. ylh!‘

,

 

w_

,<.- c

«1vv—n

y-..-

.. -hwu

A4-.. uv-v-v

.a
u.
.o- —— .
2...; A.
m——- v
.
a

-I
‘ A «‘r‘ ._.
. .. 3. «as:
a . ké'i‘ffi 3423:“:t} ,- .
5‘ u . g. g.._ ._ ' . -
3 21 .m ‘

 

fizliiii":

t.

w".

”u...-

wm. A
_..,.~—- ...- n

4»
.9:-

if
’1
A}
‘ag

.4
S
l
L

"F34

I
yu‘

~ \

 

.«4

d .«E. v .v.‘
~.._‘
"'3 "‘3”?
A*¢~

_ ,v .
l. \
”Egan "I“

":9 1

Lu, .

.rgz ;

$
u,’
:‘i.
‘1':

.«Le

:.;. 4: .
-‘ , . .
-,.,
*‘...,_:'z':a.
M

5'
.'.'_
”54* 3

2

. i a: fiflzfl
1.11; 3):]? ‘5 M; ’i’
~ 1" .l 1‘.) LI

:fiL‘m‘ 11

.‘x‘

.44

t"

A”:
{‘1‘
: 1m}:
7‘“.
39‘ I}

V.-- - .
rw- .
—.—~v

. \- Z. I :13: ‘_ 311% '

. - * _ ,J }7

v ‘. H . ‘ l A,

"l, 7 . :‘Q‘JQbU‘ _. «9% ..
v“? x‘. \ 33.1.},2‘ , j, . 4., .1.

. ‘;-v:,;
"‘27:?!
vr'”
:‘N—
.'
._..-- , _
_..I- . A

‘“‘r_h I - '
hwltgzusgzig. . , ‘. ‘ .
"*ifis';§":._ee,l .221. ‘H ~‘~‘-"“’%.‘
, K

.53?“
.. 3.2:

.p "—y
u...-

‘w u

v

Ar”:
W-

m

_-.,~
w
.....
.0...
-
'1.»

,-
I
.-

- if h .
$ in L1:
’ I‘IIJ "I (h

e-

t

A

l'ffiJ'.‘
u)

 

.-
o
‘ 0 V. -n..-. ' .

#1.-
..‘_.

. .
5....-

 

TfiL-‘LBIS

199‘)

 

1 ImmulIlil’lllllllllllllllflumull

3 1293 01812 7252

This is to certify that the

dissertation entitled

TWO ESSAYS ON REPEALINC GLASS-STEAGALL ACT:
REGULATION DISTORTION AND BANKS' ENTRY
IN SECURITY UNDERWRITING

presented by

Wei-ling Song

has been accepted towards fulfillment
of the requirements for

Ph.D. Finance

degree in

 

 

Major professor

/

Date 5707/9?

MS U is an Affirmative Action/Equal Opportunity Institution 042771

 

 

LIBRARY

Michigan State

University

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1]” WM.“

TWO ESSAYS ON REPEALING GLASS-STEAGALL ACT: REGULATION
DISTORTION AND BANKS' ENTRY IN SECURITY UNDERWRITING

By

Wei-ling Song

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Finance

1999

 

TMsc
ACIc
secur
unde
exan-
SSCU'
D€icr
bank
regul
recei
IeQU
SWum
9035
SepV
havG
prov

iSCx

ABSTRACT

TWO ESSAYS ON REPEALING GLASS-STEAGALL ACT: REGULATION
DISTORTION AND BANKS' ENTRY IN SECURITY UNDERWRITING

By

Wei-ling Song

This dissertation investigates the economic impact of repealing Glass-Steagall
Act of 1933 on financial industry. The Act prohibits commercial banks from
security underwritings. In 1987, commercial banks are allowed to engage in
underwriting and dealing certain securities on a limited basis. The first essay
examines three questions regarding commercial banks’ involvement in the
security underwriting business in the post-section 20 era: (i) how does the ten
percent revenue limitation distort the underwriting activities of commercial
banks?, (ii) how different commercial banks are from investment banks under the
regulatory and market conditions?, and (iii) which types of issuing firms do
receive benefits from the bank-entry? We find evidence consistent with the
regulatory distortion explanation that commercial banks are actively engaging in
small issue market. Compared to investment banks, commercial banks do not
possess better ability in underwriting for small issuers. In fact, they are better in
serving large clients but they can not handle large issue size. Commercial banks
have better informational advantage. The association with commercial banks
provides value to issuing firms. However, the evidence supports that the market

is concerned for conflicts of interest and informational monopoly power of

commercial banks. Due to the pros and cons of commercial banks, only issuers
with middle credit ratings receive significant benefits from bank entry. Whereas,
small issuers with high credit ratings and large issuers with low credit ratings are
suffered from the banks' expansion at the later entry stage.

In the second essay, we examine empirically the economic rationales of
coalitions of underwriters and the role of commercial banks as co-managers
when they expand into the security underwriting business. We find evidence that
commercial banks strategically select their mode of entry, either by leading the
syndicates or by participating in syndicates led by an investment bank (the hybrid
form), depending on their underwriting strength in different markets. By
cooperating with a reputable investment bank, commercial banks can effectively
reduce their conflicts of interest problem while preserving their informational
advantages. The creation of a better organizational form in conducting security
underwriting provides economic incentives for investment banks to accommodate
the entry of commercial banks. The distinct functions offered by commercial
banks facilitate their penetration into the security underwriting industry. In most
cases, the hybrid structure is superior to both the commercial bank lead
syndicate and to the pure investment bank arrangement. This paper highlights
shortcomings of previous work in this area that focus only on issues that are
commercial bank lead. By neglecting other organizational forms, such as hybrid
syndicates, previous research understates the benefits that commercial banks

bring into the security underwriting business.

Copyright by
Wei-ling Song
1999

To my parents, daughter and son.

com
Dr. C'
espec

comm

COHUHI
for offe

Wlsdon

 

ClCl‘lellt
deeply
Provide

|
A SPEC:
Care Of
famlly f
gramme

and Sat

ACKNOWLEDGMENTS

I would like to express my appreciation to the members of my dissertation
committee - Dr. Anil Shivdasani, Dr. Assem Safieddine, Dr. Naveen Khanna, and
Dr. Charles Hadlock, for their insightful guidance and encouragement. I would
especially like to thank Dr. Anil Shivdasani for all his help, patience, and
commitment of time.

I thank Dr. Richard Simonds for providing me with the opportunity to
continue my doctoral studies. I also thank Dr. Kirt Butler and Dr. Geoffrey Booth
for offering me a position as an instructor. I also appreciate the knowledge and
wisdom passed on to me by all the faculty members in the Finance Department.

I thank Dr. Miranda Lam Detzler, Aydin Yuksel, Ramana Sonti, Mike
Cichello for helping me in one way or another throughout the finance program. I
deeply appreciate my Taiwanese friends at Michigan State University who have
provided constant encouragement throughout my stay in Michigan.

I would like to thank all my family for their support, especially my parents.
A special thank to my sister, Wei-ching, who gives me peace of mind by taking
care of my parents and my daughter. I would also like to thank Mrs. Tai and her
family for providing such a loving environment for my son. Finally, my deepest
gratitude goes to my daughter, Yihua, and my son, Jinha, for their understanding

and sacrifices necessary for my completion of the doctoral program.

vi

LIST
ES

F

lni

TABLE OF CONTENTS

LIST OF TABLES ......................................................................... ix

ESSAY 1. BANK-ENTRY BENEFITS AND REGULATION DISTORTION IN

THE CORPORATE SECRURITY UNDERWRITING BUSINESS

1.1 Introduction ........................................................................... 1
1.2 Testable Hypotheses, Empirical Proxies and Prediction of
Underwriter choice ................................................................. 10
1.2.1 The Information Advantage Hypothesis ............................. 11
1.2.2 The Conflicts of Interest Hypothesis .................................. 12
1.2.3 The Bank Relationship Hypothesis ................................... 14
1.2.4 The Bank Monopoly Power Hypothesis ............................. 15
1.3 The Switching Regression Model: Motivation and Description ......... 16
1.3.1 Model Specification ....................................................... 16
1.3.2 Estimation and Interpretation of Regression Estimates ......... 19
1.4 Data and Descriptive Statistics ................................................. 24
1.5 Regulatory Distortion and Comparative Advantages of
Underwriters ......................................................................... 29
1.5.1 Regulatory Distortion of Commercial Bank Underwritings ....... 31
1.5.2 Advantages of Commercial Banks as Underwriters ............... 38
1.5.3 Disadvantages of Commercial Banks as Underwriters .......... 40
1.5.4 The Performance Comparison between Commercial
Banks and Investment Banks ................................................... 42
1.6 Conclusions ......................................................................... 45
APPENDIX 1 TABLES OF ESSAY1 ................................................. 47

ESSAY L COMPETITION AliD COAEION AMONG UNIERWRITERS:

2.1

2.2

2.3

 

THE ENTRY STRATEGY OF COMMERCIAL BANKS IN THE
POST-SECTION 29 ERA

Introduction .......................................................................... 63
The Literature on Bank Entry Strategies and Underwriting

Coalitions ............................................................................. 72
Data and Summary Statistics ................................................... 79

vii

 

24

25

26
APPE
APPE
APPE

MBLM

 

2.4 The Entry Strategies of Commercial Banks ................................. 83

2.5 The Characteristics of Hybrid Syndicates .................................... 88
2.5.1 The Determinants of Syndicate Selections ......................... 89
2.5.2 The Underwriting Abilities of Syndicates ............................ 91
2.5.3 The Performance of Syndicates ....................................... 95
2.6 Conclusion ........................................................................... 99
APPENDIX 2 TABLES OF ESSAY 2 ............................................... 101
APPENDIX A. DATA DEFINITIONS ................................................. 119
APPENDIX B. TABLES ................................................................. 123
BIBLIOGRAPHY ........................................................................... 132

viii

Tabl

Tabl

Table

Table

Table

Table ‘

 

Table 1

Table 1

Table 1

Table 2
Table 2

Table 2

Table 1.1

Table 1 .2

Table 1.3

Table 1.4

Table 1.5

Table 1.6

Table 1.7

Table 1.8

Table 1.9

Table 2.1

Table 2.2

Table 2.3

LIST OF TABLES

Summary of theoretical predictions regarding choice of
commercial bank (vs. investment bank) ............................ 48

Descriptive statistics of bond issue and firm characteristics
by year and by underwriter-type ...................................... 50

Descriptive statistics of bond issue by issue size market
and by underwriter-type ................................................ 52

Issuer credit ratings distribution and basis point spreads
by underwriter-type ...................................................... 53

Estimates of the determinants of underwriter selections
and basis point spreads on bonds underwritten by the
commercial banks or by investment banks ........................ 54

Small issue market: Estimates of the determinants of
underwriter selections and basis point spreads on bonds
underwritten by the commercial banks or by

investment banks ......................................................... 56

Large issue market: Estimates of the determinants of
underwriter selections and basis point spreads on bonds
underwritten by the commercial banks or by

investment banks ......................................................... 58

Gross benefit for firm using the service of chosen undenlvriter
rather than that of unchosen underwriter. ......................... 60

Gross benefit for firm using the service of chosen undenNriter
rather than that of unchosen underwriter for small issue
market and large issue market by subperiods. ................... 61

The frequency and percentage of bond issues by year
and by underwriter type. ................................................ 102

Descriptive statistics of bond issue and firm characteristics
by year and by underwriter type ....................................... 104

Issuer credit ratings distribution and basis point spreads
by undenlvriter type ....................................................... 106

 

 

Tabb

 

Tabb

Tabb
Tabb

Tame.

Appenq

Append

 

Append:

Appendi

Appendj;

’Table 2.4

Table 2.5

Table 2.6

Table 2.7

Table 2.8

LIST OF TABLES (CONT'D)
The percentage of bond issues by underwriter-type
in the large issue and small issue markets ......................... 108
Estimates of the determinants of syndicate selections .......... 110

Estimates of basis point spreads on bonds underwritten
by CB-Lead and Hybrid syndicates. ................................. 112

Estimates of basis point spreads on bonds underwritten
by PurelB syndicate ...................................................... 114

Gross benefit for issuing firms by credit ratings using
the service of chosen syndicate rather than that of
unchosen syndicate ....................................................... 116

Appendix B1 Estimates of basis point spreads on bonds underwritten

by the commercial banks or by investment banks
with alternative model specifications ................................. 124

Appendix 82 Small issue market: Estimates of basis point spreads on

bonds undenivritten by the commercial banks or by
investment banks with alternative model specifications ......... 126

Appendix B3 Gross benefit for firm using the service of chosen

underwriter rather than that of unchosen underwriter:
Estimated from alternative model specification ................... 128

Appendix B4 Estimates of additional basis point spreads regressions

on bonds underwritten by hybrid syndicate with
endogenous selection adjustment term ............................. 129

Appendix BS Additional estimation of gross benefit for issuing firms

by credit ratings using the service of chosen syndicate
rather than that of unchosen syndicate .............................. 131

 

 

 

 

 

£55

11
R

financi

 

to rape
prohibl"
Ofthis;
inveSUr

Th
banks:

Subsidi.

ESSAY 1. BANK-ENTRY BENEFITS AND REGULATION DISTORTION IN

THE CORPORATE SECRURITY UNDERWRITING BUSINESS

1.1 Introduction

Recently, Congress and regulators have undertaken measures to reform the
financial services industry. These efforts include, among other things, a proposal
to repeal the Glass-Steagall Act‘. The Glass-Steagall Act of 1933 effectively
prohibited commercial banks from underwriting corporate securities. As a result
of this act, all US. corporate security issues could be underwritten only by
investment banks for nearly six decades.

The relaxation of the Glass-Steagall Act since 1989 has allowed commercial
banks to engage in these investment banking activities through their Section 20

subsidiaries under firewall restrictions and a 10 percent revenue limitationz. The

 

‘ See, for instance, the bill (HR 1062) proposed by Representative Leach, Chairman of
the House Banking Committee, on February 27, 1995. In general, HR. 1062, the
Financial Services Competitiveness Act of 1995, is intended to permit affiliations
between full-service depository institutions and full-service securities companies
(Congressional Reports, 1995—1996, House Reports 104-127).

See JP Morgan & Co. Inc., The Chase Manhattan Corp., Bankers Trust New York
Corp., Citicorp, and Security Pacific Corp., 75 Federal Reserve Bulletin 192 (1989)
(hereafter, 1989 Order) for a discussion of Section-20 subsidiary, firewall and revenue
limitation. Basically, Section 20 of the Glass-Steagall Act states that a member bank of
the Federal Reserve System may not be affiliated with a company that is “engaged
principally" in underwriting and dealing in securities. The Board has established a
revenue test to determine whether a company is “engaged principally” in underwriting
and dealing for purposes of section 20. The revenue test provides that a section 20
subsidiary may not derive more than 10 percent of its total revenue from underwriting
and dealing in bank-ineligible securities. Bank-ineligible securities refer to the ones
commercial banks are allowed to deal and underwrite after the change of regulation.
Firewalls are a set of restrictions that limit the financial and information flows between
the securities and commercial bank subsidiaries in an effort to insulate potential conflicts
of interest that commercial banks may pass off bad loans to the market and use the
proceeds to pay off bank debt.

10 percent revenue limitation effectively eliminates the possibility that commercial
banks may enter the securities business by acquiring an existing investment
bank. Following the increase of the revenue limitation to 25 percent at the end of
1996‘, there were some mergers between commercial banks and investment
banks, such as Bankers Trust's merger with Alex, Brown, Inc. and Citicorp's
merger with Travelers.

In contrast to investment banks who do not have any revenue constraint,
commercial banks may be limited by this revenue restriction in the amount of
issues they can underwrite. This restriction might also cause them to deviate
from the clientele they can serve the best. Whether the revenue limitation has
material effects on the behavior of commercial banks has yet to be investigated.
Additionally, it is unclear if issuing firms are suffering from possible regulation
distortion. Besides the observed distortion of bank entry modes, we analyze in
this paper the impact of revenue restriction on the underwriting ability of
commercial banks and their client firms.

Gande et al (1997) studied commercial bank underwriting activities since
1989, and found that the issue size of commercial bank clients decreased during
the period 1993-95. The authors argue that commercial banks are better in
bringing smaller firms into the public debt market. However, they do not examine

the firm size explicitly in the their study. Although it is true that the issues of small

 

3 See Federal Reserve System Docket No. R-0841, “Revenue Limit on Bank-Ineligible
Activities of Subsidiaries of Bank Holding Companies Engaged in Underwriting and
Dealing in Securities."

firms tend to be smaller“, the matching of issuer and underwriter based on these
two dimensions of size depends on the different abilities of underwriters. The firm
size dimension is associated with the certification ability of undenrvriter, whereas
issue size is related to the distribution ability of unden/vriter. These two abilities
may not have a perfect correlation with each other, like the positive relationship
between issue size and firm size. In addition, the 10 percent revenue limitation
imposed only on commercial banks may hinder their ability in handling large
issue size, even though they may be better in serving large firms.

Using a sample of 1327 corporate bond issues from 1991 to 1996, we
formally examine the abilities of commercial banks and investment banks in
assisting issuers with different issue sizes and firm sizes. We test the hypothesis
that commercial banks actively engage in the small issue market because of
regulatory distortions rather than because of commercial banks' superior ability in
underwriting for small firms.

Since commercial banks are new entrants in the security underwriting
business, one might argue that they are at a disadvantage with respect to
distribution capabilities, because it takes time to establish distribution networks. If
this is true, we should observe commercial banks improve their distributional
ability over time. So, we examine the performance of commercial banks in
serving their clients across different time periods. We also segment the total
underwriting market by the median issue size to further investigate the

differences of commercial bank underwriting between large and small issue

 

‘ The correlation coefficient between these firm size and issue size is 0.34.

 

 

 

 

 

mark

pron:

E

banks

dbad\

 

assoc
rnonn
hflonn;
provide
Howev
banksi
locapu
Rwan(
[Ani
thhis

1933.I(

markets. The regulatory distortion from the revenue constraint should be more
pronounced in the large issue market than the small issue market.

Besides regulatory differences between commercial banks and investment
banks, there have been extensive debates regarding the advantages and
disadvantages of banks as lenders and underwriters. The advantages of bank
association have been the subject of much theoretical and empirical works. The
monitoring activity of a bank gives commercial banks a role as credible
information certifiers (Fama, 1985, Diamond, 1991). Bank relationships may also
provide insurance against credit rationing when times are tough (Holland, 1994).
However, the informational role of commercial banks comes with a cost, since
banks can acquire monopoly power over clients' information, allowing the banks
to capture a greater share of firm profits. Issues of this sort are discussed in
Rajan (1992), Sharpe (1990) and Greenbaum et al (1989).

Another disadvantage of commercial banks is a potential conflict of interest,
which is the main concern prompting the passage of the Glass-Steagall Act in
1933. Kanatas and Qi (1998) study the cost of conflicts of interest imposed on
commercial bank clients when the investors suspect that there is a potential for
the commercial banks to misrepresent the quality of their customers and to pass
off bad loans to the market, enabling the firm to use the proceeds to repay bank

debt. Despite this cost, firms may choose to use the services of commercial

 

5 See James (1987) and Billett et al (1995) for empirical evidence on the value of bank
association.

 

 

 

DBDK

 

unde

differs
banks
wesh
econo
speda

These

monop

 

VVI
dlfieren
eSIlmat

basedc

banks if there are sufficiently large scope economies in combined lending and
underwritings.

The existing empirical literature7 does not provide convincing evidence that
different types of underwriters have comparative advantages. lf commercial
banks as a group possess different abilities than those of investment banks, then
we should observe a corresponding specialization in different segments of the
economy. By using appropriate proxy variables, we further separate the test of
specialization and comparative advantages of underwriters into finer hypotheses.
These hypotheses include informational and relationship advantages, information
monopoly power, and conflicts of interest of commercial banks.

We test if there is a systematic selection of underwriters. We compare the
differences between commercial banks and investment banks explicitly by
estimating their bond pricing (net yield) equations and discuss these differences
based on the above finer hypotheses. The regression coefficient differential on
each firm or issue characteristic reveals how these two underwriters are different
from each other on this particular dimension of issuers. Given that there is a
systematic selection, this potential selection bias needs to be controlled for when

the bond pricing equations are estimated.

 

6 For more detailed discussions of the arguments for and against the repeal of Glass-
Steagall Act, see 1989 Order and Citicorp, JP. Morgan & Co., and Bankers Trust New
York Corp., 73 Federal Reserve Bulletin 473 (1987).

7 See Ang and Richardson (1994), Puri (1994), Puri (1996), Hamao and Hoshi (1997),
Gande et al (1997), and Kroszner and Rajan (1997) for more discussions of empirical
results. Most of these studies support the view that there is a net certification effect of
commercial banks. Only the study by Kroszner and Rajan (1997) showed there were
concerns for conflicts of interest of commercial banks in the market.

 

 

 

 

 

regma
chmcl
and o
bawe

bond r

 

underv
bond p
underv
clients
theiss
yield. l
previo
Unden

F
Under

That d

This leads to a natural application of the empirical model of switching
regressions with endogenous switching". The model consists of an undenNriter
choice equation, and two bond yield equations, one for commercial bank clients
and one for investment bank clients. Thus, we allow complete interactions
between issuer characteristics and underwriter identity when we estimate the
bond pricing equations to contrast the differences between these two types of
undenNriters. In addition to disentangling the forces underlying the formation of
bond prices, this model also allows us to compare the performance of
undenivriters by quantifying the benefits that different undenNriters create for their
clients. The benefit is defined as the expected mean net yield of the issuer had
the issue been underwritten by the unselected undenivriter, less the actual net
yield. Instead of comparing the net yields of different regimes of issuers as in
previous studies, we are able to compare a firm’s net yield given the selected
underwriter, to the net yield had the client chosen the unselected undenNriter.

Firms with different characteristics may have a variety of different needs for
undenlvriting services. These needs should be matched with the varied abilities
that different underwriters possess. Thus, the entry of commercial banks could
be beneficial to certain types of firms. On the other hand, the existence of
switching costs (Nanda and Warther, 1998) may prevent commercial bank clients
from using services of investment banks. Therefore, we investigate the economic
role that commercial banks play in the security underwriting business and

analyze what types of firms are better off or worse off due to the bank entry.

 

8 See Maddala (1983) for an explanation of this model.

4 .,-rm‘d arm‘-

 

 

Tobm’S q I

sma
seg
The
bank
Funhi
the fir
bank(
increa
banks
the sar
choose
inability
clients.
perform
later ent
T
Choice 0
enufifica

trade leSs

 

It appears from our results that commercial banks actively participate in the
small issue market because of regulatory constraints, rather than being in this
segment voluntarily due to any comparative advantage in assisting small issuers.
The results of the overall sample show that the underwriting ability of commercial
banks is better than that of investment banks along the firm size dimension.
Furthermore, in the large-issue market, the differences between underwriters in
the firm size dimension are more pronounced. However, these large commercial
bank clients incur huge price discounts when the size of their bond issues
increases; this is not true for investment bank clients. The ability of investment
banks to serve different size of customers, regardless of firm size or issue size, is
the same in both small and large issue markets. If commercial banks could
choose, they would have underwritten a greater number of larger clients, but their
inability to handle large issue sizes prohibits them from underwriting for large
clients. This regulatory distortion hypothesis is further supported by the
performance of commercial banks, which worsen in the large issue market at the
later entry stage.

This paper also establishes empirically that the forces underlying the
choice of undenrvriter for security issuance suggested by previous theoretical and
empirical work do exist. The informational advantage that commercial banks
have facilitates their underwriting firms that are less risky (lower leverage), that
trade less frequently, are rated lower and that have higher Tobin’s q. Since
Tobin’s q may be a proxy for investment opportunities, the commercial bank

association with high Tobin’s q clients also supports the notion that a commercial

 

bank
relati

that t

mfled
abHfiy
(finanr
increa:
investr
comme
resuhs
informa
by the s
are Sll’or

i"formati

 

bank relationship is beneficial to firms with higher financing needs. The positive
relation between lower rated issuers and commercial banks supports the views
that bank monitoring and certification are more valuable to firms in this segment.

Besides their observed specialization in underwriting for less risky, lower
rated firms with higher informational problems, commercial banks have a better
ability than investment banks in undenivriting along the dimensions of leverage
(financial risk) and Tobin's q. When the leverage and Tobin's q of client firms
increase, commercial bank underwritten bonds have lower net yields than
investment bank underwritten bonds. Further, we present evidence to show that
commercial banks can cherry pick9 from among available client firms. These
results are consistent with the hypothesis that commercial banks have
informational and relationship-based advantages. When we separate the sample
by the size of issue, the informational advantage and bank relationship effects
are stronger in the small issue market. Not surprisingly, in the large issue market,
the benefits of bank association decrease because these issuers have less of an
informational problem and better access to the capital markets.

We find that the concerns for conflicts of interest rise when firms having high
interest expense relative to operating income are associated with commercial
banks, especially when the purpose of issue is to repay bank debt. We do not
find the intent to refinance bank debt by itself a proxy for potential conflicts of

interest. In fact, when the use of proceeds is for refinancing bank debt and the

 

9 Cherry picking refers to the ability of commercial banks to select "higher quality" firms
based on their private information about clients. See Section 1.3 for a detailed
explanation of how the endogenous selection of firms ties into the cherry picking ability

 

 

 

 

 

 

 

 

than:

 

banks
mame
range
with ve
entry ii
that the
andinh
rated 5,
“IOCked.

T

 

 

hl’pothe
dhCusse
DIESents
regulatm

im'leSImE

UIIGErwrm

level of interest expense is low, there is no price discount of the bond regardless
of underwriter. We also document that the market is concerned about the
informational monopoly problem with commercial banks, especially for firms with
high stock return volatility, high profitability and higher past sales growth.

Although it appears that the benefits created by investment banks are larger
than those created by commercial banks, this does not mean that investment
banks are better underwriters than commercial banks across all segments of the
market. Consistent with Diamond (1991), firms with credit ratings in a middle
range receive the most benefits from commercial bank relations. However, firms
with very high and very low credit ratings do not benefit from commercial bank
entry in general. The results are also consistent with the theory in Rajan (1992) in
that there is a trade-off in developing relationship-specific capital between firms
and intermediaries. Negative benefits created by commercial banks for the best
rated small issuers implies that these smaller, better firms are the clients being
“locked-in”.

The rest of the paper is organized as follows. In Section 1.2, the testable
hypotheses, empirical proxies and prediction of undewvriter choice are
discussed. Section 1.3 motivates and describes our empirical model. Section 1.4
presents the data and descriptive statistics. Section 1.5 reports the effects of
regulatory distortion and the comparative advantages of commercial banks and

investment banks. Section 1.6 concludes.

 

underwriters.

cl" '-

‘Jt'iit-K‘L IO‘

 

 

 

Th
studie
1994:
subsid
and R;
befler
vauhi
find ev
Undem
inform;
SUDDOr

COUCen

1.2 Testable Hypotheses, Empirical Proxies and Prediction of

Underwriter Choice

The role of commercial banks as both lenders and underwriters has been
studied using data both prior to the Glass-Steagall Act (Ang and Richardson,
1994; Puri, 1994; Puri, 1996; Kroszner and Rajan, 1997) and the post-section-20
subsidiary era (Gande et al, 1997). The results are mixed. Except for Kroszner
and Rajan (1997), all the studies support the view that commercial banks are
better certifiers of public debt issues because of informational advantages
acquired from lending to issuing firms. Taken as a whole, these studies do not
find evidence of conflicts of interest that lead commercial banks to systematically
underwrite bad loans. Contrary to these findings, by controlling for the
informational advantages of banks, Kroszner and Rajan (1997) report results that
support the argument that when lending and underwriting are closely combined,
concerns for conflicts of interest do exist.

Nonetheless, these studies do not address the issue of how firms select
between different underwriters and do not control for the endogenous selection
problem when the net yield equations are estimated. The comparative
advantages of underwriters have never been investigated using characteristics of
issuing firms. By using appropriate proxy variables, we are able to disentangle
different forces that have material effects on bond pricing.

In the following subsections, we discuss the hypotheses related to the

comparative advantages of underwriters; the firm and issue characteristics can

10

 

fl...

 

be us
Theo
sumn
advar

banks
1.2.1

Tt
banks.
Thbinl
mmes.
with mg
banks a
with Iarg
DrodUCe
plobabili
grecliter g
In or
ones With
firm with c
c:Ommerclg

be used to test these hypotheses along with the prediction of underwriter choice.
Theoretical predictions of undenNriter choice from each hypothesis are
summarized in Table 1.1. These hypotheses include information and relationship
advantage, conflicts of interest, and information monopoly power of commercial

banks.

1.2.1 The Information Advantage Hypothesis

Through periodic monitoring and ongoing deposit histories, commercial
banks can acquire private information about their clients in a cost-effective way.
This information advantage gives banks a unique role in serving issuers with
more serious information problems (Fama, 1985; James, 1987). Therefore, firms
with higher information asymmetry receive greater benefits by using commercial
banks as their underwriters. The information problem is more severe for firms
with larger growth opportunities. Commercial banks as low cost information
producers are more beneficial for smaller firms than larger ones. Therefore the
probability of choosing a commercial bank is higher if the firm is smaller, has
greater growth opportunities, and has a lower credit rating.

ln Diamond (1991), bank monitoring provides value to firms, especially to the
ones with credit ratings in the middle range. Therefore, we should observe that
firm with credit ratings in the middle range tend to be associated more with
commercial banks than with investment banks. Both Titman and Trueman (1986)

and Chemmanur and F ulghieri (1994) model the effects of undewvriters with

11

"s ’l "I“. l 5 m1-

 

 

 

inforr

therr

 

tends
inforn'
procee
inform
Mgher
ween
Opporti
mamet
finnsfiz

depend

 

12;?

information advantages. They conclude that the higher the quality of underwriter,
the more effective it is in reducing the impact of information asymmetry and it
tends to reject a firm with a risky project. Consequently, underwriters with better
information undenivrite less risky client firms and obtain a higher price (net
proceeds) for the underwritten security. Given that commercial banks possess an
information advantage, the probability of choosing one as an underwriter is
higher if the firm is less risky. The empirical proxies for information asymmetry
are exchange listing“), volatility of equity return, equity trading volume, growth
opportunity (measured by Tobin’s q, sales growth), and firm size (defined as the
market value of equity). The proxy related to information cost effectiveness is the
firm size. The variables for credit ratings are defined in two different ways

depending on the purpose of use".

1.2.2 The Conflicts of Interest Hypothesis

The issue of conflicts of interest of commercial banks has been discussed in
both empirical and regulatory literatures‘z. Due to the perception of greater
conflicts of interest, it should be more difficult for commercial banks to convince

the public of the value of these securities than for investment banks. As

 

‘° Exchange listing is dropped in the empirical analysis because only five percent of
sample is not listed on an exchange.

1‘ See Appendix A for more information.

‘2 For detailed discussion of conflicts of interest and the concerns of commercial banks
as both lenders and underwriters, see “JP. Morgan 8 Co. Inc., the Chase Manhattan
Corp., Bankers Trust New York Corp., Citicorp, and Security Pacific Corp.,” 75 Federal
Reserve Bulletin 192 (1989), Puri (1996), and Kroszner and Rajan (1997).

12

suggested by Kroszner and Rajan (1997), organizations with greater conflicts of
interest should specialize more in issues with less information problems and less
risk. Thus, firms possessing characteristics that might be viewed as causing
conflicts of interest are less likely to choose commercial banks than investment
banks as undenNriters in order to avoid the price discount. The perceived conflict
of interest should be higher for firms with higher risk and information asymmetry
problems, with the most serious scenario being when the firm is also refinancing
bank debt (Gande et al, 1997; Kroszner and Rajan, 1997). Therefore, the
probability of choosing a commercial bank as an underwriter is smaller for firms
with higher risk and information asymmetry problems and/or when the firm is
refinancing bank debt. Given that the market is also using this information as
proxies for potential conflicts of interest, the bond price of commercial bank
clients should be lower than that of investment bank clients. The proxies used for
risk are volatility of stock return, leverage, and interest expense relative to

operating income; the predicted signs of probit estimates are negative”.

 

‘3 Although the predictions of underwriter selection are the same under both
informational advantage and conflicts of interest hypotheses, the regression coefficients
of net yield (BPS) regressions should be different. For instance, under the information
advantage hypothesis, the beta coefficients of risk variables from commercial bank BPS
regression should be smaller than that of investment bank BPS regression, because
commercial banks can certify the information better than investment bank can. The
commercial banks may reject the firm after the evaluation, so the association of
commercial bank with riskier firm is lower, in Diamond (1991), monitoring of new, low
quality borrowers can only be used as a screening device, many new borrowers will be
turned down for credit because monitoring does not provide incentive for c00perative
actions. This is an example that these three estimates in our empirical model does not
agree with each other, because the selection is not determined by the bond yield
differential. Whereas, under conflicts of interest hypothesis, the selection of underwriters
shows the same'pattern, the reverse is true for the estimates of bond pricing equations,
since investors will require a price discount when the perceived conflicts of interest is
more serious.

13

 

 

11

1.2..

and
anno
that li

again

 

hetero
loans,
to be ;
demar
select
more |

T'
andth
Dione
again:

fUtU re

1.2.3 The Bank Relationship Hypothesis

The value of bank relationships is documented in papers by James (1987)
and Billett et al ( 1995). They find that the market response to bank loan
announcements is positive. Holland (1994) points out in his case study research
that lines of credit and borrowing limits on unsecured loans provide insurance
against possible future credit rationing. But the value of the bank relationship is
heterogeneous across firms. A cash rich and profitable firm relies less on bank
loans, whereas unprofitable firms and firms experiencing high growth would like
to be associated with banks because they may anticipate higher bank loan
demand in the near future. This analysis thus leads to a lower probability of
selecting commercial banks as underwriters by firms with more cash, that are
more profitable and that anticipate slower growth.

The empirical proxy used for cash in this study is cash and cash equivalents,
and the proxy for profitability is operating income before depreciation. Growth is
proxied by 5-year sales growth and Tobin’s q. Besides the insurance value
against credit rationing, bank relationships could also provide better control of

future agency and information asymmetry problems of firms.

14

 

 

 

Inl- .

A
flexib
outsic
power
future
circurr
beher
underv
have hl

higher I
are the

profitab

 

1.2.4 The Bank Monopoly Power Hypothesis

Although banks provide the benefits of information certification and financial
flexibility, the benefits come with costs. Since firms cannot communicate with
outsiders about future prospects credibly, the bank has temporary monopoly
power over a fin'n’s private information, allowing the bank to appropriate firms'
future profits (Rajan, 1992; Sharpe, 1990; Greenbaum et al, 1989). In order to
circumvent bank power, a firm with a greater information asymmetry problem“,
better prospects”, and more free cash or higher profitability is less likely to be
underwritten by a commercial bank. However, commercial bank clients tend to
have higher cash reserve if the monopoly power of banks requires them to keep
higher compensating balances. The empirical proxies for the information problem
are the same as those in the information advantage hypothesis. The proxy for

profitability is operating income before depreciation.

 

1‘ See infonnaticn advantage hypothesis for these proxies, the predicted signs under
information monopoly hypothesis are opposite to that under information advantage
hypothesis.

‘ The proxies are Tobin’s q and sales growth, under information monopoly hypothesis,
the predicted signs of probit estimates should be negative.

15

 

 

1.3

1.3.1

It
abilitie
that 0:
bond]
select,
unden
selecti
selecti
Consid
underv
A Valid
and the
unseiec
applica‘

eXplaim
which sl
reQTeSS,‘

advanta!

1.3 The Switching Regression Model: Motivation and Description

1 .3.1 Model Specification

If commercial banks and investment banks have different underwriting
abilities, then, all else being equal, bond-issuing firms will select the underwriter
that can obtain the higher bond price. Therefore, the observed distribution of
bond prices underwritten by each type of underwriter is conditional on this
selection. The bond prices that could have been obtained by the unselected
underwriter are not observable. Thus, there is a need to control for this self-
selection when the bond pricing (or net yield) equation is estimated. Due to this
selection, commercial bank and investment bank underwritten issues may be
considered as representing two distinct regimes. The bond prices of issues
underwritten by different types of underwriters are no longer directly comparable.
A valid comparison would be the difference between the observed bond price
and the expected bond price, had the same firm been underwritten by the
unselected underwriter. The underlying selection mechanism leads to an
application of switching regression with endogenous switching. This model,
explained in detail in Maddala (1983), consists of a probit model estimation,
which shows the determinants of underwriter selection and two net yield
regressions for each type of underwriter, which contrast the comparative

advantages of different underwriters. Formally, our empirical model is as follows:

16

 

where
yield 0
banks.

VECIOTS

 

equafic

1, isao‘
tsunde,

The

Clli

The error

COIIan‘ancf

 

y.,- = X13. + u“ (1) (for CB underwritten issues)
y2, = X , ,82 + til, (2) (for IE underwritten issues)

1,’ = —(Z,7 — 5,) = —[X, (M) - (filing) (underwriter choice equation)

where CB stands for commercial banks, lB for investment banks, y, is the net
yield of bond for firm i, subscript 1 refers to issues undenNritten by commercial

banks, subscript 2 refers to those underwritten by investment banks, X ,, Z, are

vectors of exogenous firm characteristics. The standard deviation 0,,“ in

”It

equation (3) is the standard deviation of u” — uh.
Ii=1 iff li>0 (4) (yli<y2i)
[i=0 iff 1:30 (5)(yli2y2i)

I, is a dummy variable taking the value 1 if firm i chooses a commercial bank as

its underwriter, and 0 othewvise.

The observed y, is defined as

yizyli ifi 1I=1 (6)

yi=y2i ifi [1:0 (7)
(71101201:

C0V(u“,u2,,£,.) = 01202202; (8)
also-2:1

The error terms umuzna, are trivariate normal with means (0, 0, 0) and

covariance matrix defined as equation (8). Since firms will select the underwriter

17

 

 

who r

for thl

firms

The c:

the fol

 

where

distrlbu
The tes
eQuatio
advanta
We can

Case 1 ;

and '8 U

E('l.:r ) a I'

than aVe

thestmer

who can obtain the lowest possible yield, if y,, < y2,, they will have CB underwrite
for them, and we will observe y], instead of y2,. If the converse is true, then the
firm’s choice will be the IB, and we will observe yZi.

The conditional mean net yields of CB and IB underwritten issues are given by

 

the following:
¢(Zr)
_ :1 =X. __
EO’i: i1, ) ifli+0151_¢(zy) (9)
¢(Zr)
. - =0 =Xr 7 — ,
E(y21 I II ) rfl- 0-8 (D(Z}’) (10)

where ¢(-) and <D(-) are, respectively, the Standard Normal density and

distribution functions.

The tests of endogenous selectivity bias are tests for 0'” = O and 0'2: = O in
equations (9) and (10). If the endogenous selection is based on comparative
advantage, then an — an is less than zero, but a”. and 0'2.- can have any sign.
We can consider the following cases:

Case 1 : 0,5 < O, a}; > O In this case the conditional mean net yields of both CB
and IB underwritten issues are less than their unconditional means E(y,,.) and

E(y2, ), respectively. Thus, those who choose commercial banks are better off

than average firms if all firms are underwritten by the commercial banks, and
those who choose investment banks are better off than average firms if all firms
are undenlvritten by investment banks. Here, both commercial banks and

investment banks specialize in certain segments of the market that cannot be

18

 

 

explail
becau

inform.
Case 2
underv
condnh
the unc
are bet
beherc
who Chl
underw
banks c
Case 3;
Case 4:

issues a

7.3.2 E

EQU;
DamCUIa'

dimensio

explained by the publicly known firm characteristics. They coexist in the market
because they possess distinct technologies in resolving the problem of private

information of issuers. This is the best scenario for the economy.

Case 2: a"; < O, 035 < 0 Here, the conditional mean net yield of commercial bank
underwritten issues is less than the unconditional mean E(y,,), and the
conditional mean net yield of investment bank undenlvritten issues is greater than
the unconditional mean E(y2, ). Therefore, those who chose commercial banks

are better off than average firm underwritten by both underwriters, but they are
better off underwritten by commercial banks than by investment banks. Those
who choose investment banks are below average firms, but they are better off
underwritten by investment banks than by commercial banks, i.e. commercial

banks can cherry pick the firms when they provide the undenivriting services.

Case 3: 0'” > O, 0,: > O This is simply the reverse of Case 2.

Case 4: 0'“. > O, 0'25 < 0 Given the setup of the model, this is impossible,

because this implies that firms choose underwriters who underwrite the bond

issues at higher yields, or lower bond prices.

1.3.2 Estimation and Interpretation of Regression Estimates

Equation (3) is parallel to a probit estimation equation. If ,6, < ,62 for a

particular variable, then a commercial bank has a comparative advantage on this

dimension of the firm over an investment bank, so the probit estimate will be

19

 

 

 

positiv
coeffic
ofund:
advan:
exoger
selectic
explain
as a prc
measur
informer
Chi
ofunder
QGI this r
be Client

Of Ullden

 

BeSideS’

for C93. hr

Th
InStanCe’

firms! the

\

6
7? The Droc
We wm j;

positive (—(,BI — fig) > 0) for that characteristic. Comparison of these beta

coefficient differentials and the signs of probit estimates can identify the abilities
of undenNriters and the determinants of underwriter selection. The comparative
advantages of different underwriters can be then established through these
exogenous variables. The adjustment terms used to control the endogenous
selection can be used to examine the underlying selection that cannot be
explained by those exogenous variables. Therefore, these terms can be viewed
as a proxy of private information or overall ability of underwriters that cannot be
measured by the exogenous variables, which are the publicly available
information.

Of course, the model has its limitations; if the forces that influence the choice
of underwriters cannot be explained by the price effect differential, we may not
get this relationship among these three estimates. An example of this force would
be client switching costs where, due to temporary informational monopoly power
of underwriter, the client may not be able to switch to preferred undenivriters.
Besides, the ultimate decision of undenivriter choice is the net result of these

forces, not just one of them. We follow the two-stage estimation procedure
described in Maddala (1983)16 to estimate the parameters ,6,,,62, 0,5 and 02.-
The framework of two net yield regressions is essential to our study". For

instance, if commercial banks possess better ability in underwriting for large

firms, the beta coefficient of firm size from the net yield regression will be smaller

 

'° The procedure is discussed in Chapter 8 of Maddala (1983), pp.: 224-8.
‘7 We will label these net yield regressions as BPS regression for commercial bank (CB)

20

 

 

 

for co:
increa
more 1
banks
model
transla‘
cmnme

based c

andthe

 

 

for commercial bank clients than for investment bank clients. In other words,
increasing firm size reduces the net yield of commercial bank undenNritten issues
more than that of investment bank underwritten issues. Therefore, commercial
banks are better off underwriting for large firms than for small firms. Based on our
model specification, ceteris paribus, the beta coefficient differential ((31-132) should
translate into a positive relation (-(B1-Bz)>0) of firm size and selection of

commercial banks as underwriters. We classify as exogenous the selection
based on the interactions between the publicly known characteristics of the firm
and the identity of its underwriter.

If we use only one net yield equation for both types of undenlvrlters, we are
assuming there is no exogenous selection, that is, we are imposing the restriction
B1: [32 ‘8. Therefore, regardless of underwriter identity, the bond price effect of
firm size is the same. Firms receive different bond prices because they have
different sizes; this has nothing to do with the identity of their undenNriters. Thus,
an empirical model with only one net yield equation does not strictly follow from
the theoretical literature. This inconsistency might explain why previous studies

have been unable to provide a role for investment banks to exist and survive in

 

and BPS regression for investment bank (IB) for the remaining paper.

1" In the paper of Puri (1996) and Gande et al (1997), they use the framework of
Heckman’s two-step method and two-stage tobit model, respectively. The models
consist of one probit selection equation and one bond pricing equation. They use the
first-stage probit estimates to control the selection bias in the second-stage bond pricing
regression. Therefore, their models can be viewed as a restricted version of our model
that the ability of different types of underwriters in certifying public information of firms is
the same, and the only difference between these underwriters is the private information,
proxied by the probit estimates, which can be revealed only by the identity of
underwriters.

21

 

 

ther:

inves

have:
engag
benefi

isgpve

 

Elyhi

 

 

the pre-Glass-Steagall era when commercial banks were competing with
investment banks more freely.

Another advantage of this three-equation model is that, given that we do
have selection between underwriters, the benefit of allowing commercial banks to
engage in the securities unden/vriting business can be quantified. The gross
benefit for firm i underwritten by commercial bank rather than by investment bank

is given by the following equation:

¢(Zr)

E . 1::1— .:X 7+0",
(yZIl I ) yll rfl- -£1_(D(Z}’)

Ty“ (11)

The first term in equation (11) represents the expected mean net yield of
commercial bank underwritten issues, had they chosen an investment bank as
underwriter. The gross benefit for firm i underwritten by an investment bank
rather than by commercial bank can be defined similarly.

Theoretically the switching regression model can be identified if the
independent variables are the same for the three-equation system, because the
endogenous adjustment terms estimated from probit regression are nonlinear,
thus the rank condition for identification can still be satisfied. However, if the
adjustment terms are not nonlinear enough, then we may encounter serious
multicollinearity problems at the second stage when the net yield equations are
estimated”). Therefore, in the probit estimation, we use cash and reduction of
long term debt as instrumental variables (IV) to improve the identification of our

empirical estimation. In order to be a valid IV in the current study, the variable

 

‘9 See page 271 in Maddala (1983) for more discussions.

22

 

 

 

shoui

no 8X

finnc:

 

should be related to the underwriter-choice decision, and at the same time, have
no explanatory power on the yield of new bond issues after all other essential

firm characteristics are controlled for”.

 

2° As it is discussed in Opler et al (1999), firm maintains a surplus of internal funds, it
accumulates cash and pays back debt when it becomes due. Cash is simply negative
debt. Therefore, both cash and reduction of long term debt are used to proxy the level of
cash reserves of firms.

23

 

 

 

1.4

 

T
corpor
New Is

insuffic

 

comm?
study th
allowec
have m

the san

 

and nor
bOth iss
and Cor
from the

The
Usert n1

‘0 2224 .

1.4 Data and Descriptive Statistics

The data used in this paper consist of fixed-rate nonconvertible domestic
corporate bond issues from the Securities Data Corporation (SDC) U.S. Debt
New Issues database. Equity issues are not included in the study mainly due to
insufficient sample size. There are only 15 common stock issues underwritten by
commercial banks within the 12 months ended February 1997, according to the
study by Bhargava and Fraser (1998). Although commercial banks have been
allowed to underwrite corporate equity issues since 1989, the data are too few to
have meaningful results. Therefore, we only use corporate bond issues during
the sample period from January 1, 1991 to December 31, 1996. Only non-utility
and non-financial firms are used in the analysis. To be included in the sample,
both issue characteristics and firm characteristics must be available from $00
and Compustat databases respectively. Stock return data of issuers is collected
from the CRSP Stock files.

The initial data contains 2935 bond issues. After we delete issues without
useful net yield, purpose of issue and maturity data, the sample size is reduced
to 2224 observations. The requirements for calculation of volatility of stock return
eliminates an additional 604 observations; those for cash, interest expense
relative to operating income, sales growth, and total debt to total asset eliminate
57, 31, 82, and 14 issues, respectively. The market value of equity variable
reduces another 55 issues from the remaining sample. Finally, missing data for

equity trading volume and reduction of long-term debt variables eliminate 28 and

24

 

26 ct:
AHva
pnort
are m
reduct
daily cl
which
meCR

Detaile

 

Th1
for eacl
Present
banks,i
peood,
””18 of
of “WE:
COmme

level. C

26 observations respectively. The final sample consists of 1327 observations”.
All variables constructed from Compustat are measured at the end of the year
prior to the bond offering. Market value of equity and trading volume of issuers
are measured at the end of the month preceding the bond issue. The variable
reduction of long term debt is measured in the year of bond issuance. Volatility of
daily equity return is proxied by the variance of residuals from the market model,
which is estimated over 120 trading days prior to the issuing date. The return on
the CRSP value-weighted index is used to proxy domestic market returns.
Detailed definitions of each variable are listed in Appendix A.

The frequency of bond issues and summary statistics of selected variables
for each year by type of underwriter during the period 1991 to 199622 are
presented in Table 1.2. The proportion of bonds undenNritten by commercial
banks, which is 4.2% in 1991, grows to 18.6% by 1996. Over the entire sample
period, the average proportion is 13.9%. Based on a yearly comparison, client
firms of commercial banks are not significantly smaller in size compared to those
of investment banks. However, over the entire sample period, the average
commercial bank client firm is significantly smaller by about $4 bn, at a 10%

level. On the other hand, both mean and median issue amounts underwritten by

 

2‘ There are 33 observations with negative values of interest expense relative to
operating income due to negative operating income; for these observations we have
used zero to replace the negative value.

22 The frequency of commercial bank underwritten issues in 1996 is much lower than the
number of transactions reported in (Bhargava and Fraser, 1998), mainly due to the fact
that we only use industrial bond issues, and the heavy data requirement in our study
reduces the sample size to some extent.

25

 

com

onza:

BPS)
lower
BPSs
connnr

thechfl

 

butthe
yields 0
undenw
undenw
the diffs
invest”
Charact

Control]

commercial banks are significantly smaller than those by investment banks even
on a yearly basis, except the first year.

During the whole sample period, the mean net yield (basis point spread or
BPS) of commercial bank underwritten issues is 120.08, which is significantly
lower than the 137.7 of investment bank underwritten issues. The medians of
BPS show the same results as the means. The median of BPS is 71 for
commercial bank clients and 91 for investment bank clients. A similar pattern of
the difference in BPS is reported using pre-GIass-Steagall era data (Puri, 1996),
but the univariate results are not enough to draw any conclusions. The lower net
yields of commercial bank underwritten bonds could be the result of superior
underwriting abilities, but it is also possible that commercial banks may choose to
undenNrite for safer firms or firms with characteristics that reduce the BPS. Thus,
the difference in BPS between commercial bank underwritten issues and
investment bank underwritten issues is not comparable, unless the firm
characteristics and the potential endogenous selection problem are formally
controlled for.

Markets segmented by size of issue23 are also compared in Table 1.3 to
explore the possible regulatory distortion in the clientele of commercial banks.
Commercial banks underwrite 20.8% of the issues in the small issue market, and
only 7.7% of the issues in the large market. The size of issue underwritten by
commercial banks is also significantly smaller than that by investment banks in

both markets, whereas, the size of commercial bank clients in the small issue

 

23 We use the median size of issue (150 million dollars) of the full sample to separate the

26

market has no difference in mean and in median. In the large issue market, both
mean and median of firm size are significantly smaller for commercial bank
customers. Net yields (BPS), for commercial bank underwritten issues are
significantly lower although the credit ratings are not significantly better in the
small issue market. In the large issue market, the credit ratings of commercial
bank clients are significantly lower, but the net yield is not significantly higher.
The higher association of lower quality firms may imply that commercial banks
are less able to retain higher rated firms in the large issue market.

The credit ratings in Table 1.4 are categorized into 7 groups to examine
possible specialization in underwritings on this dimension. The results are
consistent with the theory proposed in Diamond (1991) that borrowers with credit
ratings toward the middle of the range have a higher demand for bank
monitoring. The effect of monitoring is also more valuable for the middle rated
firms. Commercial banks do not undenivrite any issues for the highest and the
lowest rating groups. The proportion of the sample undenrvritten by commercial
banks is the highest for the A and Baa groups. The distributions of issuer credit
ratings are significantly different at the 10% level. Basis point spreads are also
significantly lower for the A and Baa groups that are underwritten by the
commercial banks than that by investment banks, while there is no difference for
the Aa, Ba and B groups. Interestingly, category B shows higher basis point
spreads for commercial bank underwritten issues, although it is not statistically

significant. If the perceived conflict of interest increases or the value of bank

 

markets.

27

 

 

 

inonn
inves
oreve
enoug
rahngs
conflb

quesht

 

monitoring diminishes as credit rating decreases when the issuer is in the non-
investrnent grade category, then commercial bank association creates less value
or even negative value for such a firm. When the negative impact is large
enough, commercial banks will not underwrite for firms with extremely low credit
ratings such as C or below. Therefore, these results are consistent with both the
conflict of interest hypothesis and the low value of bank monitoring for

questionable or inferior quality issues.

28

 

 

 

 

1.5

T

 

onthe
impos
comps
banklr
conflic‘
The firs

underw

 

comme
ability 0
InVeshg
price (n:
The
beta Cor
arldl .7 1
Thelnde

rating, C;

1.5 Regulatory Distortion and Comparative Advantages of Underwriters

The study of the determinants of underwriter selection in this paper is based
on the certification ability of underwriters and the distortion of revenue limitation
imposed only on commercial banks. In terms of certification ability, we test four
competing but not mutually exclusive hypotheses. As stated above, they are the
bank information advantage hypothesis, the bank relationship hypothesis, the
conflicts of interest hypothesis and the bank information monopoly hypothesis.
The first two represent the relative advantages of commercial banks as
undenNriters, while the last two represent the relative disadvantages of
commercial banks as undenNriters. Due to the 10 percent revenue limitation, the
ability of commercial banks in handling large issue could be hampered. We then
investigate the possible distortion by examining the relationship between bond
price (net yield) and issue size as well as firm size.

The results of the probit model estimation of underwriter selection and the
beta coefficients of second stage BPS regressions are reported in Tables 1.5,16
and1.7 for the full sample, and the small and large issue markets respectively”.
The independent variables in all the analyses are the same except for credit

rating, cash, reduction of Iong-terrn debt, the interaction between dummy

 

2‘ In order to prevent unnecessary collinearity in estimating the BPS regressions for
these subsamples, the explanatory variable, size of issue is dropped in the small issue
market subsample, since the results in Appendix B2 show this variable does not explain
the net yield of bonds in the small issue market. For the same reason, the endogenous
adjustment terms are dropped in the large issue market.

29

 

 

variat:

 

I
exper
|

As

choice

rating t
of long-
empiric.i

The
rGillianc
oDefatin
rEgressir
between
andinter
regreSSic

term. The

 

variables representing the intention to refinance bank debt and high interest
expense”5 and the endogenous selection adjustment terms.

As in Diamond (1991), the relationship between firm credit rating and the
choice of underwriter is not linear. Firms with credit ratings in the middle of the
spectrum tend to be associated with commercial banks. So, in the probit
estimation, we use credit rating26 and non-investment grade dummy variable to
capture this nonlinear relationship. However, in the second stage BPS
regressions, we only use non-investment grade dummy to control the credit
rating to avoid too much collinearity in the linear regression. Cash and reduction
of long-term debt are the instrumental variables we used to enhance the
empirical estimation”.

The interaction between the dummy variable representing the intention to
refinance bank debt and the indicator of high interest expense relative to
operating income cannot capture the effects in the second stage BPS
regressions too well. So we use a more continuous measure, the interaction
between the dummy variable representing the intention to refinance bank debt
and interest expense relative to operating income, in the second stage BPS
regressions and report only the results of model specified with this interaction
term. The endogenous selection adjustment terms are estimated from the first

stage probit, and the variables are described in Appendix A. The second stage

 

25 The high interest expense indicator is one if the interest expense relative to operating
income is greater than the median of this variable, 0 otherwise.

2" The assigned value of credit rating is 1, if the Moody’s credit rating for the bond issue
is Aaa or Aa, the value is 2, if it is A or Baa, the value is 3, if it is non-investment grade.
27 See Section 1.3 for a detailed discussion.

30

 

 

BPS

rem ON

uncha

1.5.1

Itn

 

mshkn
this pr;
this on.
ODSEPV
Smaller
banks!
be Obs
banks'
strateg
the 10
S'eCtion
reSJUIat

tier
then CE
Further

betWEe

BPS regressions are also analyzed by alternative model specifications, which
remove some insignificant independent variables with the results essentially

unchanged. We report these additional results in Appendices B1 and B2.

1. 5.1 Regulatory Distortion of Commercial Bank Undenmr'tings

It may be plausible to assume commercial banks have disadvantage in
distributing the corporate securities they underwrite, since they were barred from
this practice for 60 years. As an entrant to the market, they need to redevelop
this crucial skill in order to distribute large issues effectively. However, we
observe that the issue size managed by commercial banks became smaller and
smaller over time during 1993 to 1995. This trend seems puzzling. If commercial
banks gain experience through completing deals, then the opposite trend should
be observed. One possible explanation of this trend would be that commercial
banks possess unique advantages in underwriting for small clients, so they
strategically position themselves in the small issue market. On the other hand,
the 10 percent revenue limitation may be the cause of this phenomenon. In this
section, we empirically explore whether the small client advantage or the
regulatory distortion is the underlying reason for the above trend.

If commercial banks possess distribution abilities similar to investment banks,
then ceten's pan'bus, there will be no observed selection based on issue size.
Furthermore, the effect of issue size on the bond prices should be the same

between these two types of underwriters. However, the results in Tables 15,16

31

 

 

 

 

bankc

dlofi:
of the

Thl
undent
condhi
bsue_-
ability (

 

and1.7 do not support this null hypothesis. In fact commercial banks show a
significant disadvantage in handling large issues. The probit estimates on issue
size in Tables 1.5 and1.6 indicate that commercial banks are significantly
associated more with small issue than with larger issues. Besides, Table 1.7
shows that the beta coefficient of issue size for BPS regressions of commercial
bank clients is highly significant and positive in the large issue market. However,
all other estimates on issue size are close to zero for BPS regressions regardless
of the identity of underwriter.

The results demonstrate that a commercial bank has to sell the bond it
underwrites at a lower price, as the issue size increases. Under current market
conditions, commercial banks lose on the ground in dealing with large bond
issue. Consequently, the clients of commercial banks suffer from the limited
ability of their underwriters. This may cause under-investment of commercial
bank customers.

Contrary to the finding by Gande et al (1997), we do not find evidence that
commercial banks facilitate smaller firms in gaining access to the capital market.
In their paper, they derive the conclusion by examining the issue size instead of
investigating the firm size directly. In fact, commercial banks can reduce net yield
of bonds more than investment banks when the size of firm is larger. Based on
the comparative advantage of commercial banks on firm size, they should
underwrite more large firms than small firms. However, the probit estimates on
firm size in Tables 1.5, 1.6 and1.7 fails to show this pattern. When issue size is

also considered, we find that the inconsistency in firm size is due to commercial

32

 

 

 

quant
mves'
some
the b.
ofbo
Wekl
Unde
unde

itScli

reDOr
resz
Appe
aCroS
COITIn

an es

banks being constrained to handle only small issues (because of their inability in
handling large issue sizes). Since firm size and issue size are positively
correlated, when one variable is restricted the other may be affected
systematically. Thus, these results do not support the small clients advantage
hypothesis.

Besides the advantage of studying the pros and cons of different
undenivriters, the application of the switching regression model allows us to
quantify the benefits that different underwriters create. Therefore, we are able to
investigate the effect of bank entry and regulation distortion at the micro-level,
something which has not been done by previous research in this area. We define
the benefits created by the chosen underwriters by the expected mean net yield
of bond issue, had it chosen the unselected underwriter minus the observed net
yield of bond issue. Thus this term will be positive if the net yield of bond issue
underwritten by selected underwriter is smaller than that by unselected
underwriters. It also means that the selected underwriter perform better to serve
its clients than the counter group.

The benefits presented in Table1.8 are estimated by using the results
reported in Table 1.5. To check the robustness of the results, we also use the
results reported in Appendix B1 to estimate the benefits and present them in
Appendix B3. The benefits created by the selected undenrvrlters are analyzed
across time, size of issue and credit ratings. The full sample results show that
commercial bank entry created insignificant benefits for clients on average, with

an estimate of 2.43 during the whole sample period. Only firms with credit ratings

33

in the middle range as a group show significantly positive benefits in Panel B of
Table 1.8.

The inefficiency of commercial bank underwritings could be driven by two
factors. First, commercial banks were barred from underwriting for nearly 60
years. As new entrants to the security underwriting business, they need time to
develop their underwriting expertise. Second, commercial banks were allowed to
undenlvrite corporate securities again under regulatory restrictions. Therefore, we
have a group of constrained underwriters competing with the unconstrained
group. Thus, the benefits created by investment banks are significantly positive
since they are “ahead” of commercial banks by 60 years and they are competing
without constraints.

In order to confirm the inefficiency that may be caused by the revenue
limitation, we continue the analysis by splitting the sample into small issue
market and large issue market and into two subperiods. years 1991 to 1993 and
1994 to 1996. We re-estimate the benefits for these four subgroups to
accommodate the changes in different issue sizes and different entry stages of
commercial banks. The results reported in Table1.9 indicate that bank-entry in
the security underwriting business is beneficial to the commercial bank clients as
a group.

When we compare the magnitude of benefits created by the selected
underwriters across different issue size markets, we find the magnitude is smaller
in the large issue market than in the small issue market. This is consistent with

the results in Tables 1.6 and1.7 that, overall, the selection of underwriters is less

 

 

 

 

 

 

 

 

 

pronounced in the large issue market. However, the patterns of undenlvriting
efficiency of commercial banks are different between markets. During the later
sub-period (during 1994 to 1996), commercial bank underwritings demonstrate a
significantly positive estimated benefit of 24.39 in the small issue market,
whereas the opposite is observed in the large issue market. In the large issue
market, all estimated benefits of commercial bank underwritings change from
significantly positive to negative numbers except the issuers with middle credit
ratings (A or Baa). However, we can only establish significant negative benefits
among the best-rated (Aaa and Aa) small issuers. The results are consistent with
the findings of Nanda and Warther (1998) that smaller firms, and those that issue
securities less often and have lower credit ratings tend to be more loyal and pay
higher fees. These are the “locked-in” clients. However, for these trapped clients,
the inefficiency occurs at the later entry stage of commercial banks.

Although bank entry is not beneficial to all issuers, we do observe that
commercial banks improve their underwriting ability over time in the small issue
market. The reduction in magnitude of estimated benefits for investment bank
clients at a later period is consistent with the above notion. The smaller positive
estimated benefits imply that the differences in underwriting abilities between
commercial banks and investment banks become smaller from the viewpoint of
investment bank clients.

If there was no distortion in the economy, as time goes by, we should
observe a similar phenomenon in the large issue market, and the market should

move towards a better equilibrium with more firms teaming up with the “right”

35

underwriters and commercial banks improving their underwriting ability. Although
we observe the above "moving towards equilibrium" phenomenon that the firms
served by investment banks at later period are the ones can receive positive
benefits, we cannot conclude that commercial banks improve their underwriting
ability. The timing of observed inefficiency is consistent with the hypothesis of
regulation constraints rather than that of lack of distribution network or expertise
because commercial banks are entrants. In addition, the large coefficient of issue
size estimated in the BPS regression for commercial bank clients in Table 1.7
points out the weakness of commercial bank underwritings in large issue market.
Besides the above evidence on supporting the regulatory distortion, we
consider two other minor modifications of the revenue limitation regulations in
1993 and in 1996. In 1993, the Federal Reserve Board allowed section 20
subsidiaries to use an alternative revenue test that was indexed to account for
changes in interest rates since 1989. The shift in interest rates causes the
eligible revenues of commercial banks to shrink, therefore, the base of eligible
income used to calculate the revenue limitation is smaller even though the level
of eligible and ineligible activities of commercial banks remains the same.
Although the new indexed test is expensive and complicated to execute, five
section 20 subsidiaries switched to the new measure”. This implied that these

section 20 subsidiaries were operating close to the 10% revenue limit around that

 

2" See Order of Federal Reserve System, Docket no. R-0841, “Revenue Limit on Bank-
Ineligible Activities of Subsidiaries of Bank Holding Companies Engaged in Underwriting
and Dealing in Securities.”

36

 

perio
1993
I
again
the in
the qr
rangir
oomrr
Steag
comm
dollar 1
that c
markel
have d
In
was a l
Operati
will be

disadVg

 

 

period. This may explain why the issue amounts are relatively smaller in years
1993 and 1994.

In 1996 the Federal Reserve Board amended the revenue limitation order
again. The new order allows commercial banks to exclude the interest earned on
the ineligible securities from the 10% revenue limit. The change would decrease
the quarterly ineligible revenue of some section 20 subsidiaries by a magnitude
ranging between 19% and 79%”. Not surprisingly, the issue amounts of
commercial bank clients increased sharply in 1996. Compared to the pre-Glass-
Steagall era, the market share of commercial banks is still small. In 1996,
commercial banks underwrote roughly 14% of the bond market based on the
dollar amount in our sample, while in 1929, Kroszner and Rajan (1997) report
that commercial banks underwrote 45.4% of the market. The relatively smaller
market share nowadays, is most likely the result of regulation constraints as we
have demonstrated.

In 1997, the Federal Reserve Board rescinded the original firewall, which
was a set of 28 rules, and consolidated the remaining restrictions in a series of 8
operating standards”. At the end of 1996, the revenue limit was raised to 25%. It
will be interesting to examine the effects of these regulation changes to see if the

disadvantage of commercial banks in handling large issue under 10% revenue

 

2" See Federal Register, September 17, 1996, “10 Percent Revenue Limit on Bank-
Ineligible Activities of Subsidiaries of Bank Holding Companies Engaged in Underwriting
and Dealing in Securities,” volume 61(181): 48953-48954.

3° See Federal Reserve System Docket No. R-0958, “Bank Holding Companies and
Change in Bank Control (Regulation Y); Amendments to Restrictions in the Board’s
Section 20 Orders.” 12 CFR PART 22_5._

37

limitation disappears and if the large issue market will move to a more efficient

matching between underwriters and client firms.

1.5. 2. Advantages of Commercial Banks as Undem/riters

The probit estimates of credit rating and non-investment grade in Tables1.5
and1.6 are consistent with the informational advantage hypothesis. Firms with
lower credit ratings tend to select commercial banks as their undenivriters. In
Table1.5, the variables related to the lnforrnation asymmetry problem are daily
equity return volatility, equity trading volume, firm size, and Tobin’s q. The results
for both daily equity return volatility and firm size are weak in supporting any
hypothesis regarding certification ability of underwriters; we do not discuss these
two variables further in this section. The association of commercial banks with
clients that have low equity trading volume supports the information advantage of
commercial bank hypothesis.

The last variable used as a proxy for the information problem is Tobin’s q,
which is also used as a proxy for investment opportunity in existing empirical
finance literature. Both probit and BPS estimations show support for the
information advantage and the bank relationship hypotheses rather than for the
information monopoly hypothesis. The positive relation of Tobin’s q with
commercial banks and the lower beta coefficient of BPS regression relative to the
estimate of investment bank underwritten issues imply that commercial banks

possess an advantage in solving the information problem and in serving the

38

financing needs of growing firms. Therefore, commercial banks may certify firms
with higher Tobin’s q better than investment banks. Similar results are found in
Houston and James (1996), indicating that among firms with public debt
outstanding, the reliance on bank debt is positively related to the importance of
growth opportunities, with bank information monopolies having less importance
for these firms. They also find that the reliance on bank borrowing is decreasing
in overall leverage, a result that we too obtain. Overall, the significant association
of commercial banks with firms that are less risky, traded less frequently, lower
rated, and having higher Tobin’s q supports the informational advantage and the
bank relationship hypotheses.

Besides the exogenous selection when firm public information interacts with
the identity of its underwriter, the signs and magnitude of endogenous selection
adjustment terms can reveal the unobservable nature of underwriter selection. In
Table1.5, the estimated signs of adjustment terms are both negative, which
supports the cherry-picking behavior of commercial banks. However, only the
estimates for commercial banks in the full sample or in the small market are
significant. This result implies that commercial banks do possess better
unobservable ability that is valued by the market, which puts investment banks at
a disadvantage in certifying the private information of their customers. Overall,
there is no evidence of endogenous selection in the large issue market, whereas,

commercial banks can cherry-pick firms in the small issue market.

39

1.5.3 Disadvantages of Commercial Banks as Undennm'ters

One important concern regarding commercial banks engaging in security
underwriting is the potential conflicts of interest. The variables used to investigate
this issue are whether the proceeds are used to repay bank debt, leverage,
interest expense relative to operating income, and its interaction with intention to
refinance bank debt. As discussed in Kroszner and Rajan (1997), if investors are
aware of the motives of an underwriter, in equilibrium the underwriter will be
forced to issue securities that are less risky and questionable to avoid the price
discount.

The probit estimates of the dummy variable for intention to refinance bank
debt and leverage are consistent with the above notion, indicating that
commercial bank clients are less risky and refinance for bank debt less
frequently. However, we do not find the support for this argument by the probit
estimates on interest expense relative to operating income and the interaction
between intention to refinance bank debt and high interest expense indicator. In
the BPS equations, interest expense relative to operating income and its
interaction with intention to refinance bank debt show large price discounts,
especially for commercial bank clients. The effect is stronger in the small issue
market but weaker in the large issue market.

In Table1.5, the estimates of interest expense relative to operating income
and its interaction with intention to refinance bank debt from the BPS regression

for commercial bank underwritten issues are nearly triple compared to the same

40

numbers for investment banks. However, the BPS is lower when the purpose of
issue is refinancing bank debt and the level of interest expense is low. Overall,
we do not find investors use the intention to refinance bank debt as a proxy for
potential conflicts of interest and impose a price discount when the undenivriter is
also a commercial bank. Instead, the effects of conflicts of interest are
demonstrated through interest expense relative to operating income and its
interaction with refinance bank debt. Thus, we need to consider the beta
coefficients of these three variables as a group when interpreting the effects of
conflicts of interest.

The variables used to test bank relationship and information monopoly
hypotheses are volatility of daily equity return, Tobin’s q, sales growth, and
profitability. Tobin’s q has been discussed previously with results that do not
support the information monopoly hypothesis. The estimates of volatility of daily
equity return, sales growth and profitability support the information monopoly
hypothesis. Although the probit estimates are not significant in Tables1.5 and1.6,
the estimates of BPS regressions indicate that commercial banks are at a
disadvantage when underwriting for firms with high volatility of daily equity return,
sales growth and profitability.

For commercial bank clients, higher growth rates will increase BPS of bonds
significantly with an estimate of 0.82 in Table1.5; conversely, for investment
bank clients, the estimate is -0.06 (insignificantly different from zero). Increasing
the daily equity return volatility increases the BPS of bond issues for the clients of

both underwriters, however, the magnitude is tripled for commercial bank clients.

41

Theoretically, while high profit is a favorable characteristic of firms, we should
observe profitability reducing the BPS of bonds. This is true for investment bank
customers, but the estimate is close to zero and positive for the commercial bank
clients. Although a bank relationship is less valuable for firms with high profits, it
should not lead to a zero beta coefficient of BPS regression. The possible
explanation is the benefit of high profit is cancelled out due to the bank’s
monopoly power to share the profit with its client in the future. The effects are
magnified in the small issue market but reduced in the large issue market. These
results provide support of a hold-up problem of commercial banks. Due to the
hold-up problem, commercial banks are less credible in underwriting for firms

with high volatility of equity return, “past” growth31 and profitability.

1.5.4 The Performance Comparison between Commercial Banks and

Investment Banks

In general, investment bank clients as a group are better off staying with
their own underwriters, while those switching to commercial banks are not
necessarily better off. When the results in Tables 1.8 and 1.9 are compared

across credit ratings, firms in the middle range of credit ratings that switch to

 

3‘ The results of Tobin’s q and sales growth are different, although we try to use them as
the proxies for growth opportunities. However, Tobin’s q represents future growth, sales
growth is realized past growth. The former requires better information technology to
certify the message it tries to convey, the latter, however, may not be a good proxy for
future growth, instead, the past growth seems to provide information similar to
profitability but it is not a complete repetition, because after operating income is
included, sales growth still possesses explanatory power in the BPS regression.

42

commercial banks are always better off, regardless of whether they are small
issuers, large issuers, at the early stage or later stage of bank-entry. The results
are consistent with the theory of Diamond (1991) that middle rated firms receive
the most benefits from bank association. For firms with other credit ratings, the
benefits are either insignificantly different from zero or significantly negative. The
groups that suffer from the bank entry are high rated small issuers and low rated
large issuers”.

Another interesting observation in Table1.9 is that if we compare the results
across subperiods, the estimated benefits for investment bank clients are all
positive at later period, but not for commercial bank clients. The results are
consistent with the explanation that as more firms that should be underwritten by
commercial banks switched from investment banks, the firms remaining with
investment banks were the ones receiving greater benefits of using the
underwriting service of investment banks.

However, the factors affecting the benefits created by commercial banks are
much more complicated. At the early entry stage, commercial banks may have
picked the firms from the top of the list, but at the same time, since they just
started developing their underwriting ability, the two effects may cancel each
other out. Similar effects may exist at the later entry stage, however the process
is reversed, because banks pick firms with lower benefit potential while their
underwriting ability improves. The 10 percent revenue limitation creates

additional complications. Thus, at the later entry stage, the performance of

 

32 The result is marginal significant if one-tailed t-test is used.

43

commercial banks worsens in the large issue market. Although one may argue
that the large issue clients could switch to investment banks, the switching cost
may not be worth the benefit, especially for the clients with low credit ratings. We

may then observe some inefficiency in the commercial bank undenlvritings.

1 .6 Conclusion

We find that commercial banks do not actively engage in the small issue
market voluntarily. In fact, based on their comparative advantage along the firm
size dimension, they should be more active in underwriting for large firms.
However, the 10% revenue limitation restricts their ability in handling large
issues. We provide evidence that this distorted behavior of commercial banks
observed in the security market during the sample period is the result of a
regulatory restriction. This contributes to the observed inefficiency in the large
issue market among the commercial bank clients at the later entry stage.

In this paper, we examine the determinants of undewvriter selection, and
establish the comparative advantages of commercial banks and investment
banks. Commercial banks are superior in terms of information advantage and
provide a better financing opportunity, thus they are associated with firms that
have higher Tobin’s q, lower leverage, are traded less frequently and lower rated.
The endogenous selection adjustment term also indicates that commercial banks
can cherry-pick better customers. Due to concerns of conflict of interest,
commercial banks clients incur larger price discounts if they have higher interest
expense relative to operating income and when the purpose of issue is to repay
bank debt. In addition, when firms with volatility of equity return, high past growth
and profitability are associated with commercial banks, the potential bank

monopoly power results in lower bond prices for these clients.

45

By using an empirical model that is consistent with theoretical work in this
area, we are able to better quantify the benefits created by the selected
underwriters for their client firms. The positive benefits created by investment
banks demonstrate that they are valuable underwriters in the economy. The
comparative advantages and specialization in security undenlvritings may also
explain why investment banks coexisted with commercial banks in the pre-Glass-
Steagall era, when both types of undenlvriters could compete freely. Therefore,
this puzzle unexplained by previous studies is, we believe, resolved.

Although the re-entry of commercial bank in the security underwriting
business is not beneficial to all firms under the current regulatory and market
environment, firms with middle credit ratings are better off on average. The
distinct abilities of commercial banks and investment banks imply their
coexistence is valuable to the economy, so clients at different segments have

more choices and may be better served by different types of underwriters.

46

APPENDIX 1

TABLES OF ESSAY 1

47

Table 1.1

Summary of theoretical predictions regarding choice of commercial bank
(vs. investment bank)

 

 

 

 

 

 

to borrower’s risk

(Volatility of stock return)
(Total debt/total assets)
(Interest expense relative
to operatinjfincome)

Hypothesis Prediction of Empirical proxy Predicted sign
probability(bank) on empirical
is: ‘ proxy
Information Positively related Equity trading volume -
Advantage to information
asymmetry Volatility of stock return +
Growth Opportunity +
(Tobin’s q)
(Sales growth)
Firm size -
(Market value of equity)
Positively related Credit ratings
to the value of (BEST) -
bank monitoring (MID) +
(LOW) -
Positively related Firm size -
to cost of
collecting
information
Negatively related Risk -

 

Positively related
to borrower’s
(unobserved)

gualitya

 

 

 

 

 

a Since the quality is unobserved, we do not have a proxy for this variable.
However, the private information can be inferred by the underwriting decision of
both undenlvriters. See Section 1.3 for a detailed discussion.

48

 

Table 1.1 (Cont'd)

 

 

 

 

 

 

 

 

 

 

 

Hypothesis Prediction of Empirical proxy Predicted sign
probability(bank) on empirical
is: PTOXY
Conflicts of Negatively related Use of proceeds includes -
Interest when bank loan is repayment of bank loan
to be repaid
Negative related to Risk -
borrower’s risk (Volatility of stock return)
(Total debt/total assets)
(Interest expense relative
to operating income)
Negatively related See the proxies in the The signs are
to information information asymmetry opposite to
asymmetry section of information those in the
advantage hypothesis Information
Advantage
Hypothesis.
Bank Positively related Cash and cash equivalent -
Relationship to need for future
bank financing Growth and investment
Opportunity +
(Tobin’s q)
(Sales growth)
Profitability
(Operating income) -
lnforrnation Negatively related See the proxies in the The signs are
Monopoly to information information asymmetry opposite to
asymmetry section of information those in the
advantage hypothesis Information
Advantage
Hypothesis
Negatively related Growth and investment -
to potential hold-up opportunity
problem (Tobin’s q)
(Sales growth)
Negatively related Profitability -
to profitable firm (Operating income)
Positively related Cash and cash equivalent +
to firm required to
keep
compensating
balances

 

 

 

 

 

49

 

03 on? on? mm? one 0m? 09 mxcmn EoEfioE.
taco? .mww 5mm 33 :63 «to? on: 9.ch ESoEEoo
Eo==E av .mEEovc: >5 cum mama. capo—2

Fméw.‘ Pods vmdt tam F node nvdom mad: 9.ch EoEuon.

tam. mm «mm .09. at.» w .mm tam Km a-amm tau. 5 3.31. mxcmn _m_o..oEEoO
ACE—=5 my hmzhémbc: >0 ®N_w gamm— 6mm:

mfow mm. 5 Eton mfnm coda Nada afimm mxcmn EoEamoZ...

Fwd? v99 mwdm 5.3. vméw modr _.N.v 9.ch _m_o..oEEoo
”>0 COECEOUCD mmzwfl *0 mmmfibmobmn—

mv Z. N8 N5. m 3 mmm mmm mom mxcmn EoEumoE.

va in vm om mm mm a 9.53 .mEoEEoO

:3 acute/coca: meson. Co .3832

 

mam w- 59. mam? mam? 39. mag Nam? 59. 50>

.69 833.53 coxoo__>> B 3338 E

8.55:6 m5 new $02 3 369 m_ memos. E 99:33:: So $9: 92 amnion mascots 9: Scream. mum 299:; m Co
569: new cmoE ccz>> EtamE oBEmano ho beacon. hammock .mj ho 29> oEm xo o5 56 259 m be cacao 29>
2cm 5 c5 .6 82an be m_ cacao £58 Emma 9:. Egon 22:5 5 333 Go o:_m> “9.5.: on. m_ use: Co 36 9F
53.305252: .3 new Emma» B octane. 9m menus econ Co mumoam ESQ mama new moum .mommucoouoa .2383: .30...

abacutEoucz 3 new use» 3 mountouuflwsu E: can can». neon .0 33333 333.530
N.-. Sam...

50

.55. 5.5.62: 5 .2 5.525535. 5.55. .5555 3 5:5 .5 .3 55 5 E55555 2... .3. ..

 

35 on 55 5x. 53 v3 :5 3 55.55 25.5552...

r... E 55 .55 5 .5n .55 55 55.55 5.5.5....50
>5 52:259.: 55:55. .0. 555.55 .505 5.55.5 5.55.2

p.23 35.53 3.53 55.53 35.53 N553 3.53 5x55 E58552:

.5553 55...: 55.53 55.5 S 55.5 E 55. 3.. N55 9.55 5.558850
.3 55:259.: 55:55. 5.. 555.55 .555 5.555 55.2

35.: 55.5 3.5 55.5 vs: 55.5 55.5 9.55 25:552..

...:.N5.5 $3.5 5nd 556 N55 .35 55.5 9.55 5.552.250
”€5.55 8 5:25.52: >5 555 2...“. 5:55. 5.55.2

v53 35.: 5:5 V5.3 3.3 55.3 55.3 9.55 258.552..

55.5 3:55.: 3.5 55.5 .. 5.5 5 3 .3 3.5 5x55 5.5.5.550
”25...... 55 5.559.: .5 555 E._u. 5:55. 555.
555 T 353 553 553 .53 553 ~53 353 55>

 

.355. «.3 2.3

51

Table 1.3
Descriptive statistics of bond issue by issue size market and by
underwriter-type

This table reports means and medians of issue size of bond, firm size (market
value of equity), credit ratings, basis point spread of bond over the ex ante yield
of US. Treasury security of comparable maturity, as well as the number of issue
and percentage of issue by underwriter-type when the market is segmented into
different sizes of issue. Credit ratings is 1 if the Moody’s credit rating for the
bond issue is Aaa or Aa, the value is 2, if it is A or Baa, the value is 3, if it is non-
investment grade. When mean and median of a variable are reported, the
difference in means between two types of underwriters are tested by t-test and
that in medians by Wilcoxon rank-sum test, respectively. The small issue market
is defined as the new-issue bond market if the size of issue is less than the
median of the issue size (150 million dollars). The large issue market is defined
as the new-issue bond market if the size of issue is greater than or equal to 150
million dollars.

 

Small Issue Market Large Issue Market

 

 

Mean Median Mean Median

Issue Size by Underwriter ($ million)

Commercial banks 45.26*** 16.5“" 207.24*** 195*"

Investment banks 79.02 100 265.8 229.15
Issuer Firm Size by Underwriter (3 billion)

Commercial banks 5.75 2.99 7.35*** 3.86***

Investment banks 6.48 2.51 13.99 6.79
Credit Ratings of Issue by Underwriter:

Commercial banks 2.08 2 2.17” 2**

Investment banks 2.08 2 2.0 2
Issue Basis Point Spread by UndenNriter:

Commercial banks 1 1 1 .75*** 71 *** 140.12 72

Investment banks 149. 36 98 128.79 85
Number of Issues Undenlvritten by:

Commercial banks 130 54

Investment banks 495 648
Percentage of Issues UndenNritten by:

Commercial banks 20.8 7.69

Investment banks 79.2 92.31

 

*, **, *** Significant at the 10, 5, and 1 percent levels, respectively, for a two-
tailed test.

52

Table 1.4
Issuer credit ratings distribution and basis point spreads by
underwriter-type

The basis points spread (BPS) is the premium of the ex ante yield spread of a
bond over the ex ante yield of US. Treasury security of comparable maturity.
The first column contains the proportion of the sample for each credit rating
category, followed by the means and median of BPS for each credit rating
category for the commercial bank undenNritten issues. Columns 4, 5, and 6
contain the same information as that in columns 1, 2, and 3 for investment bank
undenlvritten issues. Test for difference in means is a two-tailed t-test. Test for
difference in medians is a Wilcoxon rank-sum test. The last row in the table
shows the chi-square test and p-value, with the null hypothesis being that the
credit ratings distributions for issues undenlvritten by both underwriters are the
same.

 

 

 

 

 

Commercial Banks Investment Banks
Moody’s Percentage Basis Point Spread Percentage Basis Point Spread
Credit Ratings of Issue Mean Median of Issue Mean Median

Aaa 0 1.84 50.67 44.1

Aa 7.61 52.99 53 13.3 55.52 50

A 47.83 57.37*** 55*“ 40.42 80.12 76

Baa 26.09 89.26“" 80*" 25.81 128.44 1 10

Ba 9.78 234.89 233 9.71 269.78 245

B 8.7 486.97 463 8.31 427.74 396

C or below a 0 0.61 346.77 275

Chi2(6)= 10.74; p-value = .097

 

*, **, *** Significant at the 10, 5, and 1 percent levels, respectively, for a two-
tailed test.
a This category includes non-rated bonds.

53

.nocaoE 9m $5 5965 notoqm: 6: 9m 259 EEmcoo new mQEEzc 2:3» 9: .8 $5623 of. .coszzmo
Boa manage: 9: 80: 86833 2m mommfi cozoofim maocomonco o5 .9E8 2 com: 952 Eogmnfim wzocomoccm
.oEooE @5980 o. 9:99 3598 $9.95 new Eon xcmn mocmccom E 6.9 c2692... 9: w_ 3598 $295.53
Ema oocmccom .2339 $5 so 569: 9: :9: Ema. m_ 9:85 93380 2 9.6.2 3598 5925 E _m>o_
o5 : F m_ cams B363 9: 633.9 >EEau m m_ .2865 $225 :9: 693550 o .53 xcmn mEExo mocmcco. o.
m_ mama so 3093 05 t F m. 65 mfimtm> >583 m m. Emu xcmn mocmccom .938 .22 3 8233 353 so o2m> 8me
man Eon ho o:_m> xoon 05 >9 Deacon m_ c 3.58 .mmsaméo o .099 E: .o 26.3 .0 mm m_ 96$ neon 05 .2 95m.
:88 36022 on. a F m_ mama EoEsmo>Eéoz .mumzm EoEbmo>Eéoc m_ z a. .m m_ cams 9.: .95 .o < m_ z t .N 2 o:_m>
o£ .m< .o mm< m. ozmm_ 6:03 9: .8 95$ :88 @3022 9: x F 2 95m. €20 .moamm_ coztzcouca Ema EmEumoE. 05
2:0 $63.05 @383 m5 .0 new m 3828 c. .933 53:53:: Ema _m_8mEEoo >Eo $63.05 29:3 9.: .v ncm m
2.828 c. .39-. so mo:_m>-a new mcofimmzmm. Ammmv 259% E8 £me ommsmficooom ho mEo_oEooo Son of: tone. 0 2
m 28:60 .xcmn EoEbmo>E cm m_ z a 0 new .xcmn _m_o_mEEoo m m_ 5:352: nmm. 9: t w m_ Smov ofimtg Eoncoaou
9: .F 5:28 c. .cozmnoo cozoofim 5.528;: 8 “mos-“ E mo:_m>-q new mmsmEzwo 505 9.: tone. 29.28 oz: “we: 9:.

.353 ace—53>... >5 ..o 9.53 338853
2: >3 cautious: mason :o acacia 232. £23 ecu 2.2.023 Saracens: ho fleas—Egon 2.: no moan—cam.
m... win...

54

 

 

 

9: :8 R2 80.8288 .0 .02

Ed 8d mm 88.2

2d 8-888

.. .. .. .. 8.: wood 38 5.98:0. :_ 8:88”.

.. .. . . dvd «odd- $80-$88 3.2.280

Ed 8d- 8d fidd- .. . 58. E8828 88888

odd 8.va odd $.de Ed mudd- 888 58:82:82

. .. .. . odd «and 8:9 :85

omcoaxm

ddd tdm 8d 8.8 . - .882. .. s8 .8: 8:88”.

55055

.. .. . .. 5d «ad 88:. :9: .. 38 3.58 858:8.

dd 8.2 odd 8.8 8.: Sod- oE8:. 8.8888888 .882.

33%?

5d Rt- 8d 5.8- Ed Ndmd- 5.86:: 5.8 3.8: 8:88.:

8.: Rd- odd 8.:- ddd find. .88. :o 8.38.

8d 8.8- 8.: 8.8- ad dood- Ssam .0 8.9, 8.8563

8d 3d- 8d 3d. 8d wood 388 32858:. 8:980

8d 8.8 8.: 3d 8.: Ned a ”:53.

Ed 8d- 8d 8d 2.: wood. Ed. :39... 8.8

odd 3.. 2d 8d dmd modd- 88. :o 3:22...

8d 8d 8.: 8d and mad- .2828: :58. 2.38 8 2.52.5

odd and 8d dmd- odd modd- 8522, 8.8.1.55

odd Ed 8d 8d- odd :3. e88 .9058“. .93

o:_m>-n_ mumEsmm mews-n. Bagsmw mag-n. 0662mm moi—28> Eoucoaons
3:8 .8282..QO 2:8 868558.85 :85

 

m6 035...

55

.:0:3.0:. 0:0
>9: .630... :0.:009 .0: 0:0 08:0. E0.0:00 ::0 00.883: >.:00> 9.. :0: 0908:00 0...- .:0:08:00 .50.: 08.0.9: 9..
80:: 00:08:00 0:0 0000.: :0:00.00 03800080 9: 6:80 0. :00: 08:0. E08.03_:0 030:0mo::m 0800:. 05.9000
0. 02.0.9 00:00.0 .099:. ::0 E0: :80: 00:0::0m :0 8:0. :0..00:9:. 9: 0. 00:00.0 .099:_-.00: 80: 00:05:01
.0.:0.:0> 0:: :0 :0.:08 9: :0... :090. 0. 0800:. 0:..9000 9 02.0.9 00:00x0 .090.:. :0 .0>0. 0:. :. F 0. 03.0> 85.000
0:. .0_:0.:0> >883: 0 0. :0.00.::. .099:. :9... 00.3.9.8 o .50: E00 m::0.x0 00:05.0: 0. 0. 0300. :0 000930 9: :.
F 0. .0... 0.00.:0> >883: 0 0. .00: :80: 00:0:_..0m 0.0000 .90. >0 :0:.>.: >..300 :0 03.0> .0808 03.: .00: :0 03.0> 0.00:
9: >0 :0:.-.0: 0. 0 0.500.- .00.2:9:0 o .:0.0: .0: :0 30.9. :0 0m. 0. 0300. ::00 9: :0: m::9 ..:90 0.>:00.2 9: :. F 0.
0:9: E08.00>:.-:0z 0:9: E08.00>:.-:0: 0. .. :. .m 0. 03.0> 9: .00m :0 < 0. .. :. .N 0. 03.0> 9: .90. :0 00< 0. 0300. ::0:
9: :0: 0:..9 ..:90 0.>:00.2 9: :. F 0. 0::9 ..:90 .00300. :0...:.s:0::3 E0: E0888. 9: 2:0 00:20:. 0.0800 9: .o
::0 m 0:83.00 :. .00300. :0:.:2:0::3 E0: .0.0:08800 >_:0 00:20:. 0.0800 9: .v ::0 m 0:83.00 :. ..09-. :0 003.0?0
::0 0:0.00909 Amdm. 0:00:00 .500 0.00: 0m0.0-::0000 :0 0E0.0.::000 0.9. 9: :009 o 9 m 0:83.00 2:00 808.008.
:0 0. .. :. o ::0 2:00 .0.0:08800 0 0. 8.88083 :00. 9.. :. F 0. Emu: 0.:0.:0> 80::000: 9: .F :83.00 :. 80:03.00
:0..00.00 6.88083 :0 .00.-. :0 0020?: ::0 00:08:00 .590 9: :80: 0:83.00 02,. .9: 9.... 90:0..0: 85.8 cm:
0300. :0 0N.0 :0.:08 9: :9: 000. 0. 0300. :0 0~.0 9: :. .9.:08 ::0: 0300.-:50: 0F: 00 85:0: 0. .0808 0300. ._080 9. :-

.080: 808.008. :3 :o 0:80: 3.908800 0:. >0 00...:8083
0.80: :0 0:00:00 8.0.: 0.00.. 80 80:00.00 :0:_::s._0::3 :o 9:088:90: 05 :0 00:08.:0m ”.0808 0300. __08m
0... 030...

56

 

 

 

 

000 00: 000 08.020000 :0 .02

00.0 00.0 0: 00:00.00.

00.0 N0.00000:

- - - - - . .000 80:000. 0. 0000000:

- .. - . 00.0 90.0 2000-00000 0:00.500

00.0 00.0 00.0 000.1 00.0 000.0- 5:0. 0.0000200 0000000000

00.0 00.00: 00.0 00.00 0:.0 :000- 000:0 205005.002

- . - - 00.0 0:00 0000: :.00:0

00:00x0

00.0 9.00 0:.0 00.00 .. - 00:00:. - .000 .000 0000000:

5:00:05

- - - - 00.0 000.0 00:00:. 00.: - .000 .000 0000000:

00.0 00.0: 00.0 00.00: 00.0 00:0 00:80. 0000002000030 00:00:.

:0.:0.:0>

00.0 00.00- 0:.0 00.00- :00 000.0- 8:00.05. .000 0.000 0000000:

- - . - 00.0 000.0- .0000. :0 00.0.03

00.0 00.00- 00.0 0:. :0- 00.0 000.0- 00.000 :0 00.0.: 000503

00.0 00.0- 000 00.0 0:0 000.0- 0.0000 0.2.0880. 0:00:90

00.0 00.0: 00.0 00. :- 00.0 000.0 0 0.0.00:

:00 ::.0 00.0 00.0 00.0 0000. 3:00. 530:0 00.00

00.0 00.: 00.0 0:: 00.0 000.0 0000. :0 0.5.0.2

00.0 00.0 00.0 9.0: :00 000.0- 0.000.000 0:30: 2.00 :0 0...:0_0>

00.0 :0 00.0 9.0- 00.0 000.0- 00:20:, 00.00:: 2.0.00

00.0 :00 :00 9.0- 00.0 :00. 0.0000 0.20000 .06:

03_0>-n. 908:0m. 03_0>-n. 908.:0m. 03.0?0. 908.:0m 00.00.:0> ::0::000::.
0.000 2080900000 0.000 0.205.000.000 000::

 

m... 030:.

57

00020:. 0:0 >00. 0030.: 09:009
.0: 0:0 08:0: 80.0:00 0:0 00.883: >_:00> 9.. :0: 00:08:00 9. .- .:0..08.:00 :.00:0 009099.: 0... 80:: 00:08:00
0:0 0000.0 :0..00.00 03800008 9: .9800 0. 0003 08:9 808.03.:00 030:0000:m 0800:. 02.9000 0: 02.0.0:
00:00x0 .0908. 0:0 .000 800 00:02:00. :0 8:0. 8:00:08. 9: 0. 00:00.8 8998.800 .800 00:08:00 .0.00.:0>
0.0. :0 :0.:08 9.. :0... :090. 0. 0800:. 0:..9000 0. 02.0.0: 00:00x0 .0908. :0 .08. 0:. :. F 0. 03.0> 00:0.000 9:
.0.00.:0> >883: 0 0. :900.0:. .0908. 00.: 00.8080 0 ..000 .800 m::0.x0 00:0:_:0: 9 0. 0300. :0 0000:30 9.. :. F 0.
:0:. 0.00.:0> >883: 0 0. .000 .800 00:05:00. .0.0000 .90. >0 000.>.0 >:.300 :0 03.0> .0808 03.0 .000 :0 03.0> 0.000 9:
>0 00::00 0. .0 08.00... 00.8080 0 .099 .0: :0 30.00 :0 0m 0. 0300. 0:00 9.. :0: 08:0: :.00:0 0.>000$. 9.. :. F 0. 009m
.:08:00>:.-:0z .0090 .:08:00>:.-:0: 0. .. :. .m 0. 03.0> 9.. .00m :0 < 0. :. :. .N 0. 03.0> 0:. .0< :0 00< 0. 0300. 0:00 9:
:0: 0:..9 :.00:0 0.>000.>. 0:. :. F 0. 02.0: :.00:0 .00300. :0:..:>.:00:3 .800 808.008. 9.. >80 00020:. 0.0800 9: .0 0:0
0 0:83.00 :_ .00300. :0:.:2:00:3 800 .0.0:08800 >.:0 00020:. 0.0800 0:: .0 0:0 0 0:83.00 :_ :00.-. :0 003.0>-0 0:0
0:0.00909 :w0m. 0:00:00 8.00 0.000 0090-08000 :0 080.0508 0.00 9.. 8009 o 0. 0 0:83.00 2:00 808.008. :0 0.
.. :. 0 0:0 .800 .0.0:08800 0 0. :0..:800:3 000. 9: :. F 0. Emu. 0.00.:0> 800:0000 9: .F :8200 :. 80:03.00 :0..00.00
:0..:>>:00:3 :0 .09-. :0 003.0>-0 0:0 00.08.:00 :.00:0 0:. 8000: 0:83.00 02:. :0:... 0:... .9080 :0....8 003 0300. :0
00.0 :0.:08 9: 0. .0300 :0 :0... :9090 0. 0300. :0 00.0 9: :. .0808 0:00-2.0: 0:. 00 00::00 0. .0808 0300. 00:0. 00.:

.0800 808.008. >0 :0 0800 .0.0:08800 00. >0 88:800....
00:00 :0 0:00:00 8.00 0.000 0:0 0:980:00 :0:.:3:00:3 :0 0800.80.00 0... :0 00:08:0m 30808 0:00. 090..
h... 0.00...

58

 

 

000 .0 NE 08002080 :0 .02

0:0 2.0 0.: 00:00.3:

:0 N0.0.0.0000

- .. - .. 00.0 000.0 .000 5.0.002 c. 8.00000

- - .. - 3.0 000.0 3000-90000 0.2.0000

- - - - 2.0 000.0 .50: 205.0000 88000000
00.0 00.03 00.0 00.00. 00.0 000.0. 000:0 505002802
- - - - 00.0 0.0.0 08:0: :.00:0
00:00x0

00.0 00.00 0.0 00.00. - .. 000.5 - .000 0.00.. 080800
5.00.0...

.. .. .. - 00.0 {0.0 000.0. 00.: - .000 0.000 00:00:00
00.0 00.00 000 0:00- :00 000.0- 058:. 05.008.00.090 .0003.
.0.00.:0>

.00 00.0- «.0 00.0. 00.0 00.0- 5:00.05. .000 .000 00:00:00
.00 00.0 8.0 00.00. 3.0 000.0- .0000. :o 00.0.03
00.0 00.00- 00.0 00.21 00.0 03.0- .0500 :0 020:, 00.05.03
.00 :.v. .- 000 .0. 7 00.0 000.0 0.0000 0.20580. 05.0000
.00 00.0 00.0 3.00 00.0 30.0- 0 0.50.:
00.0 2.0- 00.0 00.0 00.0 000.0 2:00. 530:0 0200
00.0 K... 0:0 00.. 00.0 0.0.0- 0:00. :0 0.5.05.
00.0 00.0 00.0 00. 7 0.0 000.0- 0.000.000 530: 0.000 :0 0....0_o>
00.0 00.0 00.0 00.0 «0.0 000.0- 0522, 00.00:: 0.000
00.0 000 00.0 00.0- 00.0 000.0- 0.0000 0.05000 0.0»
03.0>-n. 0:08.:0m. 03.0>-0 0:08.:0m. 03.0>-n. 0.08..0m. 00.00.:0> .:0::0000:.

 

0.000 2060080000

0.000 0002500000

5... 0.00...

:.00:0

59

Table 1.8
Gross benefit for firm using the service of chosen underwriter rather than
that of unchosen underwriter.

The gross benefit is defined as the expected mean net yield (BPS) of bond
issues, had they chosen the unselected underwriter minus the observed net yield
of bond issues. The benefit estimated in column 1 (below the heading of
Commercial Bank Clients) is the predicted BPS of commercial bank client by
using the estimates of investment bank BPS regression reported in Table 1.5
minus the original BPS of commercial bank client. Hence, the benefit in column 3
is estimated in the same way. The p-value of t-test is reported, the tested null
hypothesis is the gross benefit is equal to zero.

Panel A. Gross benefit of full sample and by subperiods.
Commercial Bank Clients Investment Bank Clients

 

 

Years Benefit P-value Benifit P-value
1991-1996 2.43 0.63 164.62 0.00
No. of observations 184 1143

1991-1993 -0.65 0.91 171.31 0.00
No. of observations 66 664

1994-1996 4.15 0.56 155.35 0.00
No. of observations 118 479

 

Panel B. Gross benefit by credit ratings
Commercial Bank Clients Investment Bank Clients

 

 

Credit ratings Benefit P-value Benefit P-value
Aaa or Aa -16.22 0.02 181.08 0.00
No. of observations 14 173 .
A or Baa 11.98 0.00 156.88 0.00
No. of observations 136 757

Ba or below -28.08 0.26 178.79 0.00
No. of observations 34 213

 

Panel C. Gross benefit by the issue size (in million dollars)
Commercial Bank Clients Investment Bank Clients

 

 

Issue Size Benefit P-value Benefit P-value
>= 150 -3.7 0.74 170.79 0.00
No. of observations 54 648

< 150 4.98 0.36 156.55 0.00
No. of observations 130 495

 

60

 

 

 

hm v. an N. mcozmzmmno ho .02

and mod? ..o.o mmdm cod and? mad 3.91 26.8 .0 mm

mmm 2 SF vm macsmzomno .0 .oz

006 omhm- 3.0 NNdP cod mméor mNd mad mam .5 <

mm v mm v mcostmmno “—0 .02

mod 3...” :.o 3.9 00.0 mwem omd N...NN- m< .o mm<

Fov Pm mom mv mcoszmmno .0 .oz

cod omém- cod mmém cod m v .8? vmd om? 3:29 __<

m:_m>-n. Eocmm m:.m>-n. Emcmm o:_m>-n. Eocom o:_m>-n. Eocom 3:33. .690
9:95 3:0..0 3:25 3:25

.53 E9532: .53 3.056800

xcmm 2.958%. .53 3.055800

 

$me 2.3. 35..

reins. 25m. =me

 

.32 2 Saw actsn macaw. «:38 .3 «cocoa 39.0 .< .25“.

.93 o. :33 m. Emcee 306 oz. m. £8509... =3: no.8. on. .notoao. m. 69.. .6 m:.m>-a on... .notoam.

.0: 9m 20.893. mam 32% 9.2.8 co__..E 0...... o. .33 .0 cm... .235 9:2. mum 5.3 333. 3.3.2... $.me 03mm.
mmcm. on... .33 06mm 9: c. noumEzmo 9m m o. m 9.828 :. mEocmn m5 .ooco... .Em_.o xcmn 6.96668 .0 mam 3598
on. 95.8 $5.8 8...? cm? cm... $9. 6me 38. =me 5,. 9.392 56369 69 29m... :. cotoao. .86QO9

mam xcmn E68695 kB $66.60 9: 9%: >9 Em=o xcmn 3.05888 .0 mam 8.0.8:. o... m. .3535 xcmm _m_o.oEEoo
.o 9.699. o5 26.3. r :828 5 868.60 50:2. 9: .333. econ .6 Be... am: 82830 on. 35:. 53.53::
62023:: on. 5on >9: no: .333. neon .o .95. 2o... .0: come vouooaxo 9: mm noccou m. Emcee $20 on...

anotoansm .3 3qu 03mm. 3.5. can 3qu came.
=2.» .5. Butane—cc: 535:: .o as: 55 .352 3553:: nonoso .o 3.23 05 9:»: SE so. «cocoa «no.0

m... 2an

61

 

00 I.
_.0.0 50.00 0N0 V0.0-

v0w 0N
P00 dev 00.0 mmém

0N N

n00 0V0? 00.0 50.3-

tum mm
00.0 00.3 050 00.0..
03.970 2.330 03.0>-n. E030

Nv NP

5.0 03K .30 00.00
00? 00

00.0 5.00 00.0 00.0w
hm v

00.0 00.50 #00 5.90-

NmN 00

00.0 5.: 00.0 mméw
03.03. Emcmm 03.030 50:00

 

06:0 38:0
.30 EoEummé. xcmm .0.03EE00

35:0 98:0
.30 .3536. 0.500 .0.03EE00

 

30.6.2 033. 040.00..

30:05. 0:00. .680

30.620000 6 .02
26.00 6 00
30.002030 .0 .oz
30 .0 <
30.620000 6 .oz
m< .o mm<

30.620000 6 .oz
85.9 __<
00:30.. 0.00.0

 

.83 o. 32 0550 00:02 0.09.0 .3 Enema 320 .m .25.

.0280. 0.. 03¢

62

ESSAY; C EMPEtTITIQN AND COALITION AMONG UNDERWRITERS:

THE ENTRY STRATEGY OF CQMMERCIAL BANK§ IN 115
POST-SECTION 20 ERA

2.1 Introduction

During the recent debate over whether or not to repeal the Glass-Steagall

Act, investment banks have been the major opponents of the repeal. Although
investment banks have competed fiercely with commercial banks in the
legislative arena, commercial bank entry into security undenNritings has been
accommodated to some extent. Commercial banks have enteredthis market as
co-managers through participation in investment bank lead syndicates. This
particular undenivriting form accounted for only 10 percent of corporate bond
issues in 1991, but soared to 38 percent by 1996. During the same period,
commercial bank-lead syndicates underwriting bond issues increased from 4.2
percent to 18.7 percent. It is puzzling as to why commercial banks have not
competed for the lead manager roles more aggressively, having cooperated with
investment banks twice as much as they have competed. Even more puzzling is
why investment banks have not more actively deterred entry by excluding
commercial banks from participating in syndicates.

Including issues co-managed with investment banks, commercial banks
have participated in 56.4 percent of the new bond issues in 1996. Although prior

research has examined the underwriting activities of commercial banks when

63

they have led underwriting syndicates”, the role of commercial banks as junior
co-managers has yet to be examined. In this paper, we investigate the economic
rationales driving coalitions between commercial and investment banks. To the
best of our knowledge, this is the first paper examining the economic functions
and effects of commercial banks as co-managers in security underwritings.

Financial markets are often characterized as possessing an oligopolistic
structure. Anand and Galetovic (1997) show that interrnediation markets are
natural oligopolies, with entry being accommodated as long as cooperation is
feasible. In this type of market, it may be beneficial for firms of various abilities to
collude in the market. For security underwritings, it is a common practice to form
syndicates among underwriters“. We would like to examine whether the
coalitions among commercial banks and investment banks are driven
predominantly by the industry market structure or some potential benefits of
cooperation amongst underwriters.

lf coalitions are driven predominantly by the oligopolistic market structure,
i.e. if commercial banks act as perfect substitutes for investment banks in the
syndicates, we should not observe any systematic pattern of coalition formation.
Thus, the new form of underwriting activities would not differ from the existing
investment bank lead-syndicates that commercial banks do not participate in.

Following this line of reasoning, the choice of entry mode of commercial banks,

 

33 See Ang and Richardson (1994), Puri (1994), Puri (1996), Gande et al (1997), Song et
al (1999) for empirical results in this area.

3‘ As stated in Eccles and Crane (1988, p.92), "The most visible ties among investment
banks are the syndicates formed to underwrite and distribute a security offering." They
report that the top six investment banks lead-managed 6,327 domestic security issues
(excluding tax-exempt issues) from 1984 to 1986, over sixty percent of these deals were

64

either lead- or co-managing the syndicates, should not depend on the
comparative advantages of commercial banks in different segments of the
market. The percentage of commercial bank lead-managed issues relative to that
of co-managed issues should only depend on the overall market power of
commercial banks.

However, Song (1999) find that commercial banks and investment banks
represent two distinct types of underwriters. Commercial banks possess
informational advantages over investment banks due to their loan monitoring and
transactional activities with clients (Fama, 1985). Association with commercial
banks also provides clients with better financing opportunities in the future (See,
for example, Holland, 1994; James 1987). However, these benefits do come with
costs. The dual role of commercial banks as both lenders and underwriters
creates the potential for conflicts of interest. For instance, a commercial bank
may misrepresent information about its clients by overstating clients' prospects.
They could then underwrite these "bad" loans, using the proceeds to repay bank
debt (See Kroszner and Rajan, 1997; Gande et al, 1997). Additionally, the same
informational advantages may give commercial banks a temporary information
monopoly over the clients' private information, allowing banks to appropriate
firrns' future profits (Rajan, 1992; Sharpe, 1990; Greenbaum et al, 1989).

In addition to the above differences between commercial and investment
banks, Song (1999) find that the ability of commercial banks in distributing large

issues is hindered by a revenue limitation35 imposed only upon them. This

 

co—managed by another top-six investment banks.
35 The revenue limitation was set at 5 percent in 1987, was increased to 10% in 1989,

65

regulatory restriction constrains the amount of corporate issues commercial
banks can underwrite by limiting the revenues generated from certain in-eligible
security activities to 10 percent of the total revenues of section-20 subsidiaries36.
Combining these two types of undemriters with complementary abilities
could create better underwriting services. A similar positive effect is identified by
Mohanram and Nanda (1998) in their joint ventures study. They find that the
stock market reacts positively to joint ventures that involve pooling of
complementary resources. In this paper, we address the question of whether
coalitions between commercial and investment banks are motivated by economic
incentives. We do so by testing the hypothesis that coalition arrangements of
commercial and investment banks mitigate disadvantages of both types, yielding
higher bond prices for their clients. In other words, we argue that the cooperation
is motivated by mutual benefits. When an investment bank is the lead underwriter
and a commercial bank participates as a co-manager, we define this new
structure as a Hybrid syndicate. The syndicates led by commercial banks are
defined as CB-Lead syndicates”. Those coalitions that do not include any

commercial banks are referred to as Pure/B syndicates.

 

and was raised again to 25 percent in 1996. See Federal Reserve System, Docket No.
R-0841, "Revenue Limit on Bank-Ineligible Activities of Subsidiaries of Bank Holding
Companies Engaged in Underwriting and Dealing in Securities."

36 See JP Morgan & Co. Inc, The Chase Manhattan Corp., Bankers Trust New York
Corp., Citicorp, and Security Pacific Corp., 75 Federal Reserve Bulletin 192 (1989)
(hereafter "1989 Order") for discussions on the regulatory environment that commercial
banks are allowed to underwrite corporate securities. The section 20 subsidiary is the
organizational form in which commercial banks are permitted to conduct the ineligible
security activities. Ineligible activities refer to those not permitted before the relaxation of
Glass-Steagall Act in 1989.

37 Commercial bank lead syndicates may or may not have other investment banks
participate as co—managers. Although commercial banks may also cooperate with other
commercial banks as a fourth type, the sample size representing only 2.2% of the bond

66

When testing the above hypothesis, we also focus on the role that lead
managers play in alleviating the moral hazard problem, i.e. the potential conflicts
of interest inherent in commercial bank security underwritings. The literature
provides both theoretical and empirical support for this problem. Pichler and
Wilhelm (1998) provide a monitoring role for the lead undenrvriter to improve
incentives and mitigate the moral hazard problem of syndicate participants.
Nanda and Yun (1997) find that lead managers share a larger proportion of the
damage to reputation arising from an unsuccessful offering. Jain and Kini (1999)
find evidence consistent with the demand for lead underwriter monitoring in initial
public offerings. In the context of this paper, investment banks are not subject to
the aforementioned conflict of interest problem faced by commercial banks. They
should be more credible in certifying security issues when the purpose of an
issue is to repay bank debt and when the level of interest expense is relatively
high. Thus, we test whether the reputational concerns of an investment bank will
alleviate concerns by the market for conflicts of interest. We propose and find
evidence that the Hybrid syndicate serves this purpose.

Despite disadvantages it has in underwriting for large issues and for issues
with higher potential for conflicts of interest problem, commercial banks are found
to possess better abilities in serving clients with certain attributes, such as firms
with middle credit ratings38 (Song, 1999). It would be implausible for commercial

banks to not take advantage of their superior abilities when they expand into

 

issues, is too small to conduct multivariate analysis. For this analysis, the data is lumped
in with other commercial bank lead syndicates. Thus, we focus our discussions on the
three types of syndicates mentioned in the text.

3" The issues with middle credit ratings are those with Moody's ratings of either A or Baa.

67

these segments of the market, given that lead managers gain a larger share of
the proceeds. Thus, we hypothesize that commercial banks strategically select
their entry mode - leading the syndicate or participating as a co-manager - based
on their strengths in different segments of the market.

Using a sample of 1327 corporate bond issues from 1991 to 1996, we
examine how issuing firms select among these three syndicates and how hybrid
syndicates are different from CB-Lead and PurelB syndicates in terms of their
underwriting abilities. We also analyze the performance of syndicates by
subtracting the net yields of bond issues that selected syndicates obtain for their
clients from predicted yields of unselected syndicates for the same clients. We
find that in circumstances when the coalition arrangement is less beneficial to the
commercial banks, they behave more aggressively in the market, having a higher
tendency of choosing to lead a syndicate rather than to just participate in it. In the
small issue market, where commercial banks display less distributional
disadvantages, commercial banks lead-managed three times as many bond
issues as they underwrote in the large issue market. They cooperate with
another lead investment bank half as often in the small issue market as they do
in the large issue market. Similar aggressive behavior is observed when
commercial banks underwrite for firms with middle quality. They lead-underwrite
a higher percentage of bond issues in the middle quality segment than in the high
and low segments. These results support the hypothesis that commercial banks

strategically select their entry modes.

68

We also find that the hybrid syndicate is distinctly different from both CB-
Lead syndicates and PurelB syndicates on many dimensions. Clients of hybrid
syndicates tend to be large issuers but not large firms. Hybrid clients are more
profitable, have lower leverage, and are less likely to have an informational
asymmetry problem. The purpose of their issues is more likely to be refinancing
bank debt.

The superiority of a hybrid syndicate is demonstrated through its ability to
achieve higher bond prices for different clientele. Most of the beta coefficients in
the net yield equation for hybrid clients fall between those of CB-Lead and
PurelB syndicates. These estimates tend to fall closer to estimates of one or the
other type of syndicate. The hybrid coefficient estimates on leverage support the
informational advantage argument due to the involvement of commercial banks.
Meanwhile, the involvement of investment banks in the hybrid syndicate alleviate
part of the informational monopoly problem inherent in CB-Lead syndicates, as
demonstrated by the estimates on sales growth and operating income being
closer to those of PurelB syndicates.

In addition, there are synergies with regards to volatility of equity return and
Tobin's q. The hybrid estimates of the net yield equation for these two variables
are the smallest of any group. Hybrid estimates of the net yield equation on
equity trading volume and firm size resemble the averages of the other two
parties. However, the hybrid syndicate performs the worst in the area of maturity

of issue, with the estimate being the largest among the syndicates. Although both

69

CB-Lead and PurelB syndicates demonstrate cherry-picking ability39 over hybrid
syndicates, there is no endogenous selection bias identified for hybrid
syndicates. One possible explanation is that hybrid clients are the group with less
of an informational problem. Thus, the identity of a syndicate can not provide
additional value when all exogenous characteristics of firms have been
considered.

Overall, we find that the hybrid structure can preserve the advantages of
commercial banks while mitigating many of their disadvantages. By cooperating
with an investment bank, which lead-manages the bond issue, potential conflicts
of interest of commercial banks can be reduced. The reputation of an investment
bank and the more remote distance between lending and underwriting activities
make the certification of the hybrid form more credible. Therefore, hybrid client
firms that issue bonds for the purpose of refinancing bank debt will suffer less of
a price discount than commercial bank-lead customers will.

We provide evidence that the hybrid structure is a superior organizational
form for many firms under current regulatory and market conditions. However,
the existence of client switching costs enables undenrvriters to have temporary
monopoly power over their clients. Not all of the customers have the privilege of
enjoying better syndicate services. Consistent with the characteristics of client
firms mentioned before, hybrid clients tend to have less informational asymmetry

problems, and, thus, lower switching costs.

 

3" The cherry-picking ability refers to the informational advantage or unobserved ability of
underwriters possess in that they are able to select clients they can serve better among
firms with the same public known characteristics.

70

Our analysis of gross benefits, in terms of net yield differentials of bond
issues, achieved by a chosen syndicate versus an unchosen syndicate supports
the previous discussion. In the eyes of PurelB clients and most of the CB-Lead
clients, the hybrid structure is a preferred choice. However, due to the existence
of switching costs, these firms are still retained in less favorable syndicates.
Among the three segments of issuing firms - high, middle, and low quality40 -
middle quality firms that are lead managed by commercial banks are the only
group which do not prefer the hybrid syndicate. Middle quality firms are better off
staying with a CB-Lead syndicate than switching to the hybrid form.

The remainder of the paper is organized as follows. Section 2.2 presents a
review of the literature on entry of commercial banks into different financial
markets and coalitions among underwriters. Section 2.3 describes the data and
summary statistics. Section 2.4 discusses the entry strategies of commercial
banks. Section 2.5 describes the characteristics of Hybrid syndicates in

comparison to the other two syndicates. Finally, Section 2.6 concludes the paper.

 

‘° High quality firms are those with Moody’s credit ratings of Aaa and Aa; middle quality
firms are those with A and Baa ratings; low quality firms refer to those with Ba or below.

71

2.2 The Literature on Bank Entry Strategies and Underwriting Coalitions

Deregulation in the banking industry has provided numerous opportunities
for commercial banks to expand their scope of operations. Bhargava and Fraser
(1998) find positive abnormal returns for commercial banks when the Federal
Reserve Bank began authorizing commercial banks to provide limited forms of
security undenrvritings“. However, Bhargava and Fraser found that market
reaction to increased securities undenrvriting powers in 1989 was negative”.
Nonetheless, the expansion of commercial banks into security underwritings has
grown rapidly since 1987. As of February 1, 1999, there were 51 section 20
subsidiaries. Among them, 41 were also authorized to conduct Tier II powers,
while 2 were limited to underwriting corporate debt issues“.

Boot et al (1998) propose a theoretical framework to explain the
expansion of scale and scope in banking as an optimal strategy. They do so in
an environment of "strategic future skills uncertainty", where it is not known
whether the bank has the skills to compete effectively in new markets. In
contrast to merging with an investment bank, adding a section 20 subsidiary
allows commercial banks to acquire superior information regarding its

underwriting ability and to ascertain whether it has the skills to compete in the

 

‘" In 1987, three bank holding companies were granted the power to underwrite
municipal revenue bonds, private mortgage-backed securities, commercial paper and
consumer—receivable-related securities. These powers have come to be known as "Tier I
section 20 powers". See Citicorp, JP. Morgan & Co., and Bankers Trust New York Corp,
73 Federal Reserve Bulletin 473 (1987)(hereafter "1987 Order").

‘2 The corporate security underwriting powers allowed in the 1989 Order have become
known as "Tier II section 20 powers."

‘3 For the list of section 20 subsidiaries, see "Section 20 Secruities Subsidiaries."

72

new business without an excessive up-front investment. However, due to current
regulations, adding a section 20 subsidiary is the only feasible method for banks
to enter the security underwriting business“.

Given that commercial banks have entered the security undenrvriting
business, they face two choices - competing or cooperating with investment
banks. Since lead-managers may grab a large share of the underwriting
compensation and build reputation for future underwriting business, it seems
reasonable to assume that commercial banks would compete more aggressively
in acquiring the lead-underwriter role when they possess stronger market power.
Conversely, when they are a smaller player in a given market, it seems rational
for them to cooperate with others who may complement their weakness. In the
following paragraphs, we explore possible circumstances under which
commercial banks may need to cooperate rather than to compete.

Kanatas and Qi (1998a) compare the level of information production and
the amount of a client firm's investment in specialized versus integrated
institutions. They find that the differences depend on the extent of borrower moral
hazard, and the size of information costs. In related work, Kanatas and Qi
(1998b) demonstrate that the incentive conflict of intermediaries who both lend
and underwrite securities may impose a cost on their customers seeking to raise
capital and that the regulatory separation of lending and underwriting may be
optimal. They also find that when both the opportunity cost of the incentive

conflict and the savings from economies of scope are sufficiently large, there is a

 

(Onlinel Available rmJ/www.bog.frb.fed.us/qenerglinfo/sectionZOI, February 25, 1999.
For a review of the consolidation of the financial services industry, see Berger et al

73

role for reputation building in mitigating the incentive conflict problem. This can
be achieved without the need for regulation.

Another possible mechanism commercial banks may use to enhance their
credibility is the choice of an organizational form that distances the operations of
lending and underwriting. Kroszner and Rajan (1997) study the underwriting
activities of commercial banks before the Glass-Steagall Act. In the absence of
other distortions, their results suggest that in comparison to the in-house
department underwriting, the organizational structure of a separate affiliate is an
effective commitment mechanism to address the conflicts of interest problem.
They argue that market pressures force banks to adopt an internal structure that
might mitigate these concerns. In a similar study by Gompers and Lerner (1998),
the authors examine the underwriting of initial public offerings by investment
banks that hold equity in a firm through a venture capital subsidiary. Consistent
with the rational discounting hypothesis, investors require a greater discount at
the time of the offering to compensate them for potential adverse selection. The
investment bank-affiliated venture firms address the potential conflict of interest
by investing in and subsequently underwriting less information-sensitive issues.

Similar issues regarding the choice of organizational forms in dealing with
different types of agency costs in the property-liability insurance industry are
investigated by Cummins et al (1997). Their results support the hypothesis that
the market will sort organizational forms into market segments where they have
comparative advantages in minimizing the costs of production. By the same

token, commercial banks may enter the security underwriting business by

 

(1 999).
74

cooperating with a reputable investment bank to enlarge the distance between
underwriting and lending when the concern for conflicts of interest perceived by
the market is serious.

Pichler and Wilhelm (1998) provide a model explaining the economic
functions of the syndicate organizational form. This form provides a monitoring
role for the lead undenrvriter that improves incentives and mitigates the moral
hazard problem. It is through the delegation of additional powers and, thus,
additional responsibilities that the lead banker is effectively forced to put at risk
the reputational capital on which its future earnings power depends. Nanda and
Yun (1997) find that lead managers share a larger proportion of the damage to
reputation arising from an unsuccessful offering. Jain and Kini (1999) find
evidence consistent with the demand for lead underwriter monitoring in initial
public offerings (lPOs). The post-issue performance of IPO firms is positively
associated with the lead bank reputation.

In relation to our study, an investment bank without a conflict of interest
problem is more credible in certifying a security issue when the purpose of the
issue is to repay bank debt. It may pay for a commercial bank to participate in a
syndicate as a co-manager and to allow a reputable investment bank to refinance
the client's bank debt. By putting its reputation capital on the line, the lead
investment bank may capture a larger share of the proceeds, and, thus, increase
the probability of the bank leading future offerings or participating in syndicates

that it does not lead.

75

The discussions in the previous paragraphs demonstrate that there are
strong theoretical grounds for positive economic effects of commercial bank entry
into the security underwriting business. There are two factors that may induce
commercial banks to cooperate with investment banks in securities
undenNritings. First, the aforementioned Federal Reserve revenue limitation
hinders commercial banks from underwriting large issues. Second, investment
banks have much more experience at these underwritings, putting commercial
banks at a competitive disadvantage vis-a-vis investment banks. While the
reasons for commercial banks to cooperate with investment banks are apparent,
the reverse may not be so obvious. In the remainder of this section, we review
the literature regarding the economic functions of syndicates and the possible
reasons for coalitions among undenrvriters in the security underwriting business.

Anand and Galetovic (1998) provide an explanation for oligopolistic behavior
in the investment banking industry when property rights over information are
weak. They study the effect of entry on equilibrium market structures and prices,
and show that collusion is robust to information free riding and entry; in contrast,
monopoly is not robust to either. When inputs are non-excludable, cooperation
between firms may be necessary for the market to exist. The authors also derive
ranges of the size and number of intermediaries in equilibrium. If intermediaries
are too small, then incentives to free ride become too large. Conversely,
intermediaries can not be too large, since they would reduce others' market
shares to the point that cooperation would no longer be sustainable. This

framework is applicable to entry of commercial banks into the security

76

underwriting market, since commercial banks can offer services similar to those
of investment banks.

Similar behavior of accommodating an entrant is observed in the discount
department stores industry. Khanna and Tice (1998) study the expansion pattern
of Wal-Mart, finding that the markets with high Herfindahl indices are the ones
WaI-Mart can penetrate faster. These markets are dominated by large firms that
tend to interact in multiple markets with Wal-Mart. Given this multi-market
interaction, there is less of an information problem, and, thus, once entry takes
place, incumbents are likely to cooperate with entrants.

Bhattacharyya and Nanda (1998) discuss similar issues regarding
cooperative arrangements among investment banks in the context of the
introduction of innovative financial products. In their study, they show that an
asymmetric distribution of market shares will tend to enhance innovation activity
not only because the large bank has greater incentives to innovate but also
because it is the natural partner for smaller banks in cooperative arrangements.
In our paper, commercial banks are much smaller than investment banks in
terms of market share. Alternatively, White (1996) argues that innovations can be
organizational innovations involving new ways of doing things. In addition, Song
(1999) find that commercial banks and investment banks possess differential
comparative advantages in underwriting. Therefore, the formation of hybrid
syndicates, which combine the complementary abilities of different undenrvriters,

may provide a new breed of underwriting services. This new way of underwriting

77

would have never been available without the entry of commercial banks and the
accommodation of investment banks.

Mohanram and Nanda (1998) study the stock market reaction to joint
venture announcements in an effort to identify which types of joint ventures
create value. They find that the stock market reacts positively to joint ventures
that pool complementary resources together. Analogous to our study, if the
cooperative arrangement between commercial banks and investment banks is a
valuable form of conducting security underwriting, the prices of bonds
undenrvritten by this hybrid form should be higher than if the same clients had
been underwritten by either a commercial bank lead syndicate or a pure
investment bank syndicate. Although the cooperation between commercial banks
and investment banks could have existed before the Glass-Steagall Act, we are
not aware of any study that investigates this interesting issue during the pre-

Glass—Steagall era.

78

2.3 Data and Summary Statistics

The sample consists of 1327 public debt offerings45 between January 1,
1991 and December 31, 1996. Data after 1996 is not included due to the
increase in the revenue limitation imposed on commercial bank underwriting
activities”. Financial companies and utilities are excluded from the sample.
Characteristics of the issues are obtained from the Security Data Corporation
(SDC) U.S. Debt New Issues Database, while firm characteristics are drawn from
Compustat and CRSP. In order to be included in our analysis, an issuing firm
must have data available on SDC, CRSP and Compustat. Definitions of variables
used in the analysis are listed in Appendix A.

The frequency and percentage of bond issues by year and underwriter types
during the period 1991 to 1996 are reported in Table 2.1, Panels A and B,
respectively. Over the six-year period, commercial banks lead managed 184
bond issues. Of the issues led by commercial banks, 155 issues did not have
other commercial banks involved as co-managers. These commercial bank lead
issues (commercial bank-lead syndicates) may or may not have had investment
banks involved. Twenty-nine issues did have other commercial banks involved as
co-managers (commercial bank-coop syndicates). There are no observations of

commercial bank-coop syndicates in our sample for 1991, 1992 or 1994. Ideally,

 

‘5 Equity issues are not included in the sample due to an insufficient number of
observations of commercial bank lead underwritings.

‘6 The revenue limitation was raised from 10 percent to 25 percent at the end of 1996.
See Federal Reserve System Docket No. R-0841, "Revenue Limit on Bank-Ineligible
Activities of Subsidiaries of Bank Holding Companies Engaged in Underwriting and
Dealing in Securities.”

79

we would like to analyze the commercial bank-coop category separately, but the
number of issues is too few for multivariate analysis. Thus, in this section, we
analyze these 29 issues as an independent category, but when we conduct the
multivariate analysis in next section, these issues are lumped in with other
commercial bank-lead-managed issues. In the next section we relabel this new
category as CB-Lead syndicate. Additionally, there are 345 issues underwritten
by the hybrid syndicate form. The hybrid form is the syndicate organizational
form where an investment bank is the lead manager while a commercial bank is
a co-manager. Finally, 798 issues are undertaken by pure investment bank
syndicates (PurelB).

Over the sample period, the hybrid syndicate form has increased
dramatically from underwriting 9.8 percent of bond issues in 1991 to 37.7 percent
in 1996. This increase is at the expense of pure investment bank syndicates, who
underwrote 86 percent of bond issues in 1991, but only 43.6 percent in 1996.
Although the number of commercial bank-lead managed issues increased over
the sample period regardless of whether other commercial banks were involved
as co-managers, the percent of the total issues in the sample is relatively small.
In fact, commercial bank-lead issues as a percent of total issues actually
decreased in 1996.

The results indicate that commercial banks enter the market under two
possible forms - as lead underwriters, or as co-managers. The latter form
occurred twice as often as the former did. Thus, it appears that investment banks

choose to accommodate entry rather than to deter it. One interesting question is

80

if a commercial bank is just another investment bank. Therefore, the observed
accommodation of entry is the process of a natural oligopolistic market structure.
Or there are special characteristics of commercial banks that cause them to stay
competitive as lead undenrvriters when they possess advantages over investment
banks and to ally themselves with investment banks as co-managers when they
have weakness in certain markets.

The descriptive statistics reported in Table 2.2 demonstrate that commercial
bank-lead syndicates underwrite the smallest issues in terms of the mean and
median of issue size. Conversely, hybrid syndicates underwrite the largest
issues. There is no observable difference of issue size between pure investment
bank syndicates and commercial bank-coop syndicates. We do not observe the
same pattern among different syndicates in terms of issuing firm size. Pure
investment bank syndicates underwrite the largest mean and median size firms,
while commercial bank-coops underwrite the smallest firms. Both commercial
bank-lead and hybrid syndicates underwrite for middle size firms. The former
underwrites relatively smaller firms than the latter on average, but there is no
difference in median statistics. Thus, although hybrid syndicates underwrite the
largest issue size, they do not serve the largest clients. In fact, pure investment
bank syndicates serve the largest client firms. The smallest client firms are
assisted by commercial bank-coops but their issue sizes are significantly larger
than the issue sizes of the commercial bank-lead clients.

The net yields, or basis points spread (BPS), of bond issues highlight

another pattern. While the bond issues of commercial bank-lead clients have the

81

lowest BPS, those of commercial bank-coops have the highest BPS. Although
the BPS of commercial bank-coop underwritten issues is significantly higher than
that of commercial bank-lead underwritten issues, it is not significantly different
from those of hybrid or pure investment bank underwritten issues. The latter two
syndicate forms do not differ from each other in terms of mean and median of
BPS, but the BPS of their client bond issues are both significantly larger than that
of commercial bank-lead syndicates. Due to the limited nature of univariate

analysis, we will analyze the interesting patterns observed in this section further

via multivariate analysis in Section 2.5.

82

2.4 The Entry Strategies of Commercial Banks

In this section, we continue our univariate analysis to identify systematic
differences between syndicate forms. We then proceed to multivariate analysis in
the next section. Table 2.3 displays the distributions of issue credit ratings by
different types of syndicates. Regardless of the pairs we compare, all of the
distributions are significantly different from each other. In the high quality
segment, where the Moody's credit ratings are either Aaa or Aa, the dominant
players are the hybrid and pure investment bank syndicates. The BPS of bond
issues do not differ between these two syndicates nor among all syndicate types.
The commercial bank-lead syndicates only underwrite 14 out of 187 (7.5%) high
quality issues, indicating that they can not penetrate this particular market very
well as a lead underwriter. There are 46 out of 187 (24.6%) high quality issues in
which commercial banks participate as co-managers, which is slightly lower than
for the overall sample (26%) when the market is not segmented by credit ratings.
No issues in this segment are underwritten by commercial bank-coop syndicates.

In the middle quality category, where the Moody's credit ratings are either A
or Baa, commercial bank-lead syndicates are able to lower BPS significantly
more for their clients than hybrid or pure investment bank syndicates are able to.
Commercial bank-lead syndicates underwrite 13.4% of these issues. This
percentage is almost double that of the high quality firm segment. In fact, if we
add the issues underwritten by commercial bank-coops, the number more than

doubles. Among all the commercial bank-lead underwritten issues, 77.5 percent

83

of them are middle-rated firms, while for the hybrid underwritten issues and pure
investment bank underwritten issues, the percentages are 62.4 and 67.9,
respectively.

For those with B credit ratings, hybrid syndicates and commercial bank-coop
syndicates show a comparative advantage over commercial bank-lead and pure
investment bank syndicates in lowering BPS for their clients. In these low quality
segments, i.e. Ba and below, the percentages of issues underwritten by hybrid
and commercial bank-coop syndicates are 34 and 5.26, respectively. The
percentage of issues underwritten by commercial bank-coops in the middle-rated
category (1.79%), is one third that of the non-investment grade category.
Conversely, hybrid syndicates underwrite 10 percent more issues in the non-
investment grade segment than in the investment grade category.

The figures in Table 2.3 demonstrate a pattern of systematic selection of
different syndicates by client firms. Investment banks continue to be the
dominant players in the high quality segment. Commercial banks appear to be
able to penetrate this market by cooperating with a lead investment bank.
Commercial banks also lead-manage a larger proportion of the middle quality
segment, where they show a comparative advantage in lowering the net yield of
their clients' bond issues. In the non-investment grade market, where commercial
bank-lead syndicates do not show a distinct advantage, cooperative
arrangements with either investment banks or another commercial bank are the
dominant strategies of entry for commercial banks. The specializations and

different forms of entry by commercial banks indicate that a commercial bank is

84

not another "pure investment bank". If there were no difference between
commercial banks and investment banks, we should not observe these
systematic patterns of entry and selection.

In Table 2.4, we show the percentages of bonds underwritten by different
types of syndicates in the small and large issue markets. These markets are
segmented by the median size of issues of the entire sample. Thus, the small
issue market contains issues that are less than 150 million dollars, while the
large issue market has issues that are greater than or equal to 150 million
dollars. In the small issue market, the percentage of commercial bank lead-
managed bond issues is 18.6 percent on average, which is more than triple that
of bond issues lead-managed by commercial banks in the large issue market.
Conversely, the percentage of hybrid managed bond issues is on average 17.6
percent in the small issue market, while the percentage doubles for the large
issue market.

The above pattern of entry into different issue size markets selected by
commercial banks is consistent with the findings in Song (1999) that commercial
banks show a disadvantage in handling large issues. If we combine the
percentage of bond issues of hybrid syndicates and commercial bank lead
syndicates, we find that commercial banks are actively engaging in the large
issue market, but mainly by cooperating with an investment bank that may have
a better distributional network and is not restricted by the 10 percent revenue

limitation imposed on commercial banks.

85

Effective on November 12, 1996, commercial banks are allowed to exclude
the interest earned on the ineligible securities from the 10 percent revenue limit".
This amendment would decrease the quarterly ineligible revenue of some section
20 subsidiaries by a magnitude ranging between 19 percent and 79 percent.
Since revenue limitations are calculated over a rolling eight-quarter period,
commercial banks that anticipated this change may increase their underwritings
in 1996. As shown in Table 2.4, there is a shift of underwriting activities by
commercial banks from the small to the large issue market. Issues managed by
CB-Lead and hybrid syndicates in the small issue market declined dramatically in
percentage terms from 1995 to 1996, while the opposite occurred in the large
issue market.

Although the overall mean and median of the issue size of hybrid syndicate
underwritten bonds are significantly larger than for any other form of syndicate,
the firm size of hybrid syndicate clients is significantly smaller than that of pure
investment bank syndicates in both mean and median. Compared to commercial
bank lead syndicates, the firm size of hybrid syndicate clients is larger for means
but no different for medians. However, examining the results of probit regression
in Table 2.5, which analyze the pairwise syndicate selection, we find hybrid
clients are significantly larger in firm size and in issue size than CB-Lead clients,
but there is no difference between hybrid clients and PurelB clients in these two

variables.

 

‘7 See Federal Register, September 17, 1996, "10 Percent Revenue Limit on Bank-
lneligible Activities of Subsidiaries of Bank Holding Companies Engaged in Underwriting
and Dealing in Securities," volume 61(181):48953—48954.

86

Tables 2.6 and 2.7 report the results of net yields (BPS) regressions that
demonstrate the undenrvriting ability of syndicates. we compare the beta
coefficients of firm size, proxied by log of market value of equity. The coefficient
estimate of firm size for hybrid syndicates is -27.32, which is smaller than that of
PurelB syndicates in Table 2.7, but considerably larger than the CB-Lead
estimate in Table 2.6. Thus, hybrid syndicates do not maintain the advantage
that CB—Lead syndicates have shown in assisting large firms in lowering their net
yields of bond issues. This may explain why hybrid syndicates do not underwrite
for the largest firms even though its clients have the largest issue size.

In summary, the results support the hypothesis that the entry modes of
commercial banks are strategically chosen. In the segment of middle quality firms
where commercial bank show better ability, they behave more aggressively to
acquire the lead-manager role. In the large issue market, where they are unable
to handle large issues due to existing regulations, commercial banks tend to

cooperate with investment banks.

87

2.5 The Characteristics of Hybrid Syndicates

The specialization of different syndicates mentioned in the previous sections
implies that underwriting abilities are different amongst the various syndicates. In
this section, we examine how these abilities are different based on firm and issue
characteristics and how the differences affect client firms. Due to the multiple
choices of syndicates and a potential endogenous selection problem,
polychotomous choice selectivity model would be the preferred choice of model.
However, parameter estimates in this model are very sensitive to distributional
assumptions“. Because of this lack of robustness in polychotomous choice
selectivity models, we apply a framework of switching regression analysis with
endogenous switching that involves analysis of two choices. We use pairwise
comparisons of different syndicates in the first-stage probit analysis. Thus, we
have three pairs of syndicates - CB-Lead versus Hybrid, CB-Lead versus PurelB,
and Hybrid versus PureIB - we need to consider. In the second stage, we
estimate the bond pricing equations, that is the net yield or BPS equations, for
each type of syndicate using an endogenous selection adjustment term

estimated from the first-stage probit”.

 

‘8 Chung (1997) applies polychotomous choice selectivity models, which allow for
different distributional assumptions, to male wages in four sectors. The multiple
selections of sectors are controlled for when the male wage equations are estimated.
The results are quite different for different models and estimation methods.

‘9 See Maddala (1983) for detailed discussions of both the switching regression with
endogenous switching model and the two-stage estimation method, and Song et al
(1999) for an application in the underwriter selection problem and discussions of
variables used in the analysis.

88

2. 5.1 The Determinants of Syndicate Selections

We analyze the selection of syndicates by probit models, presenting
coefficient estimates in Table 2.5. We focus on the differences in selecting CB-
Lead versus Hybrid and Hybrid versus PurelB syndicates, attempting to identify if
the characteristics of hybrid syndicate clients are similar to CB-Lead, PurelB, or
neither. The overall results indicate that these three types of syndicates are
different from each other and that the hybrid syndicate is not the simple average
of a commercial bank lead syndicate and a pure investment bank syndicate.

Hybrid syndicates are similar to commercial bank lead syndicates in terms
of leverage, past 5-year sales growth, operating income, and cash reserves. Both
syndicates are associated with firms that have lower leverage, grow slower, have
higher operating income and higher cash reserves. All of the estimates of these
variables are not significant for the CB-Lead vs. Hybrid regression, but they are
all significant for the Hybrid vs. PureIB regression.

Hybrid syndicates resemble pure investment bank syndicates on the
dimensions of equity trading volume, Tobin's q, firm size, issue size, and the
interaction between high interest expense and the purpose of issue repaying
bank debt. In comparison to CB-Lead clients, hybrid clients tend to have lower
Tobin's q, to be larger firms, to be traded more frequently in the equity markets.
Their clients also tend to issue bonds with larger size and fewer bonds with high
interest expense for the purpose of refinancing bank debt than CB-Lead clients

do.

89

A hybrid syndicate resembles neither a commercial bank lead syndicate
nor a pure investment bank syndicate when maturity of issue, the purpose of
issue and credit ratings are considered. These three syndicates are not
significantly different in volatility of stock return, interest expense relative to
operating income and reduction in long-term debt.

Based on the above results, we may characterize the clients of hybrid
syndicates as firms with better quality, having higher profitability and cash
reserves, lower leverage, and less of an informational problem, having lower
Tobin's q, higher equity trading volume and larger firm size. They issue bonds of
significantly larger size compared to issues underwritten by commercial bank
lead syndicates. Hybrid syndicates' clients are more likely to issue debt to
refinance existing bank debt with lower interest expense relative to operating
income. Hybrid syndicate issues have longer maturity and they have a higher
tendency to be non-investment grade.

The distinct characteristics of hybrid clients sheds some light on the
possible economic rationales of cooperation between commercial banks and
investment banks, and yielding an understanding of the role of commercial banks
as co-managers in the security underwriting business. Given the aforementioned
evidence on systematic selection of hybrid syndicates, a more thorough analysis
of comparative advantages of hybrid syndicates over other forms is warranted.
Therefore, we estimate the bond pricing, or BPS equations for different

syndicates, analyzing their comparative advantages in the following subsection.

90

2. 5.2 The Underwriting Abilities of Syndicates

We investigate the underwriting abilities of syndicates by estimating bond
pricing equations in Tables 2.6 and 2.7. Regression results of the BPS equations
for different syndicates are reported. Since we use net yields as the dependent
variable in the BPS equation, the larger the beta coefficients, the lower the bond
prices the syndicates can achieve for their clients.

In Table 2.6, we report two BPS regressions of CB-Lead syndicates with
different endogenous selection adjustment terms, one estimated from the pair
comparison of CB-Lead vs. PurelB (CB-Lead (1 )), and the other from that of CB-
Lead vs. Hybrid (CB-Lead (2)). The third set of regression results in Table 2.6 is
for hybrid syndicates. Since there is no evidence of selection bias, we report only
the estimates without the control of endogenous selection. For completeness,
additional BPS regressions for hybrid syndicates with the control of selection bias
are presented in Appendix B4. In Table 2.7, the estimates of all three possible
regression specifications are presented for the PurelB syndicates.

We first examine if hybrid syndicates preserve the advantages of
commercial banks. The proxy variables used in studying the informational and
relationship-oriented advantages of commercial banks with their clients are
leverage (total debt/total asset), equity trading volume, volatility of equity return,
Tobin's q and an endogenous selection adjustment term. The estimate of
leverage for hybrid clients is very similar to that for commercial bank lead clients,

with both being negative and statistically insignificantly. This results indicate that

91

hybrid syndicate preserves the advantage of a commercial bank on the
dimension of leverage. Conversely, the estimate for pure investment bank clients
is positive and statistically significant. Based on the equity trading volume
variable in Table 2.6, hybrid syndicates obtain bond prices for their clients that
are very close to the simple averages of CB-Lead and PurelB syndicates. As for
the estimates of volatility of equity return and Tobin's q, hybrid syndicates can do
better than the other two types of syndicates. The estimates of these two
variables for hybrid syndicate are the smallest among all types of syndicates.
The estimates of the endogenous selection adjustment terms for commercial
bank lead syndicates are both negative and statistically significant regardless of
the pair comparisons. When compared to CB-Lead syndicates, these estimates
for both hybrid and PureIB syndicates are not statistically significant. One way to
interpret the adjustment term of endogenous selection is as a proxy of private
information extracted from the identity of syndicates. The negative estimates of
CB-Lead syndicates50 indicate that they can lower the net yields of their clients'
bond issues further even though all other exogenous characteristics of firms have
been controlled for. Thus, commercial banks can cherry-pick the clients they can

serve better based on endogenous selection.

 

¢(Zr)
I — <D(Zr)
when its assigned value as the dependent variable in the probit estimation is one. This
term is estimated from the first stage probit regression. Since all numbers of this form
will be positive, a negative regression coefficient for the BPS regression means a
lowering of the net yield of bond issues. The adjustment term for the other syndicate is

- ¢(Zr)
¢(Zr)
Therefore, in order to lower the net yields of bond issues based on endogenous

5° The endogenous selection adjustment term has the form for the syndicate

when its assigned value as dependent variable in the probit estimation is zero.

92

In Appendix B4, we report two possible model specifications for the BPS
equation of hybrid syndicate clients. None of the estimates for the endogenous
selection terms are statistically significant, whereas, the estimate of this term in
Table 2.7 - model PurelB (2) - is significantly positive. This indicates that PurelB
syndicates demonstrate superior cheny-picking ability over hybrid syndicates.
Therefore, PurelB syndicates seem to underwrite for clients that they can lower
the net yield of bond issues due to the identity of syndicates. Thus, in terms of
certifying private information that can only be revealed by the choice of
underwriters, hybrid syndicates act like neither CB-Lead syndicates nor PurelB
syndicates. One possible explanation for failing to establish an endogenous
selection bias of hybrid syndicates is that the identity of underwriters is of less
importance for hybrid clients. This is consistent with client characteristics that we
identified in the previous subsection, who appear to be firms with less of an
informational problem.

The second dimension of hybrid syndicates we would like to examine is their
ability to reduce the monopoly power of information that banks have over client
firms. Due to the fact that a commercial bank may appropriate its client's profits,
a commercial bank is less credible than an investment bank in underwriting firms
with higher growth and profitability. Thus, the clients of a hybrid syndicate are not
subject to monopoly power as are clients of CB-Lead syndicates. Estimates from
the BPS regression of hybrid syndicates on sales growth and operating income

are in between those of CB-Lead and PurelB syndicates, leaning towards

 

selection, the regression coefficient for the BPS regression should be positive.

93

estimates from the PurelB syndicate regression. The results support the view
that hybrid clients are free from the bank informational monopoly power.

The varying distances between lending and underwriting activities for
different syndicate types may cause different levels of concern about conflicts of
interest. There is no conflict of interest problem for a pure investment bank
syndicate. However, there is serious concern about conflicts of interest under a
CB-Lead syndicate. A hybrid syndicate should raise an intermediate amount of
concern. The BPS regression estimates of the proxies for conflicts of interest
demonstrate this pattern. It a hybrid client is issuing bonds for the purpose of
refinancing bank debt, there is no increase in the net yield of a bond issue. By
cooperating with a reputable investment bank that leads a syndicate, a
commercial bank may effectively reduce these concerns. Nonetheless, at high
levels of interest expense, both hybrid and commercial bank lead-managed
issues have higher net yields. The hybrid syndicate is still less credible than the
PurelB syndicate, which has no conflicts of interest problem like a CB-Lead
syndicate does.

Consistent with the literature regarding the role of lead managers of a
syndicates‘, the reputation of investment banks may mitigate conflict of interest
problems that commercial banks have. Therefore, the coalition arrangement
reduces the disadvantage that commercial banks have as underwriters and
allows bond-issuing clients with high interest expense to refinance bank debt,

thus partially avoiding a large price discount. This benefit of coalitions allows

 

5‘ See Pichler and Wilhelm (1998), Nanda and Yun (1997) and Jain and Kini (1999) for
discussions of the economic functions of a lead underwriter for the syndicate.

94

hybrid syndicates to underwrite more issues of bonds for the purpose of repaying
bank debt than commercial bank lead syndicates could alone. The association
with investment banks also allows firms to circumvent the bank informational
monopoly power. Thus, bond-issuing clients reap the benefits of their association
with commercial banks, while having the investment bank to lighten the negative
impacts of bank relations.

In summary, we find that the characteristics of coalitions between
commercial banks and investment banks are different from those of CB-Lead
syndicates and the PurelB arrangement. The hybrid structure creates synergies
beyond the simple average of the other two organizational forms, which should
be beneficial to many firms in the economy. We continue our analysis by
comparing the net yields that syndicates have obtained in serving existing clients

to predicted numbers if clients were served by other syndicates.

2. 5.3 The Performance of Syndicates

In Table 2.8, we compare the performance of different types of syndicates,
presenting the gross benefits for clients utilizing services of selected syndicates.
Gross benefits are defined as the expected mean net yield of bond issues had
clients chosen an unselected syndicate minus the observed net yield of bond
issues. If the resulting number is positive, then the chosen syndicate can acquire
a lower net yield of bonds for its client than the unchosen syndicate could. If the

net yield comparison is the only factor that is utilized in the selection of

95

syndicates, we should observe that all benefits are either positive or
insignificantly different from zero. Unfortunately, information asymmetry in the
capital markets may prevent clients from switching among different underwriters
freely. The existence of switching costs causes some firms to be captured by
their underwriters even though there are other types of syndicates that may
achieve better bond prices for them. Thus, gross benefits may be negative before
the consideration of switching costs even when better choices of underwriters
exist in the market.

The results shown in Table 2.8 are the products of results from Tables 2.6
and 2.7. The PurelB (1) model, which controls for selection bias, in Table 2.7 is
used to estimate the benefits of CB-Lead clients with respect to PurelB
syndicates”. Panel A of Table 2.8 displays the results for CB-Lead clients had
they used the services of hybrid and PurelB syndicates. Compared to the service
of hybrid syndicates, the overall benefits for CB-Lead clients on average is
negative and statistically insignificant. The negative benefits arise only for clients
with high and low credit ratings. Thus, clients with high and low credit ratings
would be better off if they were underwritten by hybrid syndicates. However, the
group with middle credit ratings enjoys significant positive benefits by using the
services of commercial bank lead syndicates. The results are consistent with the
hypothesis that hybrid forms possess superior underwriting abilities in serving

some CB-Lead clients, while CB-Lead syndicates are better in assisting middle

 

52 We also repeat the same procedure by using model PurelB (3), which does not control
for selection bias, and report the estimated benefits in Appendix B5. These results are
essentially unchanged from those in Table 2.8.

96

quality firms. Thus, there seems to be a need for both syndicate forms to exist in
order for different types of firms to acquire higher bond prices.

Compared to the services of PurelB syndicates, CB-Lead clients on average
are significantly better off staying with their own underwriters. If a PurelB
syndicate were used, only the estimated benefit for high quality clients is
marginally significant. Additionally, all of the estimated benefits are larger than
those for which a hybrid syndicate is used. Thus, from the point of view of CB-
Lead clients, the hybrid structure is better than the pure investment bank form.

The results in Panel B of Table 2.8 show that hybrid clients appear to be a
group of privileged customers. They fully enjoy the benefits of cooperation
between commercial and investment banks, as of all the estimated benefits in
Panel B of Table 2.8 are significantly positive. These results are robust to
different model specifications. In Appendix B5, we do not control for the
endogenous selection problem if a PurelB syndicate were the underwriter, with
the results still supporting the view that a hybrid syndicate is a better choice for
its clients.

Our analysis shows that the clients of pure investment banks on average
should not use the services of a commercial bank lead syndicate, but should
switch to a hybrid syndicate. The estimated benefits with respect to hybrid
syndicates are all negative. Only non-investment grade firms are indifferent
between these two types of syndicates. Compared to the characteristics of hybrid

clients, PurelB clients are less profitable and have larger information asymmetry

97

problems. Thus, the switching costs for these firms may be higher than for those
who switched to the hybrid syndicate.

In the eyes of PurelB clients, the underwriting ability of the hybrid syndicate is
superior to that of a pure investment bank arrangement. Hence, the cooperation
arrangement is not only beneficial to commercial banks, but also to investment
banks. The creation of a new organizational form allows the incumbent
investment banks to enhance their services further by accommodating the entry
of competitors, who possess distinct comparative advantages. The economic
motivation of investment banks explains why commercial banks are able to
penetrate the security underwriting business so quickly by participating in a

syndicate that is lead-managed by an investment bank.

98

2.6 Conclusion

In this paper, we add a new dimension to the research on commercial bank
undenivriting. Unlike the extant literature, which focuses only on underwriting
activities when commercial banks are lead managers of a syndicate, we
investigate the role of commercial banks as co-managers. We provide evidence
consistent with the hypothesis that commercial banks strategically select their
entry modes. In the segments where the underwriting abilities of commercial
banks are hindered by the concerns for conflicts of interest and revenue
constraints, they tend to ally with investment banks. Otherwise, they compete
more aggressively in acquiring the lead role.

The benefits of bank entry are by no means limited to the cases when
commercial banks lead the syndicates. The creation of a valuable hybrid
organizational form is impossible without the participation of commercial banks
as co-managers. By considering only commercial bank lead syndicates, previous
studies on the effects of commercial banks' increased underwriting activities have
likely understated the benefits of bank entry.

At the same time, hybrid syndicate is shown to preserve many advantages
of the other two types of syndicates. In some cases, there are synergies created
by a coalition arrangement. This superior organizational structure explains why
commercial banks may penetrate the market so quickly while investment banks
are willing to accommodate this entry. The cooperative arrangement has strong

economic motivations, particularly for commercial banks, which strategically

99

choose the modes of entry depending on their underwriting strength in various
markets.

The results in this paper also highlight the need for the existence of an
independent investment bank that is free of the moral hazard and adverse
selection problems that plague commercial banks, particularly where the
underwriting and lending activities are closely combined. Without the existence of
an independent investment bank, firms falling into this category are forced to pool
with those having conflict of interest problems, thus incurring a price discount that
they would otherwise not incur. Consequently, these type of firms are suffering
from the entry of commercial banks into the security underwriting business. Thus,
the structure of the financial system is evolving in a way that is beneficial to many

issuing firms under the current regulatory and market conditions.

100

APPENDIX 2

TABLES OF ESSAY 2

101

 

mm? mm won 000 02 000 w- F00?

 

 

0mm 0? 00F mm mm 0mm?
0mm 2. 00 E 00 0000
mm? 0 vs 3. 0m 0000
00m m 009 00 mm 009
mvm 0 0: mm mm N02.
3m 0 00.. N 0 50?
0000 x000 000..

0003-..... 0.000 30.08800 «0080002.. 0.00 0.52.. 8.000 .0.0.08800 .00>
00000. 0000 .0 6000005 00... .< .0000

000.0 0050200000 05 .0 000 .0 .000008

-00 0 00 0029:. 00.000 .0.0.08800 .050 000 00.0.5 000 .000 0.208800 0 0. .0.0200000 000. 05 00.0.5 0. 00.00.00?
08000.00. 0000-02.00 0.808800 0.000 808.008. 00 0. 0.03.0000 000. 05 000 0029:. 00.000 .0.0.08800 00

0.0 0.05 00.55 0. 0.00.050 0. 0.000 808.008.. 0.01 .0000. 00 0. 00208. 0.0 00.000 .0.0.08800 .050 000 .0.0200000
000. 05 00 0.000 808.002.. 00 0>00 00.03 0900.050 08000.00. 0.000.. .0000. 05 0. 00208. 0. x000 .0.0.08800
.050 00 .00 0.080000 000. 05 00 0.000 .0.0.08800 0 0.00 00.02, 0080.050 000000.00. 0003.000 50.08800 .003
.0.0200000 >0 000 .00.» >0 5.0500000. .0 000 < .0000 0. 000000. 0.0 00000. 0000 .0 000080800 000 0.00800 .000...

00.: .00....30000 .3 000 .00» >0 00000. 0000 no 005000.00 000 600000.: 00...
fin 0.00...

102

 

 

 

00.. NN v.00 0.0N N: ommviwmmv
00.. 0.0 0.00 5.50 0.N_. 000w
00.. 0.0 0.0.0 5N0 0.0? 000—.
Dov 0.0 0.00 v.00 0.0.. 30—.
Dow w.—. v.00 0.0m 0.0? 000..
00.. 0.0 0.00 v._.N 9.0—. Nmmw
cow 0.0 0.00 0.0 N0 900—.
0000 0.000 000..

0003...? 0.000 30.08800 «0080002.. 0.00 000.... 0.000 30.08800 .00>
00000. 0000 .0 00.900030 0:... .m .0000

.350. 3. 0.0a»

103

 

 

.8 00.0 00.0 5.0 . 00.0 . . 8.8.0200 00.328
2 00. z 0 F .0? 3.0 0.0.2 om. : R? 20, €00 0:05.095 050
m 0.0 R? 00.0 N: 3.0 0: 00.: 25>:
00.0 00.0 8.0 00.0 00.0 02: 5.0 82-022 00.05500
A0250 8 0902:0003 >0 0N5 FEE .5300. 000—2
802 9: o: . on? 0. . 8-8-0.ch 00.05.80
3 of 2: 03 2: on. 02 a? 0.53 20.5095 050
“0: com of o2 o: 80 SN 25>:
V0 2: 0? 2 2: 2: 8. 820.50 00.05500
A0057: 3 00:02:00.0: >0 0N5 0000. 0050—2
803 at 02 . om. . . 8.8.0000 00.0558
2 0: 03 02 02 02 SN 000 0.53 202095 050
~80 000 new com «5 0.0 80 25>:
mm .5 S 00 00 5 3: 000.022 00.08500
ACE—=5 $0 00:02:09.3 >0 0N5 0300. .505.
002-32 002 008 32 88 $2 32 30>

 

2:208 0.0000800 00 E5000 03000.... .00 00 22> 080 x0 05 .05

0:00 0 00 000.00 0_0_> 080 x0 05 00 8280.0 05 0_ 000.00 0.500 0000 00... 00.60 00050 E 0500 00 30> 000008
000 0_ 08:0 00 000 00k .000 00088000: >0 0:0 00> >0 000000. 0.0 00:00_ 0:00 00 0000.00 .500 0000 000 0000 000.

00>“ 00:02:00,... >0 0:0 00> >0 0000000000000 8.5 0:0 0:00_ 0:00 00 0000080 020000000
«.0. 0.00...

104

.aooo-xcmm 36.08500 new 255 Emgmgs 05m c0928 2:69: :o :59: S 855:6 EmoEcgm .

.Eooo-xcmm _m_o:mEEoov 253 80869:. 2.5 new 25>: cmmzamn 3:69: :0 game 5 macemtfi E8535 :3:

.Eooo-xcmm $688500 .xcmm

E9535. 22:: 25>: new 832 832:8 38.58800 5923 3:69: :o 2:08 :_ moan-5:6 EmoEcgw GEM

 

 

u :: :2 v: - a: - - 8-82:2 358558
a .8 :: m: :N 8 N2 :2 2:2 88:25 :5:
a :: m: :5 8 N: 5 v: 25»...
N: :: :: F: 2 :: :: 822:2 322.58
>n cmttEmucD mmawfl .9 Omaha—w “Eon m_mmm agnos—
u :9 :2 :3 - F2 - . 8-82:2 68.2.58
3 :2 2: :9 :2 SP :8 E 2:2 E2825 :5:
a :2 our 8. x: NE E :3 :22:
E 2. :: :3 v: a: N: 822:2 522.58
>0 :mEScmocD m03m£ .0:— Ummaw “Eon m_wmm cam—2
:8 2.: 8w ::.: - Se - . 8-82:2 86.2.58
3 ::.v 8.: :5: F: 8.: _ 8.: 2.: 2:2 22.88:. as:
2.: 2.: 2.: 2.: EN vim 2.: 88:
8.: 5.: :3 :3 8.: 3.: :3 822:2 36.2.58
”32:3 3 $5253: >9 mum ctr. 53mm. 52005.
$2.32 :8. :8. 32 22 82 32 28>

 

5.80. Na :32

105

2:00 808::08. :0: m. .2000 .0.0:08800 :0: :0:0:: mo .080: 05 0:0 ::0:.:>>:00c:
£00 >0 00088000: :0:::_ :0: 80.30086 :0::0: :.00:0 05 :05 00.00 :_:0£00>:_ =2: 05 5.3 .:0:0£c0:00
c. 03.0>-0 0:0 ::0: 0:000:80 0:: :50800 0.00: 0:: c. :30: 00:5 ::0_ 0E- .>.0>.:000:0: .:0:::. 000020000: 0000-0800
.0.0:08800 0:0 2:00 808::08. 0:00 :0: :0::08 080: 05 c. 005:00 0:0 2:80.00 008.080: 0;... .:0:::. 000020000:
0.:0>.. :0: m 000 .N .F :08200 c. :05 :0 00:08:08. 080: 05 3.0800 0 000 .m .0 9:83.00 .:0:::_ 000020000:
000.2800 .0.0:08800 05 :0: 300200 9:00: :.00:0 2000 :0: mmm. :0 00608 000 :c008 0;: >0 00326: 500060
9:00: :.00:0 0000 :0: 0.080: 0...: :0 0060090 05 8.0800 083.00 :2: 0:.» 30308 03200800 :0 >830: 33:00:...
.m.: :0 0_0_> 080 x0 0;: :0>0 0:00 0 :0 000:0: 0.0_> 080 x0 05 :0 82.890 05 :_ 0.0-0m. 000:0: :860 :_:00 0::

00>: 3:03.600: >0 0009.0: 8.00 :.000 0:0 00300506 09:00.: :.00:0 .8300.
nN 0.00...

106

20000-2000 .0.0:08800. 2000 808808. 0:30 000 0.:0>... 0002000 :00608 :0 :0008 0. 0000:0::6 8000.005 0:
20000-2000 .0.0:08800 .2000

808808. 0:30. 0.:0>I 000 :03::. 0003-2000 .0.0:08800 0002003 :00608 :0 :0008 0. 00090.20 8000.006 .83
:0000 00:0:000 :003.00. 8000.00 :.0... :

 

 

 

 

 

 

 

80.0: 00.00 80.0. 3.3 .000. 9.0m 0009mm: .:>
.30. 00:: 82:: :0:: 0.8:: .:>
.80. 8.8 22>: .:>
20.0.00
.. .. 0.0 mmm mmm 0.0 00.: 000 0.0 2 .. 0.0 «30.00 :0 o
u 000 n 000 0.0: :09. 8000 0.: a 000 a 000 00.. ~00 mm: 0.: m
0:0 «mm 0. :0 8000 803 0.0 mmm 3w 0.0: «K 00w 0.: 0m
s 00 8 K. 0:0 85: 803 Yam :0 00: Now N0 00 0.8 00m
:0 v: NC 80> 800 0N0 : >0 2 Nu >00 mm 00 0.00 <
.. 2 0.0 00 0: m.) :0 mm 0.0: mm mm 0.0 0<
.. .. 0.0 :0 >0 0.: 00 00 0d . .. 0.0 00<
0060.2 000.2 0060.2 000.2 0060.2 000.2 0060.2 000.2 000000
000:0w 03::. 000:0w 03::. 000:0w 03::. 000:0w 03::. :.00:0
8.00 :.:00 :0 .2: 8.00 :.:00 :0 .x. 8.00 2:00 :0 :8 8.00 :.:00 :0 .x: 0.600:
00002.00m .0.0:08800 200m «005008. 0:30 23>... 0032.00m .0.0:08800

 

QN 030...

107

 

 

 

OOF 7N mdm m.mm oh mam—3.60..
co.- o.m NNN odm md mom.-
oow 7m mdm va 6v mmmw
co.- od mdm 0.5m 0d V09.
00.- vé 5.00 Pdm Q? Maw—-
00.. Cd v.00 NdN vb NmmP
cow 0.0 mém 0N.- Nd ram—-
0000 200m 000..

000>...-..< -2000 .0.0:08800 808.008. 0:30 0.:0>... .200m .0.0:08800 :00>
.0808 0300. 0.01:0. 05 0. 00300. 0000 :0 $0809.00 00... .< .0000

030:0 0000200003 00. :0 000 :0 :000008

-00 0 :0 00208. 02000 .0.0:08800 :080 :00 00.03 000 2000 .0.0:08800 0 :. :0880003 000. 00. 00.03 0. :0.00.00>:
080:0:00: 0000-2000 .0.0:08800 .2000 808.008. 00 :. :0880003 000. 00. 000 00208. 02000 .0.0:08800 00

0:0 0:00. 00.02, 0. 0.0060>: :. 2000 8002008. 0:30 .0300. 00 0. 00208. 0:0 02000 .0.0:08800 :050 000 :0880003
000. 00. :0 2000 808.008. 00 08... 00.03 00.00.00? 0.000080: 0.002.. .0300. 00. 0. 00208. 0. 2000 .0.0:08800
:050 00 .30 :0880003 000. 00. :0 2000 .0.0:08800 0 080 00.02, :0:00.00>: :.00:0:00: 000.2000 .0.0:08800
.0.080: 0:.80 0.... :0 030:. :0 00.: $8.60 00....8 0:: 00608 00: >0 0080800: 0. 00.: 030:. 00:. 2028030: 0:02:08
03:0. :08: 000 03:0. 00:0. 00. :0: m 000 < .0000 0. 000000: 0:0 00220880003 >0 0030:. 0000 :0 000080200 0....-

80800: 0300. :08: 000 0300. 00:0. 00: 0. 003230200003 >0 00300. 0000 :0 00080200 02...
0.0 0.00h

108

 

 

 

00: «.0 o. :0 0.0: 0.0: 000750:
00: 0.0 ~00 :.:: N0: 000:
00: N0 0.00 0.00 0.00 000:
00: 0.0 0.00 :.00 0.0: 000:
00: 0.0 0.00 0.0: 0.0: 000:
00: 0.0 :.:: :.:: 8:: ~00:
00: 0.0 0.00 0.0 00 :00:
0000 0.00m 000..

000>:.-..< -200m .0.0:08800 0000:0008. 0:30 0.:0>... 0.00m .0.0:08800 :00>
00000.0 0300. =0E0 05 0. 00300. 0000 :0 000000200 00... .m .0000

.0800. 0.0 200:

109

0003.00. 0:0 >00.
00300. 000000: .00 0:0 08:0. 80.0000 000 00.8830 800.. 00. :0: 00.08.00 00... .08000. 00000000 0. 02.0.0: 000090
.0008. 000 .000 2000 00000000 :0 8:0. 00:00:08. 00. 0. 0000020 .00:08...000 2000 00000000 .0.00.:0> 0.0. :0 00.008
00. 000. 00:0. 0. 08000. 0000000 0. 02.0.0: 000090 .0008. :0 _0>0_ 00. :. : 0. 03.0> 0000.000 00. .0_00.:0> >8830
0 0. 5.00.00. .0008. 00.: 00.20008 0 ..000 2000 000000 0000000: 0. 0. 03:0. :0 0:00:30 00. :. : 0. .00. 0.00.:0> >8830
0 0. .000 2000 00000000 .0.0:00 _0.0. >0 000.20 3.300 :0 03.0> .02:08 03.0 .000 :0 03.0> 2000 00. >0 000000 0. 0
0.0.00 :. 00.20006 0 00.0: .00 :0 2.200 :0 00 0. 03:0. 0000 00. :0: 08.0: .000 0.8005. 00. :. : 0. 000:0 808.008.-002
000:0 808.008.-000 0. .. :. .m 0. 03_0> 00. .000 :0 < 0. .. :. _N 0. 03.0> 00. .0< :0 00< 0. 0300. 0000 00. :0: 0000: :.00:0
0.8005. 00. :. : 0. 00:0: ..00:0 00208. 0:0 02000 808.008. 200 .00. 0.00.00>0 00. :0: 0000.: 0.050 >02. 0800 00.
0. 000000 0:0 0083.00 008.080: 00. .0000... .02000 .0.0:08800 0..2. 00.0:00000 .. 000 2000 808.008. 00 0. 08200003
002 05 0. .00. .052 m 0. ._ :_ o 0:0 .253 00:00:58 0 0. 0.0808: 000 00. :_ : :_ €03-00. 030:9, .5880:
00. .: 083.00 0. 00000300 0000200 0.0200003 :0 .00.-. :0 :03_0>-0 000 00.08000 000:0 00.20.00 00. 00000: 0.00.0.0...

000000.00 0.00.00>0 :0 8000.80.00 00. :0 00:00:..0m
m.~ 0.00...

110

 

 

 

00.. 000 000 08000008 .0 .oz

00... 0.0 3.0 8.8800

.00 .80- 0.0 08.0- 0.0 08.0- 500 5.0.002 0. 02.8000
00.0 80.0 8.0 08.0 00.0 .80 Emma-.0000 0.2.2000
.00 05.0 0.0 002.0- .00 .200- 000:0 005.095-82
00.0 000.0 8.0 000.0 3.0 30.0 05.0: .090
5.8.00.

00.0 .02.- 00.0 3.0 00.0 00.0 .090... .0: - .00.. .00.. 8000000
:0 000.0- 3.0 08.0- R... 0.0.0 00:80. 0009002000006 .0900.
.0.00.:0>

8.0 000.0 .00 000.0- 8.0 0.0.0- 0.800.. .00.. .03 8000000
.00 08.0- 8.0 000.0- 8.0 00.0- .0300. :0 000.03
00.0 0.0.0 «0.0 0.....- No.0 «0.0. .0500 :o 020:, 0005.03
8.0 0.0.0 .0... 2.0.0 0.0 .000- 0.0000 0.020580. 05.2000
00.0 0.0.... 8.0 at... 0...: 00.0 0 0.050.
8.0 .80- 2.0 30.0- .00 30.0 .0>-0. 526:0 0200
00.0 08.0 .00 0.0.0- 80 Rod- 0:00. :0 0.5.0.2
00.0 .000- 000 .000- 00.0 0.0.0- 0030.09. 0:30: 0.300 :0 0___.0_o>
00.0 .80 00.0 30.0- .00 08.0- 0.022, 00.00:. 0500
No.0 .80- 8.0 0.0.0- 00.0 08.0- 0.0000 0.02.000 _0.0.
03.0>-n. m.mE_.mw 03.0>-n. m.mE_.mm 03.0>-n. 908....wm mm_nm_:m> .chchmOE
.. n 0000.. .. H 83.00. .. u 0003-00. 03009 00000000

 

0.0.30 .0> 0002..

0050 .0> 083.00

0.5.: .0> 003-00

0." 030...

111

.3385 En >9: 2955 nmtoam. 6: 2m
9.53 Emwmcoo ncm mm_EE:n >=mm> m5 .8 8.9233 ms... .mEooE 9:5an 2 @252 mmcmaxm 6295 new Emu xcmn
8:2ch *0 gm: 8:09.35 2: 2 $598 $922.28 xcmn wocmccmm .mmsamfio o .Emu xcmo 95:8 @2856. 2
w_ 952 B 8092. 9: t F 2 “m5 033.? 2:88 m m_ Emu xcmo mocmccmm 23$ .82 >2 822v 353 n.0 m:_m> 5me
man Emu ho m:_m> xoon m5 3 nmccmn m_ a 3.50» .mmEcmcE o .83. 8: 5 26.09 5 mm m_ 03$ neon m5 .8 95m.
:85 @3022 9: t F m_ mama EmEummETcoz .mnma EmEummETco: m_ g .._ .m m_ m2m> 9: .95 .o < 2 g a .N m_ m:_m>
m5 .m< .o mm< w_ mzmfl neon m5 .2 mczm. :88 {6002 m5 2 F w_ 95m. .690 SE“ EmEfiivm mzocmmoncm 25.05
8: $8 56859 9: van mmzmg cmESCmncsfitgc 9: 2:0 $3.05 29:3 9: .0 new m 9.828 :. .mmfiofiiw
25>: .w> nmm..-mo 3 “m5 933 Amy tam..-mo .3522? 92:; .m> umm._-mo “6 28:32.8 :3 m5 3% A: ummfifio
.coszzmm 305 39ng 9: E0: $65.30 9m mmmmfi cozoflmm maocmmoccm m5 .058 2 8m: £59 EmEuwzfim
msocmmoucm .6mm..-mov 6:352: 93. 9: m_ xcmn _m_8mEEoo a 5:2, $33 .8385 99:3 05 .v 2 P 2.5.8 5
.69; B mm:_m>-a 98 255.852 Ammmv mummaw Eon $me 3299.83 ho £58508 38 m5 roam. o 2 F 28200

6335:?» 25>... can two..-mo 3 205.3332: £23: :0 acacia “Eon mine no 3353mm
m.~ min...

112

 

 

113

 

mvm X: X: 225388 3 .oz
mud mmd mmd mm 853.6(-
- - mod 8.8- 86 3.2- :5. 23.8.8 88385
23 m P .9: cod 8n: 2:. 3.8. 2.95 55395.82
mmcmaxm
Ed 3.2 5.0 8.0a No.0 3. z: 3925 - 5% was 8cmémm
8d 8.3 So 3.: mod 3.8 9:85 839805283 399:.
Ami—25>
and 2.8- ood 2.3- So BB- 58.2: 5% EB Beacemm
2d 8.0 8.0 o; So m3 833 ho 3353
8d «QR- 8d 5.2..- go 8.8- $58 ho was, 552.53
Nod R. F- 8.0 So. Ed who 33$ @2385 9:230
mad m3 and 52 93 mom a 3.38
$6 9.0 cod o: cod Rd 33.8 €85 83w
8d «3 2.0 m3- and and 2.3 3 £552
8d SN cod ~20- ood 8m $353: 539 35am .6 a___§o>
So 2d 2.0 was. omd emo- meae, 9.6m: :55
:.o and. 86 3o- mud 3.? 93% _Sosgmu EB
m:_m>-n_ QmEzmw mam-Si mumEzmw mews-n. 295mm mofimtm> Emncmamns
23;: Q 83.8 E 83-8

 

QN «En-_-

.8365
9m >9: canon. nmtoam. .9. 9a 959 E938 new $3.83 >_.mm> m5 .0. 999.33 05 .coszzwm .305 mom-am
-6... 9: E9. 888:8 9m 9..me c2823 msocmmoucm 9: 62:8 2 8m: £59 EmEfiafim mzocmmoncm .0885
9599.0 o. @269 $598 699... new 59. xcmn mocmccmm .o 8.9 820995 on. w. mmcmaxm $995.53 xcmo
mocmccmm .mm_2cm£o o .59. xcmn 9.32.8 8559 o. m_ 95m. .6 309:5 9: t F m. 65 033.9 >EEzn m m_ 500 xcmn
mocmccmm .233 92 .5 8220 353 .o m2m> .352. men Emu .o m:_m> x89 9: >3 3:ch m_ a 3.58 .mm_>>.m£o
o .399 .0: .o .529. .0 mm m_ 96m. 9.09 9: .2 959 .690 $6022 9: ._ F m_ 9.90 EmE.mm>c_-coz .muma EmE.mm>:_
.5: m. z t .m m. m:_m> 9.. 6%. .o < a g t .N m_ m2m> 93 dd. .0 mm< m_ mamfl 9.02 9: .o. 9.39 .690 3600.2 9.: t
F m_ 9.29 .690 .E.9 EmEfiafim mnocmmoncm 9: 9.205 .0: $8 8. 99:... .wmfiofigm m_9:a .m> 25>: .6 65 8mg
AN. 99.5 .8522? 99am .m> FEB-mo .o 52.9.2.8 ._ma 9: $3 E 993.. .coszzmm .505 mafia-.9: 9: So...
8662mm 9m 835 5.623 msocmmoocm 9.. 6.2.8 9 8m: 2:9 E9536... maocmmovcw .69... .0 mm:_m>-a new
986:? m_9:n_ .0 9.23909 Ammmv 959% .58 mama mama-0:83 .0 9:20:38 99. 05 team. 0 o. F 9.8200

.3352?» 993.. .5 53:33:: 3:2. :0 «39% «Son «Ban .0 3qusz
EN snub

114

 

 

 

m2 m9. 8. 253298 .0 .02
NS NS 9... 9 8.3.2
- - So 58 8d 9.... EE .559? 38985
8d 2.9: 8.0 8+9 8d :8: 89¢ 23395-82
mmcmaxm
5o E...” :3 5mm :3 «mm-m $925 - 5% 5:2 8:253.
:3 an... «.2. 8.8 :3 8... 9:85 9.6.8285qu 7.995
339.?
8... wim- ood $.8- 5... 5. Fm- .9865. 5% .58 8:959.
8.0 ovo- Ed 9d- 8d 93. 583 .o 3353
8d 58. cod 9. a- cod 8.8- 3.28 .0 029, .95563
8.0 85- cod N2”- 8d mod. 383 393585 05.980
8.0 9.2 cod 3.9 8.0 52 a 558»
two «to- NE. 9d .2. NS- 659 5:65 3.8
8.0 N: 5.0 8.0 8... o: 93. .6 5.252
8... 3m 8... v3 cod 9... E3585 532 5:8 .o 35:59
cod and 8... end 8... ~00 use? 9.8.. 53m
cod 8. F 8d a? 2:. BF 388 .9058“. ES
mag-n. 9mE=mw mag-n. 99.5mm mag-n. 9mE=mm mm_nm_.m> .cmucmamns
a. 993. a. 99.... E 993..

 

h-N 03¢...

115

Table 2.8

Gross benefit for issuing firms by credit ratings using the service of
chosen syndicate rather than that of unchosen syndicate.

The gross benefit is defined as the expected mean net yield (BPS) of bond
issues, had they chosen the unselected syndicate minus the observed net yield
of bond issues. The benefit estimated in column 1 (below the heading of Hybrid)
is the predicted BPS of CB-Lead client by using the estimates of hybrid BPS
regression reported in Table 2.6 minus the original BPS of CB-Lead client.
Hence, the benefit in column 3 is estimated in the same way. CB-Lead stands for
a commercial bank as the lead underwriter. Hybrid represents an investment
bank as the lead undenivriter while other commercial banks are involved in an
issue. PurelB stands for the underwriter syndicate includes only investment
banks. The p-value of t-test (two-tailed) is reported, the tested null hypothesis is

the gross benefit is equal to zero.

Panel A. Gross benefit for clients of CB-Lead syndicate

 

 

 

Hybrid

Credit ratings Benefit P-value Benefit P-value
All ratings -3.63 0.46 8.23 0.10
No. of observations 184 1 84

Aaa or Aa -21.46 0.03 -8.75 0.14
No. of observations 14 14

A or Baa 3.20 0.25 17.68 0.00
No. of observations 136 136

Ba or below 2364 0.33 -22.59 0.36
No. of observations 34 34

 

Panel B. Gross benefit for clients of hybrid syndicate

 

 

 

CB-Lead
Credit ratings Benefit P-value Benefit P-value
All ratings 173.42 0.00 101.96 0.00
No. of observations 345 345
Aaa or Aa 202.26 0.00 102.78 0.00
No. of observations 46 46
A or Baa 168.39 0.00 100.98 0.00
No. of observations 215 215
Ba or below 170.50 0.00 104.03 0.00
No. of observations 84 84

 

116

Table 2.8 (Cont'd)

Panel C. Gross benefit for clients of PurelB syndicate.

 

 

 

Hybrid CB-Lead

Credit ratings Benefit P-value Benefit P-value
All ratings -11.10 0.00 137.08 0.00
No. of observations 798 798

Aaa or Aa -17.98 0.00 142.39 0.00
No. of observations 127 127

A or Baa -8.82 0.00 128.93 0.00
No. of observations 542 542

Ba or below -13.88 0.24 166.11 0.00
No. of observations 129 129

 

117

APPENDICES

118

APPENDIX A

DATA DEFINITIONS

A.1 Issue Characteristics

CBL: A dummy variable that is 1 if the lead undenNriter is a commercial bank; 0
otherwise. The commercial banks represent the section 20 subsidiaries
of bank holding companies, which are permitted to underwrite corporate
securities by the Federal Reserve Board.

Commercial Bank-Lead: If the lead underwriter is a commercial bank and there is
no other commercial bank involved in the underwriting, then the assigned
value is 1. Otherwise, it is 0. (There may or may not be an investment
bank involved as a co—manager.)

Commercial Bank-Coop: If the lead underwriter is a commercial bank and there
are other commercial banks involved in the underwriting, then the
assigned value is 1; 0 otherwise.

CB-Lead: If the lead underwriter is a commercial bank with or without other
commercial banks involved in the underwriting, the assigned value is 1; 0
otherwise. (This variable includes both Commercial Bank-Lead and
Commercial Bank-Coop.)

Hybrid: If the lead underwriter is an investment bank and there are other
commercial banks involved in the underwriting, the assigned value is 1; 0

otherwise.

119

PurelB: If the lead undenNriter is an investment bank and there are no other
commercial banks involved in the underwriting, the assigned value is 1; 0
otherwise.

BPS: The basis points spread is the premium of the ex ante yield spread of a
bond over the ex ante yield of a US. Treasury security of comparable
maturity.

Size of issue: The size of issue is the principal amount in millions of dollars.

Maturity: It is the years to maturity.

Non-investment grade: A dummy variable that is 1 if the Moody’s credit rating for
the bond issue is Ba or below or not rated; 0 otherwise.

Credit rating: If the Moody’s credit rating for the bond issue is Aaa or Aa, the
assigned value is 1; if it is A or Baa, the value is 2, if it is non-investment
grade, the assigned value is 3.

Refinance bank debt: A dummy variable that is 1 if the purpose of the issue is to

refinance existing bank debt; 0 otherwise.

A.2 Firm Characteristics

Total assets: Total asset (AT) in millions dollars.

Interest expense relative to operating income: Interest expenses (XINT) divided
by operating income before depreciation, depletion and amortization
(OIBDP).

Firm size: Natural log of market value of equity (MKVAL) in millions dollars.

Operating income: Operating income before depreciation, depletion and

amortization (OIBDP) divided by total assets, multiplied by 100.

120

Exchange listing: 1 means the stock of the company is listed on either NYSE or
AMEX; 0 otherwise.

Leverage: It is defined by total debt (DT) divided by total assets, multiplied by
100.

Volatility of equity return: It is proxied by the variance of residuals from the
market model, which is estimated over 120 trading trades prior to the
issuing date. The return on the CRSP value-weighted index is used to
proxy domestic market returns.

Equity trading volume: Average trading volume (3 years average, CSHTR3)
divided by common shares outstanding (CSHO).

Tobin’s q: It is defined by the book value of debt (DT) plus market value of equity
divided by total assets.

Sales growth: The five-year lease square growth rate of sales (GSALES).

Cash Reserves: It is defined by cash and cash equivalent (CHE) divided by total
asset (AT) net of cash.

Reduction of long-term debt: Reduction of long term debt (DLTR) in the year of
bond issuing divided by total assets multiplied by 100.

Refinance bank debt*high interest indicator: A dummy variable that is 1 if the
purpose of issue is for refinancing bank debt and interest expense
relative to operating income is above its median value.

Refinance bank debt*interest expense: The interaction term between refinance

bank debt and interest expense relative to operating income.

121

DUM91: Yearly dummy for 1991. Hence, all other years are defined in the same

way.

A.3 Estimated adjustment terms for the second stage regressions

¢(Z7)

, which is estimated
1 - <D(Zr)

M1: The adjustment term in Equation (9), that is,

from the first stage probit regression for the subsample when its assigned

value as dependent variable in the probit estimation is one.

- ¢(Z7)

, which is estimated
(NZ?)

M2: The adjustment term in Equation (10), that is,

 

from the first stage probit regression for the subsample when its assigned

value as dependent variable in the probit estimation is zero.

122

APPENDIX B

TABLES

123

00.0205 20 >2: :92: 02.002
:0: 20 252 E2800 0:0 m0_EE:0 >:00> 2: :0: 00:08:00 0:: 00:25:00 :.00:0 002022: 0.: E0: 00:02:00 20
00005 :0:00_0m m:0:0m00:0 2: _0::00 2 00m: 252 52500.60 0:0:0moncm .0E00:_ 0::2000 o: 02:22 00:00x0
7.225 ncm 500 5:00 00:0::0m :0 E22 8:00:25 2: m_ 00:00x0 6225.500 xcmn 00:09.0”. 0.00:? 0:: :0 :062:
2: :2: 2090. m_ 0E00:_ 0.52000 2 02:22 00:00x0 622:. :0 _0>0_ 05 t F m_ 03.0> 00:90.00 2: 0.00:? >580:
0 .5 68065 :0225 :9... 0052050 0 .30: 5:00 026:0 00:05.2 2 m_ 030$ :0 000050 2: t F m_ :2: 0.00:?
>EE00 0 0_ E00 €00 00:0::0m 20000 .22 >0 000E: >300 :0 0:_0> :2:0E 03.0 300 :0 03.0> x000 2: >0 005.00
0_ 0 0.508 005550 0 .022 :0: :0 2,200 :0 mm 0_ 0:02 ::00 2: :0: 052 E20 m.>000_2 2: t F 0_ 002m E0E:w0>:_
:02 .0020 ::0E:m0>:_-:0: 2 z t .m m_ 03.0> 2: .00m :0 < m_ g t .N m_ 0:_m> 2: .0< :0 00.0. m_ 0:005 0:00 0:: :0: 05:2
2020 m.>000_2 2: t F 0_ 05:2 5020 00:05 :0::2:00:: 0E0: ::0E:m0>:_ 2: >_:0 000205 0.0E00 2: .m 2 m 02:20.:
:_ 00:02 :0::2c0ncz xcmn _0_20EE00 25 000205 0.0800 0.: .v 2 F 0:55.00 :_ 202-: :0 002.0?0 0:0 0:22.052
3di 000200 E60 0300 0020-0:0000 :0 2:06.208 200 0.: :002 w 2 P 055.00 220:0: :0____E on: 0:05 :0
020 :062: 2: 2 .0000 :0 :2: .2020 0_ 0:05 :0 020 2: : :0::0E 0:00-30: 0:: mm 0050: m_ :0::0E 0:02 090. 2:

2.0335000» .002: 05052.:
5:5 9.53 E08302: >0 .0 0x000 .222:an 05 >0 :0§.E00:: 20:00 :0 000050 “500 2000 :o 00:0Ezmm
.0 5.2.3.3

124

 

8: 8: v2 :2 28:28:: 8 .oz

2.: 2.: 8.: 8.: mm 8822

.. .. :0: 8.0- 3.: 8.:? 8.: 8.::- 58. .8888 880880
8.: 3.8: 8.: R8: 8.: 2:3 8.: 2:3 08.: 85825-82
00:00x0

8.: 8.8 8.: 2.8 8.: 5:2 8.: 8.8 7.925 .. so: 0.58 88:80
8.: 8.2 3.: 8.2 8.: 8.2. B: 8.8 0585 05828202802” 8.925
m_nmtm>

8.: 5.2- 5.: 8.:- 8.: 3.8. 8: 8.8- .2865: 50: 0.8: 8:28:
8.: 8.: . . 8.: 8.: . . 8:5 8 0508..
8.: 8.8- 8.: 8.::- 8.: 8.8- 8.: 8.8- 858 8 03.0> 888:3
8.: 8.:- 8.: 8.0- . . 8.: :.:- 288 088585 8:980
8.: 2.2 8.: 8.2 . . 8.: 8.: a 058»
8.: 8.:- 8.: 8.:- 8.: 8.: 8.: 8.: :8: :59: 8.00
8.: 8.: 8.: 8.: 3.: 8.: 2.: 8.: 085 :o 5.252
3.00200:

8.: 8.: 8.: 8.: 8.: 8.: 8.: 8.: 522 5:8 8 5.8.8
8.: 8.: 8.: 8.: . .. 2.: mm:- 0522 :58: 580
8.: 2.: 8.: 8.: . . 5.: 8.:- e88 .5288 .28

 

00_0>-n_ 20E:0w

0:_0>-n_ 20E:wm

0:_0>-n_ 208:0w 0:_0>-n_ 208:0m

 

xcmm ::0E:m0>:_
c: 000

xcmm E0E:m0>:_
8v mam

xcmm _0_20EE00
>3QO

.0 8.2.83

xc0m .0_20EEoo

€QO

m0.00:0> E03303.

125

0000.05 20
>2: :0:0:: 02.002 :0: 20 05:0: ::20:00 ::0 00.55:: >..00> 0:: .0: 00:05:00 0: H .:0.:0E:00 :.00:0 000:0-:0:: 0::
80.: 00:25:00 20 0000.0 8.8200 0:0:000::0 0:: 2:80 0: :00: 0E0: 3020200 00800005. 00:00:. 05:22.0
0: 02:02 00:00.0 0225 0:0 :00: ::00 00:05:00. :0 E0: 5:00.25 0:: 0. 00:00x0 0225-50: ::00 00:05:00.
.0.00.:0> 0.:: :0 :0.:0E 0:: :0:: .000. 0. 2:005 05:0.000 0: 02:52 00:00x0 :0225 :0 _0>0_ 0:: :. F 0. 03.0> 020.000
0:: .0.00.:0> >EE:: 0 0. 8:00.05 :0225 :0... .00.2:0::0 o .:00: :.:00 050.0 00:0::2 0: 0. 0:00. :0 000050
0:: :. F 0. :0:: 0.00:0> >EE:: 0 0. :00: 5:00 00:05:20 0:0000 .20: >0 :0:.>.: >:.:.00 :0 0:_0> :0:..0E 00.0 :00: :0 0:.0>
0.000 0:: >0 :0::0: 0. 0 0.500..- 00.3.2.6 o .022 :0: :0 26.00 :0 0m 0. 0:00. ::00 0:: .0: 05:2 :.:20 0.>:00.2 0:: :.
F 0. 0:20 ::0E:00>5-:0z .0020 ::0E:00>5-:0: 0. :. :. .0 0. 0:.0> 0:: .00m .0 < 0. :. :. .N 0. 0:.0> 0:: .20. .0 00< 0. 0:00.
::00 0:: .0: 05:2 :.:20 0.>:00_2 0:: :. F 0. 05:2 :.:20 00:00. :0:..8005 ::00 502008. 0:: 25 0000.05 0.05.00
0:: .m :E:.00 :_ 00:00. :0:..8005 0.:00 .0.0:0E500 >.:0 00:50:. 0.0000 0:: .F :E:.00 :. .:00:-: :0 000.0?0 0:0
0:0.00202 $02 000200 :500 0.000 000:0-000000 :0 2:22:08 0:00 0:: 0002 v 0: F 0:63.00 .256: :0.__:: 0va
0:00. :0 0N.0 :0.:2: 0:: :0:: 000. 0. 0:00. :0 0.00 0:: :. :0:..0E ::00 0:00.050: 0:: 00 :0::0: 0. :0:..0E 0:00. ..0E0 0: H

00000050000 .002: 0>.:0E0:_0 5.3 0:000 ::00-000E.
>0 :0 00.000 .0.0:08800 0:: >0 02:...250003 00:00 :0 0:00:00 :50: 0.000 :0 00:0E.:0w ”:0:...05 0:00. =0Ew
«0 8:58,:

126

 

 

 

000 00: 08002008 :0 .oz

000 00.0 «a 00:00:}:

0. 0 No.0 ::00. 8:0: ::08:0:_:0 009.0085
00.0 00.00: 8.0 00.00: 000:0 505002.802
00:00x0

00.0 00.00 Rd 2.3. 00.25 - 500 0:00 00:00.00:
00.0 3.0 F 00.0 050$ 08005 05:0:000:00:00x0 :00:0:5
:0.:0.:0>

00.0 00.00. 000 00.0- 5:00.05: :00: €00 00:00:00
:00 :30 Rd 00.0- .0000. :0 00.0-:03
8.0 00.00. 80 000:- .0500 :o 020) 00:05:03
8.0 00.? - - 0:00.00 0:20.085 0:00.000
00.0 05; - . 0 0.508
00.0 «to 00.: ~00 .0>-0: 52,90 0000
00.0 0: 8.0 :0: 0:00 :0 0020.2
8.0 00.0 8.0 000: 30:200.: 50:0: >530 :0 >:___:0_o>
00.: 0F: . :- 08:.0> 05:0: 355
8.0 00.0 . - 0:00.00 0:20:00 08F
0:_0>-n. 0:08.:0m. 00.03: 0:08.:0m 00.00.:0> ::0::000:c_

 

x000 ::08:00>c.

x000 .0.0:08800

 

00 5.2.2.3

127

Appendix B3
Gross benefit for firm using the service of chosen underwriter rather than
that of unchosen underwriter: Estimated from alternative model
specification.

The gross benefit is defined as the expected mean net yield (BPS) of bond
issues, had they chosen the unselected underwriter minus the observed net yield
of bond issues. The benefit estimated in column 1 (below the heading of
Commercial Bank Clients) is the predicted BPS of commercial bank client by
using the estimates of investment bank BPS regression (model 3) reported in
Appendix B1 minus the original BPS of commercial bank client. Hence, the
benefit in column 3 is estimated in the same way by using model 1 reported in
Appendix B1. The p-value of t-test is reported, the tested null hypothesis is the
gross benefit is equal to zero.

Panel A. Gross benefit of full sample and by subperiods.
Commercial Bank Clients Investment Bank Clients

 

 

Years Benefit P-value Benifit P-value
1991-1996 3.56 0.48 128.39 0.00
No. of observations 184 1143

1991-1993 0.55 0.93 133.54 0.00
No. of observations 66 664

1994-1996 5.24 0.46 121.25 0.00
No. of observations 118 479

 

Panel B. Gross benefit by credit ratings
Commercial Bank Clients Investment Bank Clients

 

 

Credit ratings Benefit P-value Benefit P-value
Aaa or Aa -15.09 0.02 139.21 0.00
No. of observations 14 173

A or Baa 13.08 0.00 121.74 0.00
No. of observations 136 757

Ba or below -26.85 0.28 143.22 0.00
No. of observations 34 213

 

Panel C. Gross benefit by the issue size (in million dollars)
Commercial Bank Clients Investment Bank Clients

 

 

Issue Size Benefit P-value Benefit P-value
>= 150 -2.33 0.84 133.63 0.00
No. of observations 54 648

< 150 6.00 0.27 121.52 0.00
No. of observations 130 495

 

128

 

00:0.05 0:0 >00: :000:: 000000: :0: 0:0 08:0: ::0:0:00 ::0 00.880: >.:00> 0:: :0: 00:08:00 0: :- .08005 050000
0: 02:0.0: 00:00x0 :00:0::. ::0 :00: :.:00 00:05.0”. :0 8:0: :0.:00:0:5 0:: 0. 00:00x0 :00:0:5...:00: :.:00 00:05:0m
.00.>>:0::0 o .:000 ::00 05:0.x0 00:0:00: 0: 0. 0000. :0 000800 0:: :. F 0. :0:: 0.00.:0> >880: 0 0. :00: :.:00 00:05:01
0:0000 .06: >0 000.2: >:_000 :0 00_0> :0::08 00.0 :00: :0 00_0> 0.000 0:: >0 005:0: 0. 0 0.500... 00.3050 0 .:0:0:
:0: :0 26.00 :0 0m. 0. 0000. ::00 0:: :0: 05:0: :.00:0 0.>:00.2 0:: :. F 0. 000:0 ::08:00>5-:0z 000:0 ::08:00>5-:0:
0. :. :. .m 0. 00_0> 0:: .000 :0 < 0. :. :. .N 0. 00_0> 0:: .00. :0 00< 0. 0000. 0:00 0:: :0: 05:0: :.00:0 0.>:00.2 0:: :. F 0. 05:0:
:.00:0 .00:00.::>0 0.05m .0> :.:0>I :0 :00.:00800 0:: 80:: 00:08:00 0. AN: :.:0>I :0: :05 .AF: 0.:0>_._ :0: 0:00.::>0
:.:0>... 0: 00:00800 0. 0:00.::>0 0000-00 :0:; :0.:08.:00 :.00:0 000:0-:0:: 0:: 80:: 00:08:00 0:0 0000.0 :0.:00_00
000:000::0 0:: .2800 0: :00: 08:0: ::08:00._:0 000:000::m. 00000. :0::.:2:0::0-:.:0>: 0:: >80 0000.05 0.0800 0:...
.:00:-: :0 000.0>-0 ::0 0:0.000:00: $0.0: 0:00:00 :500 0.000 000:0-::0000 :0 0::0.0_..:000 0:00 0:: ::000: v 0: F 0:80.00

::0: 2050300 00:00.00 000:000::0
5.3 0:00.0:>0 0.:0>: >0 005:3:0000 00:00 :0 008009.00: 0000.50 5.00 0.000 3:00.000 :0 00:05.:0m
«0 50:08.4.

129

 

 

 

130

0:0 00 0000020000 :0 .oz

000 0:..0 .0. 00:00.00.

00.0 :00: 00.0 00.00- :00: 50500.00 0000000000
00.0 00.00: 00.0 00.00: 000:0 2050005002
00:00x0

00.0 00.0: 00.0 0000 :00:0::. . :000 ::00 02:05:00.
00.0 00.00 00.0 00.00 0800:. 0:.:0:000:00:00x0 :00:0::.
.0.00.:0>

00.0 000:- 0:.0 00.00- 5:02:00 :00: 0.000 00:00:00
00.0 00.0 00.0 00.0 .0000. :0 00.0.03
00.0 00.00. 00.0 0:. :.0- .0.000 :0 0200 00:05:03
00.0 :0. :- :00 00. :- 0:0000 00:00:80. 0000000
00.0 00.: 00.0 00.0 0 0.0.00:
00.0 0:.0 00.0 00.0 ..:>-0: 526:0 00_00
:00 0:: 00.0 00.0 0000. :0 0020.2
00.0 3.0 00.0 0:.0 0000.00: 50:0: 0.000 :0 >:___:0_0>
00.0 0:.0 00.0 00.0 05200 00.00:: 0.000
:00 00.0- 00.0 00.0. 0:0000 0:00:00: 08:
m:_m>-n_ 3mE=ww m:_m>-n_ 99::sz 0055.5) 50950.20:—

 

.0: 25>: .:: 25>:

00 0000000.

Appendix BS

Additional estimation of gross benefit for issuing firms by credit ratings
using the service of chosen syndicate rather than that of unchosen

syndicate.

This table reports additional estimation of gross benefit by using the BPS
regression of PurelB syndicate without the endogenous adjustment term. The
benefit estimated in column 1 (below the heading of CB-Lead clients) is the
predicted BPS of CB-Lead clients by using the estimates of regression PurelB (3)
reported in Table 2.7 minus the original BPS of CB-Lead clients. Hence, the
benefit in column 3 is estimated in the same way for Hybrid clients. CB-Lead
stands for a commercial bank as the lead underwriter. Hybrid represents an
investment bank as the lead underwriter while other commercial banks are
involved in an issue. PurelB stands for the underwriter syndicate includes only
investment banks. The p-value of t-test (two-tailed) is reported, the tested null
hypothesis is the gross benefit is equal to zero.

 

 

 

CB-Lead Clients Hybrid Clients
Credit ratings Benefit P-value Benefit P-value
All ratings 9.84 0.05 6.92 0.05
No. of observations 184 345
Aaa or A3 -7.24 0.22 10.00 0.38
No. of observations 14 46
A or Baa 19.30 0.00 5.59 0.05
No. of observations 136 215
Ba or below -20.97 0.39 8.65 0.45
No. of observations 34 84

 

131

BIBLIOGRAPHY

132

 

Bibliography

Anand, Bharat N. and Alexander Galetovic (1997), Information, non-excludability,
and financial market structure, Working paper, Yale School of
Management and Universidad de Chile.

Ang, James S. and Terry Richardson, 1994, The undenivriting experience of
commercial bank affiliates prior to the Glass-Steagall Act: A re-

examination of evidence for passage of the act, Journal of Banking and
Finance 18, 351-395.

Beatty, Randolph P. and Jay R. Ritter, 1986, Investment banking, reputation, and

the underpricing of initial public offerings. Journal of Financial Economics
15, 213-232.

Berger, Allen M, Rebecca S. Demsetz and Philip E. Strahan (1999), The
consolidation of the financial services industry: Causes, consequences,

and implications for the future, Journal of Banking and Finance 23, 195-
214.

Bharat, A. Jain and Omesh Kini, 1999, On investment banker monitoring in the
new issues market, Journal of Banking and Finance 23: 49-84.

Bhargava, Rahul and Donald R. Fraser, 1998, On the wealth and risk effects of
commercial bank expansion into securities undewvriting: An analysis of
Section 20 subsidiaries, Journal of Banking and Finance 22, 447-465.

Bhattacharyya, Sugato and Vikram Nanda, 1998, Client discretion, switching
costs and financial innovation, Working paper, University of Michigan.

Billett Matthew Thayer, Mark J. Flannery, and Jon A. Garfinkel, 1995, The effect

of lender identity on a borrowing firm’s equity return, Journal of Finance
50, 699-718.

Boot, Amoud W. A., Todd T. Milboum and Anjan V. Thakor 1998, Expansion of
banking scale and scope: Don't banks know the value of focus? Working

paper, University of Amsterdam, London Business School, and University
of Michigan.

133

 

Booth, James R. and Richard L. Smith, II, 1986, Capital raising, underwriting and
the certification hypothesis, Journal of Financial Economics 15, 261-281.

Carey, Mark, Mitch Post and Steven A. Sharpe, 1998, Does corporate lending by
banks and finance companies differ? Evidence on specialization in private
debt contracting, Journal of Finance, 53(3), 845-878.

Chemmanur, Thomas J. and Paolo Fulghieri, 1994, Investment bank reputation,

information production, and financial interrnediation, Journal of Finance,
49(1), 57-79.

Chung, Ching-Fan, 1997, Estimating and testing polychotomous choice
selectivity models under alternative distributional assumptions: Wages in
the public and private sectors, Working paper, National Taiwan University.

Cummins, J. David, Mary A. Weiss and Hongmin Zi, 1997, Organizational form
and efficiency: An analysis of stock and mutual property-liability insurers,
Working Paper Series of the Wharton Financial Institutions Center.

Diamond, Douglas W., 1991, Monitoring and reputation: the choice between

bank loans and directly placed debt, Journal of Political Economy, 99(4),
689-721.

Diamond, Douglas W., 1991a, Debt maturity structure and liquidity risk, Quarterly
Journal of Economics 106, 709-737.

Eccles, Robert G. and Dwight B. Crane, 1988, Doing Deals: Investment Banks at
Work, Harvard Business School Press.

Fama, Eugene F., 1985, What’s different about banks?, Journal of Monetary
Economics 15, 29-39.

Gande, Amar, Manju Puri Anthony Saunders, and lngo Walter, 1997, Bank
underwriting of debt securities: modern evidence, Review of Financial
Studies, 10(4), 1175 -1202.

Gompers, Paul and Josh Lerner, 1998, Conflict of interest in the issuance of

public securities: Evidence from venture capital, NBER Working Paper No.
6847.

134

Greenbaum, Stuart |., George Kanatas, and ltzhak Venezia, 1989, Equilibrium
loan pricing under the bank-client relationship, Journal of Banking and
Finance 13, 221-235.

Holland, John, 1994, Bank lending relationships and the complex nature of bank-

corporate relations, Journal of Business Finance and Accounting, 21(3),
367-393.

Hamao, Yasushi and Takeo Hoshi, 1997, Bank Underwriting of Corporate Bonds:
Evidence from Japan after 1994, Working Paper, Columbia University and
Osaka University.

Houston, Joel and Christopher James, 1996, Bank information monopolies and

the mix of private and public debt claims, Journal of Finance 51 (5), 1863-
1889.

Jain, Bharat A. and Omesh Kini, 1999, On investment banker monitoring in the
new issues market, Journal of Banking and Finance 23, 49-84.

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

Johnson, James M and Robert E. Miller, 1988, Investment banker prestige and

the underpricing of initial public offerings, Financial Management 17, 19-
29.

Kanatas, George and Jianping, Qi, 1998a, Underwriting by Commercial Banks:
Incentive Conflicts, Scope Economies, and Project Quality, Journal of
Money, Credit, and Banking 30(1), 119-133.

Kanatas, George and Jianping, Qi, 1998b, Underwriting by Commercial Banks:
Incentive Conflicts, Scope Economies, and Project Quality, Journal of
Money, Credit, and Banking 30(1), 119-133.

Khanna, Naveen and Sheri Tice, 1998, Wal-Mart's easy targets: A test of the

chain store paradox. Working paper, Michigan State University and Tulane
University.

135

 

Kroszner, Randall S. 1998, Rethinking bank regulation: A review of the historical
evidence, Journal of Applied Corporate Finance, Summer 1998, 48-58.

Kroszner, Randall S and Raghuram G. Rajan, 1997, Organization structure and
credibility: evidence from commercial bank securities activities before the
Glass-Steagall Act, Journal of Monetary Economics 39, 475-516.

Maddala G. S., 1983, “Models with self-selectivity", Chapter 9 in Limited
Dependent and Qualitative Variables in Econometrics, Econometrics
Society Monographs No. 3, Cambridge University Press.

Mohanram, Partha and Ashish Nanda, 1998, When do joint ventures create
value, Working paper, Harvard Business School.

Nanda, Vikram and Vincent A. Warther, 1998, The price of loyalty: An empirical
analysis of underwriting relationships and fees, Working paper, University
of Michigan and University of Chicago.

Nanda, Vikram and Youngkeol Yun, 1997, Reputation and financial
intennediation: An empirical investigation of the impact of the lPO
mispricing on underwriter market value, Journal of Financial
Intermediation, 6, 39-63.

Nayar, Nandkumar and Michael S. Rozeff, 1994, Ratings, commercial paper, and
equity returns, Journal of Finance 49(4), 1431-1449.

Opler, Tim, Lee Pinkowitz, Rene Stulz, Rohan Williamson, 1999, The
determinants and implications of corporate cash holdings, Journal of
Financial Economics 52, 3-46.

Pichler, Pegaret and William Wilhelm, 1998, A theory of the syndicate: Form
follows function, Working paper, Boston College.

Puri, Manju, 1994, The long-term default performance of bank underwritten
security issues, Journal of Banking and Finance 18, 397-418.

Puri, Manju, 1996, Commercial banks in investment banking: conflict of interest
or certification role?, Journal of Financial Economics 40, 373-401.

136

Rajan, Raghuram G., 1992, Insiders and outsiders: the choice between informed
and arm’s-length debt, Journal of Finance 47(4), 1367-1400.

Saunders, Anthony, 1999, Consolidation and universal banking, Journal of
Banking and Finance 23, 693-695.

Sharpe, Steven A., 1990, Asymmetric information, bank lending, and implicit

contracts: a stylized model of customer relationships, Journal of Finance
45(2). 1069-1087.

Song, Wei-ling, 1999, Bank-entry benefits and regulation distortion in the

corporate security undenrvriting business, Doctoral dissertation, Michigan
State University.

Titman, Sheridan and Brett Trueman, 1986, Information quality and the valuation
of new issues, Journal of Accounting and Economics 8, 159-172,

 

White, Lawrence, 1996, Technological change, financial innovation, and financial

regulation in the US: The challenges for public policy, Working paper,
Stanford University.

 

 

137

nICH
I
I

IjjIIIjIIIIIIIIjIIIIIIIIjijijI“

2? 13(31