)
Previous one-year cumulative excess stock
return is below the median / Tobin’s q is
above the median (dummy): (0
(a) x (0
Loan is secured loan and bond has low
security (dummy): (g)
(a) x (g)
Log of (Market value of equity + Book value -0.0045 -0.0037 -0.0053"' 00063"
of debt) (0.14) (0.23) (0.09) (0.05)
Intercept 0.0434“ 0.0389 0.0562" 0.0510"
(0.09) (0.13) (0.04) (0.05)
Adjusted R2 0.1127 0.1499 0.1260 0.1229
F-value 16.94 12.02 10.05 9.80
No. of observations 252 251 252 252

 

56

Table 10 (cont’d)

 

 

Independent variables (5) (6) (7)
Three-day cumulative abnormal stock returns: 0.1082" 0.1517* 0.2589***
(a) (0.05) (0.05) (0.00)
Operating income before depreciation / Total
assets is above the median (dummy): (b)
(a) x (b)
Previous one-year cumulative excess stock
return is above the median (dummy): (c)
(a) x (C)
Tobin’s q ((Market value of equity + Book
value of total debt) / Total assets) is above the
median (dummy): ((1)
(a) X (d)
Operating income before depreciation / Total 0.0271 **
assets is below the median / Tobin’s q is above (0.03)
median (dummy): (e)
(a) x (e) O.4865***

(0.00)
Previous one-year cumulative excess stock 0.0238“
return is below the median / Tobin’s q is (0.03)
above the median (dummy): (0
(a) x (t) 01638“

(0.10)
Loan is secured loan and bond has low -0.0170
security (dummy): (g) (0.31)
(a) x (g) -0.0010
(1.00)

Log of (Market value of equity + Book value -0.0060"”" 00068" -0.0076
of debt) (0.04) (0.03) (0.13)
Intercept 0.0506" 0.0571 ** 0.0736"

(0.04) (0.03) (0.08)
Adjusted R2 0.2179 0.1349 0.0985
F-value 18.42 10.79 6.16
No. of observations 251 252 190

 

57

Figure 1
Cumulative abnormal returns from day -10 to day +10 around the loan
announcement.

The sample consists of 254 bank loan announcements from 1995 to 2003. The sample
is compiled by searching Factiva for loan announcements. To be included in the
sample, borrowing ﬁrms must have daily bond return data in DATASTREAM, daily
stock return data in CRSP, ﬁnancial data in COMPUSTAT, and bond characteristic
data, such as rating and bond security, in the Fixed Income Security Database. Financial
and utility companies are excluded from the sample. Firms for which bond prices do
not move at all during the event window (days —170 to +10 relative to the bank loan
announcement) and ﬁrms that release other important information with the bank loan
announcement are also excluded. Abnormal stock returns are computed using the
market model. The market model is estimated by using 150 trading days of stock return
data ending 20 days before the loan announcement. The CRSP value-weighted return is
used as the benchmark. Abnormal bond returns (ABR) are computed using the
following model: ABRit = BRit — (ai + [9 1 x Chg_Spreadt + ,82 x Chg_Slopet), where
Brit is the daily bond return of borrower i at time t, Chg_Spreadt is the change in

aggregate credit spread at time t, which is deﬁned as the change in the Baa- over Aaa-
rated bond yields, and Chg_SIopet is the change in the slope of the yield curve at time t,

which is deﬁned as the change in the difference between 10-year Treasury over l-year
Treasury yields. This model is estimated using 150 trading days of bond return data
ending 20 days before the loan announcement. When a borrowing ﬁrm has multiple
bonds outstanding, the ABR is estimated as the weighted average of the bonds’
abnormal returns, where the weights are assigned according to the face value of each
bond. The daily abnormal stock (bond) returns are accumulated to obtain cumulative
abnormal return (CAR) from day -t to day +t relative to the announcement date.

58

 

Eo>m 250:6 >60

 

 

 

o._.-

 

 

 

£639.62. 65 “2... x08» tea 2.2. «5
.8315: 2:36.. .2533 o>_uu_:E:o

(%) uvo

 

 

. 6.66....

S9

BIBLIOGRAPHY

60

REFERENCES

Best, Ronald, and Hang Zhang, 1993, Alternative information sources and the
information content of bank loans, Journal of Finance 4, 1507-1523.

Billett, Mathew T., Mark J. Flannery, and Jon A. Garﬁnkel, 1995, The effect of lender
identity on borrowing ﬁrm’s equity return, Journal of Finance 50, 699-718.

Billett, Matthew T., Tao-Hsien Dolly King, and David C. Mauer, 2004, Bondholder
wealth effects in mergers and acquisitions: New evidence from the 19803 and
19903, Journal of Finance 59, 107-135.

Campbell, Tim S., and William A. Kracaw, 1980, Information production, market

signaling, and the theory of ﬁnancial intermediation, Journal of Finance 35, 863-
882.

Chen, Long, David Lesmond, and Jason Wei, 2007, Corporate yield spreads and bond
liquidity, Journal of Finance 62, 119-149.

Diamond, Douglas W., 1984, Financial intermediation and delegated monitoring, Review
of Economic Studies 51, 393-414.

Dimson, Elroy, 1979, Risk measurement when shares are subject to infrequent trading,
Journal of F manczal Economics 7,197-226.

F ama, Eugene F. 1985, What’s different about banks? Journal of Monetary Economics
10, 10-19.

Fama, Eugene F., and Kenneth R. French, 1993, Common risk factors in the returns on
stocks and bonds, Journal of Financial Economics 33, 3-56.

James, Christopher, 1987, Some evidence of the uniqueness of bank loans, Journal of
Financial Economics 19, 217-235.

Kang, Jun-Koo, and Wei-Lin Liu, 2007, Bank incentives and suboptimal lending
decisions: Evidence from the valuation effect of bank loan announcements in Japan,
Journal of Banking and Finance, forthcoming.

Lummer, Scott, and John McConnell, 1989, Further evidence on the bank lending process
and the capital market response to bank loan agreements, Journal of Finance
Economics 25, 99-122.

Masulis, Ronald W., 1988, The debt-equity choice, Institutional Investor Series in
Finance (Ballinger Press).

61

Maxwell, William F., and Clifford P. Stephens, 2003, The wealth effects of repurchases
on bondholders, Journal of Finance 58, 895—921.

Maxwell, William F ., and Ramesh P. Rao, 2003, Do spin-offs expropriate wealth from
bondholders? Journal of Finance 58, 2087-2108.

Mikkelson, Wayne, and Megan Partch, 1986, Valuation effects of securities offerings and
the issuance process, Journal of Financial Economics 15, 31-60.

Modigliani, Franco and Merton H. Miller, 1963, Corporate income taxes and the cost of
capital: A correction. American Economic Review 53, 433-43.

Ongena, Steven, Viorel Roscovan, and Bas J .M. Werker, 2007, Banks and bonds: The
impact of bank loan announcements on bond and equity prices, Working paper,
Tilburg University.

Peek, Joe and Eric S. Rosengren, 2005, Unnatural selection: Perverse incentives and the
misallocation of credit in Japan, American Economics Review 95, 1144-1166.

Preece, Dianna C., and Donald J. Mullineaux, 1994, Monitoring by ﬁnancial

intermediaries: Banks vs. nonbanks, Journal of Financial Service Research 4, 191-
200.

Scholes, Myron and Joseph Williams, 1977, Estimating betas from nonsynchronous data,
Journal of Financial Economics 5, 309-327.

Smith, Clifford W., Jr., 1986, Investment banking the capital acquisition process, Journal
of Financial Economics 15, 3-29.

62

I11

11

1111ij

3062 9

H
H
IN.|

.1

IIHHHHIHHLIIQJIIW

31