Content uploaded by Serdar Dinc
Author content
All content in this area was uploaded by Serdar Dinc on Jul 20, 2014
Content may be subject to copyright.
3057
(Journal of Business, 2006, vol. 79, no. 6)
䉷2006 by The University of Chicago. All rights reserved.
0021-9398/2006/7906-0011$10.00
I. Serdar Dinc¸
Northwestern University
Monitoring the Monitors: The
Corporate Governance in Japanese
Banks and Their Real Estate
Lending in the 1980s*
The modern financial intermediation theory empha-
sizes the monitoring role of banks over nonfinancial
companies as part of their lending to and, wherever
allowed, equity holding in, the latter companies.
1
Yet
what is much less understood is the impact of these
activities on monitoring the banks, in particular, the
bank managers themselves. Does monitoring the man-
agers of nonfinancial companies make disciplining
bank managers themselves more complicated? Does
corporate governance of banks become more difficult
if the bank also lends to its shareholders and/or holds
shares in its own shareholders? Put another way, how
different is the corporate governance in banks from
* An earlier version of this article was circulated with the title
“The Shareholding Structure of Japanese Banks and Their Real
Estate Lending in the 1980s.” I started this research when I was
visiting the Faculty of Economics and the Center for International
Research on the Japanese Economy at the University of Tokyo. I
am grateful for the hospitality and financial support. I would also
like to thank the referee, Masahiko Aoki, Akiyoshi Horiuchi, Har-
umi Ito, Takuma Nakatsuka, Tetsuji Okazaki, Masahiro Okuno-
Fujiwara, Katsutoshi Shimizu, and seminar participants at numerous
institutions and Western Finance Association, for comments and
discussions; several practitioners who prefer to remain anonymous,
for insights; Julian Atanassov, Makoto Itani, Patrick McGuire,
Kamran Parekh, Alexander Serebrinsky, Junko Takahara, Shinako
Tsugawa, and the staff of the Economics Library at the University
of Tokyo, for their help with data and Japanese. Contact the author
at s-dinc@kellogg.northwestern.edu.
1. For a recent survey, see Gorton and Winton (2003), who also
argue that corporate governance in banks remains poorly understood.
The corporate governance
role of banks in “bank-
centered” countries like
Japan has been well stud-
ied. This article studies
the corporate governance
in Japanese banks. It
shows that large share-
holders restrained bank
managers from real estate
lending in the 1980s.
However, this effect was
absent for the sharehold-
ers who belonged to the
same keiretsu (corporate
grouping) as the bank.
Relationship banking and
cross shareholding pre-
vented these shareholders
from disciplining the
bank managers. In finan-
cial systems where banks
play a large role in cor-
porate governance, the
more effective the banks
are in monitoring other
companies, the more dif-
ficult it may become to
monitor bank managers.
3058 Journal of Business
that in nonfinancial companies due to the particular role played by banks in
the economy? This article studies these questions in the context of Japan,
where banks play a very large role in the economy and where these problems
seem to be especially acute.
The role of Japanese banks in monitoring other companies through their
lending and shareholding has been well demonstrated.
2
These relationships
are considered particularly, though not exclusively, pronounced in corporate
groupings called (financial) keiretsu. Some of the largest Japanese banks are
at the center of these keiretsu, and many of the largest shareholders of these
banks are themselves members of banks’ keiretsu. Furthermore, Japanese
banks increased their real estate lending substantially in the 1980s even though
the Japanese bank managers knew about the higher risks involved with such
lending, as documented in the next section. Real estate lending in this period,
almost half of which was made by keiretsu banks, subsequently proved to be
detrimental for the banks. Did the monitoring capabilities of some Japanese
banks over their shareholders through loans and cross shareholdings prevent
these shareholders from restraining bank managers from real estate lending?
This article studies this question by focusing on the role of shareholders in
the real estate lending of Japanese banks in the 1984–89 period, which includes
the “Japanese bubble.”
This article shows that shareholders, on average, had a restraining effect
on Japanese bank managers regarding real estate lending. Banks decreased
their real estate as the shares held by their largest five shareholders increased.
However, this effect was absent when the shareholders belonged to a keiretsu
led by the bank itself. This result is robust to controlling for bank size, income,
capital ratio, bank fixed effects, and time fixed effects, as well as to using
different measures of real estate lending.
By the end of 1989, real estate loans as a proportion of total loans exceeded
an average of 11.6% across the banks in the sample, with some banks lending
to the real estate sector more than 25% of all their loans. Moreover, these
numbers understate the banks’ exposure to real estate, for many of the loans
to other sectors had land as collateral. When commercial land prices in major
urban areas decreased by 58% from 1990 to 1995, banks found themselves
in serious difficulties.
Keiretsu banks were responsible for 45% of real estate loans outstanding
by all the banks in the sample at the end of fiscal 1989. Similarly, their total
lending was more than 40% of all outstanding loans by the sample banks.
Hence, the corporate governance problems in keiretsu banks are likely to have
2. See, e.g., Kaplan (1994), Kaplan and Minton (1994), Kang and Shivdasani (1995, 1997),
Morck and Nakamura (1999), and Morck, Nakamura, and Shivdasani (2000) for empirical evi-
dence. Aoki and Patrick (1994) and Hoshi and Kashyap (2001) provide detailed treatments of
Japanese banking.
Corporate Governance in Japanese Banks 3059
exacerbated the financial crisis in Japan through the very large role played
by these banks in the Japanese economy.
3
The closest work to this article is that of Gorton and Rosen (1995), who
study the role of corporate governance in risk taking by U.S. banks in the
1980s. They first provide a theory to demonstrate that, when there is excess
capacity in banking and some banks—and their managers—have to exit bank-
ing, bank managers have incentive to take more risks than optimal from outside
shareholders’ perspective. They also provide empirical evidence that the en-
trenchment through managerial shareholding led to risk-taking behavior in
U.S. banks in the 1980s, a period of shrinking lending opportunities for these
banks due to deregulation.
4
The 1980s were also a time of deregulation in
Japan, and, as the number of subsequent mergers and bank failures suggests,
there was excess capacity in Japanese banking: 21 nationwide banks in 1989
coalesced into just seven financial holding companies or banks as of October
2005. However, managerial shareholding was negligible in Japanese banks.
Instead, shareholdings in keiretsu banks by same-keiretsu shareholders were
very important: among the largest five shareholders of keiretsu banks, keiretsu
members held, on average, more shares than non-keiretsu shareholders did.
This article focuses on the entrenchment effect created by the shareholders
over which the banks had control through keiretsu ties.
Relative to the research on corporate governance in nonfinancial companies
in Japan, there are few empirical studies that focus on disciplining bank
managers. Horiuchi and Shimizu (2001) find that the banks that offered jobs
to retired bank regulators enjoyed less strict monitoring by regulatory agents.
Anderson and Campbell (2004) show that the top managers of poorly per-
forming Japanese banks were more likely to be replaced in the 1990s. Hanazaki
et al. (2004) study the role of financial companies that are shareholders of
other financial companies. These articles do not consider the influence bank
managers have on some of the bank shareholders through lending and cross-
shareholding ties.
The findings in this article also complement—and contrast with—thestudies
that demonstrate distortions in bank lending caused by the influence of con-
trolling shareholders. La Porta, Lopez-de-Silanes, and Zamarripa (2003) study
how the controlling shareholders of Mexican banks channel resources to them-
selves through “related lending” at the expense of minority shareholders and
the government. Sapienza (2004), Dinc¸ (2005), and Khwaja and Mian (2005)
show how the politicians use government-owned banks for their own political
purposes. This article focuses instead on the inability of certain shareholders
3. See Gibson (1995) and Kang and Stulz (2000) for the effects of bank health on their
borrowers in Japan in the 1990s.
4. See also Saunders, Strock, and Travlos (1990), Knopf and Teall(1996), Demsetz, Saidenberg,
and Strahan (1997), and Anderson and Fraser (2000), among others, about the effects of corporate
governance on risk taking by banks in the United States. This literature is discussed in more
detail in the next section.
3060 Journal of Business
to discipline bank managers where their inability is caused by the bank’s own
relationship banking activities.
Section I presents the theoretical, empirical, and institutional background.
The next section discusses the main hypotheses, while Section III introduces
the data. Section IV presents the methodology and provides the regression
analysis. Section V contains the robustness checks. The conclusion follows.
I. Background
Despite the worldwide importance of banking crises, bank risk taking seems
to be best studied in the U.S. context. As the main insights are valid for other
countries as well, this section first reviews the theoretical background and
related empirical evidence in the U.S. context before discussing the institu-
tional background in Japan.
A. Theoretical Background and Empirical Evidence from the United
States
The moral hazard created by fixed-cost deposit insurance and low capital
values can be an important source of risk-taking behavior in banking, as
discussed by Merton (1977) and Keeley (1990), among others. When banks
have low capital values, they have incentives to take advantage of fixed-cost
deposit insurance by increasing the risk they undertake. Since they have only
low capital and the deposit insurance does not depend on their activities, they
have little to lose if the risk they undertake does not pay off; however, they
stand to gain substantially, if it does.
While the logic of this moral hazard argument is relatively straightforward,
it assumes that the shareholders make the lending decisions in banks. However,
if the bank managers make the loan decisions and disciplining the managers
is costly, the agency problems between shareholders and managers must also
be considered. In terms of risk taking, this agency problem may manifest
itself in two ways. The firm-specific investment made by managers may make
them more risk averse than is optimal from the shareholder perspective because
managers would lose their jobs and associated control benefits in bankruptcy.
On the other hand, as Gorton and Rosen (1995) argue, conservative behavior
may not be sufficient for the managers to maintain their jobs when their
investment opportunities have deteriorated and an “excess capacity” exists in
banking even if bank assets are still healthy. In that case, some banks have
to exit the industry; hence, some managers lose their jobs even if they behave
conservatively. Consequently, managers take more risk than is optimal from
the shareholder perspective. Intuitively speaking, if the risk pays off, they
keep their jobs; if not, they were going to lose their jobs anyway. Hence, in
an environment characterized by excess capacity in banking, the risks taken
by the managers increase as the cost of disciplining them increases.
Saunders et al. (1990) provide empirical evidence that the risks taken by
Corporate Governance in Japanese Banks 3061
U.S. banks in the period 1979–82 increased in the shares held by the managers,
especially in times of deregulation. Gorton and Rosen (1995) show that, as
their theory predicts, managerial entrenchment through insider ownership
played a more important role in the risk taking by American commercial
banks in the 1980s than the moral hazard problem created by the deposit
insurance did. Similarly, Knopf and Teall (1996) demonstrate that the risks
taken by U.S. thrifts before 1989 increased in the shares held by the managers
but decreased in the shares held by institutions. Demsetz et al. (1997) show
that insider shareholding increased the risks taken in U.S. banks with low
charter value during the relatively stable banking environment of the 1990s
but not in the banks with high charter value. Similarly, Anderson and Fraser
(2000) find that managerial shareholding in the United States is positively
related to risk taking in the 1980s but negatively related in the 1990s.
Hence, both theoretical arguments and empirical evidence suggest that the
shareholding structure affects the risks taken by a bank. In particular, the risks
taken by managers increase with the managerial entrenchment if the lending
opportunities deteriorate after deregulation.
B. Japan
This article focuses on the lending behavior of Japanese banks from 1984 to
1989.
5
In this period, the Japanese banking industry was healthy, unlike the
U.S. thrifts in the late 1970s.
6
However, it was also clear by 1984 that the
lending opportunities of Japanese banks had deteriorated due to the low de-
mand for bank funds because the traditional bank borrowers had become able
to generate large funds internally to finance their investment needs, which
were already decreased with the overall slowdown in growth. Furthermore,
capital market deregulation allowed corporate borrowers to use bond markets,
which caused an additional decrease in the lending opportunities of banks.
7
Regulators, however, limited banks from entering new businesses,in particular,
from securities underwriting.
8
Hence, the conditions Gorton and Rosen (1995)
analyze, as discussed above, existed for the Japanese banks during this period.
In fact, there was a large subsequent decrease in the number of banks in Japan
after 1989 through mergers, acquisitions, and failures. Twenty-one nationwide
banks in the sample analyzed in this article had decreased to seven banks and
bank holding companies as of October 2005.
The 1984–89 period also contains the so-called “Japanese bubble” years
during which stock and real estate prices increased substantially. Japanese
5. For a more detailed discussion of the financial environment in Japan during the 1980s, see
Hoshi and Kashyap (2001, esp. chaps. 7 and 8).
6. U.S. thrifts were hit by rising interest costs on their deposits in the late 1970s while their
main assets, mortgage loans, carried a fixed rate, which led to a rapid deterioration of their
balance sheet (see, e.g., Kane 1989; Barth 1991; and White 1991).
7. See, e.g., Hoshi, Kashyap, and Scharfstein 1993; Horiuchi 1995.
8. The slow and incoherent deregulation has often been seen as one of the main sources of
banking problems in Japan. See esp. Hoshi and Kashyap 1999.
3062 Journal of Business
banks also increased their real estate lending significantly during this period.
Stock prices started their slide in 1990, followed by real estate prices. From
1990 onward, a large number of Japanese banks could no longer be considered
financially healthy, and the first use of funds from the Deposit Insurance
Corporation took place in 1991.
9
This study focuses on the real estate lending of banks. It is clear ex post
that Japanese real estate was very risky. However, Japanese bankers seem to
have been well aware of the high risks involved in the real estate loans relative
to their other loans before the market crash even if they were unlikely to have
foreseen the magnitude of the eventual crash. For example, in 1987, more
than three years before the real estate prices started their eventual decline,
Kenichi Komiya, then the president of Mitsui Bank as well as the president
of the Federation of Bankers Associations of Japan, publicly stated that “the
[Federation of Bankers] association recognized the importance of limiting the
amount of money lent to finance land purchases. The association recommended
its members [to] tighten up their lending procedures and . . . carefully in-
vestigate the uses to which the land in question is to be put.”
10
Furthermore,
Milhaupt and Miller (1997) describes how the Japanese banks used mortgage-
lending companies called jusen in channeling commercial real estate loans
they deemed too risky. Naturally, this does not imply that the economic agents
had foreseen the timing and magnitude of the eventual market crash. However,
this study only requires the managers and shareholders to be aware of higher
risks involved in real estate lending relative to the traditional loans of these
banks.
The shareholding structure of Japanese banks is particularly important for
this study. The owner-manager model does not represent the Japanese banking
sector. In fact, managerial shareholding in the Japanese banks examined in
this study is virtually nonexistent. No manager is among the top five share-
holders, and the fifth largest shareholder usually has less than 3% of the shares
outstanding.
11
Hence, corporate governance problems and the effects of share-
holders must be taken into account.
The role of shareholding, however, is more complicated in Japan than in
the United States because shareholding is often part of a wider business
relationship between banks and their shareholders. The bank often acts as the
main bank of its shareholders, so there is a lending relationship between them.
There is also cross shareholding; the bank itself is often a major shareholder
of its own shareholders. Furthermore, the source of power a bank may have
over its shareholders is not limited to lending and cross shareholding. For
example, even though most of the life insurance companies that are among
the large shareholders of banks are organized as mutual companies rather than
9. See Hoshi and Kashyap (2001, 272–73) for the rescue of Toho Sogo in 1991 and other
early signs of banking troubles.
10. Kyodo News International 1987 (October 27).
11. Low shareholding by the management is also typical for nonfinancial corporations in Japan
(see Kaplan 1994).
Corporate Governance in Japanese Banks 3063
as joint-stock, a large bank often has leverage over a shareholding life in-
surance company through the life insurance coverage it purchases for its own
employees.
12
Finally, these ties are especially strong in a financial group called
a (financial) keiretsu.
13
It is clear that these ties have the potential to create managerialentrenchment
as the bank managers may be able to influence their shareholders through
them. As this managerial entrenchment is different from the one typically
studied in the U.S. context, the hypotheses tested must be modified for the
Japanese context. This issue is discussed in more detail in the next section.
II. Hypotheses
The main question studied in this article is “Did large shareholders of Japanese
banks restrain bank managers from real estate lending in the 1980s?” As the
previous discussion suggests, not all shareholders were likely to be as effective.
In particular, a shareholder that borrows from the bank and/or has the bank
among its own shareholders may be prevented from exerting any discipline
on bank managers. The analysis below uses keiretsu ties to identify such
shareholders. Hence, the second question of this analysis is “Were keiretsu
shareholders as effective in disciplining bank managers as other shareholders?”
There are several advantages for using keiretsu ties to identify shareholders
that have relationship banking ties with the banks of which they are a share-
holder. First, the keiretsu are often led by the banks themselves and are
characterized by financial ties among their members and the keiretsu banks.
Second, keiretsu membership makes it possible to capture the leverage the
keiretsu banks have over the keiretsu life insurance companies. These insur-
ance companies are typically among the largest shareholders of banks but are
organized as mutual companies so no cross shareholding is applicable. Finally,
keiretsu membership is relatively easily identifiable.
Unfortunately, the keiretsu membership is not a perfect proxy for two
reasons. First, cross shareholdings and relationship bank lending also occur
outside keiretsu membership so the bank may have some power over non-
keiretsu shareholders as well. This may be especially important for the na-
tionwide banks that do not lead a keiretsu but still have lending, cross share-
holding, and other ties with their shareholders. Second, the strength of these
ties is considered to have been weakened during the sample period, which
12. Komiya (1994) describes how banks purchase life insurance policies for their employees
from a shareholder life insurance company according to the magnitude of its shareholdings.
13. There are two types of industrial groups in Japan, and both are called keiretsu. Roughly,
the financial keiretsu is a group of companies from diverse industries with a large nationwide
bank among the nucleus companies (e.g., Mitsubishi, Mitsui, Sumitomo); the production keiretsu
is a vertically integrated group of companies in one or few related industries (e.g., Toyoto, Nissan,
Matsushita). Unless otherwise noted, all the references to keiretsu in this paper are to the financial
keiretsu. While relationship lending and bank shareholding are not exclusive to companies within
a financial keiretsu, the membership in a keiretsu is often a good proxy for the identity of a
company’s main bank and the strength of its bank ties.
3064 Journal of Business
decreased the influence bank managers may have over their keiretsu share-
holders. However, the imperfect nature of the criteria used biases the analysis
against finding any differences between keiretsu shareholders and other share-
holders. The methodology used in testing these hypotheses is discussed after
the data are presented.
III. Data
A. Overview
The sample covers fiscal years (April to March for all the banks) 1984 through
1989. It includes all of the 84 banks that were listed continuously in the first
section of the Tokyo Stock Exchange and that did not take part in an acquisition
or merger during this period.
14
As no bank failed in that time period, the
sample is free from survivorship bias. The data include both nationwide (city,
long-term credit, and trust) and regional (both tier one and two) banks. All
the balance sheet data are obtained from the Nikkei Electronic Economic Data
System (NEEDS) and include the trust operations of the trust banks. As in
previous studies of Japanese companies, the balance sheet data are
unconsolidated.
The 12 city banks in the sample are large commercial banks with a na-
tionwide branch network that is especially extensive in large cities. They
primarily lend to large companies and, through their large shareholdings, play
an important role in the corporate governance of large companies in Japan.
Nationwide banks in the sample also include three long-term credit banks and
six trust banks. Out of the 21 nationwide banks in the sample, five city banks
and five trust banks belong to one of the six large financial keiretsu,as
determined in Industrial Groupings in Japan by Dodwell Marketing Con-
sultants (1986).
15
There are 63 regional banks in the sample; each of these
banks operates in one or a few neighboring prefectures and focuses on lending
to small- and medium-size companies. None of them belongs to a keiretsu.
16
B. Balance Sheet Variables
Table 1 presents sample statistics on the main variables used in this study.
The average size of banks based on the book value of their assets during the
sample period is 8.3 trillion yen, which is about $69 billion at $1 p¥120.
However, the standard deviation is very large, which reflects the large size
difference between the nationwide banks and the regional banks. Indeed, the
14. The latter requirement is not very stringent however. The only excluded bank is Sumitomo
Bank, which acquired Heiwa Sogo Bank in 1986.
15. They are Mitsubishi Bank and Mitsubishi Trust (Mitsubishi group), Mitsui Bank and Mitsui
Trust (Mitsui group), Sumitomo Trust (Sumitomo group), Fuji Bank and Yasuda Trust (Fuyo
group), Dai-ichi Kangyo Bank (Dai-ichi Kangyo group), and Sanwa Bank and Toyo Trust (Sanwa
group).
16. For more details about Japanese banks in the 1980s, see Suzuki (1987) and Hoshi and
Kashyap (2001, chap.7 especially) .
Corporate Governance in Japanese Banks 3065
TABLE 1 Sample Statistics
All Banks
Nationwide
Banks Only
Keiretsu
Banks Only
Assets (in trillion yen)
Mean 8.3 26.6 35.5
SD 12.7 13.7 12.1
Real estate loans/assets (%)
Mean 5.12 4.34 3.67
SD 2.91 2.43 0.73
Real estate loans/total loans (%)
Mean 9.90 12.28 12.54
SD 5.23 5.76 3.67
Loans to keiretsu real estate com-
panies/total loans (%)
Mean 0.25
SD 0.28
Total keiretsu loans/total loans (%)
Mean 5.43
SD 1.12
Capital ratio (%)
Mean 2.83 1.79 1.59
SD 0.93 0.63 0.68
Income/assets (%)
Mean 0.48 0.40 0.42
SD 0.14 0.13 0.16
Market-to-book
Mean 1.06 1.09 1.09
SD 0.04 0.05 0.05
Top5(%)
Mean 18.1 16.5 16.9
SD 5.4 3.1 2.4
Same-keiretsu shareholders (%)
Mean 10.2
SD 4.1
Nonaffiliated shareholders (%)
Mean 6.6
SD 4.6
N84 21 10
Note.—The table presents the sample statistics for Japanese banks from fiscal 1984 to fiscal 1989 with
annual observations. All the values are book values unless stated otherwise. Loans to keiretsu real estate
companies pthe loans to the real estate companies that belong to the same keiretsu as the bank and are traded
in the Tokyo Stock Exchange (TSE); total keiretsu loans pthe loans to the companies that belong to the same
keiretsu as the bank and are traded in TSE. Both variables are set to zero for non-keiretsu banks. Capital ratio
pthe book value of equity and reserves (except loan loss reserves) divided by total book value of assets.
Income pthe ordinary income. Market-to-book pthe sum of market value of equity and the book value of
liabilities, divided by the book value of assets. Top 5 pthe total shares held by the largest five shareholders.
Same-keiretsu shareholders pthe total shares held by the shareholders that are among the largest five share-
holders and are members of the same keiretsu as the bank, and nonaffiliated shareholders pthe total shares
held by the shareholders that are among the largest five shareholders and are not members of the same keiretsu
as the bank; both variables are zero for non-keiretsu banks. SD pstandard deviation.
average size for nationwide banks is 26.6 trillion yen. The nationwide banks
that belong to a keiretsu are even larger; their average size is 35.5 trillion
yen.
The biggest asset for banks is their outstanding loans, and this study focuses
on the loans to the real estate sector. The loan data are unconsolidated and
do not include the loans made by overseas branches. Of all bank loans, 9.9%,
3066 Journal of Business
or 5.1% of all assets, are to the real estate sector. At 12.3%, nationwide banks
have a larger share of loans to the real estate sector. However, due to these
banks’ large financial assets, including their shareholdings in other companies,
the real estate loans constitute a lower proportion of their assets, at 4.3%. At
12.5%, the share of real estate loans in total loans is even higher for keiretsu
banks, yet its share in total assets is lower, at 3.7%.
Table 1 also presents sample statistics for the capital ratio of banks, cal-
culated as the book value of equity plus special reserves (excluding loan loss
reserves) and divided by the book value of assets. A striking feature is that
Japanese banks had a very low capital ratio in this period. The average capital
ratio of all the banks during the sample period is 2.8%. The average capital
ratio of nationwide banks is lower, at 1.8%. The keiretsu banks have even
lower capital ratio, at only 1.6%. Japanese banks did not have a high market-
to-book ratio in this period. The average was only 1.06 with nationwide banks
and 1.09 with keiretsu banks.
C. Shareholding
The data on the identity of and the shares held by each of the top five
shareholders of each bank in the sample are obtained from the Japan Company
Handbook by Toyo Keizai (various issues). The data are collected for each
year in the sample period to construct a panel. Table 1 presents samplestatistics
on the total shares held by these shareholders. On average, the top five share-
holders hold a combined 18.1%. This average is lower for nationwide banks
and keiretsu banks, with 16.5% and 16.9%, respectively.
The sample statistics provided in table 1 demonstrate the importance of
same-keiretsu shareholders for keiretsu banks. On average, the same-keiretsu
shareholders that are among the top five shareholders hold a combined 10.2%
of the keiretsu banks. This compares with the total average holding of 16.9%
of the top five shareholders in these banks. In other words, if the managers
of keiretsu banks can influence their same-keiretsu shareholders, they can
neutralize any monitoring discipline the other shareholders may provide.
IV. Regression Analysis
A. Methodology
The following analysis aims to uncover the effects of large shareholders on
the real estate lending of Japanese banks. To isolate these effects from the
other determinants of a bank loan portfolio, the analysis controls for bank
characteristics such as size, bank capital ratio, cash flow, market-to-book ratio,
and bank type. However, there are also several factors that are not so easy to
observe but can affect a bank’s real estate lending. First, banks may differ in
the credit-screening ability of their loan officers. In fact, it may be optimal
for a bank with highly able officers to focus on riskier sectors like real estate.
Although this difference across banks is not likely to change over a short
Corporate Governance in Japanese Banks 3067
time period given the long-term employment practices of Japanese banks, it
may still be reflected in the overall riskiness of a bank’s loan portfolio.
Second, the level of shareholding of each shareholder is partially determined
by historical reasons, especially for shareholdings that are part of wider busi-
ness relationships as in a keiretsu. These shareholders adjust their portfolio
only slowly, due to tax reasons, as they may face a large capital gains tax
when they sell shares bought long ago at much lower prices.
17
Third, and
related to the second factor above, relationship bank lending practices are
very important in Japan. Hence, the existing lending relationships at the be-
ginning of the sample period will affect the bank lending during the years
covered in the sample.
Although it is very difficult to control for these factors in a cross-sectional
analysis, they can be controlled for in a fixed-effects analysis that the panel
data allow. However, two issues complicate the panel analysis. First, banks
often make loans with maturity longer than one year—the sampling frequency
used in this analysis—so they may wait until the maturity of a loan to im-
plement general policy changes on a specific loan. This suggests that, when
the dependent variable is based on the loans extended by the bank, lagged
dependent variables must be included as explanatory variables.
18
Second, the
explanatory variables are likely to be only predetermined with respect to the
error term but not strictly exogenous. In particular, the shareholders may adjust
their shareholding based on the bank’s past lending policies. As ordinary least
squares would give inconsistent estimates with either of these problems, Ar-
ellano and Bond’s (1991) panel estimation approach, which is based on Gen-
eralized Method of Moments (GMM) and uses the past values of explanatory
variables as instruments, is adopted in the analysis below (Wooldridge 2002).
B. Main Results
Table 2 presents the results of the regression analysis where the dependent
variable is the total loans to the real estate sector normalized by the total
assets (book) of the bank. All the explanatory variables are as of year t⫺1
and instrumented with their past values. The regressions also include lagged
dependent variable, bank fixed effects, and year dummies.
All the regressions include as control variables log (assets), which is the
logarithm of total (book) assets; capital ratio, which is the shareholder equity
(plus special reserves) divided by total assets; income, which is cash flow
divided by total assets; and market to book, the market-to-book ratio. The
first regression does not include any shareholding variables and serves as a
benchmark. Log (assets) has a positive coefficient while capital ratio, income,
and market-to-book all have negative coefficients. However, with the exception
17. The potential adverse price reaction that a large shareholder may cause when it liquidates
its portfolio would also lead to a slow adjustment of shareholdings.
18. Indeed, the lagged dependent variable tends to be highly significant.
3068 Journal of Business
TABLE 2 Large Shareholders and Bank Real Estate Lending: The Role of
Keiretsu Ties
(1) (2) (3) (4) (5) (6)
RE ( )t⫺1 .789*** .761*** .702*** .756*** .701*** .752***
(.077) (.076) (.071) (.072) (.069) (.069)
Log (assets) .002 .002 .008 .001 .009 .003
(.007) (.007) (.008) (.006) (.008) (.006)
Capital ratio ⫺.078 ⫺.193 ⫺.100 ⫺.162 ⫺.122 ⫺.201
(.283) (.274) (.279) (.267) (.279) (.269)
Income ⫺.845 ⫺1.027* ⫺.890 ⫺1.070* ⫺.811 ⫺.997
(.579) (.606) (.590) (.591) (.612) (.614)
Market-to-book ⫺.030 ⫺.029 ⫺.028 ⫺.023 ⫺.022 ⫺.015
(.034) (.032) (.031) (.030) (.033) (.033)
Top 5 (a) ⫺.096* ⫺.098* ⫺.089* ⫺.093* ⫺.091*
(.049) (.051) (.050) (.050) (.051)
Top 5 #nationwide-
non-keiretsu bank .150 .137
(.250) (.233)
Same-keiretsu share-
holders (b) .340** .331**
(.152) (.144)
Nonaffiliated
shareholders ⫺.079 ⫺.084
(.102) (.103)
Bank fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
(a) ⫹(b) .241 .238
(.163) (.160)
p-value of F.000 .000 .000 .000 .000 .000
Z⫺.056 ⫺.076 .154 ⫺.074 .085 ⫺.147
Number of banks 84 84 84 84 84 84
Note.—The dependent variable is RE (t), the total loans to the real estate sector divided by the bank’s total
assets (book). Log (assets) pthe logarithm of the book value of total assets. Capital ratio pthe book value
of equity and reserves (except loan loss reser ves) divided by total book value of assets. Income pthe ordinary
income divided by total (book) assets. Market-to-book pthe sum of market value of equity and the book
value of liabilities, divided by the book value of assets. Top 5 pthe total shares held by the largest five
shareholders. Nationwide-non-keiretsu bank pa dummy variable that takes the value one for nationwide
banks that are not among the core banks in the six largest financial keiretsu. Same-keiretsu shareholders p
the total shares held by shareholders who are among the largest five shareholders and are members of the
same keiretsu as the bank, and nonaffiliated shareholders pthe total shares held by the shareholders that are
among the largest five shareholders and are not members of the same keiretsu as the bank; both variables are
zero for non-keiretsu banks. All explanatory variables are lagged by one year. The estimation procedure is
Arellano-Bond (1991) one-step GMM panel estimation with bank fixed effects and time dummies. Hetero-
scedasticity-robust standard errors are in parentheses. Fis an F-statistic for the hypothesis that all the explanatory
variables except time fixed effects are jointly zero. Zis a (standard) normally distributed test-statistic for the
lack of second-order correlation.
* Denotes the coefficients that are different from zero at significance level 10%.
** Denotes the coefficients that are different from zero at significance level 5%.
*** Denotes the coefficients that are different from zero at significance level 1%.
of income in some later specifications, none of these coefficients is statistically
significant.
The second regression includes “top 5,” the total shares held by the top
five shareholders. The coefficient of this shareholding variable is negative and
significant at the 10% level, which indicates that the large shareholders re-
strained bank managers from lending to the real estate sector on average.
Corporate Governance in Japanese Banks 3069
However, the effect of shareholders that were members of the bank’s keiretsu
was very different, as shown in the next regression. Regression 3 includes
both the top five and the same-keiretsu shareholders, which is the total shares
held by the shareholders that are among the top five shareholders and belong
to the same keiretsu as the bank.
19
The coefficient of same-keiretsu share-
holders is positive and statistically significant at the 5% level. The sum of
the coefficients of the top five and the same-keiretsu shareholders, measuring
the total effect of keiretsu shareholders, is positive but statistically insignifi-
cant. This indicates that the shareholders that belonged to the bank’s keiretsu
did not restrain bank managers from real estate lending.
The different role played by same-keiretsu shareholders has to do with their
keiretsu ties rather than with the type of bank in which they hold shares, as
the next three regressions show. In regression 4, same-keiretsu shareholders
are replaced with nonaffiliated shareholders, which indicate the shares held
in a keiretsu bank by the shareholders that are among the largest five share-
holders but do not belong to the bank’s keiretsu. The coefficient of nonaffi-
liated shareholders is negative though not statistically different from zero.
The different roles played by the same-keiretsu shareholders and the in-
dependent shareholders are also beyond the differences between the share-
holders of nationwide banks and those of regional banks. Regressions 5 and
6 include the interaction term “top 5 #nationwide-non-keiretsu bank,” where
“nationwide-non-keiretsu bank” is a dummy variable that takes the value of
one for nationwide banks that are not part of the six main keiretsu. The
coefficient of same-keiretsu shareholders continues to be positive and signif-
icant at the 5% level while the coefficient of nonaffiliated shareholders remains
negative and statistically insignificant.
These results indicate that large shareholders tried to restrain Japanese banks
from lending to the real estate sector, on average. However, the keiretsu ties
of big keiretsu banks provided protection to these banks as the same-keiretsu
shareholders did not discipline the banks of their keiretsu. The results confirm
that the keiretsu banks have leverage over their keiretsu shareholders. Notice
that the keiretsu shareholders are likely to be less tolerant to risks taken by
their banks as these banks are their main bank. Any financial distress in their
main bank can adversely affect their ability to raise funds for their needs, as
was the case in the 1990s.
20
Hence, these lending ties bias our analysis against
obtaining the findings above. Finally, notice that the regressions have bank
fixed effects and year dummies so they are unlikely to capture other effects
specific to a given bank or to a given year. The next section provides further
robustness checks.
19. Same-keiretsu shareholders and nonaffiliated shareholders are trivially set to zero for banks
that do not belong to one of the six keiretsu.
20. See Gibson 1995; Kang and Stulz 2000.
3070 Journal of Business
V. Robustness
A. Real Estate Lending to Keiretsu Companies
One possible concern is whether the keiretsu effect reflects the loans of keiretsu
banks to the real estate companies of their keiretsu. To explore this possibility,
data on the amount borrowed by each listed keiretsu real estate company from
each bank are obtained from NEEDS. Data show that the loans to the real
estate companies that are members of the bank’s keiretsu are only a small
fraction of the total real estate loans for these banks. Such loans are less than
7% of a bank’s total lending to the real estate sector in any given year in the
sample period with an average across all the keiretsu banks of approximately
3% or less.
21
In fact, only one keiretsu real estate company (Mitsui Fudosan)
is ever among the top five shareholders of a keiretsu bank. When total loans
made by keiretsu banks to the real estate companies are subtracted from their
total real estate lending and the regressions in table 2 are repeated, the qual-
itative results remain unchanged, as reported in table 3.
B. Keiretsu Lending and Real Estate Loans
Hoshi (2001) finds that decreases in loans to keiretsu companies were asso-
ciated with increases in real estate loans in the 1980s, which is consistent
with the risk taking by banks that lost their traditional customers. One natural
concern is then whether the effects due to the shareholding by keiretsu com-
panies merely capture a similar loss of traditional borrowers rather than the
effects related to corporate governance. To distinguish between the two effects,
data on the borrowing of each listed keiretsu company from each bank are
obtained from Nikkei NEEDS, and the total lending of each keiretsu bank to
its own keiretsu companies is calculated.
Table 4 presents the regression analysis when “keiretsu loans,” the annual
change in the fraction of keiretsu loans to the total loans, is included in the
regressions. Keiretsu loans are negative but not statistically significant. The
coefficient of same-keiretsu shareholders, however, continues to be positive
and significant. These results indicate that keiretsu shareholding is not merely
a proxy for the lost lending opportunities of keiretsu banks to their traditional
borrowers.
22
21. It is interesting to note that keiretsu real estate companies had a larger share in keiretsu
banks’ real estate lending in 1980–82, before the study sample period.
22. As an additional evidence that keiretsu shareholding is not just a proxy for the lost customers
of keiretsu banks, notice that Hoshi (2001) also finds that the banks that had a higher decrease
in the fraction of loans to the listed companies had a higher increase in the fraction of real estate
loans. If keiretsu shareholding were a proxy for the lost customers of keiretsu banks, one would
expect a similar effect from the nonaffiliated shareholders of keiretsu banks as well because most
of those shareholders are also listed companies.
Corporate Governance in Japanese Banks 3071
TABLE 3 Keiretsu Shareholders and Bank Real Estate Lending: Loans to Keiretsu
Real Estate Companies Excluded
(1) (2) (3) (4) (5) (6)
RE—no keiretsu ()t⫺1 .791*** .764*** .703*** .759*** .702*** .755***
(.077) (.076) (.071) (.072) (.069) (.069)
Log (assets) .001 .002 .008 .001 .008 .002
(.007) (.007) (.008) (.006) (.008) (.006)
Capital ratio ⫺.078 ⫺.193 ⫺.101 ⫺.164 ⫺.122 ⫺.202
(.284) (.274) (.279) (.267) (.279) (.268)
Income ⫺.859 ⫺1.042* ⫺.900 ⫺1.088* ⫺.817 ⫺1.020*
(.579) (.607) (.591) (.593) (.612) (.616)
Market-to-book ⫺.030 ⫺.029 ⫺.028 ⫺.023 ⫺.022 ⫺.015
(.034) (.032) (.031) (.030) (.033) (.033)
Top 5 (a) ⫺.095* ⫺.098* ⫺.089* ⫺.093* ⫺.091*
(.049) (.051) (.050) (.050) (.051)
Top 5 #nationwide-
non-keiretsu bank .149 .137
(.250) (.232)
Same-keiretsu share-
holders (b) .343** .336**
(.153) (.145)
Nonaffiliated
shareholders ⫺.084 ⫺.089
(.102) (.103)
Bank fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
(a) ⫹(b) .245 .243
(.163) (.161)
p-value of F.000 .000 .000 .000 .000 .000
Z⫺.068 ⫺.090 .142 ⫺.088 .074 ⫺.163
Number of banks 84 84 84 84 84 84
Note.—The dependent variable is RE—no keiretsu (t), the total loans to the real estate sector except to the
real estate companies that belong to the same keiretsu as the bank, divided by the bank’s total assets (book).
Capital ratio pthe book value of equity and reserves (except loan loss reserves) divided by total book value
of assets. Income pthe ordinary income divided by total (book) assets. Market-to-book pthe sum of market
value of equity and the book value of liabilities, divided by the book value of assets. Top 5 pthe total shares
held by the largest five shareholders. Nationwide-non-keiretsu bank pa dummy variable that takes the value
one for nationwide banks that are not among the core banks in the six largest financial keiretsu. Same-keiretsu
shareholders pthe total shares held by the shareholders that are among the largest five shareholders and are
members of the same keiretsu as the bank, and nonaffiliated shareholders pthe total shares held by the
shareholders that are among the largest five shareholders and are not members of the same keiretsu as the
bank; both variables are zero for non-keiretsu banks. All explanatory variables are as of . The estimationt⫺1
procedure is Arellano-Bond (1991) one-step GMM panel estimation with bank fixed effects and time dummies.
Heteroscedasticity-robust standard errors are in parentheses. Fis an F-statistic for the hypothesis that all the
explanatory variables except time fixed effects are jointly zero. Zis a (standard) normally distributed test-
statistic for the lack of second-order correlation.
* Denotes the coefficients that are different from zero at significance level 10%.
** Denotes the coefficients that are different from zero at significance level 5%.
*** Denotes the coefficients that are different from zero at significance level 1%.
C. Top 1 and Top 3 Shareholders
The main analysis is repeated by focusing only on the largest shareholder or
the three largest shareholders instead of the largest five. The results are pre-
sented in table 5, panels A and B. The coefficient of same-keiretsu shareholders
remains positive and statistically significant. The magnitude of the coefficient
is also similar to the magnitude when the analysis is focused on the largest
3072 Journal of Business
TABLE 4 Keiretsu Shareholders, Loans to Keiretsu Companies, and Bank Real
Estate Lending
(1) (2) (3) (4) (5) (6)
RE (t⫺1) .767*** .750*** .709*** .742*** .704*** .741***
(.070) (.071) (.066) (.068) (.061) (.065)
Log (assets) .003 .003 .007 .003 .008 .004
(.006) (.006) (.008) (.006) (.008) (.007)
Capital ratio ⫺.022 ⫺.162 ⫺.106 ⫺.150 ⫺.145 ⫺.187
(.282) (.264) (.268) (.256) (.267) (.254)
Income ⫺.936 ⫺1.135* ⫺.979* ⫺1.094* ⫺.868 ⫺1.046*
(.575) (.595) (.568) (.588) (.581) (.607)
Market-to-book ⫺.021 ⫺.020 ⫺.025 ⫺.017 ⫺.019 ⫺.013
(.030) (.028) (.029) (.028) (.030) (.030)
Keiretsu loans ⫺.116 ⫺.147 ⫺.156 ⫺.159 ⫺.167 ⫺.174
(.206) (.203) (.197) (.194) (.211) (.205)
Top 5 (a) ⫺.094* ⫺.096* ⫺.088* ⫺.097* ⫺.090*
(.049) (.051) (.050) (.050) (.050)
Top 5 #nationwide-
non-keiretsu bank .196 .178
(.216) (.200)
Same-keiretsu share-
holders (b) .242** .258**
(.116) (.119)
Nonaffiliated
shareholders ⫺.057 ⫺.065
(.082) (.084)
Bank fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
(a) ⫹(b) .146 .161
(.127) (.133)
p-value of F.000 .000 .000 .000 .000 .000
Z⫺.109 ⫺.152 .041 ⫺.142 ⫺.033 ⫺.226
Number of banks 84 84 84 84 84 84
Note.—The dependent variable is RE (t), the total loans to the real estate sector divided by the bank’s total
assets (book). Log (assets) pthe logarithm of the book value of total assets. Capital ratio pthe book value
of equity and reserves (except loan loss reser ves) divided by total book value of assets. Income pthe ordinary
income divided by total (book) assets. Market-to-book pthe sum of market value of equity and the book
value of liabilities, divided by the book value of assets. Keiretsu loans pthe change in the fraction of loans
to the companies that belong to the same keiretsu as the bank; it is set to zero for non-keiretsu banks. Top 5
pthe total shares held by the largest five shareholders. Nationwide-non-keiretsu bank pa dummy variable
that takes the value one for nationwide banks that are not among the core banks in the six largest financial
keiretsu. Same-keiretsu shareholders pthe total shares held by shareholders that are among the largest five
shareholders and are members of the same keiretsu as the bank, and nonaffiliated shareholders pthe total
shares held by shareholders that are among the largest five shareholders and are not members of the same
keiretsu as the bank; both variables are zero for non-keiretsu banks. All explanatory variables are as of t⫺
. The estimation procedure is Arellano-Bond (1991) one-step GMM panel estimation with bank fixed effects1
and time dummies. Heteroscedasticity-robust standard errors are in parentheses. Fis an F-statistic for the
hypothesis that all the explanatory variables except time fixed effects are jointly zero. Zis a (standard) normally
distributed test-statistic for the lack of second-order correlation.
* Denotes the coefficients that are different from zero at significance level 10%.
** Denotes the coefficients that are different from zero at significance level 5%.
*** Denotes the coefficients that are different from zero at significance level 1%.
Corporate Governance in Japanese Banks 3073
TABLE 5 Keiretsu Shareholders and Bank Real Estate Lending: The Effect of
Top 1 and Top 3 Shareholders
(1) (2) (3) (4) (5)
A. The Effect of the Top Shareholder
RE ( )t⫺1 .784*** .774*** .766*** .766*** .756***
(.075) (.071) (.071) (.072) (.067)
Log (assets) .002 .006 .003 .006 .002
(.007) (.006) (.007) (.006) (.006)
Capital ratio ⫺.104 ⫺.144 ⫺.162 ⫺.161 ⫺.168
(.272) (.254) (.258) (.265) (.269)
Income ⫺.902 ⫺.784 ⫺.696 ⫺.785 ⫺.611
(.625) (.600) (.613) (.612) (.616)
Market-to-book ⫺.033 ⫺.019 ⫺.032 ⫺.011 ⫺.013
(.032) (.031) (.031) (.034) (.033)
Top 1 (a) ⫺.152 ⫺.140 ⫺.174* ⫺.147 ⫺.152
(.105) (.092) (.101) (.096) (.099)
Top 1 #nationwide-
non-keiretsu bank .025 .095
(.174) (.227)
Same-keiretsu share-
holder (b) .231*** .222***
(.069) (.080)
Nonaffiliated
shareholder ⫺.250*** ⫺.199***
(.080) (.049)
Bank fixed effects Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
(a) ⫹(b) .091 .075
(.108) (.114)
p-value of F.000 .000 .000 .000 .000
Z⫺.092 ⫺.188 ⫺.066 ⫺.215 ⫺.123
Number of banks 84 84 84 84 84
B. The Effect of the Top Three Shareholders
RE ( )t⫺1 .762*** .779*** .744*** .769*** .751***
(.074) (.080) (.074) (.073) (.073)
Log (assets) .002 .003 .003 .003 .002
(.007) (.006) (.008) (.006) (.007)
Capital ratio ⫺.150 ⫺.189 ⫺.144 ⫺.214 ⫺.174
(.275) (.253) (.270) (.271) (.269)
Income ⫺.999* ⫺1.053* ⫺1.009* ⫺.935 ⫺.801
(.596) (.582) (.595) (.626) (.629)
Market-to-book ⫺.027 ⫺.009 ⫺.023 ⫺.009 ⫺.013
(.031) (.031) (.030) (.032) (.033)
Top 3 (a) ⫺.104** ⫺.107** ⫺.108** ⫺.106** ⫺.108**
(.051) (.049) (.051) (.052) (.053)
Top 3 #nationwide-
non-keiretsu bank .245** .287*
(.103) (.168)
Same-keiretsu share-
holder (b) .318* .197*
(.164) (.115)
Nonaffiliated
shareholder .077 .053
(.168) (.158)
Bank fixed effects Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
3074 Journal of Business
TABLE 5 (Continued)
(1) (2) (3) (4) (5)
B. The Effect of the Top Three Shareholders
(a) ⫹(b) .211 .091
(.169) (.114)
p-value of F.000 .000 .000 .000 .000
Z⫺0.114 ⫺0.077 ⫺0.128 ⫺0.127 ⫺0.187
Number of banks 84 84 84 84 84
Note.—The dependent variable is RE (t), the total loans to the real estate sector divided by the bank’s total
assets (book). Log (assets) pthe logarithm of the book value of total assets. Capital ratio pthe book value
of equity and reserves (except loan loss reser ves) divided by total book value of assets. Income pthe ordinary
income divided by total (book) assets. Market-to-book pthe sum of market value of equity and the book
value of liabilities, divided by the book value of assets. Top 1 and Top 3 pthe total shares held by the largest
one and the largest three shareholders, respectively. Nationwide-non-keiretsu bank pa dummy variable that
takes the value one for nationwide banks that are not among the core banks in the six largest financial keiretsu.
Same-keiretsu shareholders pthe total shares held by shareholders that are among the largest one or three
shareholders depending on the regression and are members of the same keiretsu as the bank, and nonaffiliated
shareholders pthe total shares held by the shareholders that are among the largest one or three shareholders
depending on the regression and are not members of the same keiretsu as the bank; both variables are zero
for non-keiretsu banks. All explanatory variables are as of . The estimation procedure is Arellano-Bondt⫺1
(1991) one-step GMM panel estimation with bank fixed effects and time dummies. Heteroscedasticity-robust
standard errors are in parentheses. Fis an F-statistic for the hypothesis that all the explanator y variables except
time fixed effects are jointly zero. Zis a (standard) normally distributed test-statistic for the lack of second-
order correlation.
* Denotes the coefficients that are different from zero at significance level 10%.
** Denotes the coefficients that are different from zero at significance level 5%.
*** Denotes the coefficients that are different from zero at significance level 1%.
five shareholders. This may be due to the fact that most of the same-keiretsu
shareholders are either the largest or second-largest shareholders of their kei-
retsu bank, as the median rank of the same-keiretsu shareholders among the
largest five shareholders is two.
D. Nonlinear Effects of Shareholding
Gorton and Rosen (1995) find evidence of nonlinear effects in their study of
bank shareholding and real estate lending. Main regressions are repeated by
including the squared terms of “top 5,” the total shares held by the top five
shareholders, and of same-keiretsu shareholders, the total shares held by the
shareholders that are among the top five shareholders and belong to the same
keiretsu as the bank. The results are presented in table 6. No evidence of
nonlinear effects is found. In particular, the coefficients of both the linear and
squared terms of same-keiretsu shareholders are positive, with the latter being
statistically significant.
The lack of nonlinear effects comparable to those found in studies focusing
on the United States may be due to the fact that the holdings by most of the
shareholders of Japanese banks are much smaller than the levels at which
nonlinear effects are detected for U.S. banks. With Japanese banks in thissample,
95% of each shareholding-year is 6.1% or smaller. For same-keiretsu share-
holders, the largest shareholding in the sample is only 8.7%. Despite the lack
of comparable nonlinear effects of shareholding, the effect of same-keiretsu
shareholding remains similar to earlier results and statistically significant.
Corporate Governance in Japanese Banks 3075
TABLE 6 Keiretsu Shareholders and Bank Real Estate Lending: Nonlinear Effects
(1) (2)
RE ( )t⫺1 .746*** .690***
(.072) (.069)
Log (assets) .000 .007
(.007) (.008)
Capital ratio ⫺.205 ⫺.117
(.278) (.281)
Income ⫺.961 ⫺.826
(.603) (.569)
Market-to-book ⫺.029 ⫺.023
(.032) (.030)
Top 5 ⫺.232 ⫺.226
(.182) (.181)
(Top 5)
2
.282 .268
(.328) (.327)
Same-keiretsu shareholders .034
(.149)
(Same-keiretsu shareholders)
2
1.166*
(.633)
Bank fixed effects Yes Yes
Year fixed effects Yes Yes
p-value of F.000 .000
Z⫺.087 .054
Number of banks 84 84
Note.—The dependent variable is RE (t), the total loans to the real estate sector divided by the bank’s total
assets (book). Log (assets) pthe logarithm of the book value of total assets. Capital ratio pthe book value
of equity and reserves (except loan loss reser ves) divided by total book value of assets. Income pthe ordinary
income divided by total (book) assets. Market-to-book pthe sum of market value of equity and the book
value of liabilities, divided by the book value of assets. Top 5 pthe total shares held by the largest five
shareholders. Same-keiretsu shareholders is the total shares held by the shareholders that are among the largest
five shareholders and members of the same keiretsu as the bank; it is set to zero for non-keiretsu banks. All
explanatory variables are as of . The estimation procedure is Arellano-Bond (1991) one-step GMM panelt⫺1
estimation with bank fixed effects and time dummies. Heteroscedasticity-robust standard errors are in paren-
theses. Fis an F-statistic for the hypothesis that all the explanatory variables except time fixed effects are
jointly zero. Zis a (standard) normally distributed test-statistic for the lack of second-order correlation.
* Denotes the coefficients that are different from zero at significance level 10%.
** Denotes the coefficients that are different from zero at significance level 5%.
*** Denotes the coefficients that are different from zero at significance level 1%.
E. Other Measures for Real Estate Lending
The analysis presented in the previous section is repeated with five different
measures of real estate lending: (a) shares of real estate loans in total lending;
(b) real estate loans normalized by the market value of the bank, which is
defined as the sum of the market capitalization of the bank and the book value
of bank liabilities; (c) real estate loans normalized by the book value of bank
assets in 1982; (d) real estate loans normalized by the total loans in 1982;
and (e) the real estate loans normalized by the market value of the bank in
1982. Normalization by the market value of the bank takes into account the
increase in the market value of bank equity and the amount of funds the bank
can raise by an equity issue. The use of 1982 values, on the other hand,
corrects for any effect peculiar to the deregulation or large increases in the
equity prices during the sample period. Table 7 presents the regressions with
3076 Journal of Business
TABLE 7 Keiretsu Shareholders and Bank Real Estate Lending: Alternative Measures of Real Estate Lending
REL
(Real Estate Loans /
Total Loans)
REMKT
(Real Estate Loans /
Market Value of Bank)
REA82
(Real Estate Loans [t]/
Total Assets [1982])
REL82
(Real Estate Loans [t]/
Total Loans [1982])
REMKT82
(Real Estate Loans [t]/
Market Value of Bank [1982])
Dependent variable (t⫺1) .662*** .708*** .938*** .774*** .937***
(.060) (.078) (.180) (.176) (.185)
Log (assets) .018 .007 .018 .066 .021
(.018) (.007) (.024) (.053) (.028)
Capital ratio .127 ⫺.079 ⫺.431 .085 ⫺.457
(.579) (.246) (.598) (1.045) (.603)
Income .995 ⫺.664 ⫺1.638 ⫺1.093 ⫺1.641
(.889) (.520) (1.161) (1.995) (1.153)
Market-to-book ⫺.050 ⫺.011 ⫺.049 ⫺.026 ⫺.053
(.063) (.028) (.071) (.149) (.071)
Top 5 (a) ⫺.062 ⫺.082* ⫺.238* ⫺.359* ⫺.249*
(.075) (.046) (.128) (.210) (.134)
Same-keiretsu shareholders (b) .813** .259** .774** 1.800*** .809**
(.350) (.121) (.369) (.649) (.406)
Corporate Governance in Japanese Banks 3077
Bank fixed effects Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
(a) ⫹(b) .751* .177 .535 1.441** .559
(.385) (.129) (.360) (.688) (.389)
p-value of F.000 .000 .000 .000 .000
Z⫺1.019 .599 .836 .840 .816
Number of banks 84 84 84 84 83
Note.—REL pthe loans to the real estate sector divided by the bank’s total loans. REMKT pthe loans to the real estate sector divided by the sum of market value of bank’s equity
and the book value of bank’s liabilities. REA82 pthe loans to the real estate sector in the current year divided by the bank’s total assets in 1982. REL82 pthe loans to the real estate
sector in the current year divided by the bank’s total loans in 1982. REMKT82 pthe loans to the real estate sector in the current year divided by the sum of market value of bank’s equity
and the book value of bank’s liabilities in 1982. Log (assets) pthe logarithm of the book value of total assets. Capital ratio pthe book value of equity and reserves (except loan loss
reserves) divided by total book value of assets. Income pthe ordinary income divided by total (book) assets. Market-to-book pthe sum of market value of equity and the book value of
liabilities, divided by the book value of assets. Top 5 pthe total shares held by the largest five shareholders. Same-keiretsu shareholders pthe total shares held by shareholders that are
among the largest five shareholders and are members of the same keiretsu as the bank; it is zero for non-keiretsu banks. All explanatory variables are as of . The estimation proceduret⫺1
is Arellano-Bond (1991) one-step GMM panel estimation with bank fixed effects and time dummies. Heteroscedasticity-robust standard errors are in parentheses. Fis an F-statistic for the
hypothesis that all the explanatory variables except time fixed effects are jointly zero. Zis a (standard) normally distributed test-statistic for the lack of second-order correlation.
* Denotes the coefficients that are different from zero at significance level 10%.
** Denotes the coefficients that are different from zero at significance level 5%.
*** Denotes the coefficients that are different from zero at significance level 1%.
3078 Journal of Business
these alternative real estate lending measures. The effect of keiretsu share-
holders relative to other shareholders remains positive and statistically
significant.
VI. Conclusion
This article studies the effects of bank shareholding structure on the real estate
lending of Japanese banks in the 1980s. It shows that large shareholders
restrained the managers on average and that the real estate lending of banks
decreased as the total shares held by the top five shareholders increased.
However, this effect was absent when the shareholders belonged to a keiretsu
led by the bank itself. The additional business ties between the banks and
these shareholders, such as cross shareholding and bank borrowing, weakened
the ability of the shareholders to discipline bank managers.
The effect of poor corporate governance in borrowing companies on fi-
nancial crises has been shown (Johnson et al. 2000; Mitton 2002). The results
in this article indicate that corporate governance problems in banks are also
likely to have an important role in financial crises. With 45% of all real estate
lending and more than 40% of total lending by the sample banks, the sheer
size of keiretsu banks suggests that their corporate governance problems may
have increased the magnitude of the financial crisis in Japan.
The results also suggest avenues for future research. This article only shows
the link between corporate governance problems in banks and their real estate
lending. While it is clear from bankers’ statements made in the 1980s that
banks regarded these loans riskier than their average loans, it would still be
helpful to extend the analysis using nonperforming loan measures. Unfortu-
nately, data on nonperforming loans of Japanese banks are very unreliable
due to perverse incentives of banks, among other reasons, as shown by Peek
and Rosengren (2003). Hence, a more fruitful direction might be to extend
the analysis to loan level by using the data from borrowing companies.
The findings in this study have further implications for countries where
banks play a central role in the finance and governance of large companies.
While these countries may be capturing the benefits of relationship banking,
they may be particularly vulnerable to problems in the corporate governance
of banks themselves, especially after deregulation and integration in the credit
markets.
23
Furthermore, the impact of these problems are also likely to be
transmitted to other countries, as found by Peek and Rosengren (1997, 2000).
To the extent that relationship banking prevents effective disciplining of bank
managers, as this article demonstrates for Japan, financial systems with a large
23. See Aoki and Patrick 1994; Allen and Gale 2000. Weinstein and Yafeh (1998) and McGuire
(2002) argue that most of the benefits from relationship banking in Japan were disproportionately
captured by the banks in the 1980s, while Dinc¸ (2000) provides a theory of relationship banking
and capital market competition where the division of benefits is a function of capital market
competition. See Aoki (2001) for a theory of institutional change through increased integration
and competition.
Corporate Governance in Japanese Banks 3079
role of corporate governance for banks may have a built-in fragility to de-
regulation, an issue little discussed in the literature.
The results presented in this article also highlight questions for future re-
search about corporate governance in banks. One question is how the share-
holders acted in the 1990s when the banks were no longer financially healthy.
Did the shareholder pressure facilitate restructuring or further risk taking?
Anderson and Campbell (2004) find that the top managers of poorly per-
forming Japanese banks were more likely to be replaced in the 1990s. Another
question is who will provide corporate governance in banks after the crisis.
Some of these banks are very large but currently very few financial players
in the Japanese economy are a source of discipline for the postcrisis banks.
The role of regulatory incentives and politics in dealing with failing banks
has been well demonstrated (Kroszner and Strahan 1996; Brown and Dinc¸ 2005).
Another interesting question is whether cross shareholdings also ledto regulatory
forbearance because a bank failure can adversely affect the biggest nonfinancial
companies through the loss of value of their cross-shareholding investments.
This may help explain why so few large banks were allowed to fail in Japan.
References
Allen, Franklin, and Douglas Gale. 2000. Comparative financial systems: Competition versus
insurance. Cambridge, MA: MIT Press.
Anderson, Christopher W., and Terry L. Campbell II. 2004. The corporate governance of Japanese
banks. Journal of Corporate Finance 10, no. 3:327–54.
Anderson, Ronald C., and Donald R. Fraser. 2000. Corporate control, bank risk taking, and the
health of the banking industry. Journal of Banking and Finance 24, no. 8:1383–98.
Aoki, Masahiko. 2001. Toward comparative institutional analysis. Cambridge, MA: MIT Press.
Aoki, Masahiko, and Hugh Patrick. 1994. Japanese main bank system. Oxford: Oxford University
Press.
Arellano, Manuel, and Stephen Bond. 1991. Some tests of specification for panel data: Monte
Carlo evidence and an application to employment equations. Review of Economic Studies 58:
277–97.
Barth, James R. 1991. The great savings and loan debacle. Washington, DC: AEI Press.
Brown, Craig O., and I. Serdar Dinc¸. 2005. The politics of bank failures: Evidence from emerging
markets. Quarterly Journal of Economics 120:1413–44.
Demsetz, Rebecca S., Marc R. Saidenberg, and Philip E. Strahan. 1997. Agency problems and
risk taking at banks. Research paper no. 9709, Federal Reserve Bank of New York.
Dinc¸ , I. Serdar. 2000. Bank reputation, bank commitment, and the effects of competition in the
credit markets. Review of Financial Studies 13:781–812.
———. 2005. Politicians and banks: Political influences on government-owned banks in emerging
markets. Journal of Financial Economics 77:453–79.
Dodwell Marketing Consultants. 1986. Industrial Groupings in Japan (1986–87). Tokyo: Dodwell.
Gibson, Michael S. 1995. Can bank health affect investment? Evidence from Japan. Journal of
Business 68:281–308.
Gorton, Gary, and Richard Rosen. 1995. Corporate control, portfolio choice, and the decline in
banking. Journal of Finance 50:1377–1420.
Gorton, Gary, and Andrew Winton. 2003. Financial intermediation. In Handbook of the economics
of finance, vol. 1A, Corporate finance, ed. George Constantinides, Milton Harris, and Rene´
Stulz. Boston: Elsevier Science.
Hanazaki, Masaharu, Jung Wook Shim, Toshiyuki Souma, and Yupana Wiwattanakantang. 2004.
Do large shareholders monitor or collude with banks in Japan. Working paper, Asian Devel-
opment Bank Institute.
Horiuchi, Akiyoshi. 1995. An evaluation of Japanese financial liberalization: A case study of
3080 Journal of Business
corporate bond markets. In Financial deregulation and integration in East Asia, ed. Takatoshi
Ito and Anne Krueger. Chicago: University of Chicago Press.
Horiuchi, Akiyoshi, and Katsutoshi Shimizu. 2001. Did Amakudari undermine the effectiveness
of regulatory monitoring in Japan? Journal of Banking and Finance 25, no. 3:573–96.
Hoshi, Takeo. 2001. What happened to Japanese banks? Monetary and Economic Studies 19,
no. 1:1–29.
Hoshi, Takeo, and Anil Kashyap. 1999. The Japanese banking crisis: Where did it come from
and how will it end? In NBER macroeconomics annual 1999. Cambridge, MA: National Bureau
of Economic Research.
———. 2001. Corporate financing and governance in Japan: The road to the future. Cambridge,
MA: MIT Press.
Hoshi, Takeo, Anil Kashyap, and David Scharfstein. 1993. The choice between public and private
debt, an analysis of post-deregulation corporate financing in Japan. Working Paper no. 4421,
National Bureau of Economic Research, Cambridge, MA.
Johnson, Simon, Peter D. Boone, Alasdair Breach, and Eric Friedman. 2000. Corporate gover-
nance in the Asian financial crisis, 1997–1998. Journal of Financial Economics 58:141–86.
Kane, Edward J. 1989. The S&L insurance mess: How did it happen? Washington, DC: Urban
Institute.
Kang, Jun-Koo, and Anil Shivdasani. 1995. Firm performance, corporate governance, and top
executive turnover in Japan. Journal of Financial Economics 38:29–58.
———. 1997. Corporate restructuring during performance declines in Japan. Journal of Financial
Economics 46:29–65.
Kang, Jun-Koo, and Rene M. Stulz. 2000. Do banking shocks affect borrowing firm performance?
An analysis of the Japanese experience. Journal of Business 73:1–23.
Kaplan, Steven. 1994. Top executive rewards and firm performance: A comparison of Japan and
the United States. Journal of Political Economy 102:510–46.
Kaplan, Steven, and Bernadette Minton. 1994. Appointments of outsiders to Japanese boards:
Determinants and implications for managers. Journal of Financial Economics 36:225–58.
Keeley, Michael C. 1990. Deposit insurance, risk, and market power in banking. American
Economic Review 80:1183–1200.
Khwaja, Asim Ijaz, and Atif Mian. 2005. Do lenders favor politically connected firms? Rent
provision in an emerging financial market. Quarterly Journal of Economics 120:1371–1411.
Knopf, John D., and John L. Teall. 1996. Risk-taking behavior in the U.S. thrift industry:
Ownership structure and regulatory changes. Journal of Banking and Finance 20:1329–50.
Komiya, Ryutaro. 1994. The life insurance company as a business enterprise. In Business en-
terprise in Japan: Views of leading Japanese economists, ed. Kenichi Imai and Ryutaro
Komiya. Cambridge, MA: MIT Press.
Kroszner, Randall S., and Philip E. Strahan. 1996. Regulatory incentives and the thrift crisis:
Dividends, mutual-to-stock conversions, and financial distress. Journal of Finance 51:
1285–1319.
Kyodo News International. 1987. Global stock plunge a natural correction, banker association
head declares. Kyodo News International–Japan News Wire, October 27.
La Porta, Rafael, Florencio Lopez-de-Silanes, and Guillermo Zamarripa. 2003. Related lending.
Quarterly Journal of Economics 118, no. 1:231–36.
McGuire, Patrick M. 2002. The changing nature of Japanese firm bank relationships. Ph.D. diss.,
University of Michigan.
Merton, Robert C. 1977. Analytic derivation of the cost of deposit insurance and loan guarantees:
An application of modern option pricing theory. Journal of Banking and Finance 1:3–11.
Milhaupt, Curtis J., and Geoffrey P. Miller. 1997. Cooperation, conflict, and convergence in
Japanese finance: Evidence from the “Jusen” problem. Law and Policy in International Business
29, no. 1:1–78.
Mitton, Todd. 2002. A cross-firm analysis of the impact of corporate governance on the East
Asian financial crisis. Journal of Financial Economics 64:215–41.
Morck, Randall, and Masao Nakamura. 1999. Banks and corporate control in Japan. Journal of
Finance 54, no. 1:319–39.
Morck, Randall, Masao Nakamura, and Anil Shivdasani. 2000. Banks, ownership structure, and
firm value in Japan. Journal of Business 73, no. 4:539–67.
Peek, Joe, and Eric S. Rosengren. 1997. The international transmission of financial shocks: The
case of Japan. American Economic Review 87:495–505.
Corporate Governance in Japanese Banks 3081
———. 2000. Collateral damage: Effects of the Japanese bank crisis on real activity in the
United States. American Economic Review 90:30–45.
———. 2003. Unnatural selection: Perverse incentives and the misallocation of credit in Japan.
Working Paper no. 9643, National Bureau of Economic Research, Cambridge, MA.
Sapienza, Paola. 2004. What do state-owned firms maximize? Evidence from the Italian banks.
Journal of Financial Economics 72:357–84.
Saunders, Anthony, Elizabeth Strock, and Nickolaos G. Travlos. 1990. Ownership structure,
deregulation, and bank risk taking. Journal of Finance 45:643–54.
Suzuki, Yoshio, ed. 1987. The Japanese financial system. Oxford: Clarendon.
Toyo Keizai. Various issues. Japan company handbook. Tokyo: Toyo Keizai.
Weinstein, David E., and Yishay Yafeh. 1998. On the costs of a bank-centered financial system:
Evidence from the changing main bank relations in Japan. Journal of Finance 53:635–72.
White, Lawrence J. 1991. The S&L debacle: Public policy lessons for bank and thrift regulations.
Oxford: Oxford University Press.
Wooldridge, Jeffrey M. 2002. Econometric analysis of cross section and panel data. Cambridge,
MA: MIT Press.
