Financial Contracting: Banks versus Banks

lasagnaseniorΔιαχείριση

28 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

267 εμφανίσεις


1

Financial Contract
ing
: Banks versus Banks


Bill Francis

Lally School of Management and Technology,

Rensselaer Polytechnic Institute

110, 8th Street, Pittsburgh Building

Troy, New York 12180
-
3590

Telephone: 518 276 3908

E
-
mail:
francb@rpi.edu


Iftekhar Hasan

Schools of Business
,

Fordham University

1790 Broadway, 11
th

floor

New York, NY 10019

and Bank of Finland

Telephone:
646 312 8278

E
-
mail:

ihasan@fordham.edu


LiuLing Liu


Lally School of Ma
nagement and Technology,

Rensselaer Polytechnic Institute

110, 8th Street, Pittsburgh Building

Troy, New York 12180
-
3590

Telephone: 518 8676310

E
-
mail:
liul5@rpi.edu








Corresponding author: please send all correspondence to
liul5@rpi.edu


2

Abstract

We use a sample of 1,602 syndicated loans

issued to 495 commercial banks from 1995 to 2007
in 42 countries to examine the impact on a bank’s borrowing cost if the lender is another bank.
We find that the cost of bank loans to banks is significantly influenced by banking regulations
and the banki
ng environment within each country. Our results further demonstrate that the effect
of bank regulations on reducing the cost of banks’ financing is higher if the lead lender is from a
different country than the borrower. This relationship is most pronounce
d among borrowers
without any previous lending or cooperation relationship, and borrowers without rating
s
. Finally,
we find that the regulation effect on the reduction of banks’ borrowing cost is more pronounced
in a highly concentrated banking industry or

in countries with weak legal systems.



3

1. Introduction

Over the past two decades, bank loans have grown as a large and increasingly important source
of debt financ
ing in the global capital market.
With this growing importance a
large body

of
research h
as
been devoted to understanding bank loan contracts
(e.g., Strahan, 1999; Qian and
Strhan, 2007; Bharath et al., 2008; Graham et al., 2008; Bae and Goyal, 2009; Bharath et al.,
2009; Chava e
t al., 2009; Hao et al., 2011). However, most of this literature
focuses on corporate
borrowing, which largely ignores another important type of borrowers
-

notable banks. Given the
significant differences between banks and other corporate borrowers it raises the following
questions: How different are financial contracts

if the borrower is a bank?
1

As banks increasingly
engage in borrowing activities in addition to their traditional lending activities, this question
begins to emerge as increasingly poignant. In this paper we examine the factors, the national
contracting
factors in particular, that determine the structure of financial contracts when the
borrower and lender are both banks.

In order to answer this empirical question, we
collect the

international bank loans

made
to the banking sectors

from the LPC’s DealScan
.

We notice that the bank
-
to
-
bank loans also
exist in the traditional interbank market, however, we believe that those syndicated bank loans
form DealScan are likely to serve our research interest better for the following reasons.
First, in
most countries
the interbank loans are quoted through OTC, and therefore, the bilateral amounts
and identity of each borrower and lender are usually not available.
2

However, in the syndicated
loan market, loans are made directly between bank borrowers and lenders (or syn
dicate lenders),
hence we can track the contracting environment in both borrowing and lending countries. Second,
in the traditional interbank loans, the amount and interest rate are majorly quoted on a simple
overnight money market basis (or short maturity

of less than one month), while the bank
-
to
-
bank
loans usually have much longer maturity (average of 30 months). Therefore, our bank
-
to
-
bank
loan setting allows us to investigate the loan contracts as a whole by
analyzing
the determinants



1

That banks are different and special is not debatable. Numerous papers have established that this is in fact the case. In fa
ct in
the corporate literature empirical studies routinely remove banks from the sample due to these differences.

2
The exemption
is the studies that use Italy’s e
-
MID data, which includes the bilateral amounts and the identity of each borrower
and lender in interbank transactions because of the data requirements of banking supervision
(
Cocco et al.,2009; Affinito, 2011,
Angelini et

al.,2011
)
.







4

of not only the
price but also the non
-
price contract terms.
More importantly
, the global
interbank market is largely un
-
integrated, where different transaction rules are adopted in
individual countries
. Therefore, it is

impossible for us to analyze
how
the county
-
level
contracting factors shape the bank
-
to
-
bank loan contracts

using the traditional interbank market
setting
.


We believe
that the answer to

this question is

important for several reasons.
First, as the
most important credit providers
, banks


own ability to o
btain financing has significant impact on
the finance and investment policies of firms

and the growth of economies

(Faulkender and
Petersen, 2006; Lemmon and Roberts, 2011)
.
3

Therefore, an understanding of how the
availability and cost of banks’ funding re
sponds to the contracting environment is crucial for
both the suppliers and the users of capital. As pointed out above,

in addition to retail deposits
and short
-
term interbank funds, banks increasingly borrow from the
bank

loan market which
serves as an i
mportant channel to solve the liquidity needs of banks, particularly during a crisis
period (e.g., the Asia crisis in 1997).

4


T
hus, research on the determinan
ts of the cost of banks’
financing

through the
bank
loan market could be economically meaningful
.

Second, examining the determinants of the bank
-
to
-
bank loan contract furthers our
understanding

of bank loan contracts.
Banks are
arguably significantly
more opaque than
nonfinancial firms (Slov
in et al., 1992; Morgan, 2000).
The fact that depositors h
ave the right to
withdraw at any time at par
highlights

the frag
ile capital structure of banks.
To prevent bank
panics, banki
ng systems are highly regulated. All

of th
e
se special features of the banking
industry
imply
that
in all likelihood that
loan con
tracts
are significantly
differen
t for banks and
corporate borrowers.
This further suggests that factors that impact corporate loan contracts could
have a significantly different impact on bank
-
to
-
bank loan contracts, or there might be some
banking indust
ry specific factors exists in shaping these contracts.
5

Thus,
a thorough



3
When banks face difficulties in
acquiring

financing, they become more conservative by cutting back on lending and

by

charging
borrower
s

higher rates (Kashyap and Stein, 2000; Hubbard, Kuttner, and Palia, 1999), which in turn, affects the finance and
investment policies of firms (Faulkender and Petersen, 2006; Lemmon and Roberts, 2011).

4
For example, the new issuance volume suddenly jumps
by 150% in 1997, when the financial crisis hit the entire region

initially
.
This implies that banks tend to recover through the channel of borrowing from peers.

5
In an unreported table, we compare different aspects of the characteristics
of loans to bank b
orrowers and loans to non
-
financial
borrowers,
such as loan size, loan maturity, interest rates
,

and fees paid. We find that loans issued to bank borrowers are
significantly smaller, are characterized by shorter maturities,
and
enjoy substantially lower sp
reads and fees
as
compared to those
loans
issued
to

non
-
financial firms. In addition, covenants, collateral
,

and performance pricing provisions are almost non
-
existent
for the bank
-
to
-
bank loan contracts, while all three features of loan contracts are argu
ed within the extant literature to serve as
monitoring mechanisms in loans to non
-
financial firms.


5

understanding
of
bank
-
to
-
bank loan contracts

would be helpful in furthering our knowledge of
debt financial contracts
, which arguably is the most important financial contract
.

Third,

the use of bank loans as a source of financing by banks is pervasive and is
growing in importance. The

statistics compiled by LPC’s Deal
S
can
show

that over the 1995 to
2007

time
-
period
, the banking industry obtained
, on average,
$130 billion

in

new loans each year,
which represents 15% of the aggregate number
,

or 6% of the aggregate dollar amount of loans
issued in the global
bank
-
loan market. In addition, over our sample period,

banks’ borrowing in
the
global

loan market has increased by 120%
in terms of the number of loans and 80% in
terms
of the
total volume of loan issuances (Figure 1)
. This phenomenon suggests that our sample is
representative for the interbank borrowing and lending, which allows us to investigate in a global
setting,
the
determinants of the bank
-
to
-
bank loan contract
s
, specifically
the cost of this
particular funding source
.

B
ank regulations and market structure
are the key elements in the banking literature
.
There are theoretical reasons, as well as empirical evidence, su
ggesting that

t
he major function
attributed to bank regulation is that it changes the operating structure and riski
ng
-
taking behavior
of the banks, and hence impacts
the
stability of the financial system
and individual bank
performance

(
Barth, Caprio, and
Levine,
2004, 2006, and 2008,
BCL henc
eforth;
Demirguc
-
Kunt
et al.,2004
,

Laeven and Levine ,2009
).
M
arket structure of the banking sector
also
has a
nontrivial impact on
the price and non
-
price of bank product, bank efficiency and profitability,
risk taking,

the stability of the financial system, the financing of firms and

overall economic
growth in a cross
-
section of countries (see
for example literature reviews by

Berger
, Demirg
ü
ç
-
Kunt, Levine, and Haubrich, 2003)
.

To test the effect of regulatory environm
ent and banking sector structure on the cost of
banks’ debt financing,
u
sing a sample of 1,602 loans issued to 495 commercial banks
from
42
countries
over the
1995 to 2007

time
-
period
, we examine how the pricing term (all in spread
drawn, AISD), nonprice terms (loan maturity), and syndicate structure (the number of lenders)
vary with banking regulations, industry
environment and other country
-
level variables such as
in
stitutional quali
ty (as measured by information sharing, property rights and creditor rights) and
a country’s level of financial and economic development. In our tests, we also control for bank
borrower and loan characteristics that are likely to shape contract terms thro
ugh variations in
credit risk or loan demand. However, we view bank regulations as the most important

6

determinants of the bank
-
to
-
bank contracting environment, and hence it is the main focus of our
paper.

W
e draw several broad conclusions from our
finding
s.
First,

we find that bank
regulations,
banking environment
and
institutional development all significantl
y impact banks’
borrowing cost.
Specifically, we find that commercial banks operating in countries with
regulations favor traditional banking (or r
estrict the financial conglomerates)

and
more
transparent

accounting disclosure, pay significan
tly lower average loan spreads.
Regarding
banking environment, we find that loans issued to

the

banking industry with a high fraction of
foreign
presence

or
a
m
ore concentrated

environment
, on average, pay sig
nificantly lower loan
spreads. Consistent with Qian and Strahan (2007) and Bae and Goyal (2009), t
he institutional
quality related to information sharing
,
creditor rights

and property rights

in
the
borrower
’s

country are negatively and significantly associated with loan spreads

An important
fact that should be noted regarding

these results is that lenders may give
loans to their
ownership connect
ed bank borrowers for various rea
sons that go against economic
rules. T
herefore, the
effect of country
-
level factors

that
we
just
iden
tified may not be as
important. To be specific, s
ince both p
arties in the bank
-
to
-
bank loan contracts

are from
the
banking

industry,
it is highly likely that a borrower is, in fact, t
he branch

or
subsidiary

of
a
certain

lender
in their bank loan syndicate
(or vice
-
versa). In those cases,
the s
yndicate
d

loan
market may simply become a platfor
m for internal capital transfer, where pricing formation and
contract design are irrelevant to n
either the contractual environment nor the riskiness of the
borrower.
To address
the above
concern,
we re
-
run all our estimations by excluding those
ownership
-
connected loans, and our results

remain
substantially
unaffected
.

Second, looking beyond just t
he various regulations, banking structure
s

and institutional
quality
in the borrower country, we also explore whether differences

(or gaps)

in the
country
-
specific

determinants between borrower and
lead
lender countries influence the

cost of bank
-
to
-
bank loans.
Our results strongly indicate that the distance
s

in bank regulations, banking
environment
and institutional development are also important

for the bank
-
to
-
bank loan contracts.
Importantly, the signs of the coefficients are gen
erally consistent with our base
-
model results,
suggesting that if the circumstances in their
own
countries are weaker than

those of the borrower
country,

lenders place additional value on
regulations encouraging traditional banking and

transparent accounti
ng disclosure, banking environment with high

foreign
presence

and

7

concentrated
structure, and institutions with strong information sharing, creditor rights and legal
rights. Additionally, we also control for the culture distance with regards to language,
geographic region, and colony tie when we estimate the “Gap” model. We find that lenders
charge significantly higher loan spreads when the borrower is culturally distant which is
consistent with the first
-
order pricing effect of home bias.

Third, bank re
gulations affect domestic lenders and foreign lenders differently.

Specifically, we find that the benefit of the regulations, that restricting financial conglomerates
and encouraging transparent accounting disclosure, on reducing the cost of banks’ financi
ng is
higher if the lead lender is from a different country than the borrower. In addition,
this
relationship is most pronounced among borrowers without any previous lending or cooperation
relationship, and borrowers without
a
credi
ble rating from S&P or
Moody’s.
These results
suggest that foreign lenders

rely more on formal regulation mechanisms to mitigat
e

the
information asymmetry problems
due to their
monitoring

disadvantage, particularly

when
they
lend

to highly opaque counterparties,
which is
consis
tent with finding by Mian(2006) and Qian
and Strahan(2007).

Despite the above results, we further investigate the economic context of regulation
benefits by conducting a series of additional tests to explore whether the level of banking
environment and th
e legal and institutional environment inf
luence the degree of regulation

effect
on the bank
-
to
-
bank loan contracts.

We find that regulations that encourage traditional banking
and transparency reduce banks’ borrowing cost further in a high monopoly bankin
g industry. In
addition, o
ur results
also
suggest that bank regulations and country institutions are substitutes in
determining the cost of banks’ loan financing. Specially, marginal improvements in bank
regulations produce greater reductions
on

the cost
of loan
s

for banks from weak legal systems
than those from well
-
d
eveloped institutional systems.

Finally, with regard to the non
-
price terms of bank
-
to
-
bank loan contracts, we find that
loans in countries

that restrict financial conglomerates and encourage

transparent bank financial
disclosure, as well as in countries that have a
more concentrated banking sector str
ucture, have
longer maturities. We find that higher foreign presence and a more concentrated banking
environment
are associated with larger syn
dicate size
, although we do not observe a significant
impact of bank regulations on the bank
-
to
-
bank syndicate structure.



8

Our paper is broadly related to the growing literature regarding the various funding
sources of banks, and specifically the interban
k market that justifies the lender
-
of
-
l
ast
-
resort role
of central banks.
6

Some studies investigate whether there are some bank credit quality effects,
apart from Central Bank money policy action, on the interbank interest rates.
Furfine (2001)

use
unique
overnight interbank market data in
the Fed Funds market

find that
larger banks bo
rrow
and lend at more favorable
terms.
They also

find that more profitable banks lend less, and banks
with a higher proportion

of non
-
performing loans lend more and borrow
less. Thus ba
nks that
have better investment
opportunities tend to be net borrowers.

The main weakness of this study
is that the sending and receiving banks identified in the Fed Fund dataset may not be the actual
transaction counterparties.
Angelini et al
. (2011) analy
ze

the Italian interbank market during the
period preceding and following the crisis
. Focusing on the longer terms interbank loans
(
maturities of 1
-
week or beyond
) of 21 contracts, t
hey find that before the financial crisis
interbank rate wer
e insensitive to bank
-
specific characteristics, whereas afterwards they became
sensitive to borrowers' creditworthiness.

Other studies investigate the lending relationships in the
interbank market. For example,
Affinito (2011) matches the bilateral amounts

and the

identity of
each interbank borrower and lender
and
show that interbank customer relationships

exist in Italy,
persist over time, and
functioned well over the crisis
.

Cocco, Gomes, and
Martins (2009) use the
Portuguese interbank market data and fin
d that smaller
banks, banks with lower return on assets,

banks with a higher proportion
of non
-
performing loans, and banks subject to more vola
tile
liquidity shocks rely more
on

the interbank lending relationships. Similar to our data source,
Hale (2010) u
ses
the loan
-
level bank
-
to
-
bank loans from the Loan Analytics database and adopts

a network approach to investigate the impact of the crisis on banking relationships.
She
documents a negative impact of the global financial crisis of 2008
-
09 on the formatio
n of new
customer relationship. Our paper extends our understanding on banks’ inter
-

borrowing and



6

Most of the literature has concentrated on the overnight segment of the interbank market. See the seminal contributions by
Spindt and Hoffmeister (1988), Hamilton (1996, 1997), and, more recently, Bartolini, Bertola, and Prati (2001, 2002), and Pra
ti,
Bart
olini, and Bertola (2003). Similar to the interbank loans, the bank
-
to
-
bank loans in our study are rarely collateralized and
therefore expose the lending institution to the risk of default by the borrowing institution. The theoretical models of inter
bank
m
arket highlight the importance of interbank markets as distributors of liquidity (Ho and Saunders, 1985; Bhattacharya and Gal
e,
1987; Allen and Gale, 2000; Freixas et al., 2000).






9

lending by showing the importance of country
-
level contracting environment in shaping the
bank
-
to
-
bank loan contracts.

Second, this paper is related to a st
rand of literature that examines the international bank
loan contracts (Esty and Megginson, 2003; Esty, 2006; Qian and Strahan, 2007; Bae and Ghoyal,
2008). Although, we are not the first study that
examine the pricing of international syndicated
loans and

the contract features of these loans (see, e.g.,
Qian and Strhan, 2007; Bae and Goyal,
2009
) our paper, to our knowledge, is the first to examine the impact of borrowers being banks
on the cost of bank loans. Third, we show that there is an implicit trus
t among borrower and
lender in bank
-
to
-
bank loans that are not present in loans to non
-
bank borrowers as important
monitoring mechanisms that are present in the latter type of loans are not present in bank
-
to
-
bank
loans. Finally
, our paper contributes to
the law and finance literature, pioneered by La Porta,
Lopez
-
de
-
Silanes, Shleifer,and Vishny (1997, 1998; LLSV hereafter).

We find that not only the
quality of the
borrower but also
country
-
specific
risks due to un
-
favorable bank regulations,
banking stru
cture, or weak legal institutions

play an important part in determining the contract
features of
bank
-
to
-
bank loan
s
.

The rest of the paper proceeds as follows. Section 2 reviews the literature and develops
the hypotheses. Section 3 describes our data and
sample construction procedure. Section 4
presents our methodology, major results, and robustness tests. Section 5 concludes.

2. Determinants of Bank
-
to
-
Bank Loan Contracts: Bank Regulations, Banking
Environment, and Other Factors

In this section we define
our measures of bank regulations and banking environment.
W
e review the various components of above country factors and discuss their potential effects on
banks’ private debt financing.
In addition, we provide a description of
other country
-
level
variables

such as
in
stitutional quality (as measured by information sharing, property rights and
creditor rights) and a country’s level of financial and economic development that we also
incorporated in our empirical models.


2.1. Bank Regulations

2.1.1. Measuring

Bank Regulations


10

We obtain bank
regulations

information from three World Bank surveys conducted by
Barth, Caprio, and Levine (BCL henc
eforth,
2004, 2006, and 2008). The first survey was
conducted in 1998 and covered 117 countries. The second survey was co
nducted in 152
countries in 2003

and

reflect
s

the regulatory situation at the end of 2002. The latest survey
was
updated in 2007,
and
characterize
s

the regulatory status of 142 countries. The three surveys
provide
a
comprehensive picture of
various aspects

of
bank
regulation

and supervision

across
countries
over the past decade,
and
we focus on the sample period of 1995
-
2007. Specifically,
our regulatory variables for the period 1995 to 1999 use the values
from

the first

survey
. The
regulatory variables for

the period 2000 to 2003 use the values
from

the s
econd survey
. The
regulatory variables for the period 2004 to 2007 use the values
from

the
third survey
.
The two

regulations
we focus on are regulatory

on bank
ing
-
commerce link

(
i.e. Financial Conglomerates

Restriction
) and bank accounting disclosure (
i.e. Financial Statement Transparency
)
.

To be specific
,

the index of
Financial Conglomerates Restriction

measures the extent to
which banks may own and control nonfinancial firms, the extent to which nonfinanc
ial firms
may own and control banks, and the extent to which nonbank financial firms may own and
control banks. For each question,
the
value equals 1 if “Unrestricted”, 2 if “Permitted”, 3 if
“Restricted”,
and

4 if “Prohibited”. Higher values of the index
indicate more
favors of traditional
banking.

We use

Financial Statement Transparency

to measure the degree to which banks face
regulatory restrictions on their accounting disclosure. It is constructed from Barth et al. (2006)
based on the following five q
uestions: (1) whether the income statement includes accrued or
unpaid interest or principal on non
-
performing loans; (2) whether banks are required to produce
consolidated financial statements, including non
-
bank financial affiliates or subsidiaries; (3)
w
hether the off
-
balance sheet items are disclosed to the public; (4) whether banks’ directors are
legally liable for misleading or erroneous information; and (5) whether the penalties have been
enforced. The indicator potentially ranges from 0 to 5, where h
igher values indicate greater
restrictions.

2.
1.2
. Bank Regulations
and the Cost of Bank
-
to
-
Bank Loans

The extent to which banking and commerce should be allowed to mix is a subject of
long
-
lasting public policy debate. The most cited potential risks from allowing the mixing of

11

banking and commerce often are the concerns about conflicts of interest, expansi
on of the safety
net

and too big to discipline
.

To be specific
, the conflicts of interest may arise when a bank
refuse
s

to lend to the affiliate’s competitors

or

grand credit preferentially

to its commercial
affiliate(s)
. In either case, banks’ income redu
ces. Safety
-
net issues may arise when the parent
organization shift bank
funds to its nonbank affiliates e.g. require the

bank

to

buy assets at
inflated prices

from the affiliate
,

or inject

capital

to the affiliate through a cheap loan. As a result,
the pa
rent could shift potential losses to the bank, ultimately threatening the safety and
soundness of banks.
In addition, allowing for mixing banking and commerce may lead to
financial conglomerates

that are

too politically and economically powerful to discipl
ine

(Laeven
and Levine, 2005).

From this perspective, prudential regulatory that separate banks from
commerce
should

reduce the likelihood that commercial affiliates misuse banks’ fund, and hence
lower

credit risk of bank borrowers. As argued above, due to

the lower risk that bank
-
commerce
link hurts the safety and soundness of bank borrowers, the monitoring costs should be reduced
when lenders make loans to those countries that restrict the level of banking and commerce mix.
Lenders may choose to share, or

pass on, these savings to its bank borrowers through reduced
borrowing cost, more flexible loan contract terms, or a combination of both.

Among the potential benefits of mixing banking and commerce are operational
efficiencies (including economies of sca
le, economies of scope), informational efficiencies, and
diversification benefits. To the extent that the presence of banking
-
commerce link reduce

the

information asymmetry in corporate lending
, enhanc
e

banks’ monitoring
and contract
enforcement
ability (S
antos, 1999; Barth, Brumbaugh, and Wilcox, 2000, and Haubrich and
Santos, 2005), we expect that, a bank in a country
that encourages the mix of banking and
commerce may

have better
performed lending portfolio
, which
could

transfer
in
to lower default
rate w
hen it
comes to their own borrowing
.
7

In contrast to these conflicting effects regarding restrictions of financial conglomerates,
regulatory encouraging transparent bank accounting disclosure appears to be universally
beneficial. Specifically, a
restrictiv
e regulation that forces accurate information disclosure



7
Existing empirical studies provide mixed results regarding to the efficiency of such bank
-
commerce link regulation. Literature
supporting the presence of
universal banking suggest
s

that unrestricted banks are more operationally efficient (Vennet, 1999),
e
njoy positive diversification benefits ( Eisenbeis and Wall, 1984; Kwan and Laderman, 1999), and

are

associated with a lower
probability of banking crisis (Barth, Caprio, and Levine, 2001a, Barth, Caprio, and Levine (1999). However, Beltratti and
Stulz(20
10) find that the restriction on universal banking is positively associated with the performance of banks during the crisis,
although they find no evidence that such restrictions reduce banks
’ risk before the crisis.


12

reduc
es

the
information asymmetry between bank borrowers and their industry peer lenders
(Bharath et al., 2008),

enhanc
es

bank

borrowers’

excess return

and growth (Cole et al., 2008),
and

promot
es

the stability of host countries’
bank
ing

sector

(B
arth et al., 2004).

Therefore, we
expect that the bank
-
to
-
bank loans issued to a country with
more
restrictive regulations on bank
-
accounting disclosure, all else being equal, will have lower loan spreads a
nd
more relaxed
contract

terms.


2.2
.
Banking Environment

2.2
.
1.

Measuring
Banking Environment

In addition to bank regulations
, we also investigate how the cross
-
country variations in
the structure of the banking sector

impact banks’ own borrowing.
Our focuses are two
perspectives:
Banking Concentration

and

Foreign Presence
.

We

measure
Banking C
oncentration

as the fraction of bank assets held by the three
largest commercial banks in the country
. It is computed using the BankScope database.
Consid
eri
ng that the coverage of BankS
cope increases over the sample period, the change in
coverage might drive the change in concentration measure. To mitigate such biases we use an
alternative measure of concentration in an unreported robustness test by averag
ing

the annual
concentration value over the sample period 1995
to

2007. In addition, our results stay unaffected
after using other different measures of concentration
,

such as the fraction of bank deposits held
by the five largest commercial banks, or the HHI

of bank assets (deposits) in a given country.
We use
Foreign
Presence
, obtained from BCL (2006, 2008), to control for the ownership
composition of the banking industry.
It measures the
share of banking system assets
that
are held
in foreign
-
owned banks
.


2.2
.
2.

Banking Environment

and the Cost of Bank
-
to
-
Bank Loans

Economic theory offers conflicting predictions about the relationship between

banking
concentration and the cost of banks’ private debt financing. On the one hand, the
“concentration

stability” view

(see
Allen and Gale, 2004
)

suggests that
banks in more concentrated markets use
their market power to extract rents from customers while paying significantly lower rates on their
retail deposits, which in turn boosts bank pro
fit and provides a “buffer” against adverse shocks

13

(Berger and Hannan, 1989; Hannan, 1991). The boosted bank profits would increase banks’
franchise value, reduce their incentive to take excessive risk, and consequently lower the chances
of financial distr
ess (Hellman et al., 2000; Besanko and Thakor, 1993; Boot and Greenbaum,
1993; Matutes and Vives, 2000)
.
8

Proponents of this view
also argue that
in a more competitive
environment,
the free
-
riding problem
reduces banks’

incentives for

screening

and

monitor
ing
,
which worsens the quality of banks’ portfolios, and leads to the financing of less efficient
investment projects (Gehrig, 1998; Petersen and Rajan, 1995
;

Boot and Thakor 1993, Allen and
Gale 2000
).

9

Additionally
,
a concentrated banking system with a
small number of banks

is easier
to monitor and supervise compared to competitive banking systems with a large number of banks.

On the other hand,
the “concentration
-
fragility” view argues that
banks with greater
market power tend to charge higher interest

rates to firms, inducing them to take on greater risk, and
hence increasing the fragility of the financial system (Boyd and De Nicolo, 2005).

In addition, banks
in a more concentrated banking environment frequently receive “too
-
big
-
to
-
fail” subsidies fro
m
safety net policies, inducing them to take on greater risk and fragile the financial system as a whole
(Anginer and Warburton 2011, Kane 1989). Finally, as the global financial crisis has amply
illustrated, large banks also can be more difficult to super
vise given their complexity, and their
ability to politically capture their supervisors (Johnson and Kwak, 2010).

10


The process of financial globalization since the 1990s has brought a dramatic change to
the banking industry across the world (i.e., a sign
ificant increase in the share of bank assets held



8
The l
it erat ure also suggest s t hat
t he

s
upervision of banks in a concent rat ed banking sect or is easier and more effect ive
; t hus,
t he
risk of cont agion will be less pronounced in
a
concentrated banking sector.

A growing body of literature provides evidence that
concentration is not
a
sufficient

measure of competition (Beck, Kunt, and Levine
, 2003; Kunt, Laeven, and Levine, 2004
).

9
A

bank will have incentive to establish lending relationship with a young firm if it can share the initial information costs in

their
future stream of profits. However
, in highly competitive banking sector, a bank knows that it may not able to maintain a long
-
term relationship with the successful firms since those firms will seek the lowest cost supply of credit in the banking secto
r once
they become mature. Therefore,
the competitors that didn’t invest initial funding sources would have a cost advantage by offering
better contract terms than the bank that attempts to recoup the initial cost(Free
-
riding problem).

In more intense competition
environment, banks switch to m
ore risky, opaque borrowers (Dell’Ariccia and Marquez, 2004); and acquire less information on
borrowers (Hauswald and Marquez, 2006).

10

Given t he current policy relevance of t he t opic and confli ct ing t heoret ical predict ions, t here is also a growing
empirical
lit erat ure on t he impact of bank compet it ion on bank fr agil it y. Individual count ry st udies
-

most ly for t he US
-

have not come up
wit h conclusive findings.Cross
-
count ry analyses have shown t hat more concent rat ed banking syst ems are less likely t o

suffer a
syst emic banking cr isis but so are more compet it ive banking syst ems, pot ent ially suggest ing a st abilizing effect for having a

cont est able banking syst em (Beck, Demir guc
-
Kunt, and Levine 2006, Schaeck, Cihak, and Wolfe 2009). Ot hers have shown t ha
t
banks in more competitive banking systems hold more capital, which could explain the positive effect of bank competition on
stability (Schaeck and Cihak 2010, Berger, Klapper, and Turk
-
Ariss, 2009). Beck, De Jonghe, and Schepens (2011) find a
positive co
rrelation between accounting measures of bank soundness and market power, and examine the cross
-
country
heterogeneity in this relationship, identifying links with regulatory and institutional features.


14

by foreign banks).

11

There is an ongoing debate on the benefits and costs of foreign ownership.
But a larger body of literature supports a positive view and suggests that foreign bank presence is
associated

with greater efficiency for domestic banks (Claessens et al., 2000, 2001; Claessens
and Lee, 2003; Bayraktar and Wong, 2004) and stronger competition in the host country’s
banking sector (Claessens and Laeven, 2004; Gelos and Roldos, 2004). We expect that

loans
issued to a country with a higher fraction of foreign bank presence have a lower average cost of
bank borrowing.


2.3
.
Institutional Development

While we focus our analysis on the impact of bank regulations and banking
environment on the bank
-
to
-
bank loans, we also test how information sharing registries, creditor
rights and legal enforcement affects banks’ cost of borrowing. Previous literature su
ggests that
these country
-
level institutional variables are potentially important
determinant
s

of financial
development and loan contracting (La Porta et al., 1998, 2000; Qian and Strahan, 2007; Bae and
Goyal, 2009
;
Jappelli

and Pagano, 2002; Djankov,

McLi
esh, and Shleifer, 2007).


Information Sharing
:
The existence of

I
nformation sharing mechanisms

reduce
s

the monitoring
costs of lenders
, and therefore, is an important determinant

of credit market development (e.g.,
Jappelli and Pagano, 2002; Djankov,

McLi
esh, and Shleifer, 2007). Our data source

of
information sharing

is Djankov,

McLiesh, and Shleifer (2007) (DMS, hereafter)
,

and the World
Bank “Doing Business” dataset. DMS (2007) collect information on private and public credit
institutions in 129 countri
es between 1978 and 2003
, and the
World Bank “Doing Business”
dataset updates information annually since then.
12

In our empirical model, we create a dummy
variable,
Information Sharing
, that equals 1 if the country where the bank borrower operates has
an in
formation sharing agency (either public registry or private bureau) at the time of loan
origination, and zero otherwise.




11
Such a phenomena is even more pronounced for
developi
ng count ries,
where
t he share of bank asset s held by

forei gn banks has
risen from 22 percent in 1996 t o 39 percent in 2005(Claessens, Gurcanlar, Merca
do Sapiani, and Van Horen 2008
)
.


12

There are t wo types of informat ion sharing mechanisms: public r egist ri
es and privat e bureaus. Bot h public regist ries and
privat e bureaus collect t he credit wort hiness of borrowers and make dat a availabl e t o banks and financial inst it ut ions. The ma
jor
difference bet ween t he t wo is t hat a public regist ry is owned by a public au
t hority (usually t he cent ral bank or banking
supervisory aut horit y), while a privat e bureau is owned by privat e commercial firms or nonprofit organizat ions.



15

Creditor Rights
:

The strength of creditor rights determines the legal rights that lenders have in
reorganization and liquidation proced
ures, which is an important consider
ation in bank loan
contracting (see Qian and Strahan, 2007). W
e include the credit rights index,
Creditor right,
from DMS (2007) in our regression. It measures the power of secured lenders in bankruptcy in
129 countries
between 1978 and 2003. Following La Porta, Lopez
-
de
-
Silanes, Shleifer, and
Vishny (LLSV, 1997), the index consists of four components
: (1) whether there are restrictions
when a debtor files for reorganization
(
e.g.
,

creditor consent
)
; (2) whether there is

no automatic
stay or asset freeze, thereby

allowing

secured creditors to seize their collateral after the petition
for reorganization is approved; (3) whether secured creditors are paid first compared with other
creditors (e.g.
,

the
government or
employee
s
) in the liquidati
on of

a bankrupt firm; and (4)
whether the management does not stay in control of the business during the reorganization.
Creditor right

is the aggregated score, ranging from 0 to 4, with a higher value indicating
stronger credit rights.

We extend the values of 2003 for
the years

2004 to 2007.

Property Rights
:
The
extent to which creditor rights are enforced is also important for loan
contracting. Poor enforcement, which lowers recovery rates and increases the time spent in
repossessing
collateral in the event of default, increases lenders’ monitoring costs and hurts their
ability to recontract (
Bae and Goyal, 2009; Esty and Megginson, 2003)
.
To control for judicial
efficiency, we use the property rights index from the Index of Economic F
reedom compiled by
the Heritage Foundation/Wall Street Journal since 1995. A higher value indicates a better
contract enforcement environment.

2.4
.
Other Country Factors

In our multivariate analysis, we also control for the following variables reflecting

the
stability

of
banking sector and the financial and economic development.

Explicit Deposit Insurance
: The explicit deposit insurance scheme
is universally adopted as a
meant to prevent banking crisis and promote financial stability.
13

Our measure of dep
osit



13
The adopt ion of
explicit
deposit insurance scheme is an import ant feat ure in
t he
banking indust r
y. However, its desirabilit y is
debat ed. On t he one hand, t he presence of explicit deposit insurance reduces t he severity of bank runs and output costs in a
financi al crisis (Diamond and Dybvig, 1983; Angkinand, 2009), enhance
s

banks’ chart er value

(Gon
zál
e
z, 2005), r educe
s

banks’
risk t aking incent ives (Gropp and Vesala,

2004)
,

and increase
s

deposit

(Chernykh and Cole,

2010). On t he ot her hand, a lar ge
body of researchers accept t hat deposit insurance induce
s

moral hazard
s

and excessive risk t aking
by

bank
s (Mert on, 1977;

Bhat t acharya and Thakor, 1993; Bhatt acharya et al., 1998; Hendrickson and Nichols, 2001; Demirguc
-
Kunt and Kane, 2002;

16

insurance is based on World Bank’s
C
omprehensive Deposit Insurance
A
round the World
Dataset, compiled by Kunt,
Kane,

and Laeven

(2008).
From various country sources and surveys
to officials of deposit insurance institutions and central banks, t
his database
documents detailed
information on the deposit insurance scheme across 190 countries between the years 1934 and
2003.
14

The information in Barth, Caprio, and Levine (2008) is then considered to identify the
countries that have adopted deposit in
surance schemes between 2004 and 2007.

We construct a
dummy variable,
Explicit
D
eposit

Insurance
, which equals 1 if the country where a bank
borrower operates has explicit deposit insurance scheme at the time of loan origination year.

Banking Crisis:

Our i
nformation on banking crisis is obtained from the dataset complied by
Laeven and Valencia (2008)
. We construct a dummy,
Banking Crisis
,
which takes

on

the

value
one
if the country where a bank borrower operates is
experienc
ing

a
systemic
banking crisis
at

the time of loan origination year
, whereas it is zero
.


Sovereign Debt Ratings:

We control for the overall country risk by including Standard and
Poor’s ratings on the long
-
term sovereign bonds for the country of borrower. We converted the
rating to a num
erical score. A lower value indicates a worse rating. For countries without a
sovereign debt rating, we assign a value of zero for them, and also construct an indicator for
those missing rating observations. The log
transformation of the sovereign

rating i
s used
in the
regression specification.

Economic development:

Regulations, banking structure and stability, and institutional quality of
a country are often correlated with its economic development (La Porta et al., 1998). To control
for the difference in
country economic development, we include GDP growth, the natural log of
GDP

per capital, and the natural log of inflation, obtained from the World Development
Indicators database (WDI).







Hovakimian et al., 2003; Houston et al., 2010), increase
s

the likelihood of banking crisis (Demirgüc
-
Kunt and Detragiache, 2002;
Barth et al., 2004),
and
lower
s the

credit ratings of banks (Pasiouras et al., 2006). Additionally, Barth (2004) suggests
an
insignificant relation between deposit insurance scheme
s

an
d bank development and efficiency. Cull et al. (2005) find that
deposit insurance has a negative impact on financial development and growth in the long run, except in countries with strong
legal and regulatory institutions.

14
The det ailed infor mat ion inclu
des
est ablishment

year,
deposit insurance coverage, coverage rat ios, co
-
insurance, t he l evel of
co
-
insurance requirements, percentage of the value of deposit covered, and whether the payments are per depositor or

per
depositor per account.



17

3
. Backgrounds and Sample Selection

3.1. Backgrounds of Bank
-
to
-
Bank

Loans

Since our study is the first international study
that investigates

the determinants of the
bank
loan contracts obtained by bank borrowers, a clear picture
of

the bank
-
to
-
bank loan

market
would provide
some
useful background knowledge.

We collect the
syndicated
loan data from
the
Loan Pricing Corporation’s DealScan (LPC) database, which contains detailed information on
individual loan facilities (such as loan spreads, maturity, collateral, covenants, performance
pricing, loan types, loan

purposes
,

and lender information).
15

The database also identifies the
place where each loan was made.

Figure 1 plots the time series
trend
of bank
’s

borrowing in the
syndicated

loan market in
terms of both the total number of loan facilities and the total
face value of loan facilities. The
figure shows that banks’
private
borrowing
displays
an overall increasing trend, with
a
120%
increase in
the
total number of loan issuances, and
an
80% increase in
the
total volume of loan
issuances.
16

Noticeably
, t
he tota
l volume jumps significantly by 150% in 1997 when
the
financial crisis

initially

hit the whole region,
highlighting

the importance of
the
syndicated loan
market in serving the liquidity needs of banks.

[Insert Figure 1 here]

Table 1 shows the average loan

volumes distribution among bank borrowers, corporate
borrowers
,

and non
-
bank financial borrowers across different regions between 1995 and 2007.
17

Similar to the study of Carey and Nini (2007) on the non
-
financial borrowers, the most
prominent markets (me
asured by
average loan volume

each year) for bank borrowing are
Asia,

Europe
,

and the
U.S. markets.
To be specific, Asia has the larges
t annual loan volumes

issued to
its banking industry (54 billion) and represents 16% of the total syndicated loans in thi
s region.



15
St rahan (1999)

and Graham et al.

(2008) provide a good description of the LPC Deal
s
can database.

16
In an unreport ed figure,
we also plot s t he t ime series of bank borrowing in different regions, such as t he U.S., Europe, and Asia
Pacific, in t erms of bot h
t he t ot al number

of loan facilit ies and t he t ot a
l face value of loan facilit ies from 1995
-
2007. We received
a simil arly increasing t rend.
Specif ically, volumes grew st ably in t he Asian market from 26 bill ion in 1995 t o around 50 bil lion in
2007, reaching a peak of 89 bil l
ion in 1997,
t hus
implying t hat banks t end
ed

to recove
r

from the Asian crisis through the channel
of borrowing from peers.

17
The st at ist ics are compiled from LPC Deal Scan dat abase. It
allows us
t o
ident ify t he count ry or region
where

each borrower
operat es, which defines t he market where a loan was issued.


18

Europe is the second largest market for bank borrowing.
18

The case of U.S. is slightly different.
Even though the U.S. has the third largest loan volume issued in the banking sector

each year
,
this only accounts for 1% of its overall syndicate
d

loan
s due to the much less frequency of bank
borrowing compared to the corporate borrowing in the U.S.


[Insert Table 1 here]

A

clear picture
of

the bank
-
to
-
bank loan flows
in the global market

is provided in
Table2
. Panel A
presents

how bank borrowers in

each region allocate their loan issuance across
different regions
.
It reveals

that bank borrowers located in the U.S., Europe
,

and Asia usually
issue loans in the regions they operate (i.e. stay home), as indicated by fractions in excess of 5
0%
in

Panel A
, and
that Europe is usually the
issue place of
choice

for borrowers located outside the
three
regions (i.e.
those located in
Middle East
,

Africa,

and

Latin America)
.
The statistics are
similar to
those

of the non
-
financial borrowers in Carey and Nini (
200
7).

Panel B shows the allocation of lenders’ lending portfolio
s

across

different regions.
Generally

speaking
, among the
global syndicated loan portfolio

to the
banking industry, 46% is
from the Europe, 33% is from Asia, and 18% is from the U.S.

It appears

that lenders allocate a
substantial percentage of their loan portfolio
to

their home
region,
display
ing

substantial home
bias
.
For example, lenders
from

the U.S. channel 42.5%, 27.8% and 15.9% of their syndicated
lending volume to borrowers located in the

U.S., Asia
,

and Europe, respectively.

[Insert Table 2 here]

3.2
. Sample Selection Procedures

I
nitial
ly,

t
he DealScan data consists of 13,690 loan facilities issued to 1480 commercial
banks in 107 countries from 1995 to 2007.
19

We screen the data using the following criteria: 1)
the interest rate

(
i.e.
,

all
-
in
-
spread drawn
)

is not missing;
20

2) LIBOR is the base rate; 3) we
exclude bonds, notes, leases, loan
-
style floating rate notes, bond
-
style floating rate notes, fixed
-
rate C
D, loan
-
style floating rate CD, bridge loans, guarantee, standby letters of credit, step



18
Not iceably, in East ern Europe,
almost 40% of all t he syndicat e loans are invest ed in t he banking indust ry.


19
Our number of loan facil it ies from LPC’s Deal Scan is comparabl e t o
Hale (2012), who use
Dealogi c's Loan Analyt ics dat abase

and obt ain a t ot al of 15.324 loans bet ween 1980 t o 2009.

20
76% of t he facilit ies
do not have informat ion on

all
-
in
-
spread draw
n.


19

payment leases, limited lines, traded letters of credit, multi
-
option facilities, and other or
undisclosed loans. The above screening process leaves us with a sample o
f 2,519 loan facilities.
Our analysis is conducted at the loan facility level since the contract terms and the identity of
the
lenders can differ across facilities within a deal. The results are similar if we aggregate individual
facilit
ies

into a deal
-
lev
el based on the weighted average loan amount.




We then carefully match the name of commercial bank bor
rowers to Fitch
-
IBCA Ltd’s
BankS
cope by a combination of algorithmic matching and manual checking.
21

Bank
S
cope
provides a good source of balance sheet
and income statement information for both public and
private banks across a wide range of countries.
22

The unconsolidated financial reports are used
whenever available,
and
if
they are
not then consolidated financial reports are used. To ensure
the accounti
ng information is publicly available at the time of a loan origination, the borrower
s’

financials are measured at the year prior to the loan initiation date.

We then add in
country

variables from various sources. Countries are dropped if any of
the key ba
nk
regulations
,
banking environment,
institutional development, and macroeconomic
factors are not available. We also drop countries whose number of observations
is

less than 6.
The above procedures leave us with a sample of 1,602 loan facilities associated

with 1,436 loan
deals that are borrowed by 495 commercial banks from 1995 to 2007 in 42 countries.

3.3. Summary Statistics

Table

3 provides the
calendar
-
time
distribution

of our final loan sample.
Table 4
presents summary statistics regarding our key var
iables u
sed in the regression analysis.
In order
to mitigate the impact of outliers or mis
-
recorded data, all variables are winsoriz
ed at the 0.5%
and 99.5% level.
In our sample, the average loan spread is 95 basis points (and median is 50
basis points),
which is significantly smaller than the spread paid by non
-
financial companies.

23

The mean loan size is $189 million

(median is $100 million)

with a maturity of 30 months

(median is 24 months)
.
Interestingly, the traditional monitoring mechanisms in the corporate loan



21
The mat ching bet ween

Dealscan and Bank
S
cope dat abases
is
based on b
ank name and
a seri es of

ident ifi cat ion infor mat ion
such as count ry, st at e, cit y, zip code, fax number
,

websit e
,

et c.


22
Fit ch
-
IBCA Lt d. Bank
S
cope Dat abase

covers 198 count ries hist orically.

23
In Hao et al.,(2011),

t he average loan spread for a sample of 1
2,468 loans t o t he non
-
financial borrowers in 30 count ries is
149.13 (medi an is 125). In Bae and Goyal (2008), t he medi an loan spread for a sample of 17791 loans t o t he non
-
financial fir ms
in 38 count ries is 82.5 basis point s.


20

contracts rarely exist in such long
-
term bank
-
to
-
bank loan contracts.
To be specific, the
likelihood
on average
of having covenants is 0.03, the likelihood

on average

of having loan

pricing tied to borrower performance is 0.04, and
there is an
18%
chance
that the loans
are

likely
to be secured.

The mean of the syndicate size is 14 (median is 13). Foreign banks represent an
important position in these syndicates. As a fraction of our

final sample, loans arranged by
foreign lead banks represent a mean of 78% of the sample.

In terms of bank borrower characteristics, we find that
the borrow
ers

are mainly
medium
-

or small
-
sized
commercial banks

with
mean (median) book value assets of $1
2

(
median
is
$2

billion
)
billion
.
On average
, the sample bank borrowers have
investment to asset ratio of
0.44,
ROE of 0.14 and
capital

ratio of 0.12.
The mean nonperforming loan to loan ratio is 0.04,
the mean log of zscore is 3.1, and the mean deposit rat
io is 0.64.
The table also shows great
variations in terms of the bank regulations, banking environment, and other country
-
level
variables.


In an un
-
reported table, we compared different aspects of the loan characteristics
between loans to bank borrowers
and those to non
-
financial borrowers. We find that bank
-
to
-
bank loans are significantly different from corporate loans. To be specific, bank borrowers on
averages receive loans that are significantly shorter by 14 months. The loan spreads charged are
113
basis points lower in terms of AISD and 13 basis points lower in terms of AISU. Also, they
are 13 and 25 basis points lower in terms of commitment fees and upfront fees as charged by the
lenders. The comparison of the bank loan contracts issued to bank bo
rrowers and corporate
borrowers across different points of time indicates a similar pattern


[Insert Table 3 here]

[Insert Table 4 here]

4. Empirical Results

4.1. Pricing of Loan Facilities

We first test how country level bank
regulations
, banking environment
and other
country level factors
affect the loan spreads of bank
-
to
-
bank loans. We use a regression model of
the following form:


21

Log (AISD) = f (Bank Regulations, Banking Environment,
Institutional Development, Other
Country Factors
, B
ank Borrower Characteristics, Loan Characteristics)
(1)


w
here the dependent variable is the natural log of
AISD

(i.e.
,

all
-
in
-
spread drawn), which is
measured as the number of basis points over LIBOR, indicating the interest rate spread banks
charge for each dollar drawn.
24

We use
Financial Conglomerates Restriction

to measure
banking
-
commerce link

regulations
, and
we
use
Financial Statement Transparency

to measure
the effectiveness of bank accounting disclosure restrictions. Banking Environment includes
Foreign
Presence

and
Bank Concentration
.

W
e control for institutional
quality of the borrower
country using
Information S
haring,

Creditor Rights

and

Property Rights
.
Explicit Deposit
Insurance

and

Banking Crisis

are used to control for the degree of banking stability.

In addition, we also control for the overall country risk by including Standard and
Poor’s ratings on the
long
-
term sovereign bonds for the country of borrower. We converted the
rating to a numerical score. A lower value indicates
a
worse rating.
To those

countries without
a
sovereign debt rating we assign a value of zero, and
we
also construct an indicator fo
r those
missing rating observations. The log transformation of the sovereign rating is used in the
regression specification.
Regulations, banking structure and stability, and institutional quality of
a country are often correlated with its economic develop
ment (La Porta et al., 1998).
To control
for the difference in country economic development, we include the
GDP growth, log of inflation
and the
natur
al log of GDP per capital

over the 1995
-
2007 period.

To identify the effect of

bank
regulations
and banki
ng environment
on

the bank
-
to
-
bank
loan contracts
, it is important to control for bank borrower risk characteristics and loan
characteristics. Therefore, all of the regressions include
Bank Assets
,
Investment Ratio
,
ROE
,
Capital ratio
,
Deposit ratio
, and
Bank Risk

(
measured by
Zscore

and Non
-
performing Loan
ratio
). The spread regressions also include non
-
price loan terms such as the log of loan amount,
the log of loan maturity, collatera
l, missing collateral indicator.

The existence of information asymmetr
y (both adverse selection and moral hazard
problems) affects lenders’ lending decisions and the efficiency of loan contracts. Such
informational frictions can be reduced if there are strong past relationship between lenders and
borrowers (Boot, 2000; and B
harath et al., 2009). Therefore, following Dahiya et al.,(2003) and



24
To ensure appropriat e compa
rabilit y in currency and a benchmark for pricing loans, we only include US$ loans and
t he London
Int erbank Offered Rat e (LIBOR)

benchmark. The benchmark informat ion is obt ained from DealScan’s “Base Rat e and Margin”.


22

Bharath et al., (2009), we construct a dummy variable,
Relationship Lending
, which equals to
one if there is prior lending by the same lead banks over the previous 5
-
year window, and zero
otherwise.

In addition, since

our borrowers are commercial banks
that

are usually
syndicate lenders,
it is very likely that
a
certain fraction of
the loans in our sample
is made among the previous
syndicate partners. In other words, in the bank
-
to
-
bank lo
ans, some borrowers obtain loans from
lenders with
whom

they used to serve in the same syndicate (regardless of lending to non
-
financial or financial firms). We classify those cases as
Syndicate partner loan
.
We expect that
previous cooperation could also
reduce the information asymmetry problem faced by lenders and
hence lower the cost of bank
-
to
-
bank loans.
In our sample, we identify that
42
% of the loans are
from borrowers’ previous syndicate partners.

The other control variables include dummy variables

for loan origination year, loan
purpose
,

and loan type. We also create a dummy variable

called

“Investment Grade
,
” which
equals one if the bank borrower
has

an S&P senior unsecured debt ra
ting equal or higher than
BBB,
and zero
otherwise.
The detail
ed

des
cription of the variabl
e has been provided in
Appendix
1.

In our sample, more than half of the borrowers have multiple loans in the same year.
Treating each loan independently may lead to biased standard errors and inference
, as

similar
factors within the b
orrower may drive the origination of each of those loans. To address this
issue, it is important to adjust standard errors to account for correlations within a cluster (see
Petersen
,
2007). Therefore, in all
of
the regressions we report robust t
-
statistics

that adjust for the
clustering at the bank borrower level.
25

We do not include either country
-
level or firm
-
level fixed
effects in our estimations. This is the case because there is limited time variation in
our

regulations and

banking environment variable
s,
and i
nclusion of either firm
-
level or country
-
level fixed effects would render it impossible to identify the country effects of those institutional
variables on the bank
-
to
-
bank loan contracts.





25
Similar ly, loans in a given count ry c
an also not be t reat ed as independent observat ions if t here are unobservable common
count ry fact ors. Adjust ing t he st andard errors for het eroskedast icit y and clust ering wit hin a count ry renders our results
unaffect ed.




23


[Insert Table 5 here]

The regression results are
presented in Table 5. First, consider the bank regulations

variables.
The results suggest that a higher level of activities restrictiveness in the borrower
country reduces bank risk and hence lower
s

the loan spreads.
In other words,
holding other
things co
nstant, a borrower country whose regulations
most
favor traditional banking (i.e.
,

Traditional Banking Index equal to 6) is likely to have 27 (=0.0736*95*4) basis points lower
loan spreads relative to the borrower country that
most
favor
s

universal banking

(i.e.
,

Traditional
Banking Index equals to 2).

Lending to countries that require

high standard

bank
-
accounting
disclosure reduce
s

the information asymmetry faced by lenders. Therefore, we find that loan
spreads decline with the improvements in bank
-
accoun
ting transparency. The estimates indicate
that, holding all else constant, the l
oan spreads decrease by around 11

basis points, for every unit
increase in the index of bank accounting
regulations
.

Next, consider banking environment. Our results show that
both foreign
presence

and
bank concentration are important determinants of bank
-
to
-
bank loan contracts. First,
one
standard deviation increase in
the foreign presence is associated with
a
5 basis points decrease in
interest rate. This is consiste
nt with th
e argument that
a
banking system
with a high fraction of

foreign banks tends to be
more efficient
. Thus, lenders respond

to

the superior banking
environment by reducing loan spreads. Second, bank concentration negatively and significantly
at
the
1% level i
mpact
s

the loan spreads. Specifically, a one standard deviation increase in bank
concentration, o
n average, is associated with 15% (or 14

basis points) lower loan spreads.
26


In summary, the results indicate that lenders charge smaller spreads when bank
bor
rowers are in countries with more restrictive financial conglomerate
regulations
, relaxed bank
entry barrier
s
, stringent bank accounting disclosure
regulations
, better
-
developed credit rights
and information sharing system
s
, and more concentrated or foreig
n dominated banking
environment
s
.


4.2. Concerns on Ownership Connected Loans




26

The loan spread regr essions
include loan mat urit y and loan size as predet ermined variables. Result s are similar if predict ed
loan sizes and predict ed loan mat urit ies are used inst ead.



24

A concern
regarding the

above results is that some borrowers might be branches or
subsidiaries of certain lenders in their bank loan syndicate
,

or vice
-
versa.
27

In those cases,
the
syndicated loan market may simply become a platform
for

internal capital transfer.
28

Therefore
the effect of
the
country
-
level factors we just identified may not be as important. To identify
those ownership
-
connected loans, we first obtain the sharehold
er information of the borrowers
and lenders at the year of loan origination. We use BankScope’s annual shareholder data
from
the years
2001, 2003, 2005 and 2007

(based on the CD versions).
29

To make the ownership data
time series we fill in
the
year
s

1994
-
2001 with
the
ownership value at 2001. We fill
the
year
s

2002
-
2003 with
the
ownership value at 2003. We fill
the
year
s

2004
-
2005 with
the
ownership
value at 2005. And we fill
the
year
s

2006
-
2007 with
the
ownership value at 2007.
30

We then
match the name
s

of borrowers’ shareholder
s

with the name
s

of lenders

and
match the name
s

of
lenders’ shareholder with the name
s

of borrowers, and
we
even match the shareholders of
borrowers with
the names of
lenders. A dummy variable, Ownership connected loan, is
construc
ted to indicate whether a loan is made among ownership related parties. In our sample,
we identify that 11% of the loans are borrowed by commercial banks that have ownership
connection with at least one of their syndicate lenders. We then re
-
run our estima
tion model
(Column 6)
by excluding the sample of ownership
-
connected loans. Our results remain
quantitatively consistent
.


4.3. Distance of Regulations and
Banking Environment between Lender and Borrower Country

In this sub
-
section, we explore the specific

connections between borrower and lender
countries. In particular, we
investigate

whether differences

(or Gaps)

in
regulations and the



27

For example,
B
ank
A
der at
PLC
(id169090) borrow
s

from
t he
syndicat e loan market. Two of t he syndic
at e members are Bank
Melli Iran(BMI) and Bank Sepah Int ernat ional
PLC
, who
are
bot h owned by Iran. And bank borrower
B
ank
A
derat
PLC

is also
owned by Iran.

28

For inst ance,
a
lender might
inject capital to rescue

those
illiquid

affiliated
subsidiaries/branc
hes by lending at an unusually
low interest rate.


29

Even t hough we also have t he
B
ankscope CD of 1999, we did not use t he ownership informat ion of t hat year since most of t he
dat a

is not as accurat e

as t he dat a from t he lat er years.

30

We not ice t hat

in t he syndicat e loan market, inst it ut ional lenders become a si gnifi cant proport ion. It is possible
t hat
t hose
inst it ut ional lenders and bank borrowers are sharing t he same shareholders. But t his requires det ailed global ownership dat a,

which we do not ha
ve. But we ident ify
all t he ownership
-
connect ed cases when

banks are t he leader arrangers. We are aware t hat
wit hout ident ify
ing

t he common shareholder cases bet ween inst it ut ional lenders and bank borrowers
,

we
mi ght under est imat e t he
effect of borrower
-
le
nder connection on loan contracts. But we doubt the possibility
is so significant
that it
would
flip our results.



25

banking environment

between borrower and
lead
lender countries influence the

cost of bank
-
to
-
bank loans.
We est
imate the
following regression:


Log (AISD) = f (Bank Regulations Gap, Banking Environment Gap, Institutional Development
Gap,
Culture Distance
, Bank Borrower Characteristics, Loan Characteristics, Other Control
Variables)



(2)




The Gap measures are calculated as the absolute difference between the levels of country factors
in the borrower country and those in the lender count
ry.
The value of Gap measures go from
negative to positive.
A

large positive
regulatory gap indicates more restrictive
b
ank regulations
in the borrower country relative to those in the lead lender country
, and a
negative
gap
means the
regulatory restrictio
ns are more stringent in the lead lender country than
in the
borrower country.
To be specific, a high value of
Traditional Banking Gap

suggest
s

that

borrower country has a
much higher restriction on universal banking (or that they favor traditional banking
) as compared
with the lender country. A large
Transparency Gap
suggest
s

that

the borrower country has a
much higher standard of accounting disclosure as compared with the lender country. A

positive
value of
Bank Concentration

indicates that
the
borrower c
ountry has
a
more concentrated
banking sector
as
compared
with

the lead lender country.
Similarly, a p
ositive
value of
Foreign
Presence

indicates that the borrower country has

a more foreign dominated banking industry
than the lender country
.

We control for the
institutional development gap
and explicit deposit insurance gap in
our estimation.
In addition,
a combined measure of
Culture Distance

is also included
to control
for potential
cultural
differences between the borrower and lead lender c
ountries.
This measure
involves
three
components:
Geographic D
istance,

L
anguage

Distance,

and
C
olonizer

Distance
.
Geographic D
istance

is measured as
a dummy, which equals one if the lender and borrower are
not from the same geographic region and 0 otherwise.
L
anguage

Distance

and
C
olonizer

Distance

indicate

whether the borrower and lead lender countries

share a common official
language, or have ha
d a c
ommon colonizer after 1945 (equals 1 if they do not share common
language or common colonizer tie; 0 otherwise).

The
Culture Distance

sums up these three
components. A larger value indicates a larger culture distance.


26


[Insert Table 6 here]


Table 6 provi
des strong evidence that the distance in bank regulations

and
banking
environment between the borrower and lead lender countries also matters for the bank
-
to
-
bank
loan contracts. Moreover, the directions of the coefficients are highly consistent with our
p
revious findings in Table 5. Specifically, we find that lead lenders from countries
that are less
supportive of traditional banking and transparent accounting disclosure
charge significantly
lower loan spreads when lending to
a market where bank regulation
s favor traditional banking
and accounting transparency
. In terms of banking environment, the results suggest that lenders
from
a country with less of a foreign presence
or
a
less concentrated banking industry lower the
interest rates dramatically for bank

borrowers from
countries with a high
foreign
presence

or
a
highly concentrated banking industry. Regarding institutional quality, we find that lenders in
countries with poorly developed institution
al

system
s

charge significantly lower loan spreads
when le
nding to countries with well
-
developed creditor rights and information sharing.
We also
find that a large pricing discount is given if the borrower country has an explicit deposit
insurance scheme and the lender country does not.

In addition, the observed

results provide us some evidence
in support of

the theory of
“Home Bias
.
” The “Home Bias” theory suggest
s

that the concern of familiarity causes lenders
to
limiting their credit exposure in foreign count
ries, which
implicitly implies that unfamiliarity
wo
uld push up interest rates charged by foreign lenders due to information frictions

(
Karolyi an
d
Stulz, 2002)
.

Our paper, in the global bank
-
to
-
bank loans context, provides direct empirical
evidence that lenders charge significantly higher loan spreads when

the borrower is culturally
distant, which is consistent with the first
-
order pricing effect of “Home Bias.”

4. 4. Foreign Lending vs. Domestic Lending




In the global loan market, foreign lenders are playing an increasingly important role
(Demirguc
-
Kun
t and Levine, 2001). This is especially the case in the bank
-
to
-
bank loan market.
Our statistics in Table 4 indicate that around 78% of bank
-
to
-
bank loans have foreign banks as
lead banks in the syndicate. This brings up an interesting
question

of

whether
the effect of
regulations on the loan spreads various across foreign and domestic lending.




27


The l
iterature

related to
“Home bias” suggests that foreign
banks have

a monitoring
disadvantage in terms of investment

and are therefore more sensitive to countr
y risk than
domestic investors. In line with this view, empirical literature on corporate loan contracts finds
that foreign banks

are more likely to resort to formal legal

system (
Qian and S
trahan, 2007; Mian,

2006
;

Haselmann et al., 2010
)
. Given the fact
that
the traditional monitoring mechanisms are
largely absent

i
n
bank
-
to
-
bank loan contracts, and the relative dis
advantage

held by foreign
lenders

in overcoming information and control problems
as compared to

domestic

lenders,

we
therefore expect that the

benefits of regulations on reducing banks’ borrowing costs should be
more pronounced if the loan is made by foreign leaders, especially when the loans have severe
information asymmetry problems.


To simplify our estimation, we create an
aggregate
d measure

of bank regulations by
summing up the traditional banking regulations and transparency regulations. Our results are
robust in either the aggregated or component measures of regulations. To test if there exists any
cross
-
sectional variation of regulation e
ffect on loans spreads across foreign and domestic
lending, in the first column of Table 7, we docompose the
Regulation

variable in two parts by
interacting it with
Foreign

Lead

and
Domestic Lead

dummies.
Foreign Lead

equals 1 if the lead
banks operate in
a different country from
the borrower, while
Domestic

Leader

equals 1 if the
lead banks operate in
the same

country
as

the
borrower
.
31

Interacting
Regulation
with each of
these variables is equivalent to running separate regressions for each of the foreign and domestic
loan subgroups. The results show that effects of bank regulations on loan spreads are more
pronounced when the banks borrow from their foreign p
eers.



In column 2 and 3 we
test if above observed relationship is more pronounced when
the
loans have severe information asymmetry problems.
We measure the degree of information
asymmetry by two proxies.
The first proxy is
Non Relationship Loan
, measure
d as a dummy that
equals 1 if there is neither a prior lending relationship nor a syndicate partnership between the
lead lender and the borrower; it equals 0 otherwise. The second proxy
is

Not Rated
, which equals
1

if either S&P or Moody’s does not rate th
e bank borrower
and 0

otherwise.
To conduct the test,



31

If t here are mult iple lead lenders in t he syndicat e, our measure requires t hat all t he lead lenders be foreign.





28

we interact our proxies of information asymmetry with our existing foreign and domestic loan
dummies, thus the
Regulation

variable is decomposed into four parts by the 4 interaction
categories. Our resu
lts suggest that
foreign banks rely
heavily

on restrictive
regulations

that can
reduce the risk of the bank borrowers
, especially when it comes to risky loans
.


[Insert Table 7

here]


4. 5
. Bank Regulations: Does Banking Environment and Institutional Quality Matter?

Despite the above results, we further investigate
whether the level of banking
environment and the legal and institutional environment
s

inf
luence the degree of the regulation

e
ffect on the bank
-
to
-
bank loan contracts.

Strong regulations actually signal quality and stability.
A natural question
emerges:
would bank regulations still matter if these countries do have strong
institutional and banking environment
s?

To better understa
nd the economic context of
the
regulations benefits, we conduct a series of additional tests to explore whether the level of
banking environment and the legal and institutional environment influence the degree of
regulations effect