CEIS Tor Vergata

RESEARCH PAPER SERIES

Vol. 6, Issue 5, No. 118 – March 2008

Bank Shareholding and Lending: Complementary

or Substitution? Some Evidence from a Panel of

Large Italian Firms

Emilio Barucci and Fabrizio Mattesini

Bank shareholding and lending: complementarity or

substitution? Some evidence from a panel of large Italian

firms

∗

Emilio Barucci* and Fabrizio Mattesini**

*Politecnico di Milano

Dipartimento di Matematica

Via Bonardi 9, 20133 Milan, Italy

**Universit´

a di Roma “Tor Vergata”

Dipartimento di Economia e Istituzioni

Via Columbia 2, 00133 Rome, Italy

January 10, 2008

Abstract

The paper studies the motivations behind banks’ shareholding of non-financial

firms using a panel of large Italian companies in the period 1994-2000. Empirical

evidence shows that banks are shareholders of companies that are less profitable,

have experienced slower growth, are more indebted and are endowed with collateral

and have hard time to repay their debt out of current income. Banks are more

likely to hold shares in companies they lend to. Overall the evidence suggests that

there is complementarity between bank equity holding and lending. A plausible

explanation is the shareholder-debtholder conflict, the evidence is weakly compatible

with governance and information hypotheses.

Keywords: Lending, cross shareholding, conflict of interest.

JeL classification: G21, G24.

∗

We thank Bob Chirinko and two anonymous referees for useful comments. The usual disclaimers

apply.

1

1

Introduction

Economists have long debated the reasons behind various types of bank-firm relation-

ships and their effects on firm performance. The debate has concentrated mainly on the

dichotomy between Anglo-Saxon type, market centered systems, where banks are rela-

tively unimportant and have arms’ length relationships with firms, and German-Japanese

bank centered systems where banks are allowed to invest directly in non-financial com-

panies, and companies are often related to a bank through exclusive relationships, the

so called main bank liaison. Other systems which fall between these two extremes have

been subject to less intense scrutiny. During the 1990s, however, a pervasive financial

deregulation around the world has taken place, giving rise to important reforms, such as

the Second Directive on Banking of 1993 in the European Union and the Gramm-Leach-

Biley Financial Services Modernization Act of 1999 in the US. Thanks to these pieces of

legislation, the gap between these two systems has closed significantly. Many restrictions

on bank ownership of equity in non financial firms have been removed in the US and in

many European countries, and this has rendered the issue of banks’ involvement in the

capital of non financial firms quite relevant also in systems traditionally characterized

by a strict separation between banks and firms

1

.

In this respect the Italian case is quite interesting. Even though the Italian system

can be defined as a ”bank based” system where banks play a central role in firms’ external

financing and security markets are relatively underdeveloped, traditionally the Italian

financial market has been characterized by arms’ length banking relationships and by

a rigid separation between commercial and investment banking

2

. These characteristics

developed as the result of the dramatic regulatory changes induced by the banking law

of 1936, which was approved as a response to the collapse of the large ”German type”

1

The financial deregulation on bank-non financial companies relationship occurred during the 1990s

was probably motivated by the positive governance effect associated with bank shareholding in Germany

and Japan as detected at that time, for empirical evidence see [Gorton and Schmid, 2000, Flath, 1993,

Prowse, 1992].

2

For

a

characterization

of

the

Italian

financial

system

see

[Cobham, Cosci and Mattesini, 1999].

2

universal banks and the subsequent rescue by the Italian State. This crisis was largely

the consequence of the enormous equity stake that these banks had acquired on the major

Italian corporations. During the 1990s, however, the regulatory framework was radically

reformed. Banks were transformed into joint stock companies and then privatized, bank

branching was liberalized and the Bank of Italy adopted a system of ”prudential” (as

opposed to ”structural”) supervision. Among these provisions there was a significant

removal of restrictions on banks’ investment in non-financial firms. This important

change occurred with the approval in 1994 of the Uniform Banking Code (Testo Unico

Bancario) which introduced in the Italian legislation the norms contained in the Second

EU Directive on banking, designed to encourage the transformation of European banks

into universal banks. According to the new code, banks are allowed to invest up to

15% of their capital in the capital of non financial firms and up to 3% of their capital

in a single firm; the Bank of Italy has the power to relax these limits for banks with a

large capital or that specialize in long term activities, or in case of restructuring of firms

with financial difficulties. These limits are much more restrictive than those imposed by

the Second EU directive but represent, nevertheless, a significant novelty relative to the

previous situation.

3

The new regulation allows banks to play a crucial role in the Italian financial mar-

ket. Due to the short period elapsed, it is difficult to obtain a very clear assessment of

the evolution of the Italian financial system. If we measure the role of financial insti-

tutions through the financial intermediation ratio,

4

we observe that the ratio in Italy

has increased during the last ten years (from 0.34 in 1995 to 0.38 in 2004), but if we

restrict the ratio to bank assets we see that it has remained constant around 0.27 (see

[Banca d’Italia, 2006] and [Bartirolo and De Bonis, 2005]) remaining at an intermediate

3

The directive simply establishes that a bank cannot invest more than 60% of its capital in

shares of non-financial firms and sets at 15% of a bank’s capital the ceiling on the acquisition by

a bank of shares of any single non-financial company.

4

The financial intermediation ratio is defined as the ratio of assets held by financial corporations (the

central bank, banks, other financial intermediaries, insurance corporations and pension funds) over total

assets.

3

level between the one of Germany (0.37) and that of the US (0.14). This seems to in-

dicate that in Italy banks have interpreted the opportunities offered by the new system

by enlarging their business towards non traditional banking services (asset management-

bancaassurance). However, if we look at bank equity holdings, we can detect a significant

increase in the last few years. While in 1998 banks owned about 3 billion euros of shares

of non financial firms (excluding the companies offering direct services to banks), in 2006

banks’ equity holding amounted to more than 8.3 billion.

To assess the effect of the new regulation, we turn to the micro level and we analyze

the rationale behind Italian banks’ equity investments in non-financial firms

5

. The aim

is twofold. First, we provide new evidence on bank-firm cross-shareholding in a context

which is quite different from those usually analyzed in the literature i.e., a system where

banks do not tend to act as ”main banks” with respect to companies as in Germany and

Japan and a system that is also quite different from the Anglo-Saxon model. Second,

we evaluate what is the rationale behind banks’ holding equity in non-financial firms in

relation to the traditional activity of banks, i.e. that of providing credit to firms. In

particular, we ask whether there is complementarity between bank shareholding and bank

lending and we try, with the available data, to provide some interpretation of the nature

of this relationship. The issue is quite important because bank-equity relationships

may be crucial in determining the health of a country’s financial system. Some recent

evidence on cross shareholding and relationship banking in Japan, for example, shows

that this phenomenon has been quite negative, being responsible for some of the problems

that have affected the Japanese financial system in the nineties.

6

This stands in clear

contrast with the literature developed in the eighties and in the nineties which viewed

the involvement of banks in non-financial firms as a strength of the Japanese financial

5

Due to the short period of time elapsed from bank shareholding liberalization, the evidence on the

Italian financial market is rather limited. [Bianco and Chiri, 1997] analyzing bank shareholding by the

end of 1996 show that only 20% of the total banks equity holdings of non-financial companies is due to

pure investment, while the vast majority of equity shareholding seems to be due to debt restructuring.

6

Note that Italy and Japan share a weak degree of creditor protection, see [La Porta et al., 1998],

and banks may be interested in detaining a stake to ensure debt repayment.

4

system.

We analyze a panel of the largest Italian non financial companies (listed and not

listed) considering data on balance sheets, ownership structure and detailed information

on firms’ long term debt relationships. The dataset is unique because we have information

on specific bank-company equity and lending relationships, and this allows us to look at

the interplay between lending and banks’ holding equity stakes in non-financial firms.

The empirical analysis indicates that bank shareholding is strongly and positively

correlated with the amount of long term debt granted to the firm by the bank. Moreover,

banks seem to hold equity mainly in companies that are relatively less profitable and

less dynamic, have more debt, have a large fraction of tangible assets and are less tightly

controlled. These results indicate that banks tend to hold shares of companies which are

current or prospective borrowers and that there is a complementarity between banks’

shareholding and lending. Evidence that banks have acquired equity to improve the

governance of firms, mitigating the asset substitution effect or agency costs, or to gather

an information advantage to allocate credit in an efficient way is very weak. Although

we cannot exclude the banks hold shares of non financial firms to participate to the

potential upside performance of the firm, the results are consistent with the hypothesis

that banks hold equities to influence the management of the company towards debt

repayment (sharehodler-debtholder conflict).

The paper is organized as follows. In Section 2 we discuss the literature on banks’

equity investment focusing on the Japanese experience. In Section 3 we provide a de-

scription of the dataset. In Section 4 we present our empirical results on banks’ equity

investment. In Section 5 we analyze institutional investors’ shareholding.

2

Banks’ equity holdings of non financial companies: the-

ory and evidence

The literature on bank-firm relationships concentrates on the debt side and, in particu-

lar, on the effects of the so-called main bank relationship which arises when a bank acts

5

as a main lender to a firm and provides assistance on the governance and on the man-

agement of the company. A well established result shows that companies with a main

bank relationship are less liquidity-constrained (see, among others, [Hoshi et al., 1991,

Weinstein and Yafeh, 1998] for Japan and [Petersen and Rajan, 1994] for the US), but

the effect of a main bank-firm relationship on performance is mixed. [Weinstein and Yafeh, 1998]

show that firms with a main bank relationship in Japan are characterized by lower prof-

itability, lower growth rates and a higher cost of capital.

Also the evidence on the relationship between banks’ shareholding in non finan-

cial companies and their performance is mixed. In Japan, a nonlinear relationship

(decreasing and then increasing) between Tobin’s q and bank shareholding has been

detected by [Morck, et al., 2000]; [Yafeh and Yosha, 2003] show that ownership concen-

tration and debt reduce agency costs and managerial private benefits, but no effect is

associated with a main bank relationship; [Lichtenberg and Pushner, 1994] find a pos-

itive relationship between banks’ shareholding and increments of total productivity of

borrower firms. In Germany, [Gorton and Schmid, 2000] have shown that firm perfor-

mance benefits from concentrated ownership and in particular from bank shareholding;

[Chirinko and Elston, 2003], instead, obtain different results: bank control negatively

affects company profitability but statistical significance is weak or absent.

Our study differs from the ones quoted above in two major respects. First it tries to

ascertain whether also in the Italian financial system, which is quite different from both

the German and Japanese main bank systems, it is possible to detect a link between

bank shareholding and lending. In particular, it tries to establish whether, in the Italian

case, bank shareholding is a complement or rather a substitute of bank lending. Second,

we investigate whether available empirical evidence allows us to say something about the

motives behind the banks’ decisions to hold equity in non-financial firms.

On this last point it is important to be extremely careful, because inferring motives

from the data may be tricky. Banks hold equity mainly for three reasons: i) to partici-

pate to the potential upside performance of the company instead of receiving the more

conservative debt payoff; ii) to better monitor and condition the management of the firm,

6

reducing agency costs (management-shareholder conflict) and limiting the asset substi-

tution effect (shareholder-debtholder conflict); iii) to obtain information advantages in

the allocation of credit. In this last respect, bank shareholding may be viewed as a part

of relationship banking, see [Diamond, 1984].

According to these motivations, bank shareholding can be a substitute or a comple-

ment of lending: while the informational advantage argument leads to complementarity,

the first and the second motivation are ambiguous on this point: in the first case, the

choice between debt contract and equity depends on risk aversion and on the possibility

to hedge firm specific risk; in the second case it depends on the effectiveness of outside

blockholding and debt as monitors. If bank shareholding is driven by the desire to limit

the asset substitution effect, which arises when the existing assets of the firm are sub-

stituted with riskier assets to increase the likelihood of the potential upside in favor of

shareholders and against debtholders, then complementarity is expected.

The effects of bank shareholding on company performance are ambiguous: the second

and the third motivations may be associated with a positive performance because the

bank holding equities acts as a strong monitor to limit agency and asset substitution

costs but the effect becomes negative if the bank is also a creditor of the company and

simply aims to debt repayment in conflict with shareholders’ interests. Also the effect

of equity holding on bank lending is ambiguous: on one hand we have the classical

information advantage argument, on the other we have a change of incentive. The bank

wants to participate to potential upside and therefore is more reluctant to provision for

non performing loans, to terminate non profitable investments and to reduce loans to

troubled companies, i.e., we have the soft budget hypothesis

7

.

We can conclude that the second and the third motive can easily generate a cost rather

than a benefit both for the bank and the non financial company. Moreover the three

motives are not mutually exclusive and conclusive on complementarity/substitutability

between shareholding and lending.

Particularly relevant, for the purpose of this paper, is the literature on bank-firm cross

7

For a theoretical foundation of this hypothesis see [Dewatripont and Tirole, 1994, Berlin et al., 1996].

7

shareholding in Japan, see [Prowse, 1992, Flath, 1993, Sheard, 1994, Miyajima et al., 2003,

Miyajima and Kuroki, 2006, Franks et al., 2007].

8

Interestingly, the evaluation of the

main bank-firm relation and of cross shareholding in Japan has changed over time, and

is strictly correlated with the performance of the economy. [Prowse, 1992, Flath, 1993]

analyze bank shareholding in the 1980s. In the first paper, where a relationship be-

tween large equity stakes by banks and a high level of risk (represented by the return

standard error) is detected, the author claims that ”financial institutions’ equity invest-

ments are sensitive to benefits that can be obtained from exerting control over unstable

firms”. The evidence provided in the paper to support this view, however, is quite lim-

ited. [Flath, 1993] analyzes bank shareholding of non financial companies controlling for

debt endogeneity and obtains two interesting results: a) weak collateral, growth options

and high R&D expenses lead to more shareholding by the main bank and less leverage,

suggesting that agency/asymmetric information problems inhibit borrowing and induce

banks to detain shares; b) greater overall borrowing and borrowing from the main bank

elicit greater shareholding by the main bank and vice versa.

According to these results there is complementarity between lending and sharehold-

ing by the main bank, although the two phenomena represent a different response to

agency/asymmetric information problems. Moreover, the assessment of bank holding

equity in non financial firms is positive

9

.

8

[Chirinko and Elston, 2003] analyze bank shareholding in Germany, they show that bank share-

holding is a substitute for control: the likelihood of observing bank control is decreasing in ownership

concentration, the dividend yield and the cash flow over interest expenses ratio.

9

These results have been confirmed only in part by the analysis of bank blockbuilding after the second

world war. [Miyajima et al., 2003] analyzing the 1964-1969 period have shown that the change in bank

equity holdings in non financial companies was positively sensitive not only to lending relationships but

also to firm performance (return on assets, interest coverage ratio), growth opportunities and financial

conditions. [Franks et al., 2007] suggest that insiders’ shareholding increases in the post war period

are due to less effective creditor protection; analyzing incumbent shareholding (banks and other firms)

changes in the 1950-1955 and 1962-1967 periods, they provide evidence confirming this claim: high

leverage/financial distress cause an increase in bank shareholding through debt-equity swap; the relation

with financial health and performance, instead, is rather weak.

8

The cross shareholding system remained stable for a long time but things appear to

have changed considerably, coinciding with the deep economic and financial crisis that

hit Japan during the 1990s. With the increase in foreign shareholding, cross sharehold-

ing between banks and non financial firms began to dissolve: the fraction of shares ex-

changed in the stock market owned by banks went from 15.7% in 1990 to 7.7% in 2002, see

[Miyajima and Kuroki, 2006]. Analyzing the phenomenon, [Miyajima and Kuroki, 2006]

show that the likelihood of unwinding is small if there is a main banking relation-

ship and cross-shareholding.

The decision to sell shares by banks, instead, turned

out to be positively related with growth options and expected rates of return (shares

that are easy to sell) and only weakly negatively affected by credit risk. This be-

havior lead to a deterioration of banks’ portfolios. Confirming [Kang and Stulz, 2000,

Weinstein and Yafeh, 1998], these authors also show that bank ownership and main bank

relationships have had a negative effect on firm performance during the 1980s and 1990s.

This evaluation is supported by the evidence on banks’ lending behavior during the

Japanese financial and economic crisis in the 1990s. [Peek and Rosengren, 2005] show

that during the 1990s domestic bank loans in Japan continued to increase until the mid-

1990s: instead of reducing (increasing) loans to troubled (safe) firms, banks have mainly

increased loans to troubled borrowers (banks’ forbearance). Credit was not allocated in

a proper way, and this phenomenon was mainly concentrated on companies with strong

corporate affiliations, i.e. on firms belonging to a keiretsu, and when the bank belonged

to the same keiretsu of the company. We can conclude that bank-firm liaisons have

contributed to a bad allocation of credit and delayed the restructuring process of non

financial companies. An explicit statistically significant relation between bank holding

shares of non financial companies, non performing loans and loan loss provisions has

been detected by [Dewenter et al., 2007]

10

.

This evidence casts serious doubts on the rationale behind banks’ shareholding in

non financial companies and on its effects on firm management and credit allocation. In

10

A similar result has been obtained at the international level for relationship banks opposed to trans-

actional banks, see [Dewenter et al., 2006].

9

the case of Japan, empirical evidence has detected a strong complementarity between

lending and equity holding and that the informational/governance advantages tradition-

ally associated with banks’ equity holding have turned out to be instead a cost both for

banks and non financial companies.

3

Data and variable definitions

We analyze a dataset collected by Ricerche & Studi of Mediobanca on the largest finan-

cial and non financial Italian companies, including holding companies with consolidated

balance sheets. We restrict our attention to non financial companies. We consider seven

years of data (1994-2000). The sample varies over time, on average it includes 190 large

companies (82 companies are listed in the Italian stock exchange); on the whole we have

1450 observations. The dataset includes balance-sheet data, ownership structure data

as well as information on bank-firm debt relations. Since only a part of the companies is

listed in the Italian stock market, we cannot use information on market valuation. The

dataset is free of survivorship bias: a company belongs to the dataset depending on its

size, in case of a company exiting the dataset at some date previous observations are

included in the dataset.

To evaluate the rationale for bank shareholding and its degree of complementar-

ity/substitutability with leverage, we consider variables addressing specific aspects of

the company.

The most interesting aspect of our data-set is that we have information on bank-firm

debt relationships, and this allows us to get a clear picture on bank specific shareholding-

lending behavior. In the first part of our empirical work we consider a dummy variable

indicating whether a shareholding bank is also a lender to the firm (partloan), in the

second part we consider directly the percentage of the of debt that a firm has with a

shareholding bank (loanshare). Obviously a positive relationship between sharehodling

and either partloan or loanshare indicates complementarity with lending.

Although extremely indicative of the connection between shareholding and lending,

however, these variables are mute on the rationale of the connection. In order to obtain

10

some insights on this important issue, we include in the regressions a series of controls

that could be indicative of the motivations behind banks decisions to hold equity in non

financial firms.

The first control we use is company size, which is measured through the logarithm

of total assets (lasset), the number of employees (employees) and total sales (sales). It

is difficult to assess the relationship between size and bank shareholding. If the bank

acts as a monitor of the company or holds stocks to gather more information, then bank

shareholding is more likely in small firms.

A second, important control variable is firm performance. In general, the relationship

between bank shareholding and firm performance is ambiguous. High profitability is a

good signal for future returns but a bank can also be involved in a debt restructuring

processes (debt for equity swap) and in deals aiming at turning around poor performers.

Moreover, we expect equityholding to be negatively correlated with performance if the

bank acquires equity to protect its own interests as creditor or to monitor the company:

the less profitable is the company, in fact, the lower is the probability of debt repayment.

As indicators of profitability we consider the return on equity (roe) or the return on

assets (roa). Similar arguments hold for indicators of firm’s growth. We consider the

growth in the number of employees (growthempl ) and in the level of sales (growthsales)

with respect to the previous year. As the sample includes unlisted companies, we cannot

employ forward indicators of growth opportunities such as the market value over book

value (mvbv ) or the Tobin’s q (tobinq).

Further results on the relationship between bank shareholding and lending can be

obtained by considering the total debt/total assets ratio (leverage) and financial condi-

tions through the ratio of current earnings to the cost of debt (ebitda/debtcost). If bank

shareholding is positively related to leverage and negatively to ebitda/debtcost, then the

evidence is in favor of complementarity with lending rather than substitution. The ratio-

nale can be traced back to a governance motivation (banks hold shares to limit the asset

substitution in the interests of shareholders and to act as an effective monitor) or to a

negative motivation for the shareholders of the company (banks induce the company to

11

undertake more conservative policies directed toward debt repayment). It is difficult to

disentangle these two motivations.

To assess the relevance of the debtholder-shareholder conflict, we use the ratio be-

tween tangible assets and total assets (tangible/assets). Firms with fewer tangible assets

are likely to be more opaque, to be characterized by higher asset substitution costs and

higher bankruptcy costs in the event of financial distress and the shareholder/debtholder

conflict is more intense. Tangible assets can be used as collateral and therefore are as-

sociated with a higher leverage ratio. As a consequence a positive relation between the

fraction of tangible assets and bank shareholding provides evidence of complementarity

with lending and that the debtholder-shareholder conflict is not intense.

As it is widely documented, in the Italian financial market ownership of companies is

highly concentrated, see [Bianchi et al. (2001)]. In Table 1 we provide data on ownership

structure of companies listed in the Italian stock market. Note that the stake of the

largest shareholder is decreasing over time and the fraction of listed companies controlled

de jure (the largest shareholder holds a stake above 50%) has decreased significantly by

15%. However, ownership concentration is still high compared to other financial markets

and it is also interesting to note that banks’ and institutional investors’ shareholding is

quite limited.

To take into account these peculiarities of the Italian financial market, in our analysis

we consider ownership structure through the fraction of stocks of the largest shareholder

(lrgshr ) which is defined as the percentage of voting rights held by the largest share-

holder, or by the coalition of shareholders controlling the company. Notice that a bank

cannot be the largest/controlling shareholder of a non financial company and therefore

lrgshr does not include bank shareholding. In some regressions we consider the dummy

variable control assuming a value equal to 1 if the company is controlled (the largest

shareholder/coaltion holds a stake above 30%) and zero otherwise.

The variables lrgshr and control give us a measure of the degree of control exercised

by the largest shareholder over the firm, which is likely to influence the decision of a bank

to hold a stake. If the intention is to influence the management of the firm to favor the

12

Table 1: Ownership structure of listed companies

year

lrgshr outsid ins

bank

found inves

cont

contf

coalit

noncon

1996

50.4

10.7

1.9

4.3

3.8

0.8

66.8

12.2

4.8

16.2

1997

38.7

8.4

2.2

5.1

3.1

0.1

48.1

12.4

6.3

33.2

1998

33.8

9.7

2.5

4.8

5.2

0.1

32.3

21.7

7.4

38.6

1999

44.2

8.2

1.5

5.3

4.5

0.2

55

16.7

10.8

17.5

2000

44

9.4

3.2

5.9

5.0

0.3

51.4

18.5

9.6

20.5

Note: Data, from statistics by the Stock Exchange Authority (Consob), concern all the stock market

(including financial companies) and refer to stakes of shareholders holding stakes above 2% weighted by

the stock market capitalization. Columns 2-3 report the stake of shareholders controlling the company

(lrgshr ) and of non controlling relevant shareholders (outsid ), columns 4-7 report the stake detained by

insurance companies (ins), banks (bank ), foundations (found ), mutual/private equity funds (inves); the

last four columns report the fraction of companies controlled by absolute majority (cont), controlled de

facto

, i.e., the largest shareholder controls the company with a stake below 50% (contf ), controlled by a

coalition of shareholders (coalit), non controlled companies (noncon).

creditor, then tightly controlled firms are not very attractive and therefore the incentive

to acquire equity in this type of firms is smaller. On the other hand, with concentrated

ownership, private benefits of control (including exploitation of debtholders) are high,

and therefore banks have an incentive to hold shares and to monitor the company.

11

As a measure of investment liquidity we consider a dummy variable assuming value

equal to 1 if the company is listed and zero otherwise (listed ). If a bank holds equity

in order to condition management and solve governance problems, it will more likely

prefer unlisted firms, as these problems are likely to be more relevant in firms less open

to market scrutiny. The shareholder/debtholder conflict, instead, does not have a clear

relationship with the variable listed. On one side, in fact, the conflict is greater if a firm

is unlisted. On the other side, in case of distress, swapping debt with equity is easier if a

11

Ownership structure variables could be used, as in [Chirinko and Elston, 2003], to test for bank

stockholdings as a means of control substitution: obviously a large stake by the largest shareholder

is likely to be associated with a small stake by minority shareholders (including banks) but a negative

dependence of bank stockholdings on lrgshr or control also provides an indicator of the attempt by banks

to stabilize the control of the company.

13

firm is listed. Obviously, if the bank holds equity as a pure investment, listed companies

are likely to be more appealing.

We finally consider, as a further proxy for bank-firm relationships, the number of

bank-lending relationships which is measured as the number of banks currently lending

to the firm (banknr ). As for the variable partloan described above, a negative relationship

between bank shareholding and banknr is indicative of complementarity between bank

shareholding and lending. If main bank relationships occur, in fact, it is more likely that

equity holding and bank debt are closely tied.

Summary statistics of all the variables described above are reported in Table 2.

Companies of the sample on average have 11,522 employees and more than 2 billion

euros in sales. Profitability, as measured by the return on equity (roe), has not been

on average very high, just 6.25% during the sample period. Total debt over total assets

(leverage) amounted to about 0.6, a large fraction of the assets is actually tangible assets

(79%).

From the data we notice two peculiarities of the Italian corporate structure. Con-

firming data on listed companies reported above, ownership is strongly concentrated,

i.e., the average stake of the largest shareholder is next to 80%, and the median is 94%

while the stake held by banks (bankshr ) and institutional investors (instshr ) on average

is less than 2%. If we consider however only the companies where banks have a positive

equity stake we see that bank shareholding is 8.7% while average shareholding by the

controlling shareholder in these companies is 48.3%. The stake detained by the largest

shareholder and by banks do not sum up to 100%, as a matter of fact there are other

outside blockholders (institutional investors, companies, etc.). Note that even a stake

below 10% is enough to play a relevant role on the governance and company decisions:

the threshold to appoint members of the board of auditors of listed companies is usually

smaller than 3% and thresholds for shareholders’ activism (e.g. to promote a derivative

suit against executive directors) are below 5%. Confirming the evidence provided in

[Detragiache et al., 2000] and in many other papers, Italian firms have multiple banking

relationships: on average, companies of the sample have obtained long term loans by

14

5.42 banks and half of the firms of the sample have more than 4 banking relationships.

Table 2: Descriptive statistics on the full sample

Variable

mean

median

st. dev.

roe

6.25

8.17

50.86

roa

0.038

0.027

0.067

sales

2,333,718

852,153

5,291,202

employees

11,522

3,875

5,776,154

growthempl

0.07

-0.006

0.69

growthsales

0.29

0.07

3.33

ebitda/debtcost

1.133

0.328

10.489

leverage

0.6

0.64

0.256

tangible/assets

0.79

0.63

0.87

lrgshr

78.2

94.4

26.68

banknr

5.42

4.00

4.62

bankshr

1.53

0

5.49

instshr

1.36

0

5.86

mvbv#

1.97

1.18

6.92

tobinq#

1.27

1.05

0.77

Note: # computation has been restricted to the subsample of listed companies (483 observations).

4

Empirical analysis

We develop our analysis in three steps. First, we compare the sample of companies with

banks among their shareholders and the sample of companies with no bank shareholding.

Then we provide an analysis of the determinants of bank stockholdings through a Tobit

analysis. Finally, we single out a set of banks and we perform a Tobit regression on

the determinants of their stockholdings looking at bank-non financial company debt

relationships.

15

Table 3: Mean and median comparison of companies with banks’ shareholding (bank)

and companies without banks’ shareholding (no bank).

Mean

Median

no bank

bank

p-value

no bank

bank

p-value

roe

7.01

-11.1

0.003

8.85

5.23

0.001

roa

0.04

0.015

0.006

0.03

0.016

0.001

assets

920,713

1,450,537

0.023

242,455

415,881

0.001

sales

1,945,937

4,123,731

0.001

774,236

1,409,153

0.001

employees

9,509

20,809

0.001

3,535

7,920

0.001

growthempl

0.08

0.04

0.44

-0.005

-.010

0.1243

growthsales

18.16

31.66

0.085

6.54

6.9

0.639

ebitda/debtcost

1.118

1.205

0.91

0.344

0.276

0.003

leverage

0.584

0.656

0.001

0.638

0.68

0.001

tangible/assets

0.77

0.87

0.12

0.61

0.71

0.03

lrgshr

84.6

48.9

0.000

100

50.9

0.000

banknr

5.16

6.48

0.001

4

5

0.005

mvbv#

2.151

1.587

0.072

1.212

1.100

0.005

tobinq#

1.337

1.185

0.002

1.071

1.021

0.002

Note: p-value refers to the probability that mean or the median of the two samples are different. #

computation has been restricted to the subsample of listed companies (483 observations).

16

4.1

Univariate analysis

The univariate comparison of companies with bank shareholding and no bank share-

holding offers some interesting insights. The results of this comparison, with simple

univariate tests on the distribution of the variables (mean and median), are reported

in Table 3. In the column no bank we refer to the sample of companies with no bank

shareholding, while in the column bank we refer to companies with bank shareholding.

Companies with bank shareholding are less profitable - with a negative mean roe- rela-

tive to companies with no banks among shareholders. They are also characterized by a

larger size, as expressed by the level of sales, the number of employees and total assets.

While variables assessing past growth provide conflicting evidence with weak statistical

significance, forward looking growth indicators (computed on the subsample of listed

companies), such as tobinq (market value of assets over the book value) and mvbv (mar-

ket value over the book value), signal that bank shareholding is more likely in companies

with poor growth opportunities. We observe that companies with bank shareholding, on

average, have a higher leverage ratio and a higher fraction of tangible assets, control is

less concentrated and the number of bank relationships is higher.

4.2

Tobit analysis

A large number of firms in our sample does not have banks among their shareholders. Our

dependent variable, the fraction of a firm’s equity owned by banks (bankshr), is therefore

only partly continuous with positive probability mass at zero. The estimation method

that is most commonly used, in this case, is the Tobit model. In Table 4 we provide a

Tobit analysis of our panel. Building on the discussion about the motives behind banks’

decision to hold equity in a firm and on the univariate analysis provided above, we

first analyze a reference model where we include, as regressors, lasset, ebitda/debtcost,

leverage, tangible/assets, listed, control, partloan, roe, growthempl (see column 1). In

the second column, we use an alternative indicator of firm growth (growthsales) and in

the last two columns we include banknr as exogenous variable, restricting our attention

to a subsample.

17

We choose to use, for these regressions, a random effect estimator rather than a

pooled estimator. The use of a pooled estimator, in fact, is not indicated when the

proportion of the total variance contributed by the panel-level variance component (the

Rho in the last two rows of the table) is different from zero, which is actually the case

in our estimates. We have excluded the fixed effects estimator because of the large

reduction in the number of observations that this would imply

12

. For each regression

we compute the maximum value of the likelihood function L(Ω). The chi2 statistic is

χ

2

= 2 {ln L(Ω) − ln L(ω)} where L(ω) is the value of the likelihood function under the

assumption that all coefficients except for the constant are zero. As we can see from our

estimates, all the regressions pass the chi2 test.

A first, very important result of the analysis is that the dummy variable partloan,

that takes the value of 1 if the bank that acquires a stake in the capital of the firm is also

a creditor of the firm and 0 otherwise, not only is positive, but it also turns out to be the

one with the highest degree of statistical significance. There is therefore, a very strong

and significant complementarity between bank shareholding and lending: if a bank is a

lender to the firm, then is also more likely to hold equity.

Beside the variable partloan, there are only two variables that are highly significant

in all the regressions: the roe which enters with a negative coefficient, and the ratio of

debt to total assets (leverage) with a positive coefficient. Given the particular meaning of

our dependent variable, this implies that banks not only tend to acquire equity in firms

that are unprofitable and highly indebted, but also the size of their stake is negatively

correlated with profitability and positively correlated with leverage.

Consistently with this interpretation, the indicator of debt sustainability (ebitda/debtcost)

is negatively correlated with our dependent variable, although not always statistically

significant. As far as growth perspectives are concerned, the variable growthempl is never

statistically significant, but the coefficient of the alternative growth indicator (growth-

sales) is negative and significant in equation (2) although not in equation (4).

12

In a fixed effects estimation observations non associated with a change in the fixed effects are dropped

from the regression. Since in our sample banks’ shareholding, for each firm, does not change very much

across time the use of fixed effects would imply a large drop in the number of observations.

18

Table 4: Bank shareholding

(1)

(2)

(3)

(4)

lasset

0.76620

0.74310

0.22116

0.23132

(5.21)***

(5.03)***

(0.97)

(1.02)

ebitda/debtcost

-0.0001

-0.0001

-0.0001

0.0001

(1.66)*

(1.60)

(0.06)

(0.01)

leverage

4.19786

4.17689

5.09793

5.28276

(5.94)***

(5.99)***

(4.00)***

(4.25)***

tangibile/asset

0.330

0.288

1.515

1.495

(1.42)

(1.22)

(2.80)***

(2.88)***

listed

0.39664

0.33008

0.41503

0.41082

(1.90)*

(1.64)

(0.88)

(0.87)

control

-0.03047

0.00428

-3.23617

-3.33425

(0.08)

(0.01)

(4.40)***

(4.74)***

partloan

8.61651

8.06547

8.29176

8.40252

(17.50)*** (16.88)*** (12.88)*** (13.36)***

roe

-0.00595

-0.00553

-0.00933

-0.00925

(4.64)***

(4.42)***

(5.67)***

(5.77)***

growthempl

-0.00813

-0.00638

0.14003

(0.05)

(0.44)

growthsales

-0.00068

0.00042

(2.46)**

(0.17)

banknr

-0.07090

-0.07233

(1.65)*

(1.72)*

constant

0.76853

0.77768

0.49039

0.492191

(0.014)

(0.0137)

(0.0466)

(0.0439)

Observations

1160

1155

797

787

Nr. of sector

206

205

171

170

Log likelihood

-3126.1

-3089.8

-2166.7

-2141.5

Wald chi2(8)

1300.4

1237.0

347.7

384.1

p.value

0.0001

0.0001

0.0001

0.0001

Rho

0.7683

0.7776

0.4904

0.4922

St. error

0.0141

0.0137

0.0466

0.0439

Note: Tobit analysis, the endogenous variable is the stake of the company held by banks. Absolute value

of z statistics is reported in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

19

The coefficient of the dummy variable control, indicating whether the company is

controlled by the main shareholders, is negative and significant in two out of four regres-

sions and insignificant in the others, indicating a preference for the acquisition of equity

in less controlled firms where the bank may exert greater influence

13

. The coefficient of

the variable listed is always positive with an acceptable level of significance in the first

regression. The variable tangible/assets, enters all the regressions with a positive sign

and in the last two regressions is statistically significant.

In the Tobit regressions the number of banks that have credit relationships with the

firm (banknr ) is negatively correlated with the equity stake held by banks and significant

at the 10% level, which shows that banks tend to acquire equity when the company

borrows from a small number of banks.

Overall, it seems that Italian banks tend to become shareholders of the companies

they lend to, and this indicates quite clearly that shareholding and lending complement

each other. Banks seem to hold equity in companies that are highly leveraged, that do not

generate enough earnings relatively to their financial obligations and are neither the most

profitable nor the most dynamic. We cannot exclude that high leverage and bank equity

are jointly driven by a poor performance. Moreover, banks are shareholders of firms

that are less tightly controlled and that have a smaller number of credit relationships,

i.e. where it is easy to exercise some form of influence.

This evidence can be interpreted according to the hypotheses introduced in Section

2 on the rationale for bank shareholding. Results are not too informative on the first

hypothesis: weak profitability/financial conditions can be associated with turn around

deals and so are not at odds with the pure investment motive. We can exclude that

banks detain stocks to gather an informational advantage to allocate credit in a proper

way, as a matter of fact there is no evidence that banks hold equities in more opaque

companies. The second (governance) hypothesis deserves a deeper analysis.

Results indicate that banks tend to hold equity in order to influence the management

13

The result is partially due to the fact that the stake of outside blockholders in a controlled company

is smaller than in a non controlled company

20

of the company. However, this behavior does not seem to be dictated by some kind of

”virtuous” motivation, i.e., to solve agency problems, to reduce asset substitution or

to accompany the firm in a solid growth path. Rather, the fact that banks tend to be

shareholders of firms that have a lower return on equity and a large, unsustainable debt

seems to indicate that they intend to use their influence as shareholders to induce the

management to undertake actions which are favorable to creditors. This is even more

likely if bank shareholding is the result of a debt-equity swap undertaken as a means

to restructure debt. If the firm is under financial distress, in fact, the shareholder-

debtholder conflict becomes particularly acute. The only finding that is not supportive

of this interpretation is that in the last two regressions bank shareholding is positively

and significantly related to the share of tangible assets. Since this finding is limited to a

subsample of firms we cannot draw definitive conclusions on the issue. It might indicate,

however, that the shareholder-debtholder conflict is not very intense.

One problem with the estimates presented above is the possibility that the variable

partloan is not entirely exogenous, i.e. that a bank’s decision to lend to a firm is influenced

by the bank holding an equity stake in the borrowing firm. We think that this possibility

is quite remote, especially considering that partloan is a dichotomous variable, i.e. does

not refer to the size of the loan but to the choice of lending or not lending to the

firm.

A profit mazimizing bank whose main business is to grant credit will lend to

any firm independently of whether it holds equity of some specific firms and actually,

considering that banks, in our sample, seem to hold equity in the ”weakest” companies,

it is quite unlikely that the decision to grant credit is influenced by banks’ shareholding.

Nevertheless this possibility cannot be excluded a priori and therefore, in order to assess

whether partloan is indeed endogenous we have performed the Smith-Blundell exogeneity

test for Tobit regressions

14

.

14

For details on this test see [Smith and Blundell, 1986]. The test assumes that, under the null hy-

pothesis, the model is appropriately specified with all explanatory variables as exogenous; under the

alternative hypothesis, the suspected endogenous variable (in our case partloan) is expressed as a linear

projection of a set of instruments, and the residuals from this first-stage regression are added to the

model. Under the null hypothesis, these residuals should have no explanatory power.

21

To perform the test we run a set of pooled Tobit regressions where the variable

partloan has been instrumented by exogenous variables (the dummy variable bond, which

takes the value 1 if the firm finances itself also through the issue of bonds and zero

otherwise, the dummy variable ministry, which takes the value 1 if if the firm is indebted

with the Ministry of the Industry and 0 otherwise). These last two variables play an

important role in the firms’ decision to borrow from a bank since they imply the recourse

to non bank funds, but they are clearly not relevant for a bank’s decision to hold a share

of the capital of a non financial firm. In all the various specifications of the model

considered above we could never reject the null hypothesis that the variable partloan is

exogenous.

15

The regressions in Table 4, therefore, do not suffer from the inconsistency

that would be induced if indeed partloan turned out to be endogenous.

4.3

Bank specific equity holding

In the previous section we have shown that bank shareholding of non financial firms

is explained, among others, by the fact that the firm is also a borrower of one of the

shareholding banks. The finding suggests that there is a relationship between a bank’s

equity stake in a non financial firm and the fact that the company is also indebted to the

shareholding bank. If we want to investigate, at a deeper level, the positive relationship

between firm indebtedness and bank shareholding, we must try to investigate whether

there exists a direct relationship between the fraction of shares owned by a bank in one

particular non financial firm and the amount of long term loans this bank has granted

to the same firm. In order to do this, we construct a dataset where we consider the

15

Consider for example the basic specification of the model, i.e the one corresponding to column (1)

of Table 4. If the variable ministry is employed as an instrument of partloan, together with all the other

covariates, the test statistic F(1,846)= 0.0001763 with P-value = 0.9894. If the variable bond is employed

as an instrument of partloan, together with all the other exogenous variables, the test statistic F(1,939)

= 0.1518892 with P-value = 0.6968. If both the variables ministry and bond are employed as instruments

of partloan, together with all the other exogenous variables, the test statistic F(1,755) = 1.133895 with

P-value = 0.2873. Similar values are obtained using different specifications of the model, as in columns

(2) and (3) of Table 4.

22

seven most important banks (Banca Intesa, Unicredit, Capitalia, MPS, BNL, Sanpaolo

and Mediobanca) that appear in the Mediobanca survey and we consider, as dependent

variable, the equity stake held by bank i into firm j at time t. We call this variable

bankown. As in the previous analysis, we use a panel Tobit analysis with random effects.

The dependent variable is regressed on the variable loanshare, which is defined as the

amount of long term debt firm j owes to bank i at time t, as a percentage of total long

term debt of firm j, and other firm specific variables. Among the other covariates we

include total assets (asset), total sales (sales) and the variables we have used in the

previous regressions (growthempl, growthsales, ebitda/debtcost, leverage, tangible/assets

and listed ). We include both the variables lrgshr and control to capture ownership

structure. In order to account for heterogeneity accross banks we include a dummy

variable that, for each bank i = 1..., 7, takes value 1 when bank i is involved and 0

otherwise. Coefficients for these variables are not reported below.

The results of our estimates are reported in Table 5. As we can see from all the

estimates reported in the Table, the variable loanshare, which indicates the percentage

of long term debt that a firm owes to a specific bank, not only has a positive sign, but is

also the most significant determinant of the equity stake of the bank in that firm. Banks’

decision to hold capital of a non financial firm is strongly influenced, therefore, by the

amount of credit granted to the firm, confirming the strong complementarity between

bank shareholding and lending that we have found in the previous section. The sign of

the other covariates, again, seems to indicate the existence of a shareholder-debtholder

conflict: the coefficient of roe is always negative, and highly significant in two cases

out of three, the growth variable growthempl is insignificant and has a negative sign,

although the other growth variable (growthsales) in the last equation has a positive

and statistically significant coefficient, the variable leverage has always a positive sign

(although it looses significance in equations 2 and 3) indicating that banks tend to acquire

equity of indebted firms. The signs of the variables listed, lrgshr and control show that

banks tend to hold equity of firms which are more transparent and less tightly controlled.

23

Table 5: Bank specific regressions

(1)

(2)

(3)

loanshare

0.2760

0.07782

0.29393

(5.58)***

(3.23)***

(5.9)***

roe

-0.0002

-0.00008

-0.00017

(2.30)**

(2.01)**

(1.28)

asset

0.0005

0.00053

(2.56)**

(2.67)***

sales

0.00001

(2.02)**

growthempl

-0.0033

-0.00386

(0.27)

(0.73)

growthsales

0.00006

(2.16)**

ebitda/debcost

-0.0001

-0.00001

-0.00001

(0.10)

(0.46)

(0.13)

leverage

0.0913

0.01495

0.07266

(2.14)**

(0.68)

(1.66)*

tangible/assets

-0.0197

-0.01025

-0.01393

(1.02)

(1.48)

(0.54)

listed

0.3055

0.05421

0.24263

(1675)***

(3.84)***

(5.00)***

control

-0.0875

-0.05270

(1.79)*

(2.20)**

lrgshr

-0.00211

(3.11)***

Constant

-0.0080

0.06288

0.02731

(0.14)

(2.10)**

(0.32)

Observations

8261

8883

8262

Number of sector

206

206

206

Log likelihood

-9301.1

-9383.4

-9094.8

Wald chi2(8)

444.7

79.2

323.3

p.value

0.0001

0.0001

0.0001

Note: Tobit regressions, the endogenous variable is the stake held by a specific bank. Absolute value of

z statistics is reported in parentheses.* significant at 10%; ** significant at 5%; *** significant at 1%.

24

5

Institutional investors

Since results described in Table 4 suggest that banks, in deciding to hold shares of non

financial firms, do not act as normal investors focusing on profitability and financial

solidity, an interesting question is whether banks differ, in this respect, from other in-

stitutional investors. In order to address this issue, we analyze whether institutional

investors’ shareholding is related to the explanatory variables included in the previous

model. We include, among institutional investors, insurance companies, mutual and

pension funds and we explicitly exclude banks. In Table 6, we report the results of

Tobit regressions with random effects, where the dependent variable is the equity stake

of institutional investor (excluding banks) in a particular firm. Note that institutional

investors on average hold a stake equal to 1.36% (banks hold a stake equal to 1.53%).

Equations estimated in Table 6 include essentially the same explanatory variables

as the reference model of Table 4.

16

Notice that, differently from what happens in the

previous analysis on banks’ shareholding, the coefficient of roe is positive although not

statistically significant, and the coefficient of leverage is negative and non significant.

The two major determinants of institutional investors’ shareholding seem to be the size

indicator lsales and the variable listed. To a lesser extent, also the variable lrgshr, which

is the share of the larger shareholder of the company, is positive and significant.

These results suggest the existence of a significant difference between the motiva-

tions behind institutional investors’ shareholding in non-financial firms and those of

banks. Unlike banks, that tend to hold equity in non performing, highly indebted firms,

institutional investors seem to prefer large firms that are listed in the stock market and

that are less tightly controlled. These results do not support the view that banks act as

pure investors aiming at maximizing investment returns and confirm the hypothesis that

bank shareholding has different motivations, i.e., the desire to influence the management

16

Obviously we have excluded the variables referring to the bank-firm lending relationships. Moreover

we have used as indicator of size, the logarithm of total sales, and as an indicator of firm growth the

logarithm of sales growth. We have observed, in fact, that the use of lasset and growthempl creates some

collinearity problems.

25

Table 6: Institutional investors’ shareholding (excluding banks)

(1)

(2)

lsales

0.2341

0.2864

(4.07)***

(4.57)***

ebitda/debtcost

0.0001

0.0001

(0.25)

(0.29)

leverage

-0.1801

-0.2243

(0.24)

(0.30)

tangible/asset

0.0632

0.0549

(0.17)

(0.15)

listed

1.2777

0.8894

(3.68)***

(2.14)**

control

-0.3818

(0.67)

lrgshr

-0.0113

(1.68)*

roe

0.0013

0.0014

(0.72)

(0.78)

growthsales

0.0005

0.0005

(1.18)

(1.23)

Observations

1156

1163

Number of sector

205

205

Log likelihood

-3507.0

-3522.8

Wald chi2(8)

243.7

277.3

p.value

0.0001

0.0001

Rho

0.6989

0.7034

Note: Tobit regression, the endogenous variable is the fraction of shares held by institutional investors.

Absolute value of z statistics is reported in parentheses. * significant at 10%; ** significant at 5%; ***

significant at 1%.

26

of the company, or simply by debt-equity swaps.

6

Conclusions

We have analyzed, in this paper, the motivations behind banks’ shareholding of non

financial firms for a panel of large Italian firms in a period that followed immediately

the introduction, in the Italian system, of universal banking as the standard form of

organization of the banking system.

Italian banks tend to invest in firms that have a credit relationship with them, indicat-

ing the existence of a strong complementarity between bank shareholding and lending.

Moreover, unlike other institutional investors, bank hold equity of firms that are less

profitable, more indebted and less tightly controlled which seem to suggest that bank

shareholding is motivated by the need to solve, in their favor, the debtholder/shareholder

conflict, i.e., to induce companies to undertake actions that would maximize the prob-

ability of debt repayment. There is little support, in the data, for the existence of

a virtuous bank-non financial company shareholding relation associated with gover-

nance/monitoring arguments.

We can conclude that, after the deep financial deregulation process occurred in the

1990s, we have not observed a significant move of the Italian banking system toward

a more active involvement in the management of firms, to improve their governance or

to overcome informational asymmetries, i.e., the factors that were considered, especially

in the literature of the early 1990s, as the major advantage of the main bank systems

of Germany and Japan. The lack of an involvement of banks in the governance of non

financial firms is consistent with results obtained by the recent literature on the Japanese

experience and cast doubt on the positive role of banks as equity holders of non financial

companies.

27

References

[Banca d’Italia, 2006]

Banca d’Italia (2006) I conti finanziari: la storia, i metodi, l’Italia,

i confronti internazionali.

[Bartirolo and De Bonis, 2005]

Bartirolo L. and De Bonis R. (2005) The financial sys-

tems of European countries: theoretical issues and empirical evidence. Irving Fisher

Committee Bulletin, 21, July.

[Berlin et al., 1996]

Berlin, M., John, K. and Saunders, A. (1996) Bank equity stakes in

borrowing firms and financial distress. Review of Financial Studies, 9: 889-919.

[Bianchi et al. (2001)]

Bianchi M., Bianco M., Enriques L. (2001), Pyramidal groups

and the separation between ownership and control in Italy, in F. Barca e M. Becht

(eds.), The control of corporate Europe, Oxford University Press, Oxford.

[Bianco and Chiri, 1997]

Bianco, M., and Chiri, S. (1997) Le partecipazioni bancarie

nelle imprese in Italia: investimenti e ristrutturazioni finanziarie. Banca, Impresa e

Societ´

a, XVI: 393-423.

[Chirinko and Elston, 2003]

Chirinko, R. and Elston (2003) Finance, control, and prof-

itability: the influence of German banks. Forthcoming Journal of Economic Behavior

and Organization.

[Cobham, Cosci and Mattesini, 1999]

Cobham, R., Cosci, S. and Mattesini, F. (1999)

The Italian Financial System: neither Bank Based nor Market Based. The Manchester

School, n.3 vol 67, June.

[Detragiache et al., 2000]

Detragiahce, E, Garella, P. and Guiso L. (2000) Multiple ver-

sus single banking relationships: theory and evidence. Journal of Finance, 55: 1133-

1155.

[Dewatripont and Tirole, 1994]

Dewatripont, M. and Tirole, J. (1994) A theory of debt

and equity: diversity of securities and manager-shareholder congruence. Quarterly

Journal of Economics: 1027-1054.

28

[Dewenter et al., 2006]

Dewenter K., Hess, A. (2006) International evidence on relation-

ship and transactional banks as delgated monitors, mimeo.

[Dewenter et al., 2007]

Dewenter K., Hamao, Y. and Hess, A.(2007) Banks’ equity hold-

ings and the soft budget problem: evidence from Japan’s prolonged recession, mimeo.

[Diamond, 1984]

Diamond D. (1984), Financial intermediation and delegated monitor-

ing. Review of Economic Studies, 51, pp. 393-414.

[Flath, 1993]

Flath D. (1993) Shareholding in the Kieretsu, Japan’s Financial Groups.

Review of Economics and Statistics, 75: 249-257.

[Franks et al., 2007]

Franks J., Mayer C., Miyajima H. (2007) Evolution of ownership:

the curious case of Japan. mimeo.

[Gompers and Metrick, 2001]

Gompers, P. and Metrick, A. (2001) Institutional investors

and equity prices. Quarterly Journal of Economics: 229-259.

[Gorton and Schmid, 2000]

Gorton, G. and Schmid, F. (2000) Universal banking and

the performance of German firms. Journal of Financial Economics, 58: 29-80.

[Hoshi et al., 1991]

Hoshi, T., Kashyap, A. and Scharfstein, D. (1991) Corporate struc-

ture, liquidity, and investment: Evidence from Japanese industrial groups. Quarterly

Journal of Economics, 106: 33-60.

[Kang and Stulz, 2000]

Kang, J. and Stulz, R. (2000) Do banking shocks affect borrow-

ing firm performance? An analysis of the Japanese experience. Journal of Business,

73: 1-23.

[La Porta et al., 1998]

La Porta, R., Lopez-de-Silanes, F., shleifer, A., Vishny R. (1998)

Law and finance. Journal of Political Economy, 106: 1113-55.

[Lichtenberg and Pushner, 1994]

Lichtenberg, F. and Pushenr, G. (1994) Ownership

structure and corporate performance in Japan. Japan and the World Economy, 6:

239-261.

29

[Miyajima et al., 2003]

Miyajima H., Haramura K. and Enami Y. (2003) Evolution of

ownership strcuture in postwar Japan: formulation and unwinding of stable share-

holders. Financial Review 68.

[Miyajima and Kuroki, 2006]

Miyajima H. and Kuroki F. (2006) The unwinding of cross-

shareholding in Japan: causes, effects and implications. Mimeo.

[Morck, et al., 2000]

Morck, R., Nakamura, M. and Shivdasani, A. (2000) Banks, own-

ership structure, and firm value in Japan. Journal of Business, 73: 539-567.

[Peek and Rosengren, 2005]

Peek J., Rosengren E. (2005) Unnatural selection: perverse

icnentives and the misallocation of credit in Japan. The American Economic Review,

95: 1144-1166.

[Petersen and Rajan, 1994]

Petersen, M. and Rajan, R. (1994) The benefits of lending

relationships: evidence from small business data. Journal of Finance, XLIX: 3-37.

[Prowse, 1992]

Prowse S. (1992) The structure of corporate ownership in Japan. Journal

of Finance, 47: 1121-1140.

[Sheard, 1994]

Sheard, P. (1994) Interlocking shareholdings and corporate governance

in Japan, in M. Aoki and R. Dore eds. The Japanese firm: sources of competitive

strength. Oxford, Oxford University Press, 324-349.

[Smith and Blundell, 1986]

Smith, R.J. and Blundell, R.W. (1986) An exogeneity test

for a simultaneous equation Tobit model with an application to labor supply. Econo-

metrica, 54, 3: 679-685.

[Weinstein and Yafeh, 1998]

Weinstein, D. and Yafeh, Y. (1998) On the costs ofa bank-

centered financial system: evidence from the changing main bank relationships in

Japan. Journal of Finance, 53: 635-672.

[Yafeh and Yosha, 2003]

Yafeh, Y. and Yosha, O. (2003) Large shareholders and banks:

who monitors and how? Economic Journal, 113: 128-146.

30