Download the impact of changes in bank ownership structure

THE IMPACT OF CHANGES IN BANK OWNERSHIP STRUCTURE
AROUND THE WORLD
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the
Degree Doctor of Philosophy in the Graduate School of The
Ohio State University
By
Alvaro G. Taboada, CFA, MBA
The Ohio State University
2008
Dissertation Committee:
Professor G. Andrew Karolyi, Advisor
Approved by
_______________________
Advisor
Professor Isil Erel
Professor Kewei Hou
Professor Ingrid Werner
Graduate Program in
Business Administration
ABSTRACT
Large scale bank privatizations over the last ten years have resulted in vast
changes in the ownership structure of banking sectors throughout the world.
This
dissertation explores both the macro and micro level effects of these changes in bank
ownership structure. The first essay explores how changes in bank ownership structure
affect capital allocation efficiency within countries. I find that the decline in government
ownership of banks by itself does not have any impact on capital allocation efficiency;
rather, what matters is whether foreigners or large domestic shareholders acquire the
stakes relinquished by the government. Increases in domestic blockholder ownership of
banks adversely affect the allocation of capital through increased lending activity to less
productive industries, while increased foreign presence improves capital allocation
efficiency by directing credit to more productive sectors and to industries that rely more
on external financing.
In the second essay I explore how changes in bank ownership structure affect the
performance of individual banks and the banking sector. The primary contribution of this
essay is to examine the role of large domestic blockholders on bank performance. I find
that increases in large domestic blockholder ownership of banks are associated with poor
subsequent performance in terms of asset quality, profitability, and bank value. In
ii
contrast, increases in foreign ownership lead to improvements in profitability and bank
value, consistent with prior findings. Government ownership of banks continues to affect
bank performance adversely. Finally, increased presence of large domestic blockholders
in the banking sector has a positive spillover effect on banking sector asset quality and
profitability, while increased foreign presence is no longer associated with improvements
in the competitiveness of the banking sector, contrary to what prior studies have found.
iii
To my parents, Dr. Alvaro J. Taboada and Esther M. Taboada
iv
ACKNOWLEDGMENTS
I would like to thank my committee members G. Andrew Karolyi (Committee
Chair), Isil Erel, Kewei Hou, and Ingrid Werner for their invaluable comments and
continued encouragement and support throughout this process. I am also deeply indebted
to René Stulz, Roger Loh, my colleagues, and all seminar participants at The Ohio State
University for their insightful and helpful comments and suggestions.
Finally, special thanks to José F. Alvarez at Banco Sabadell, and to Mitch Gouss
at Bureau van Dijk, who provided access to key data used in this dissertation.
v
VITA
October 15, 1973…………………………………….Born in León, Nicaragua
1995…Bachelor of Business Administration, Florida International University, Miami, FL
1997……Master of Business Administration, University of Notre Dame, Notre Dame, IN
1997-2003…..International Bank Examiner, Federal Reserve Bank of Atlanta, Miami, FL
FIELDS OF STUDY
Major Field: Business Administration
Concentration: Finance
vi
TABLE OF CONTENTS
ABSTRACT................................................................................................................... ii
ACKNOWLEDGMENTS ...............................................................................................v
VITA..............................................................................................................................vi
LIST OF TABLES ..........................................................................................................x
LIST OF FIGURES ...................................................................................................... xii
CHAPTER 1: INTRODUCTION..................................................................................1
CHAPTER 2: THE IMPACT OF CHANGES IN BANK OWNERSHIP STRUCTURE
ON THE ALLOCATION OF CAPITAL .........................................................................6
2.1.
Introduction .........................................................................................................6
2.2.
Literature Review...............................................................................................10
2.2.1. Empirical Evidence on Government Ownership of Banks ..................................10
2.2.2. Empirical Evidence on Foreign Ownership of Banks..........................................13
2.2.3. Contribution of this Study ..................................................................................16
2.3.
Data and Methodology.......................................................................................17
2.4.
Changes in Bank Ownership Structure ...............................................................25
2.4.1. Current State of Bank Ownership Structure around the World............................25
2.4.2. Changes in Bank Ownership Structure and Country Characteristics ...................29
2.5.
Bank Ownership Structure and the Allocation of Credit .....................................32
2.5.1. Changes in Bank Ownership Structure & Credit Growth....................................32
vii
2.5.2. Changes in Bank Ownership Structure and Credit Allocation Efficiency............34
2.5.3. Addressing Endogeneity Concerns .....................................................................37
2.5.4. Impact of Changes in DB – Testing Alternative Hypotheses...............................39
2.5.5. Credit Growth to Industries with High Dependence on External Financing ........42
2.6.
Bank Ownership Structure and the Allocation of Capital....................................46
2.6.1. Additional Robustness Tests ..............................................................................51
2.7.
Conclusion.........................................................................................................52
CHAPTER 3: DOES THE GROWING PRESENCE OF LARGE DOMESTIC
BLOCKHOLDERS AROUND THE WORLD AFFECT BANK PERFORMANCE? ....54
3.1.
Introduction .......................................................................................................54
3.2.
Literature Review...............................................................................................58
3.2.1. Empirical Evidence on State Ownership of Banks ..............................................58
3.2.2. Empirical Evidence on Foreign Ownership of Banks..........................................60
3.2.3. Contribution of this Study...................................................................................62
3.3.
Data and Methodology.......................................................................................63
3.3.1. Bank Performance Measures..............................................................................67
3.4.
Bank Ownership and Bank Performance ............................................................68
3.4.1. Changes in Bank Ownership and Bank Performance..........................................69
3.4.2. Bank Ownership and Bank Performance – Q measure of performance...............74
3.5.
Spillover Effects of Changes in Bank Ownership Structure ................................77
3.6.
Conclusion.........................................................................................................82
CHAPTER 4:
CONCLUSION....................................................................................84
LIST OF REFERENCES...............................................................................................87
APPENDIX A CREDIT DATA SOURCES ................................................................93
APPENDIX B
DESCRIPTION OF INDUSTRIES FOR CREDIT DATA ..................98
viii
APPENDIX C EXAMPLE OF OWNERSHIP VARIABLE CONSTRUCTION ..........99
APPENDIX D
TABLES..........................................................................................101
APPENDIX E
FIGURES.........................................................................................141
ix
LIST OF TA BLES
Table 1: Summary statistics .................................................................................... 101
Table 2: Ownership of banks around the world: 1995, 2000, and 2005.................... 103
Table 3: Changes in bank ownership structure between 1995 and 2005................... 107
Table 4: Changes in bank ownership and country characteristics ............................ 109
Table 5: Which countries experienced more changes in bank ownership structure? 110
Table 6: Effect of changes in bank ownership structure on industry credit growth... 114
Table 7: Effect of changes in bank ownership structure on the allocation of credit .. 115
Table 8: Regressions using instrumental variables for changes in bank ownership .. 117
Table 9: Testing the impact of domestic blockholder ownership of banks ............... 119
Table 10: Credit growth in industries by extent of dependence on external finance . 121
Table 11: Descriptive statistics - measures of capital allocation efficiency .............. 122
Table 12: Impact of changes in bank ownership structure on capital allocationinstrumental variables approach.............................................................................. 124
Table 13: Regressions using new measure of capital allocation efficiency .............. 127
Table 14: Descriptive statistics - banks with available data by year ......................... 128
Table 15: Impact of changes in bank control on bank performance –
Heckman Model ..................................................................................................... 129
Table 16: Changes in control and bank value .......................................................... 133
x
Table 17: Spillover effects of changes in bank ownership structure – Instrumental
Variable (IV) approach ........................................................................................... 136
xi
LIST OF FIGURES
Figure 1: Changes in bank ownership structure............................................................141
xii
CHAPTER 1:
INTRODUCTION
A new wave of bank privatizations in the past decade has significantly changed
the ownership structure of banking systems around the world. These privatizations were
perhaps in part driven by the well documented findings about the poor performance of
government-owned banks (Berger, Clarke, Cull, Klapper, and Udell, 2005; Mian, 2006b;
Micco, Panizza, and Yañez, 2004), as well as by the evidence pointing to the detrimental
role of government ownership of banks on financial and economic development (Barth,
Caprio, and Levine, 2004; Galindo and Micco, 2004; La Porta, Lopez-de-Silanes, and
Shleifer, 2002a). While the impact of bank privatizations on bank performance has been
well documented in the literature, as summarized in various papers (Clarke, Cull, and
Shirley, 2005; Megginson, 2005), there is little evidence on the broader (macro level)
implications of these changes in bank ownership structure. In addition, the role of large
domestic blockholders has been ignored in most studies of bank ownership1 (Claessens,
Demirgüç-Kunt, and Huizinga, 2001; Demirgüç-Kunt and Huizinga, 1999; Micco, et al.,
2004).
1
Caprio, Laeven, and Levine (2003) study the link between governance and bank valuation, and find that
larger cash flow rights by the controlling owners boost valuation.
1
This dissertation contributes to the literature first by documenting the vast
changes in bank ownership structure over the past ten years - using a hand-collected
database on the ownership structure of the largest (top ten) banks in 90 countries - and by
examining two aspects of these changes that have not been adequately explored thus far.
In CHAPTER 2, I explore some macro level implications of the changes in bank
ownership structure. In particular, I examine how changes in bank ownership structure
affect capital allocation efficiency.
CHAPTER 3 explores how changes in bank
ownership affect the performance of individual banks and the banking sector, with
particular attention paid to the role of large domestic blockholders.
Throughout the world, banks continue to play an important role in the allocation
of capital, or more specifically, in the way credit is distributed. Given the importance of
allocating capital efficiently, which has been cited as a reason why financial development
is associated with economic growth (Beck, Levine, and Loayza, 2000b; Goldsmith, 1969;
Greenwood and Jovanovic, 1990; McKinnon, 1973; Shaw, 1973), coupled with the vast
changes in bank ownership structure over the past decade, CHAPTER 2 explores how
these changes affect the efficiency of capital allocation. I argue that if credit is allocated
efficiently, more credit should be allocated to more productive industries (those that
contribute more to GDP). I explore these questions using data on outstanding credit by
industry – collected from various Central Banks and banking regulatory authorities – and
data on industry value added (i.e. an industry’s contribution to GDP).
The large scale bank privatizations that have occurred over the past decade should
in theory reduce the politically motivated lending practices of government-owned banks,
which should translate into better capital allocation. Surprisingly, I find that the decline
2
in government ownership of banks by itself does not affect capital allocation efficiency.
What matters is who takes over the ownership stakes relinquished by the government.
When large domestic blockholders increase their stakes in banks, capital allocation
efficiency is hampered; in contrast, increased foreign presence in the banking sector has
positive effects on capital allocation efficiency.
Increased domestic blockholder (DB) presence adversely affects capital allocation
efficiency. Increases in DB are associated with increased lending to less productive
economic sectors and to industries that are not dependent on external finance, suggesting
a misallocation of funds by domestic blockholder-controlled banks. These findings are
consistent with the looting view, a pessimistic assessment of related lending practices by
banks (La Porta, Lopez-De-Silanes, and Zamarripa, 2003). Large domestic blockholders
of banks (e.g. local companies, wealthy individuals) typically have substantial interests in
nonfinancial firms as well; banks controlled by these domestic blockholders usually
direct a significant portion of their lending activities to related parties (e.g. firms
controlled by relatives), even when these firms are inefficient. This behavior, observed
primarily in developing countries with poor governance (La Porta, et al., 2003; Laeven,
2001), could thus adversely affect capital allocation efficiency.
In contrast to the adverse impact of increased domestic blockholder ownership of
banks, increased foreign presence in the banking sector improves capital allocation
efficiency. Countries experiencing increases in foreign presence in the banking sector
increase lending to more productive industries, and to those that rely more on external
finance. This new result adds support to the findings of Giannetti and Ongena (2007) who
document that in Eastern European countries, foreign banks improve capital allocation by
3
mitigating the related lending problems. Foreign banks, which have been shown to
outperform local banks, primarily in developing countries (Claessens, et al., 2001;
Demirgüç-Kunt and Detragiache, 2005; Micco, et al., 2004), are more likely to pursue
profit maximizing opportunities than government or domestic blockholder-controlled
banks, which may have ulterior motives. These foreign banks will thus be more likely to
direct investment to those firms or industries with better prospects. This will lead to the
observed positive impact of increased foreign presence on capital allocation efficiency.
CHAPTER 3 then examines the impact of changes in bank ownership structure on
individual bank performance, with particular attention paid to the role of domestic
blockholders. Large domestic blockholders may significantly affect bank performance,
just as they have been shown to affect firm value in the corporate finance literature.
There are different hypotheses that have emerged in the corporate finance literature.
According to the incentive-based view (Shleifer and Vishny, 1997), shareholders with
large cash flow ownership have an incentive to closely monitor a firm’s performance,
potentially mitigating the principal-agent problems that exist between managers and
shareholders (Jensen and Meckling, 1976). In line with this view, several studies have
shown a positive correlation between firm value and cash flow ownership of large
shareholders (Claessens, Djankov, Fan, and Lang, 2002; La Porta, Lopez-de-Silanes,
Shleifer, and Vishny, 2002b). In contrast, large blockholders may negatively affect firm
performance if they pursue their own interests at the expense of other minority
shareholders (Shleifer and Vishny, 1997). Consistent with this entrenchment-based view,
formalized by Stulz (1988), evidence has shown that firm value falls when control rights
exceed cash flow rights of large shareholders (Claessens, et al. 2002).
4
I find evidence consistent with the entrenchment-based view. Increased domestic
blockholder ownership of banks is associated with poor subsequent performance. In
particular, I find that banks changing from government to domestic blockholder control
perform poorly in terms of asset quality and profitability. In addition, these changes in
control have a detrimental impact on bank value. This evidence suggests that domestic
blockholders may be acquiring stakes in banks to extract private benefits of control,
which results in poor bank performance. These findings are also consistent with the
looting view (La Porta, et al., 2003); these banks may be directing a significant portion of
their lending to related, yet inefficient companies, which hurts performance. Consistent
with prior findings, I also find a positive impact of increases in foreign ownership of
banks on profitability and operational efficiency. I also confirm prior findings associated
with the poor performance of government-owned banks. Furthermore, increases in
domestic blockholder ownership of banks appear to improve the asset quality and
profitability of the banking sector. Contrary to prior findings, foreign presence does not
improve the competitiveness of the domestic banking sector.
5
CHAPTER 2:
THE IMPACT OF CHANGES IN BANK OWNERSHIP STRUCTURE ON THE
ALLOCATION OF CAPITAL
2.1.
Introduction
Government ownership of banks has been consistently declining since 1970. This
pattern has accelerated over the past ten years. La Porta, Lopez-de-Silanes, and Shleifer
(2002a) document that government ownership of banks was still prevalent around the
world in 1995.
Since then, the average government ownership of banks (GB) has
declined significantly, dropping to 21% as of 2005. Bank privatizations in the former
socialist countries, including Romania, Bulgaria, Hungary, and Poland, have led the way
in the latest wave of bank privatizations around the world. While privatizations in these
countries have led to an increase in foreign ownership of banks (FB), other countries such
as Belgium, Colombia, France, Norway, and Taiwan have experienced significant
increases in domestic blockholder ownership of banks (DB). Although the impact of
privatization on bank performance has received a lot of attention recently, the
consequences
of
the
resulting
changes
6
in
bank
ownership
structure on capital allocation efficiency in the corporate sector have yet to be explored.2
Specifically, no study has examined whether there is a link between a country’s banking
sector ownership structure and capital allocation efficiency.3 In addition, the role of large
domestic blockholders has been ignored in most studies of bank ownership (Claessens, et
al., 2001; Demirgüç-Kunt and Huizinga, 1999; Micco, et al., 2004).4
Allocating capital efficiently has been cited as a reason why financial
development is associated with economic growth (Beck, et al., 2000b; Goldsmith, 1969;
Greenwood and Jovanovic, 1990; McKinnon, 1973; Shaw, 1973).
Banks play an
important role in the allocation of capital, or more specifically, in the way credit is
distributed. Companies in many countries - particularly in those with less developed
equity markets and weak shareholder protection - continue to rely on bank lending for
financing (Booth, Aivazian, Demirgüç-Kunt, and Maksimovic, 2001; Giannetti, 2003).
With this in mind, this essay contributes to the existing literature by examining whether
and how the changes in bank ownership structure (including, in particular, the increased
presence of domestic blockholders) around the world affect capital allocation efficiency.
I will argue that if credit is allocated efficiently, more credit should be provided to
industries that contribute more to GDP (generate more value added). In some robustness
tests, I also use Wurgler’s (2000) definition of capital allocation efficiency as the extent
to which investment increases in growing industries and decreases in declining industries.
2
Clarke, Cull, and Shirley (2005), and Megginson (2005) provide good summaries of the existing literature
on the effects of bank privatization on bank performance.
3
Beck and Levine (2002) explore whether having a bank or market-based system matters for capital
allocation efficiency. They do not find any evidence that this matters.
4
Caprio, Laeven, and Levine (2003) study the link between governance and bank valuation, and find that
larger cash flow rights by the controlling owners boost valuation.
7
Government ownership of banks has been associated with subpar bank
performance (Dinç, 2005; Micco, Panizza, and Yañez, 2006; Sapienza, 2004), in line
with the political view, which argues that government control of financial institutions
politicizes resource allocation for the sake of advancing certain political agendas (e.g.
obtaining votes, bribing office holders), and, by pursuing such objectives, economic
efficiency is hampered (Kornai, 1979; Shleifer and Vishny, 1994). Such behavior by
government-owned banks should, in turn, hurt capital allocation, particularly in countries
with a large government presence in the banking sector; yet, this has not been explored in
the literature thus far.5 I examine whether the large declines in government ownership of
banks improve the efficiency of capital allocation. Bank privatizations should reduce the
politically-motivated lending practices that adversely affect capital allocation.
Surprisingly, I find that the decline in government ownership of banks by itself does not
affect capital allocation efficiency. What matters is who takes over the ownership stakes
relinquished by the government. When large domestic blockholders increase their stakes
in banks, capital allocation efficiency is hampered; in contrast, increased foreign presence
in the banking sector has positive effects on capital allocation efficiency.
This study further contributes to the literature by examining how the increased
presence of domestic blockholders (DB) in the banking sector affects capital allocation.
Increased domestic blockholder presence adversely affects capital allocation efficiency.
Increases in DB are associated with increased lending to less productive economic sectors
and to industries that are not dependent on external finance, suggesting a misallocation of
5
Wurgler (2000) documents that capital allocation efficiency is negatively correlated with the extent of
government presence in the economy. He does not look at government presence in the banking industry,
however.
8
funds by domestic blockholder-controlled banks. These findings are consistent with the
looting view, a pessimistic assessment of related lending practices by banks (La Porta, et
al., 2003).
Large domestic blockholders of banks (e.g. local companies, wealthy
individuals) typically have substantial interests in nonfinancial firms as well; banks
controlled by these domestic blockholders usually direct a significant portion of their
lending activities to related parties (e.g. firms controlled by relatives), even when these
firms are inefficient. This behavior, observed primarily in developing countries with
poor governance (La Porta, et al., 2003; Laeven, 2001), could adversely affect capital
allocation efficiency.
In contrast to the adverse impact of increased domestic blockholder ownership of
banks, increased foreign presence in the banking sector improves capital allocation
efficiency. Countries experiencing increases in FB increase lending to more productive
industries, and to those that rely more on external finance. This new result adds support
to the findings of Giannetti and Ongena (2007) who document that in Eastern European
countries, foreign banks improve capital allocation by mitigating the related lending
problems. Foreign banks, which have been shown to outperform local banks, primarily
in developing countries (Claessens, et al., 2001; Demirgüç-Kunt and Detragiache, 2005;
Micco, et al., 2004), are more likely to pursue profit maximizing opportunities than
government or domestic blockholder-controlled banks, which may have ulterior motives.
These foreign banks will thus be more likely to direct investment to those firms or
industries with better prospects.
This will lead to the observed positive impact of
increased foreign presence on capital allocation efficiency. This finding is also consistent
with the interest group theory of financial development (Rajan and Zingales, 2003).
9
Foreign bank presence is indicative of a more open financial system in which there is
greater competition. With increased competition, banks that rely on relationship-based
lending (incumbent financiers) will have less flexibility to pursue detrimental activities
(e.g. lending to inefficient, but related firms). This will in turn improve capital allocation
efficiency.
The rest of the chapter is organized as follows. Section 2.2 expands on the
discussion on the existing literature on bank ownership. Section 2.3 describes the data
and methodology used in the study. Section 2.4 describes the current state of bank
ownership structure around the world and explores the characteristics of countries that
have experienced the most drastic changes in bank ownership structure. Section 2.5
explores the link between bank ownership structure and the allocation of bank credit,
Section 2.6 examines the impact of changes in bank ownership structure on a broader
measure of capital allocation efficiency, and Section 2.7 concludes.
2.2.
Literature Review
2.2.1. Empirical Evidence on Government Ownership of Banks
The existing literature on government ownership of banks has documented that
this form of ownership was pervasive around the world as of 1995, is more prevalent in
poorer countries (Barth, Caprio, and Levine, 1999), and in countries with more
interventionist and less efficient governments and less secure property rights (La Porta, et
al., 2002a).
The bulk of the evidence supports the political view of government
10
ownership of banks.6
Consistent with this view, several papers document that
government ownership of banks inhibits financial development and economic growth
(Barth, et al., 2004; Galindo and Micco, 2004; La Porta, et al., 2002a). La Porta et al.
(2002a) show that higher government ownership of banks in 1970 is associated with
slower subsequent financial development and lower economic growth.
Barth et al.
(2004) examine the relationship between state ownership and banking sector
development measures. They find that government ownership of banks is negatively
related to favorable banking outcomes, and positively related with corruption. Micco,
Panizza, and Yañez (2006), Sapienza (2004), and Dinç (2005) provide further support for
the political view. Micco et al. (2006) find that the difference in public and private
banks’ performance widens during election years, supporting the hypothesis that political
considerations drive these results. Sapienza (2004) finds that lending behavior of stateowned banks in Italy is affected by electoral results of the party affiliated with the bank.
In addition, Dinç (2005) shows that government-owned banks in emerging markets
significantly increase their lending in election years relative to private banks. The author
interprets this as evidence that politicians can reward their allies and punish their
opponents through their influence on government-owned banks. Megginson (2005) has a
more complete review of this literature.
Another well-documented finding is the poor performance of state-owned banks
relative to their domestic or foreign-owned counterparts (Berger, et al., 2005; Mian,
2006b; Micco, et al., 2004). Berger et al. (2005) use data from Argentina in the 1990s to
analyze the static, selection, and dynamic effects of domestic, foreign, and state
6
This view argues that government-owned institutions pursue politically motivated objectives.
11
ownership on bank performance. They find that state-owned banks have poor long-term
performance and that those banks undergoing privatization have poor performance
beforehand, and dramatically improve their performance after privatization.
Mian
(2006b) studies 1,600 banks in 100 emerging markets and documents that government
banks perform poorly and only survive due to government support. Micco et al. (2004)
examine the relationship between bank ownership and bank performance for banks in 119
countries.
They find that in developing countries, state-owned banks have lower
profitability, higher costs, higher employment ratios, and poorer asset quality than their
domestic counterparts.
With the exception of state-owned banks having higher costs
than their domestic counterparts, they do not find evidence of significant differences
between state and domestic private banks’ performance in industrial countries. Cornett,
Guo, Khaksari, and Tehranian (2003) examine the differences in performance between
state-owned and private banks in 16 Far East countries between 1989 and 1998. They
also find that state-owned banks are significantly less profitable, have lower capital
ratios, greater credit risk, lower liquidity, and lower management efficiency.
The bulk of the evidence on state-ownership of banks suggests that it is associated
with poor bank performance and with negative economic outcomes.
There is little
evidence supporting the more optimistic development view (Gerschenkron, 1962) of
government ownership of financial institutions, which argues that governments can play a
major role in the financial and economic development of countries in which economic
institutions are not sufficiently developed.
The evidence supporting the political
motivations behind government-owned bank lending activities would support the
argument that government presence in the banking sector hinders capital allocation
12
efficiency.7
With this in mind, this essay will test whether the vast reductions in
government ownership of banks have a positive effect on capital allocation efficiency.
2.2.2. Empirical Evidence on Foreign Ownership of Banks
While government ownership of banks is associated with poor bank performance,
the bulk of the literature documents a positive impact of foreign ownership on bank
performance. Barth et al. (2001) provide data on the share of banking assets held by
foreign-controlled banks in 91 countries as of 1998.
Foreign-controlled banks hold
widely differing shares of assets across countries, but there is no obvious pattern based on
level of development. The data show that foreign-controlled banks hold the largest shares
in countries where the rule of law is well established, but where the financial sector is
less developed. There is evidence that banks’ size, efficiency and performance, and
home country restrictions play a role in determining which banks expand abroad. Several
studies find a positive correlation between bank size and internationalization (Focarelli
and Pozzolo, 2000; Grosse and Goldberg, 1991; Tschoegl, 1983). Clarke, et al. (2003)
provide a detailed summary of the evidence on foreign bank entry.
In terms of individual bank performance, Claessens et al. (2001) document that
foreign banks are more profitable than their domestic counterparts in developing
countries, but the opposite is true in developed markets. Demirgüç-Kunt and Huizinga
(1999) study banks in 80 countries over the 1988-1995 period and find that foreign banks
have higher margins and profits than domestic banks in developing countries, but the
opposite is true for industrial countries. Micco et al. (2004) also document that foreign
7
Wurgler (2000) provides indirect support for this, by finding that capital allocation efficiency is
negatively correlated with the extent of government presence in the economy.
13
banks have higher profitability, lower costs, and lower employment ratios than their
domestic counterparts in developing countries, although they exhibit higher nonperforming loans than their private counterparts. Bonin et al. (2005) examine bank
performance in six Eastern European transition economies and find that foreign banks are
more efficient in terms of cost and profit than domestic and state-controlled banks. They
also find support for the importance of privatizing banks by selling them to strategic
foreign investors. Banks privatized in such manner are more cost and profit efficient than
state-owned banks.
Majnoni, Shankar, and Varhegyi (2003) study the dynamics of
foreign bank ownership in Hungary between 1994 and 2000 and find that foreign banks,
while pursuing similar lending policies, achieve greater profitability than their domestic
counterparts.
Several studies document the impact of foreign bank entry on domestic banks.
Micco et al. (2004) find that foreign bank presence is associated with increased
competitiveness of the domestic banks (lower margins and lower overhead costs).
Claessens, et al. (2001) show that foreign bank entry diminishes the profitability of
domestic banks and reduces their non-interest income and overall expenses. When other
factors are controlled for, high profits reflect a lack of competition, while high overhead
costs, a lack of efficiency. They argue that their findings are consistent with foreign
banks improving the efficiency of domestic banks. Unite and Sullivan (2003) study how
foreign bank entry and foreign ownership of banks affect the banks in the Philippines.
They show that foreign bank entry and penetration reduces interest spreads and operating
expenses of domestic banks, making them more efficient. Barajas, Salazar, and Steiner
(2000) show that foreign entry appears to improve the efficiency of Colombian domestic
14
banks by reducing nonfinancial costs, although the increased competition may have
resulted in increased risk and deterioration in domestic banks’ loan quality. Finally,
Clarke, Cull, and Martínez-Peria (2001) find that foreign bank penetration improves
access to credit. Enterprises in countries with larger foreign presence judge interest rates
and access to long-term loans as smaller constraints on operations and growth than do
enterprises in countries with less foreign presence.
More recently, Detragiache et al. (2006) develop a model that predicts that credit
to the private sector should be lower in countries with more foreign bank penetration.
They find support for their model’s predictions using a sample of 89 low and lower
middle income countries.
argument.
Their results are explained by the “cream-skimming”
Foreign banks are better than domestic banks at monitoring “hard”
information (e.g. accounting information, collateral value), but have a disadvantage in
monitoring “soft” information (e.g. entrepreneurial ability). This leads foreign banks to
lend to safer and more transparent customers or to avoid lending to opaque firms (Berger,
Klapper, and Udell, 2001; Mian, 2006a). Once these hard information customers are
separated from the pool of other borrowers, the remaining soft information clients are left
in a worse pool of borrowers, which causes them to either pay higher interest on their
loans, or not borrow at all. This leads to an overall reduction in credit to the private
sector.
In summary, the bulk of the evidence supports the view that foreign banks
outperform their domestic counterparts and exert a positive influence on the
competitiveness of domestic banks. There is mixed evidence as to whether foreign bank
entry improves or reduces access to credit in the banking system. In addition, one paper
15
examines whether foreign bank entry affects capital allocation efficiency. Giannetti and
Ongena (2007) document that foreign presence can help mitigate connected lending
problems and improve capital allocation. They study firms in Eastern European countries
and document that foreign lending stimulates growth in firm sales and assets, and that it
is the younger firms that benefit the most from foreign bank presence. This essay
expands on their study by examining whether this positive impact of foreign presence on
capital allocation efficiency extends beyond Eastern Europe.
2.2.3. Contribution of this Study
As described in this section, the literature on bank ownership focuses primarily on
either state or foreign ownership of banks. Only three of the papers discussed above look
at state, foreign, and domestic ownership of banks. Berger, et al. (2005) examine the
effects of domestic, foreign, and state ownership on bank performance. However, that
paper is limited in scope, as it is a case study of Argentine banks, and it does not
differentiate between domestic banks’ ownership structures. This essay expands on
Berger’s study by using a more comprehensive dataset covering 90 countries. Mian
(2006b) examines the behavior of foreign, private domestic, and government banks in
100 emerging markets, and Micco et al. (2004) take a more comprehensive look at the
role of bank ownership on performance. While they study the role of state, domestic, and
foreign ownership of banks on bank performance, they also do not account for
differences in domestic banks’ ownership structure (closely-held and widely-held), and
thus fail to explore the role played by large domestic shareholders. In addition, other
than the studies on government ownership of banks (Barth, et al., 1999; La Porta, et al.,
16
2002a), only one of the above studies (Detragiache, et al., 2006) examines the broader
impact of bank ownership on economic outcomes, and that study is limited to studying
the impact of foreign bank presence in poor countries. This essay expands on these
studies by analyzing how changes in the overall bank ownership structure (including the
presence of large domestic blockholders) affect the efficiency of the allocation of capital
primarily through the banks’ role in supplying credit.
2.3.
Data and Methodology
To analyze the role of bank ownership structure on capital allocation, I follow La
Porta et al.’s (2002a) strategy and use a hand-collected database on the ownership
structure of the top 10 banks in 90 countries as of 1995, 2000, and 2005. This
methodology provides adequate coverage of banking system assets.8 To determine the
top 10 banks (commercial and development banks) from each country I use various
sources, including Bureau Van Dijk’s Bankscope, Accuity’s The Global Banking
Resource (TGBR), The Bankers’ Almanac, and Thomson Bank Directory. 9 Ownership
information is also obtained from the aforementioned sources as well as from Mergent
Online and individual bank websites and annual reports. These sources provide measures
of cash flow ownership. The extent of domestic, government, or foreign blockholder
control of a bank may exceed their respective equity ownership (e.g. via golden shares).
8
The top ten banks give adequate coverage of the banking system assets. On average, the top ten banks’
assets represent 88% of the total commercial banks’ assets as of 2005. The lowest coverage is for Japan, in
which the top ten banks’ assets represent 51% of the assets of all commercial banks.
9
Following La Porta et al. (2002a), development banks are included because they are involved in financing
long-term projects where private firms may fail. I thank Mitch Gouss at Bureau VanDijk and Jose F.
Alvarez at Banco Sabadell for providing access to the data.
17
Following La Porta et al. (2002a), I use various alternate ownership measures that
attempt to indirectly capture control of banks.
These measures classify banks as
domestic private, government, or foreign-owned when their equity ownership exceeds
certain thresholds.
To examine the impact of changes in bank ownership structure on how credit is
allocated I also use a hand-collected database on outstanding credit provided to industry
sectors in each of 57 countries. The data was collected primarily from central banks and
financial institutions’ regulatory authorities. Appendix A describes the sources of the
data. Given that the data is collected by individual countries’ regulatory authorities, there
are bound to be differences in the definition of industries and in the way the data is
aggregated (e.g. Austria collects data on all bank loans exceeding a particular threshold,
EUR 350,000). To facilitate comparison across countries, the data on industry subsectors
was aggregated to match the following broad industry definitions: agriculture;
construction; finance and insurance; healthcare; hotels and restaurants; manufacturing;
metals and nonprecious metals; mining; oil and gas; tourism; transportation and storage;
utilities; wholesale and retail trade; services, and other. (Appendix B describes the
subsectors included in each industry category). This broad definition of industries thus
lessens the potential issues associated with differences in industry definitions across
countries. In addition, the data does provide a good indication (while not complete) of
the overall credit provided to particular industry sectors.
Data on value added by industry was obtained from the United Nations Statistics
Division – National Accounts, which provides annual data on gross value added by
industrial classification of economic activity per the International Standard Industrial
18
Classification (ISIC). Gross value added is defined as the value of output less the value
of intermediate consumption; thus, industry value added measures an industry’s
contribution to GDP. The UN data includes seven broad industry classifications: 1)
agriculture, hunting, forestry, and fishing; 2) manufacturing; 3) mining and utilities; 4)
construction; 5) wholesale, retail trade, restaurants, and hotels; 6) transport, storage, and
communications, and 7) other activities.
The industries from the outstanding credit
dataset are aggregated to match the seven ISIC industry classifications, so that in the final
sample, each country will have a maximum of seven industries. Appendix B shows how
the industries from the outstanding credit dataset match the seven ISIC industries. Table 1
shows descriptive statistics of the data on industry credit and value added. As expected,
the correlation between industry value added and outstanding credit is highest among the
more developed countries. This suggests that in most developed countries, where capital
is allocated efficiently, more credit is allocated to industries that contribute the most to
GDP.
I use a number of country level control variables related to financial development.
Some of these measures are the same ones introduced by Beck, Levine, and Loayza
(2000b) and subsequently updated in Beck, Demirgüç-Kunt, and Levine (2000a). The
updated measures were obtained from Ross Levine’s website.
Other country level
control measures (e.g. lending rates, inflation) were obtained from the World Bank’s
World Development Indicators database. Finally, I use data on value added and gross
fixed capital formation from UNIDO’s Industrial Statistics database (2006) in some
additional tests, where I follow Wurgler’s (2000) approach to measure the efficiency of
19
capital allocation. The other variables used in the study will be described later in the
analysis.
Given the consolidations in the banking industries of several countries as well as
data availability issues, some countries in the sample have less than 10 banks.10
All
countries in the sample have at least four banks with available ownership information,
with the exception of Iraq (with only two banks as of 1995).11 The final 2005, 2000, and
1995 sample includes 873, 860, and 798 banks, respectively, from 90 countries. The
sample of countries in this essay is similar to the sample of 92 countries used in (La
Porta, et al., 2002a).12 The differences are due to data availability issues.
To construct the ownership variables, I follow La Porta et al.’s (2002a)
methodology.13 The first measure identifies large domestic blockholder ownership of
banks. A large domestic blockholder is any domestic shareholder (a company, or an
individual) owning more than 10 percent of the shares in a bank. I use the 10 percent
threshold following La Porta et al. (1999).14 I first calculate the total share of each bank
that is owned by large domestic blockholders (DBi) as follows:
10
For example, while ten commercial banks are identified for Singapore as of 1995, mergers and
acquisitions reduce this number to 8 and 6 as of 2000 and 2005, respectively. As a robustness test, I
replicate the main results excluding all countries that do not have information for all top 10 banks.
11
This does not represent a big problem, given that the Iraqi banking sector was state-owned as of 1995.
12
My sample does not include Afghanistan, Ecuador, Syria, and Iceland, which are part of the La Porta et
al. (2002) paper. In addition, my sample includes two countries that are not included in their study:
Macedonia, and Macau. My sample differs from La Porta et al.’s due to data availability issues.
13
While La Porta et al. (2002) constructed government ownership measures, I construct private, and
foreign ownership measures following their methodology.
14
As robustness tests, I also construct measures of domestic blockholder ownership of banks using a 5%,
and a 3% threshold to define a large domestic blockholder. Results are unchanged when these alternate
measured are used.
20
J
DB i k =
∑s
j =1
where s jd ,i > 0 .1 .
jd ,i ;
(1)
where sjd,i is the share of bank i, owned by shareholder j (a domestic shareholder); DBik is
the total share of bank i in country k that is owned by large domestic blockholders.
For
country k, the total domestic blockholders’ stake in the top ten banks (DBk) is computed
by multiplying DBik of each bank by the bank’s total assets (TA), summing this number
across all banks in a country, and dividing by the total assets of the top 10 banks.
10
DB k =
∑ DB
i =1
ik TA ik
(2)
10
∑
TA ik
i =1
DBk is thus the share of total assets of the top 10 banks in each country that is
owned by large domestic blockholders. Given that domestic blockholder control of a
bank may exceed equity ownership, the next set of variables classifies banks as controlled
by domestic blockholders if the domestic blockholders’ equity stake exceeds a minimum
threshold.
The next set of variables (DC10, DC20, DC50, DC90) attempt to capture the
extent of domestic blockholder control of the top ten banks, where control is defined
using different thresholds (10%, 20%, 50%, and 90%). Using the 10% threshold, a bank
is classified as being owned by a large domestic blockholder if the largest shareholder
owning more than 10% of the shares is a local company or individual. To compute DC10,
the assets of those banks classified as large domestic blockholder-owned using this
definition are added together and divided by the total assets of the top 10 banks in a
country.
The other measures (DC20, DC50, and DC90) are constructed in similar
21
fashion, using the respective thresholds. All of these ownership measures are highly
correlated.15 To conserve space, I will only present the results using the DB measure in
the remainder of the essay, following La Porta et al. (2002). The results are unchanged
when any of the other measures are used.
I also follow La Porta et al. (2002) in constructing the government ownership
variables (GB, GC10, GC20, GC50, and GC90). The first variable, GB, is the share of
the assets of the top ten banks that is owned by the government. It is computed as
follows:
J
GBik =
∑s
j =1
ji s gj ,
(3)
where k=1...90 indexes the 90 countries in the sample, j=1...J indexes the banks’
shareholders, i=1…10 indexes the 10 largest banks in the country, sji is the share of bank
i that is owned by shareholder j, and sgj is the share of shareholder j that is owned by the
government, and GBik is the total government’s share in bank i in country k.16 There are
several regional development banks (owned by various governments and private owners)
in the sample, and some of these banks have stakes in various banks. Following La Porta
et al. (2002a), the equity ownership in the regional bank by the local government is
estimated as the proportion of the bank’s assets that are in the country. The government’s
ownership stake in the top ten banks (GBk) is computed by multiplying each bank’s
15
For 1995 (2005), the correlation between DB and DC10 is 0.91 (0.89); the correlation between DB and
DC20 is 0.94 (0.93); the correlation between DB and DC50 is 0.93 (0.92), and the correlation between DB
and DC90 is 0.86 (0.87).
16
If a bank’s shareholder was a nonfinancial institution, various sources were used to determine that
shareholder’s ownership structure. These sources included Mergent Online, as well as company websites.
22
government stake (GBik) by its total assets (TAi) and dividing by the sum of the total
assets of the top ten banks in the country:
10
GB k =
∑ GB
ik TA ik
i =1
(4)
10
∑ TA
ik
i =1
The second variable, GC10, captures the extent of government control of the top
ten banks in the country at the 10% threshold. A bank is classified as government-owned
if the government’s share, GBik, exceeds 10 percent, and the government is the largest
known shareholder. GC10 is then computed as the sum of the assets of all governmentcontrolled banks using this definition divided by the sum of the assets of the top ten
banks in the country. GC20, GC50, and GC90 are constructed in a similar manner,
where a bank is classified as government-owned when GBik>0.2, GBik>0.5, or GBik>0.9,
respectively. All of these measures are highly correlated.17 The analysis will thus focus
on the first measure, GB. A similar approach was used to compute the foreign ownership
variables. A large foreign blockholder is any foreign shareholder owning more than 10
percent of the shares in a bank.18
J
FB
ik
=
∑
s
ji
s
, where sji*sjf>0.1;
jf
(5)
j =1
17
For the 1995 (2005) measures, the correlation between GB and GC 10 is 0.96 (0.95); the correlation
between GB and GC20 is 0.96 (0.97); the correlation between GB and GC50 is 0.97 (0.98), and the
correlation between GB and GC90 is 0.92 (0.93).
18
As robustness tests, I compute measures of foreign blockholder ownership of banks using a 5% and a 3%
threshold to define a foreign blockholder. The results remain unchanged when these alternate measures are
used.
23
where k=1...90 indexes the 90 countries in the sample, j=1...J indexes the banks’
shareholders, i=1…10 indexes the 10 largest banks in the country, sji is the share of bank
i that is owned by shareholder j, and sjf is the share of shareholder j that is owned by
foreigners, and FBik is the total share in bank i in country k that is owned by foreigners.
The foreign ownership stake in the top ten banks (FBk) is computed by multiplying each
bank’s foreign stake (FBik) by its total assets (TAi) and dividing by the sum of the total
assets of the top ten banks in the country:
10
FB k =
∑ FB
ik TA ik
i =1
(6)
10
∑
TA ik
i =1
The other measures of foreign ownership (FC10, FC 20, FC 50, and FC90) are
constructed in the same manner as the government ownership measures.19 The widelyheld measure (WIDELY) captures the share of assets of the top ten banks that are neither
government, foreign, nor domestic blockholder-owned:
WIDELY k = 1 − GB k − PB k − FB k
(7)
A similar measure is constructed at the 10% and 20% thresholds. Banks in which
neither the government, domestic blockholders, nor foreigners owns at least 10% (20%)
are classified as widely-held at the 10% (20%) threshold. The respective widely-held
measures are calculated by summing the assets of these widely-held banks, and dividing
this number by the total assets of the top ten banks in a country.
19
For the 1995 (2005) measures, the correlation between FB and FC10 is 0.94 (0.94); the correlation
between FB and FC20 is 0.95 (0.95); the correlation between FB and FC50 is 0.96 (0.98), and the
correlation between FB and FC90 is 0.91 (0.86).
24
All of these ownership measures are constructed with data from 1995, 2000, and
2005. The ownership data does not allow me to identify the ultimate owners, given that I
do not have data on multiple class shares or pyramidal structures. However, the data
does allow me to determine whether a bank’s largest shareholder is either: (1) the state,
(2) foreigners, (3) or a large domestic blockholder, where a large domestic blockholder
could be an individual, a family, or a local company. Appendix C provides a detailed
example of the calculation of these ownership measures for three banks.
2.4.
Changes in Bank Ownership Structure
2.4.1. Current State of Bank Ownership Structure around the World
Bank privatizations have significantly altered bank ownership structure around
the world. This section describes the bank ownership structure around the world in 2005,
and examines the relationship between country characteristics and the observed changes
in bank ownership structure over the past ten years.
Table 2 shows descriptive statistics of the various measures of bank ownership by
country as of 1995 (e.g. DB95), 2000, and 2005 (e.g. FB05), where countries are grouped
by the origin of their commercial laws. As described earlier, these ownership measures
represent the percentage of the assets of the top ten banks in each country that is owned
by local domestic blockholders (DB), by the government (GB), or by foreigners (FB).
Table 2 shows that as of 2005, the world mean of domestic, government, and
foreign blockholder ownership of banks is 23.12%, 20.91%, and 22.71%, respectively.
The results from Table 2 also show that government ownership of banks has been
25
consistently declining, while foreign and domestic blockholder ownership of banks have
been increasing. Most of the changes in bank ownership structure appear to happen in
the latter part of the 1990s, as reflected in the ownership measures as of 2000.
Panels B and C of Table 2 reveal that large domestic blockholder ownership of
banks is more common in developed markets, while government ownership of banks is
more prevalent in emerging markets. Interestingly, although government ownership of
banks is higher in civil law than in common law countries (Panel B), the difference is not
statistically significant and the magnitude of the difference has declined over the past ten
years. Panel D reveals that as of 1995 government (domestic) ownership of banks was
higher (lower) in countries that subsequently experienced a banking crisis.
Table 3 documents the changes in bank ownership structure that have taken place
over the past ten years. The average government ownership of banks (GB) experienced a
40.9% decline over the past ten years. GB dropped from 35.38% in 1995 to 20.91% in
2005.20 Panel B reveals that the decline in government ownership of banks was more
pronounced in emerging markets, where GB declined by 16.1 percentage points,
compared to a 9.9 percentage point decline in developed markets. As expected, Panel C
shows a significant decline in GB (16.4 percentage points) in civil law countries, where
government ownership was higher as of 1995; the decline was not statistically significant
in common law countries. The reduction in government ownership of banks has been
accompanied by large increases in foreign, and to a lesser extent, in domestic blockholder
20
This decline in government ownership of banks was similar to the decline observed between 1970 and
1995 (from 58.9% to 41.6%), per La Porta et al. (2002). Their measure of government ownership of banks
as of 1995 (GB95) was 41.6%, and differs from the GB95 measure used in this essay (35.38%) due to the
differences in the sample of countries, as documented earlier. However, the correlation between both
measures of government ownership of banks is 0.95.
26
ownership of banks. Foreign ownership of banks more than doubled in emerging markets
(FB increased from 9.9% in 1995 to 23.9% in 2005). Although FB also increased by 9.6
percentage points in developed markets, the increase was not statistically significant. FB
also increased by 15.6 (18.7) percentage points in civil law countries (countries with poor
shareholder protection).21
Finally, domestic blockholder ownership of banks (DB)
increased by 7.5 percentage points in civil law countries.
Results thus far show that while the reduction in government ownership of banks
has been widespread, the increases in foreign and domestic blockholder ownership of
banks have been concentrated in less developed countries and in countries with weaker
protection of minority shareholders (i.e. civil law countries). The increase in foreign
ownership of banks in countries with poor protection of minority shareholders may be
explained by the fact that these foreign banks are establishing a presence in these
countries to exploit future profit opportunities, ignoring the current conditions prevalent
in these countries. This view is summarized in a recent Wall Street Journal article:
Foreign banks are pouring into Turkey, spending billions to acquire stakes in
local banks despite the country’s potentially destabilizing elections, its stalled
negotiations for European Union membership, and frequent complaints of unfair
treatment for foreign investors in its courts… The attraction: Turkey is considered
“underbanked.” (Echikson, 2007)
21
Countries with poor shareholder protection are those with an Anti-Self-Dealing Index below the median
of all countries in the sample. The ASD Index was obtained from Djankov et al. (2007).
27
Whether the increased foreign bank presence is beneficial for the host countries is
a hotly debated issue. There is some evidence that foreign banking presence improves
access to credit (Clarke, et al., 2001). On the other hand, there is some evidence that the
opposite is the case (Detragiache, et al., 2006). This issue is explored further in the
remaining sections of the essay.
The motivation behind the increase in large domestic blockholders’ ownership
stakes in banks is less clear. Large domestic blockholders may want to extract private
benefits of control, and thus they may be increasing their stakes in banks precisely in
countries where minority shareholders are least protected. If this is the case, an increase
in DB should be associated with negative economic outcomes. On the other hand, the
poor protection of minority shareholders in some countries may be forcing minority
shareholders to either increase their stakes (thereby becoming large shareholders
themselves), or to relinquish their stakes to the existing large shareholders, resulting in
more concentrated ownership structures. This in turn could have a positive effect, as
these large shareholders would have strong incentives to carefully monitor the bank’s
performance (Shleifer and Vishny, 1997), thus mitigating the manager-shareholder
agency problems (Jensen and Meckling, 1976). If this view holds, there should be a
positive relation between changes in DB and economic outcomes. These hypotheses are
explored in the remaining sections of the essay.
First, I examine the relation between
country characteristics and the observed changes in bank ownership structure.
28
2.4.2. Changes in Bank Ownership Structure and Country Characteristics
The changes in bank ownership structure have not been uniform across countries.
Here I explore the relation between country characteristics and the subsequent changes in
domestic, government, and foreign blockholder ownership of banks. First, the analysis
focuses on correlations between beginning country level characteristics of financial and
economic development, and governance, and the changes in ownership over the 19952005 period (∆DB, ∆GB , ∆FB). The Kaufmann et al. (2006) governance indicators used
in this study capture six dimensions of governance: voice and accountability, political
stability, government effectiveness, regulatory quality, rule of law, and control of
corruption. All of these measures are scored on a scale from -2.5 to 2.5, with higher
scores corresponding to better governance. One of the advantages of using these
measures is that they provide a time series (the measures were published every other year
since 1996).
Panel A of Table 4 shows the correlations between the ownership measures and
the various country characteristics.
To control for differences in initial levels of
development (given that poorer countries have higher government ownership of banks), I
follow La Porta et al. (2002) and report the coefficients from regressions of changes in
bank ownership measures on the respective country characteristics, an intercept, and the
log of GDP per capita as of 1995. The results shown in Table 4 reveal that domestic
blockholder ownership of banks increased more in civil law countries and in less
financially developed countries (those with lower market capitalization/GDP).
Government ownership of banks declined more in countries that experienced a systemic
banking crisis over the period and in those that were less developed but had better
29
governance as of 1995. In particular, government ownership of banks declined more in
countries with more freedom of expression and political stability.
Finally, Table 4
reveals that foreign ownership of banks (FB) increased primarily in less developed,
politically stable, civil law countries, and in countries that experienced a systemic
banking crisis during the period. Not surprisingly, FB also increased in countries with
weaker government effectiveness (i.e. poor quality of public services). Such countries
present unexploited opportunities for foreign banks that may be able to provide better and
broader ranges of financial services than their domestic counterparts.
Next, I explore more in depth the characteristics of countries that experienced the
largest changes in bank ownership structure. To address this question, the following
regression framework is used:
∆OWN = α + β1CREDIT95 + β 2 COM + β 3GOVERN + β 4 FR + β 5CRISIS
(8)
where ∆OWN is the change in the respective ownership measures between 1995 and 2005
(e.g. DB05-DB95); CREDIT95 is the private credit/GDP as of 1995; COM is a dummy
variable for countries with common law origin; GOVERN corresponds to Kaufmann et
al.’s (2006) six governance indicators measured as of 1996, the earliest available date,
and FR and CRISIS are dummies which equal one if the country has any restrictions on
foreign bank entry or if it experienced a systemic banking crisis during the period,
respectively.22
22
The data used to construct the foreign restriction dummy was obtained from Ross Levine’s website. The
data is used in Barth, Caprio, and Levine’s forthcoming book, Rethinking Bank Supervision and
30
The results from Panel A of Table 5 show that civil law countries experienced
larger increases in domestic blockholder ownership of banks (DB). Surprisingly, none of
the governance indicators appear to explain changes in DB over the period. Neither
foreign restrictions nor having a banking crisis appear to affect DB.
Panel B of Table 5 reveals that more financially developed countries experienced
larger increases, or smaller declines in government ownership of banks (GB).23
In
contrast, countries with better governance (measured by freedom of expression, political
stability, and rule of law) experienced a more significant decline in GB over the period.
Countries experiencing a systemic banking crisis also experienced larger declines in GB,
while the decline was lower in countries restricting foreign entry.
Finally, Panel C confirms that foreign ownership of banks increased more in
countries that were less financially developed, but with better governance. In particular,
foreign ownership of banks increased more in countries with better freedom of
expression, with more political stability, and with better rule of law, regulatory quality,
and control of corruption. Given these results, in subsequent sections I will use some of
these governance measures as instruments for changes in ownership structure variables,
to address endogeneity concerns.
The impact of these changes in bank ownership
structure on the efficiency of capital allocation is explored next.
Regulation: Until Angels Govern. If a country restricts foreign entry either through acquisition, a
subsidiary, or a branch, the foreign restriction dummy is set to 1.
23
Most countries (66 out of the 90 countries in the sample) experienced a decline in government ownership
of banks over the past ten years.
31
2.5.
Bank Ownership Structure and the Allocation of Credit
2.5.1. Changes in Bank Ownership Structure & Credit Growth
Before examining how the changes in bank ownership structure affect how credit
is allocated, I first explore whether changes in bank ownership structure affect credit
growth. I will do this using the following regression framework:
CG i ,c,t = α + β ∆OWN c + γ t + δGDPGrowth c,t + ηLR c,t + φINFc ,t +
λ COM c + ε i ,c,t
(9)
where CGi,c,t is the annual growth in credit provided by the banking system to industry i
in country c from 2001-2006; ∆OWNc refers to the changes in bank ownership structure
in country c between 1995 and 2000; γt refers to year dummy variables, and the controls
include the growth in real GDP per capita, the average lending rate, LR, the inflation rate,
INF, and a common law dummy, COM, for countries with a common law origin of their
commercial laws. To mitigate concerns about discrepancies in data availability across
countries, for a country to be included in these regressions, I require a minimum of five
industries with five years of data available post 2000.24 This filter reduces the sample to
41 countries.25 While data availability prevents me from using the entire sample of 90
countries in these regressions, this subsample is representative of the full sample. The
average changes in DB, GB, and FB for this subsample (the full sample) are 4.8%
(5.5%), -9.2% (-11.1%), and 10.8% (10.6%), respectively.
To mitigate any potential
endogeneity concerns, credit growth is measured from 2001-2006, the period starting a
24
Some countries in the sample have data on a limited (three or fewer) number of industries.
25
As a robustness test, I run the regressions requiring at least three and four industries per country,
respectively, and the results remain unchanged.
32
year after the changes in bank ownership structure variables (measured from 1995-2000).
The changes in ownership variables were measured during the 1995-2000 because this
was the period in which most of the changes in bank ownership structure occurred, as
documented in Table 2.
The results from the above regressions are shown in Table 6. Results including
all ownership variables in the regressions are not shown because of multicollinearity
concerns, given the high degree of correlation between these variables.26 The results
show that industry credit growth over the 2001 through 2006 period tends to be higher in
common law countries, and in countries where lending rates, inflation, and growth in
GDP were higher over the period. These findings are not unexpected. For example,
banks are more likely to provide more credit in countries where lending rates are higher,
while the demand for credit should be higher in countries experiencing higher economic
growth; thus, countries with high lending rates and high economic growth should be
expected to experience higher growth in lending activity. With respect to inflation, the
finding is almost a mechanical one. The credit growth represents nominal growth in
credit outstanding; so, it is no surprise that countries experiencing high inflation have
higher growth in credit.
The results in Table 6 also reveal that increases in domestic (DB) and foreign
(FB) blockholder ownership of banks are associated with higher subsequent growth in
industry credit. In contrast, an increase in government (GB) ownership of banks is
associated with slower credit growth. The results are both statistically and economically
significant. For example, for the 41 countries in the sample, the average increase in FB
26
For example, the correlation between ∆GB and ∆FB is -0.61, and the correlation between ∆FB and ∆DB
is -0.33. Both are statistically significant at the 1% level.
33
was 10.7%; thus, for the average country experiencing a 10.7% increase in FB, credit
growth would be 1.61% higher than that of a country where FB did not change. For
Portugal, a country with an increase in FB close to the average, the 1.61% increase in
credit growth translates into an additional EUR 206 million of credit per year.27
2.5.2. Changes in Bank Ownership Structure and Credit Allocation Efficiency
Thus far, I have shown that increases in both DB and FB lead to increases in total
credit growth. The question remains as to whether credit is being directed to the right
industries. Defining good industries as those which generate higher value added, I now
examine whether changes in bank ownership structure affect the quality of the allocation
of credit. If countries allocate credit efficiently, productive industries should receive
more credit.
As a first step in classifying industries, I will take the average value added of an
industry over the period 1995-2000, and will rank industries in each country as high
(low) value added if the average value added generated by the industry is above (below)
the median industry value added for the country. The limited number of industries
(maximum of 7) in each country prevents me from using other classifications (e.g.
quintiles).28 The following regression framework is used:
CGi ,c,t = α + β1 HIVA + β 2 HIVA × ∆OWN + β 3 ∆OWN + γ t + δControls + ε i ,c,t
(10)
27
The average outstanding credit to industry in Portugal during this period was EUR12.79 billion. A
1.61% increase in credit growth translates into EUR 206 million.
28
In unreported results, I use triciles to classify industries as high (low) value added. The results obtained
using this classification are similar to the ones obtained using the median.
34
where CGi,c,t is the annual growth in credit provided by the banking system to industry i
in country c from 2001-2006; HIVA is a dummy variable which equals one if industry
i’s average value added over the 1995-2000 period is above the median for all industries
in a country, and 0 otherwise; ∆OWN refers to the changes in bank ownership structure in
country c between 1995 and 2000; γt refers to year dummy variables, and the controls
include the growth in real GDP per capita, the inflation rate, the average lending rate, LR,
and a common law dummy, COM, for countries with a common law origin of their
commercial laws. For a country to be included in the regressions, I require a minimum of
five industries with five years of data available post 2000.
The main variable of interest in the above regression is β2. If the looting view
holds, and domestic blockholders are directing their lending activities to related, yet
inefficient firms (or ultimately, industries) we expect the β2 coefficient (interaction
between the high value added dummy and change in DB) to be insignificant, or at worst,
negative. The latter result would indicate that domestic blockholders are providing more
credit to “poor” industries that are not contributing much to GDP (low value added
industries). Given the negative impact of government ownership of banks on individual
bank performance and economic development associated with politically-motivated
lending activities, the β2 coefficient for the regressions including the changes in GB
should also be either insignificant or negative. Finally, for the regressions including
foreign ownership of banks, the expected β2 should be positive, given prior evidence on
the positive impact of foreign ownership of banks on bank performance.
35
The results from the above regressions are presented in Panel A of Table 7. All
standard errors are clustered by country. Providing some support for the looting view, the
results show that increases in DB have a negative impact on the efficiency of credit
allocation. Results show that credit growth in high value added industries is lower than
credit growth in low value added industries in countries experiencing an increase in DB.
Annual credit growth in high value added industries is 11 basis points lower than credit
growth in low value added industries. For the average country, which experienced a
4.8% increase in DB, annual credit growth in high value added industries lagged behind
the growth in low value added industries by 0.53 percentage points.
Changes in GB do not appear to have a significant impact on the way credit is
allocated. In contrast, increases in FB do have a positive impact on how credit is
allocated. Credit growth to high value added industries is 24 basis points higher than
credit growth to low value added industries in countries experiencing a one percent
increase in FB. For the average country experiencing a 10.7% increase in FB, annual
credit growth to high value added industries is 2.57 percentage points higher than that of
low value added industries.
Thus far, I have shown that countries experiencing an increase in DB (FB)
allocate less (more) credit to industries that contribute more to a country’s GDP (high
value added industries). The way an industry was classified, based on average value
added between 1995 and 2000, could create concerns about endogeneity, given that the
changes in bank ownership structure are measured contemporaneously (1995-2000).
Ideally, the value added classification of industries should be exogenous to the changes in
bank ownership structure. To mitigate endogeneity concerns, I reclassify industries as
36
high and low value added based on value added for the period 1990-1995, prior to the
period when the changes in bank ownership structure occurred. The results from the
regressions using this new classification method are shown in Panel B of Table 7. Using
this different time period to classify industries, the results are surprisingly similar. This
alleviates some of the endogeneity concerns, and points to an interesting finding. The
results appear to indicate stability in the industries’ generation of value added across
time. Those industries generating high value added in 1990 through 1995 continue to
create high value added in the subsequent period (1995-2000). In fact, the correlation
between the two high value added dummies (1990-1995 and 1995-2000) is 0.98 and
statistically significant at the 1% level.
2.5.3. Addressing Endogeneity Concerns
As an alternative way to address potential endogeneity concerns, in this section I
will use a Two-Stage-Least Squares regression approach, using instrumental variables
(IVs) for changes in foreign and domestic blockholder ownership of banks. A proxy for
the size of the productive population (population ages 15-64 as a percent of the total
population) and the Kaufmann et al. (2006) political stability index as of the beginning of
the period are used to forecast changes in foreign ownership of banks. Countries that are
politically stable and have potentially larger work forces could attract foreign banks; yet,
these measures should not directly affect how credit is allocated in the 2001-2006 period.
Similarly, I use a proxy for bank concentration, a measure of dispersed ownership of
banks, and Kaufmann et al. (2006) rule of law index as of the beginning of the period to
37
forecast changes in domestic blockholder ownership of banks.29 Domestic blockholders
may be reluctant to increase their stakes in banks in countries with highly concentrated
banking systems, while it would be easier for them to increase their stakes in banks in
countries where dispersed ownership of banks is prevalent. In addition, if they are indeed
pursuing ways to extract private benefits of control, domestic blockholders are more
likely to increase their stakes in banks in countries where the rule of law is not well
established. Yet, these measures should not directly affect the way in which credit is
allocated to industries.
As documented by Larcker and Rusticus (2007), prior to using instrumental
variables, several tests should be performed to ensure that IV methods are indeed
preferred to traditional ordinary least squares procedures. To assess whether the use of
IVs will mitigate rather than exacerbate the impact of endogeneity, and to test the validity
of the instruments used, I follow the steps recommended by Larcker and Rusticus (2007).
Panel A of Table 8 shows the results from the first-stage regressions. The results show
that the respective instruments work well in forecasting changes in domestic (DB) and
foreign (FB) ownership of banks (the partial R2 is 0.18 and 0.19 for the regressions of
changes in DB and FB, respectively).30 The partial F-test rejects the hypothesis that the
instruments are jointly zero. In addition, the test of over-identifying restrictions fails to
reject the null that at least one of the instruments is valid. Finally, after addressing
concerns about the validity of the instruments, the Hausman test fails to reject the null of
29
Bank concentration is calculated as the ratio of the assets of the top 3 banks to the assets of all
commercial banks in the country as of 1995.
30
Given that these are cross-sectional regressions for 52 countries, the explanatory power of the
instruments appears reasonable.
38
no endogeneity problems for the changes in domestic and foreign blockholder ownership
of banks. The latter results help to alleviate concerns about potential endogeneity issues.
Panel B of Table 8 shows the results from second stage regressions, where the
changes in bank ownership structure variables have been instrumented. The results using
the IV approach confirm prior results. Increases in DB (FB) are associated with lower
(higher) growth in credit to more productive (high value added) industries. Even after
addressing endogeneity concerns in a variety of ways, the results continue to indicate that
increases in domestic (foreign) blockholder ownership of banks are associated with
deterioration (improvement) in the way credit is allocated.
2.5.4. Impact of Changes in DB – Testing Alternative Hypotheses
The results thus far show that increases in DB appear to adversely affect capital
(or credit) allocation efficiency by directing more credit to industries that do not
contribute much to a country’s GDP. These results are consistent with the looting view,
in which domestic blockholders direct a significant portion of their banks’ lending to
inefficient, but related firms, or to firms in dying industries in which these blockholders
may have large ownership stakes. An alternative interpretation of the results could be
that domestic blockholders are just continuing the politically-motivated lending policies
of the government when they acquire the banks’ stakes. In this line of thought, the
observed declines in government ownership of banks are due to pseudo-privatizations, in
which the new owners may be government officials or their related parties. If this
alternative hypothesis holds true, there should not be a significant difference in the way
credit is allocated when the government’s ownership stakes in banks are acquired by
39
domestic blockholders (who perhaps have ties to the government). To test this alternative
hypothesis, I identify countries with large declines in GB coupled with large increases in
DB.31 Using an indicator dummy, GD, I then run the following regressions:
CG i ,c ,t = α + β1 HIVA + β 2 HIVA × GD + β 3 HIVA∆OWN + β 4 ∆OWN +
δControls + ε i ,c ,t
(11)
where CGi,c,t is the annual growth in credit provided by the banking system to industry i
in country c from 2001-2006; HIVA is a dummy variable which equals one if industry
i’s average value added over the 1990-1995 period is above the median for all industries
in a country, and 0 otherwise; GD is a dummy which equals one if both the decline in GB
and the corresponding increase in DB exceed a particular threshold (e.g. 5%, 10%), and 0
otherwise; ∆OWN refers to the changes in bank ownership structure in country c between
1995 and 2000; γt refers to year dummy variables, and the controls include the growth in
real GDP per capita, the inflation rate, the average lending rate, LR, and a common law
dummy, COM, for countries with a common law origin of their commercial laws. For a
country to be included in the regressions, I require a minimum of five industries with five
years of data available post 2000.
If the domestic blockholders are just a continuation of the government in terms of
their lending policies, I expect the β2 coefficient in Equation 11 to be insignificant. This
would signify that there was no significant change in the way credit was allocated in
31
Belgium, Colombia, and Bulgaria are among the countries experiencing large declines in GB coupled
with large increases in DB.
40
countries where domestic blockholders took over the bank ownership stakes held by the
government. The results are shown in Panel A of Table 9. A country experiencing a
large decline in GB coupled with a large increase in DB was subject to a significant
deterioration in the way credit was allocated. Credit growth to high value added vis-à-vis
low value added industries was lower (by about 0.1 percentage points) in countries
experiencing large declines in GB coupled with large increases in DB. This result
provides evidence against the alternative hypothesis that the negative impact of domestic
blockholders was a result of a continuation of government policies. This is further
evidence in favor of the looting view as an explanation for the negative impact of
increases in DB on the way credit is allocated. In Panel A of Table 9, the threshold used
to define large increases in DB and large declines in GB was 10%. As robustness tests,
in unreported results I use other thresholds (5% and 15%) and obtain similar results.
Another potential explanation for the observed results on the impact of domestic
blockholders on credit allocation is that they focus more on long-term trends in value
added rather than the short-term measure used thus far to classify industries. Perhaps
using five years to measure industry value added is too short a timeframe. There may be
some industries that experienced a downturn in value added over the short-term, but they
are indeed good, productive industries. If this is the case, domestic blockholders may in
the end be targeting the right industries in their lending activities, and our classification
of industries is too short-sighted. To test this alternative hypothesis, I will use a longer
time period to establish which industries generate more value added. Using data from
1970 through 1995, to avoid potential endogeneity issues, I will use the slope of the time
41
trend in industry value added to rank industries as high (slopes above the median for all
industries in a country) and low (below median) value added, respectively.
The results are shown in Panel B of Table 9. Using this new long-term approach
to classify industries does not alter the results.
Increases in domestic blockholder
ownership of banks are still associated with lower credit growth in high value added
industries. Thus, it does not appear that domestic blockholder are choosing to lend to the
better industries, no matter whether the classification of industries is based on a shortterm or long-term approach. It should be noted that foreign blockholders continue to
have a positive impact on credit allocation even when using this long-term approach to
determine high value added industries.
2.5.5. Credit Growth to Industries with High Dependence on External Financing
Having established that increases in FB (DB) are associated with improvements
(deterioration) in capital allocation using an industry’s contribution to value added as a
proxy for good industries, I now turn to examine whether the changes in bank ownership
structure have an impact on credit growth to industry sectors that rely more on external
finance. I will use Rajan and Zingales’ (1998) approach to determine which industries
are more dependent on external finance. However, instead of relying on the use of
external finance by US firms in an industry as a proxy for the desired use of external
finance by foreign firms in the same industry, I will use data from WorldScope to
compute foreign firms’ use of external finance. This approach will capture differences in
the need for external finance for firms in the same industries across countries. This is a
reasonable approach, given that even firms in the same industry may be at different
42
stages of the product life cycle in different countries, which leads to differing needs for
external finance. As a robustness test, in unreported results, I also use US firms’ external
finance dependence as a proxy for foreign firms’ external finance dependence and obtain
similar results.
Following Rajan and Zingales (1998), a firm’s dependence on external finance
(external finance dependence ratio) is computed as a firm’s capital expenditures
(WorldScope item WS04601) minus cash flows from operations as a percent of total
capital expenditures.32 I sum each firm’s use of external finance from 1995-2000 and
divide by the sum of capital expenditures over the period. In each country, firms are
grouped by industries (requiring a minimum of three firms per industry) to match the
industry definitions of the data on outstanding credit (see Appendix B). The median
industry external finance dependence ratio is compared to the median ratio for all
industries in a country. An industry is then classified as highly dependent on external
finance if its external finance dependence ratio is above the median for all industries in its
country.
The following regression framework is used to determine whether credit is
flowing to those industries in need of external finance:
CG i ,c ,t = α + β 1 HIFD + β 2 HIFD × ∆OWN + β 3 ∆OWN + γ t +
(12)
δControls + ε i ,c ,t
32
Cash flow from operations are computed as the sum of cash flow from operations (WC04201) plus
decreases in inventory (WC04826), decreases in receivables (WC04825), and increases in payables
(WC04827).
43
where CGi,c,t is the annual growth in credit provided by the banking system to industry i
in country c from 2001-2006; HIFD is a dummy variable which equals one if industry i’s
external finance dependence ratio (capital expenditures minus cash flows from operations
divided by capital expenditures) is above the median for all industries in a country, and 0
otherwise; ∆OWN refers to the changes in bank ownership structure in country c between
1995 and 2000; γt refers to year dummy variables, and the controls include the growth in
real GDP per capita, the inflation rate, the average lending rate, LR, and a common law
dummy, COM, for countries with a common law origin of their commercial laws. Due to
data availability, the sample of countries drops to 32 in these regressions.
The results from the regressions in Equation 12 are shown in Table 10 The
results show that on average, credit growth is higher (by about 3 percentage points) in
industries that are highly dependent on external finance. Increases in DB and GB are
associated with lower growth in credit to external finance-dependent industries. For
countries experiencing a one percent increase in DB, credit growth in industries with high
external financing needs is 3.3 percent lower than more financially independent
industries. For the average country in the sample experiencing a 5 percent increase in
DB, credit growth in industries with high dependence on external finance was 16.5
percent lower. The results in Table 10 also provide some support for the political view of
government ownership of financial institutions. Increases in GB are associated with
higher credit growth in industries that are less dependent on external finance. This is
consistent with government banks pursuing politically motivated lending practices, which
result in poor credit allocation.
44
In contrast to the impact of increases in DB and GB, increases in FB are
associated with faster growth in credit to industries that are in need of external financing.
Credit growth to such industries is 10.8 percent higher in countries experiencing a one
percent increase in FB. This is yet another indication that foreign banks make better and
more objective lending decisions.
Overall, increases in DB (FB) are associated with inefficient (efficient) allocation
of credit. In countries experiencing an increase in DB, more credit is being directed to
industries that do not contribute much to GDP (generate low value added), and to those
industries that are not in much need of external financing. Thus, domestic blockholders
appear to be providing funds to firms in industries that may not need or use those funds
productively, which provides support, albeit somewhat indirect, for the looting view. In
contrast, the allocation of credit seems to be more efficient in countries where FB
increased. Increases in FB are associated with increased lending activity to industries
that are more dependent on external finance and are more productive. These findings add
further support to the findings of Giannetti and Ongena (2007), who document that
foreign bank presence in Eastern Europe improves capital allocation efficiency by
reducing related lending problems. Foreign banks, with fewer connections with local
families and governments, may indeed be better able to base their lending activities on
objective measures (e.g. firm’s growth prospects). Finally, foreign bank presence is also
indicative of a more open system, where competition may drive out poor performing
banks. In such an environment, banks that pursue relationship motivated lending may be
forced to refrain from those activities or perish. This is consistent with the interest group
theory of financial development (Rajan and Zingales, 2003).
45
Having established a link between changes in bank ownership structure and the
allocation of bank credit, the next section will examine whether these changes in bank
ownership structure affect broader measures of capital allocation efficiency.
2.6.
Bank Ownership Structure and the Allocation of Capital
Wurgler (2000) documents that financial markets improve the allocation of capital
and that more developed markets allocate capital more efficiently. He also finds that the
efficiency of capital allocation is positively correlated with the legal protection of
minority shareholders and with the amount of firm-specific information in the market,
and negatively correlated with the extent of state-ownership in the economy.
The
banking sector plays a pivotal role in the allocation of capital. After all, bank loans
continue to be a main source of financing for many firms, even in countries with welldeveloped stock markets. In fact, a recent Wall Street Journal article states that Asian
firms’ reliance on bank loans have shielded them from the recent turmoil in global credit
markets (Morse and Wright, 2007).33 With this in mind, in this section I explore whether
the vast declines in government ownership of banks around the world over the last ten
years are associated with improvements in capital allocation efficiency.
I will follow Wurgler’s (2000) approach to estimate capital allocation efficiency,
arguing that capital is allocated efficiently if investment increases in growing industries
and decreases in declining industries. The following regression is estimated for each
country using data from 1995-2004 (or the latest available date):34
33
This includes firms in countries with well-developed stock markets, such Japan and Singapore.
46
ln
I ict
V
= α +ηc ln ict + ε ict ,
I ict−1
Vict−1
(13)
where I is real gross fixed capital formation, V is real value added, i indexes
manufacturing industry, c and t index country and year, respectively.35 All of these
measures were obtained from UNIDO’s INDSTAT-3 database. Value added is the value
of shipment of goods produced minus the cost of intermediate goods; gross fixed capital
formation is the cost of fixed assets minus the value of the sale of used fixed assets; ηc is
an elasticity that measures the extent to which country c increases investment in its
growing industries and decreases investment in declining industries.
To explore whether changes in bank ownership structure affect capital allocation
efficiency, I use the following regression framework:
η c = α + β1 ∆OWNERSHIP c + β 2 FD c + β 3COM + β 4 CRISIS + ε c ,
(14)
where ηc is the estimate of capital allocation efficiency for country c from Equation 13;
∆OWNERSHIPc refers to the changes in bank ownership structure in country c between
1995 and 2000; FDc is a summary financial development measure, the log of one plus the
average sum of stock market capitalization and private credit to GDP;36 COM is a
dummy for countries with a common law origin, and CRISIS is a dummy which equals
34
There are seven countries with data available through 1998; for two countries data is available through
1999; six countries have data available through 2000; six countries have data available through 2001;
eleven countries have data available through 2002; and the remaining countries have data available through
2004.
35
The values for gross fixed capital formation and value added are in US$. Following Wurgler (2000), the
real gross fixed capital formation numbers (I) are obtained by deflating the nominal series by the US capital
goods producer price index (base year 1982), and the real value added (VA) figures are computed by
deflating the series by the US finished goods producer price index (base year 1982).
36
This measure is the same one used by Wurgler (2000). Market capitalization and private credit to GDP
ratios are obtained for the years 1990, 1995, and 2000, and these values are averaged to smooth out cyclical
variations.
47
one if the country experienced a systemic banking crisis during the period.
These
regressions use changes (rather than levels) in ownership as explanatory variables
because the question of interest is whether and how these changes (e.g. increases in
foreign presence) affect capital allocation efficiency. Data on value added and gross
fixed capital formation after 1995 is available for only 49 out of the 90 countries in the
sample. Thus, the analysis in this section focuses on this subset of countries.37 Table 11
shows descriptive statistics of the elasticity measure (ηc) as well as an alternate measure
of capital allocation efficiency used in the following analyses.
Examining Equation 14, endogeneity is a valid concern, particularly for changes
in foreign ownership of banks, and to a lesser extent, for changes in domestic blockholder
ownership of banks. For example, changes in bank regulations (e.g. deregulation), which
may have a positive impact on capital allocation efficiency, may also attract foreign
banks. Thus, one could argue that it is these changes in banking regulation that improve
capital allocation, and not the increased presence of foreign banks. To address this valid
endogeneity concern, I will use a Two-Stage-Least Squares regression approach, using
instrumental variables (IVs) for changes in foreign and domestic blockholder ownership
of banks. As in the previous section, a proxy for the size of the productive population
(population ages 15-64 as a percent of the total population) and the Kaufmann et al.
(2006) political stability index as of the beginning of the period are used to forecast
changes in foreign ownership of banks, and a proxy for bank concentration and a measure
37
To alleviate concerns of a potential bias in this sub sample, I replicate Wurgler’s (2000) experiment for
this subset of 49 countries and obtain similar results (e.g. capital allocation efficiency is positively
correlated with financial development and protection of minority shareholders).
48
of dispersed ownership of banks as of the beginning of the period are used as instruments
for changes in domestic blockholder ownership of banks.
Panel A of Table 12 shows the results from the first-stage regressions.
The
results show that the respective instruments work well in forecasting changes in domestic
(DB) and foreign (FB) ownership of banks (the partial R2 is 0.25 and 0.26 for the
regressions of changes in DB and FB, respectively).38
The partial F-test rejects the
hypothesis that the instruments are jointly zero. In addition, the test of over-identifying
restrictions fails to reject the null that at least one of the instruments is valid. Finally, the
Hausman test rejects the null of no endogeneity problems for the changes in foreign
ownership of banks; however, this test fails to reject the null of no endogeneity problems
for the changes in domestic blockholder ownership of banks.
Rather than dismiss
potential endogeneity concerns for the changes in DB, I will use instruments for both
measures of bank ownership structure in the remaining analyses.
Panels B & C of Table 12 show the results from Equation 14 regressions, where
the changes in bank ownership structure variables have been standardized to have unit
variance for ease of interpretation and comparison. Panel B shows the results using actual
values for the changes in bank ownership structure between 1995 and 2000, while Panel
C addresses endogeneity concerns using instrumental variables for changes in foreign and
domestic blockholder ownership of banks.
After controlling for financial development and legal protection of minority
shareholders, increases in domestic blockholder ownership of banks (DB) have a negative
impact on capital allocation. The results show that a one standard deviation change in
38
Given that these are cross-sectional regressions for 49 countries, the explanatory power of the
instruments appears reasonable.
49
domestic (foreign) blockholder ownership of banks results in a 0.07 (0.11) percent
reduction (increase) in ηc, which corresponds to 23% (36%) of ηc’s standard deviation,
an economically significant impact.39 As another way to interpret the impact of changes
in DB on capital allocation efficiency, a one percent increase in DB results in a 0.41
percent reduction in ηc.40 On average, DB increased by 5.5%, so, for the average country,
a 5.5% increase in domestic blockholder ownership of banks results in a 2.3 percentage
points reduction in ηc (for the average country, ηc would decline from 0.415 to 0.392).
Thus, assuming a shock to value added of 10%, countries that experienced a 5.5%
increase in DB would increase investment by 3.97% instead of the 4.15% percent
increase in investment for those countries where DB (FB) did not change.41 When
instruments are used for changes in bank ownership structure, the results remain
unchanged (Panel C).
In Panel D of Table 12 I divide the sample of countries into those with good and
poor governance using Transparency International’s Corruption Perception Index (CPI).
The results show that the negative impact of increases in DB on capital allocation
efficiency comes primarily from countries with poor governance, while foreign bank
presence has a positive impact on capital allocation efficiency in countries with good
governance.
39
The standard deviation of ηc is 0.304.
40
The standard deviation for ∆DB is 0.185.
41
Wurgler (2000) documents that a 10% shock to value added is not uncommon.
50
2.6.1. Additional Robustness Tests
Thus far, I have shown that increased domestic blockholder (foreign) presence in
the banking sector is associated with deterioration (improvement) in capital allocation
efficiency, the extent to which a country increases investment in growing industries and
decreases investment in declining industries. It could be argued that increasing lending
activities to growing industries (e.g. technology) may not necessarily be efficient. While
the capital allocation efficiency measure uses only manufacturing industries, which may
lessen such concerns, I address this issue further by focusing on whether lending is being
directed towards firms that are growing, using data on individual firms. Using this
approach will also allow me to deal with endogeneity concerns in a different manner. I
will estimate the new capital allocation measure using data from 2001-2005 to determine
how changes in ownership structure between 1995 and 2000 affect this new capital
allocation efficiency measure. Data availability issues prevent me from conducting such
an experiment using the UNIDO data.42 In this experiment, I will use firm level proxies
for gross fixed capital formation and valued added obtained from Thomson Financial’s
DataStream and WorldScope databases. The growth in real fixed assets and the annual
sales growth are used as proxies for gross fixed capital formation and value added,
respectively. Using these new proxies, I estimate the new measure of capital allocation
efficiency, ηc (new), using Equation 13 at the firm level for 46 countries with available
data. Table 11 shows descriptive statistics of this new measure. The correlation between
this new measure and the original measure of capital allocation efficiency, with both
42
There are only 17 countries with at least three years of available data to estimate the capital allocation
measure after 2000.
51
measures estimated for the period 1995-2004, is 0.48 and statistically significant at the
1% level.
Using this new measure, estimated with data from 2001 through 2005, I run the
regressions from Equation 14. The results shown in Table 13 confirm prior results.
Domestic blockholder ownership of banks adversely affects capital allocation, while
foreign ownership of banks has a positive impact on capital allocation efficiency. A one
standard deviation change in domestic blockholder (foreign) ownership of banks results
in a 0.04 (0.07) percent reduction (increase) in ηcNEW, which corresponds to 30% (53%)
of ηcNEW’s standard deviation, an economically significant impact.43 The results in Panel
B further confirm that the negative impact of domestic blockholders on capital allocation
efficiency stems primarily from countries with poor protection of minority shareholders.
2.7.
Conclusion
This essay analyzes whether the efficiency of capital allocation has been affected
by the vast changes in bank ownership structure throughout the world. Surprisingly, the
large decline in government ownership of banks by itself has not had any impact on the
efficiency of capital allocation.
The important issue is whether foreigners or large
domestic blockholders take over the stakes that are relinquished by the government.
In general, an increase in domestic blockholder presence in the banking sector
hampers capital allocation efficiency. Countries experiencing an increase in domestic
blockholder ownership of banks tend to allocate more credit to industries that are less
43
ηcNEW’s standard deviation is 0.132.
52
productive and less dependent on external financing. This is consistent with the looting
view, which argues that banks controlled by large domestic blockholders, who usually
have substantial interests in other nonfinancial firms, direct a significant portion of their
lending activities to related, yet inefficient companies. By pursuing such activities, other
firms in need of financing may be neglected and unable to obtain funding.
In contrast, increased foreign presence in the banking sector leads to
improvements in capital allocation efficiency.
Increases in FB are associated with
increased lending to more productive industries, and to those industries in need of
external financing. This finding adds support to the findings of Giannetti and Ongena
(2007), and is consistent with the view that foreign banks improve capital allocation
efficiency by mitigating the related lending problems associated with government or
domestic blockholder ownership of banks.
Most of the changes in bank ownership structure occurred in the late 1990s.
Given the short period of time that has elapsed since these changes, their full impact
cannot be determined yet. However, the preliminary evidence presented here does point
towards benefits (costs) of privatizing the banking sector to foreign (domestic)
blockholders. This evidence could serve to guide countries which have just recently
begun to privatize their banking industries (e.g. China). Only time will tell if these
countries will reap the apparent long-term benefits of increased foreign presence in their
banking industries. This of course, leaves plenty of opportunity for future research.
53
CHAPTER 3:
DOES THE GROWING PRESENCE OF LARGE DOMESTIC BLOCKHOLDERS
AROUND THE WORLD AFFECT BANK PERFORMANCE?
3.1.
Introduction
A new wave of bank privatizations over the past ten years has led to significant
changes in bank ownership structure around the world. While in countries such as
Slovakia, Bulgaria, and Romania, government ownership of banks has been replaced by
foreign ownership of banks, privatizations in other countries such as Norway, Belgium,
and Taiwan have led to an increase in large domestic blockholder ownership of banks. In
Bulgaria, for example, government ownership of banks, which decreased from 76.2% to
0.04% between 1995 and 2005, was virtually replaced by foreign ownership of banks,
which increased from 0.0% to 71.2% over the period. In contrast, Germany saw a
significant decline in government ownership of banks (from 32.7% to 16.1%) coupled
with an increase in domestic blockholder ownership of banks (from 7.2% to 33.2%).
Figure 1 shows the magnitude of these changes for a sample of countries using the
changes in the average measures of domestic blockholder (DB), government (GB), and
foreign (FB) ownership of the top ten banks.
54
The impact of increased foreign ownership on bank performance has received
most of the attention in the finance literature (Claessens et al., 2001; Demirgüç-Kunt and
Huizinga, 1999; Micco et al., 2004). In spite of their growing importance, the emerging
role of large domestic blockholders has not been adequately studied. This essay will thus
contribute to the existing literature by analyzing how large domestic blockholders affect
bank performance and by comparing the post-privatization performance of banks sold to
foreign and to domestic blockholders. In addition, I will examine how changes in the
ownership structure of the largest (top 10) banks in a country affect the rest of the
banking sector.
Previous literature on bank ownership has focused primarily on two types of
ownership: government and foreign. Overall, government ownership of banks has been
associated with weak bank performance and with negative economic outcomes. Several
studies have documented that government-owned banks have lower profitability, higher
costs, and poorer asset quality relative to private banks (Berger, Bonime, Goldberg, and
White, 2004; Berger, et al., 2005; Micco, et al., 2004).
In addition, high government
ownership of banks has been associated with slower subsequent financial development
(La Porta et al., 2002; Barth et al. 2004).
Given these findings, we should expect
improvements in bank performance post privatization.
In contrast, the evidence on foreign ownership of banks reveals that foreign banks
outperform their domestic counterparts in terms of profitability and cost efficiency,
primarily in emerging markets (Bonin, et al., 2005; Claessens, et al., 2001; Micco, et al.,
2004). Foreign bank presence also appears to improve the competitiveness of domestic
banks (Claessens, et al., 2001; Micco, et al., 2004). There is one area of ongoing debate,
55
however. While Detragiache et al. (2006) document that increased foreign presence
reduces the availability of credit to the private sector, consistent with the predictions of
their model, other studies show that foreign bank presence improves the availability of
credit (Clarke et al., 2001).
Most of the aforementioned studies compare the performance of governmentowned and foreign-owned banks to their domestic peers, without differentiating between
closely-held and widely-held domestic banks. It is important to differentiate between
these two types of banks, as large domestic blockholders may significantly affect bank
performance, just as they have been shown to affect firm value in the corporate finance
literature. There are different hypotheses that have emerged in the corporate finance
literature.
According to the incentive-based view (Shleifer and Vishny, 1997),
shareholders with large cash flow ownership have an incentive to closely monitor a
firm’s performance, potentially mitigating the principal-agent problems that exist
between managers and shareholders (Jensen and Meckling, 1976).
In line with this
view, several studies have shown a positive correlation between firm value and cash flow
ownership of large shareholders (Claessens, et al., 2002; La Porta, et al., 2002b).
In
contrast, large blockholders may negatively affect firm performance if they pursue their
own interests at the expense of other minority shareholders (Shleifer and Vishny, 1997).
Consistent with this entrenchment-based view, formalized by Stulz (1988), evidence has
shown that firm value falls when control rights exceed cash flow rights of large
shareholders (Claessens, et al. 2002).
This essay finds evidence consistent with the entrenchment-based view. I find
that increased domestic blockholder ownership of banks is associated with poor
56
subsequent performance. In particular, I find that banks changing from government to
domestic blockholder control perform poorly in terms of asset quality and profitability.
In addition, these changes in control have a detrimental impact on bank value. This
evidence suggests that domestic blockholders may be acquiring stakes in banks to extract
private benefits of control, which results in poor bank performance. These findings are
also consistent with the looting view (La Porta, et al., 2003), which argues that banks
controlled by domestic blockholders, who typically have substantial stakes in other
nonfinancial firms, tend to direct a significant portion of their lending to related
companies, even when these are inefficient. Consistent with prior findings, I also find a
positive impact of increases in foreign ownership of banks on profitability and
operational efficiency.
Finally, I confirm prior findings associated with the poor
performance of government-owned banks. Furthermore, increases in domestic
blockholder ownership of banks appear to improve the asset quality and profitability of
the banking sector. Contrary to prior findings, foreign presence does not improve the
competitiveness of the domestic banking sector.
The rest of the chapter is organized as follows. Section 3.2 reviews the existing
literature on bank ownership and performance. Section 3.3 discusses the data and the
methodology used in the study. Section 3.4 explores the relationship between changes in
bank ownership and subsequent performance. Section 3.5 explores the impact of changes
in bank ownership structure on the banking sector as a whole, and Section 3.6 concludes.
57
3.2.
Literature Review
3.2.1. Empirical Evidence on State Ownership of Banks
Government ownership of banks has been associated with poor performance.
Berger, Clarke, Cull, Klapper, and Udell (2005) use data from Argentina in the 1990s to
analyze the static, selection, and dynamic effects of domestic, foreign, and state
ownership on bank performance. They find that state-owned banks have poor long-term
performance and that those banks undergoing privatization have poor performance
beforehand, and dramatically improve their performance after privatization. Micco et al.
(2004) examine the relationship between bank ownership and bank performance for
banks in 119 countries. They find that in developing countries, state-owned banks have
lower profitability, higher costs, higher employment ratios, and poorer asset quality than
their domestic counterparts.
With the exception of state-owned banks having higher
costs than their domestic counterparts, they do not find evidence of significant
differences between state and domestic private banks’ performance in industrial
countries. Cornett, Guo, Khaksari, and Tehranian (2003) examine the differences in
performance between state-owned and private banks in 16 Far East countries between
1989 and 1998. They also find that state-owned banks are significantly less profitable,
have lower capital ratios, greater credit risk, lower liquidity, and lower management
efficiency.
More recently, Mian (2006b) explores the differences between foreign,
private domestic, and government banks in 100 emerging markets, and finds that
58
government banks perform uniformly poorly, and can only survive due to government
support.44
The problems associated with government ownership of banks extend beyond the
negative impact on bank performance. Several papers have documented that government
ownership of banks inhibits financial development and economic growth (Barth, et al.,
2004; Galindo and Micco, 2004; La Porta, et al., 2002a). By pursuing political, rather
than socially and economically optimal objectives, a large presence of government banks
can hamper economic efficiency. Consistent with this political view, several papers
document that government ownership of banks inhibits financial development and
economic growth (Barth, et al., 2004; Galindo and Micco, 2004; La Porta, et al., 2002a).
La Porta et al. (2002a) show that higher government ownership of banks in 1970 is
associated with slower subsequent financial development and lower economic growth.
Barth et al. (2004) examine the relationship between state ownership and banking sector
development measures. They find that government ownership of banks is negatively
related to favorable banking outcomes, and positively related with corruption. Micco,
Panizza, and Yañez (2006) and Sapienza (2004) provide further support for the political
view.
Micco et al. (2006) find that the difference in public and private banks’
performance (e.g. lending activity) widens during election years, supporting the
hypothesis that political considerations drive these results. Sapienza (2004) finds that
lending behavior of state-owned banks in Italy is affected by electoral results of the party
affiliated with the bank. In addition, Dinç (2005) shows that government-owned banks in
44
His paper does not differentiate between closely-held and widely-held domestic banks, however.
59
emerging markets significantly increase their lending in election years relative to private
banks. The author interprets this as evidence that politicians can reward their allies and
punish their opponents through their influence on government-owned banks. Megginson
(2005) has a more complete review of this literature.
3.2.2. Empirical Evidence on Foreign Ownership of Banks
While government ownership of banks is associated with inferior performance
and negative economic outcomes, foreign ownership of banks has been generally
associated with superior performance (e.g. profitability, operational efficiency), primarily
in emerging markets.
Claessens et al. (2001) document that foreign banks are more
profitable than their domestic counterparts in developing countries, but the opposite is
true in developed markets. Demirgüç-Kunt and Huizinga (1999) study banks in 80
countries over the 1988-1995 period and find that foreign banks have higher margins and
profits than domestic banks in developing countries, but the opposite is true for industrial
countries. Micco et al. (2004) also document that foreign banks have higher profitability,
lower costs, and lower employment ratios than their domestic counterparts in developing
countries; however, foreign banks have higher non-performing loans than their domestic
counterparts in developing countries. Bonin et al. (2005) examine bank performance in
six Eastern European transition economies and find that foreign banks are more efficient
in terms of cost and profit than domestic and state-controlled banks. They also find
support for the importance of privatizing banks by selling them to strategic foreign
investors. Banks privatized in such a manner are more cost and profit efficient than stateowned banks. Majnoni, Shankar, and Varhegyi (2003) study the dynamics of foreign
60
bank ownership in Hungary between 1994 and 2000 and find that foreign banks, while
pursuing similar lending policies, achieve greater profitability than their domestic
counterparts. More recently, Mian (2006b) documents that one of the main advantages of
foreign banks in emerging markets, which may explain their superior performance, is
their ability to tap into external liquidity through their parent bank, which lowers their
cost of funds.
Several studies document the impact that foreign banks have on domestic banks.
Micco et al. (2004) find that foreign bank presence is associated with increased
competitiveness of the domestic banks (lower margins and lower overhead costs).
Claessens, et al. (2001) show that foreign bank entry diminishes the profitability of
domestic banks and reduces their non-interest income and overall expenses. When other
factors are controlled for, high profits reflect a lack of competition, while high overhead
costs reflect a lack of efficiency. They argue that their findings are consistent with
foreign banks improving the efficiency of domestic banks. Unite and Sullivan (2003)
study how foreign bank entry and foreign ownership of banks affect those in the
Philippines. They show that foreign bank entry and market penetration reduces interest
spreads and operating expenses of domestic banks, making them more efficient. Barajas,
Salazar, and Steiner (2000) show that foreign entry appeared to improve the efficiency of
Colombian domestic banks by reducing nonfinancial costs. Finally, Clarke, Cull, and
Martínez-Peria (2001) find that foreign bank market penetration improves access to
credit. Enterprises in countries with larger foreign presence rate interest rates and access
to long-term loans as smaller constraints on operations and growth than do enterprises in
countries with less foreign presence.
61
More recently, Detragiache et al. (2006) develop a model that predicts that credit
to the private sector should be lower in countries with more foreign bank penetration.
They find support for their model’s predictions using a sample of 89 low and lower
middle income countries.
argument.
Their results are explained by the “cream-skimming”
Foreign banks are better than domestic banks at monitoring “hard”
information (e.g. accounting information, collateral value), but have a disadvantage in
monitoring “soft” information (e.g. entrepreneurial ability). This leads foreign banks to
lend to safer and more transparent customers (Berger, et al., 2001; Mian, 2006a). Once
these hard information customers are separated from the pool of borrowers, the remaining
soft information borrowers are left in a worse pool, which causes them to either pay
higher interest on their loans, or not borrow at all. This leads to an overall reduction in
credit to the private sector.
3.2.3. Contribution of this Study
This essay examines how changes in the overall bank ownership structure affect
individual bank performance and the overall banking sector. In particular, I focus on the
impact of the emergence of large domestic blockholder ownership on bank performance.
As described in the previous section, the literature on bank ownership focuses
primarily on either state ownership or foreign ownership of banks. Only three of the
papers discussed above look at domestic ownership of banks, and none of them
distinguish between closely-held and widely-held domestic banks.45 Berger, et al. (2005)
examine the effects of domestic, foreign, and state ownership on bank performance.
45
Caprio, Laeven, and Levine (2003) explore how large shareholders affect bank valuation. Using bank
ownership data from 244 banks across 44 countries, they document that larger cash flow rights by
controlling shareholders boost valuation.
62
However, their paper is limited in scope, as it is a case study of Argentine banks, and it
does not differentiate between domestic banks’ ownership structures. This essay expands
on Berger’s study by using a more comprehensive data set covering 90 countries. Micco
et al. (2004) take a more comprehensive look at the role of bank ownership on
performance. While they study the role of state, domestic, and foreign ownership of
banks on bank performance, they also do not account for differences in domestic banks’
ownership structure (closely-held and widely-held), and thus fail to explore the role
played by large domestic shareholders.
Finally, Mian (2006b) examines how
organizational design shapes banks’ behavior. He finds that domestic banks have an
advantage in lending to “soft” information firms, which allows them to lend more, while
foreign banks have lower cost of deposits because of their ability to draw support from
their parent banks. The role of large domestic blockholders is not examined in that paper,
either. This essay expands on these studies by analyzing how changes in the overall bank
ownership structure (including the presence of large domestic blockholders) affect bank
performance, and by exploring the spillover effects of these changes in bank ownership
structure on the banking sector.
3.3.
Data and Methodology
I will analyze the role of bank ownership on bank performance using a hand-
collected database on the ownership structure of the top 10 banks in 90 countries as of
1995. The performance of these banks over the 1998-2004 period will be examined.
Ownership data was collected for 1995, 2000, and 2005. To determine the top 10 banks
63
(commercial and development banks) from each country, I use Bureau Van Dijk’s
Bankscope and Accuity’s The Global Banking Resource (TGBR).46 If these sources do
not identify ten banks for a country, the list is expanded using various other sources
including The Bankers’ Almanac, and Thomson Bank Directory.
Ownership and
financial information is obtained primarily from BankScope and TGBR, and
supplemented with information from The Bankers’ Almanac, Thomson Bank Directory,
central bank websites, and individual bank’s websites and annual reports.
Given the consolidations in the banking industry of several countries as well as
data availability issues, some countries in the sample have less than 10 banks. To be
included in the sample, a country should have at least three banks with available
ownership information. While the initial sample consists of 798 banks, due to mergers
and acquisitions and data availability issues over time, the final sample drops to 628
banks.
To construct the ownership variables, I follow La Porta et al.’s (2002a)
methodology.47 The first measure identifies large domestic blockholder ownership of
banks. A large domestic blockholder is any domestic shareholder (a company, or an
individual) owning more than 10 percent of the shares in a bank. I use the 10 percent
threshold to constitute a block following La Porta et al. (1999).48 I first calculate the total
share of each bank that is owned by large domestic blockholders (DBik) as follows:
46
I thank Jose F. Alvarez, and Banco Sabadell’s management for providing access to the data, and Mitch
Gouss at Bureau van Dijk for providing access to BankScope during a trial period.
47
While La Porta et al. (2002) constructed government ownership measures, I construct domestic, and
foreign ownership measures following their methodology.
48
As robustness tests, I also construct measures of domestic blockholder ownership of banks using a 5%,
and a 3% threshold to define a large domestic blockholder. Results are essentially unchanged when these
alternate measures are used.
64
J
DB i k =
∑s
where s j d , i > 0 .1 .
j d ,i ;
j =1
(15)
where sjd,i is the share of bank i, owned by shareholder j (a domestic shareholder); DBik is
the total share of bank i in country k that is owned by large domestic blockholders.
A similar approach is used to construct the government (GBik), and foreign (FBik)
ownership measures:
J
GBik =
∑
J
s ji sgj ;
j =1
FBik =
∑s s
j=1
ji j f
, wheres ji s jf > 0.1;
(16)
where k=1...90 indexes the 90 countries in the sample, j=1...J indexes the banks’
shareholders; i=1…10 indexes the 10 largest banks in the country; sji is the share of bank
i that is owned by shareholder j; sgj is the share of shareholder j that is owned by the
government; sjf is the share of shareholder j that is owned by foreigners; GBik is the total
government’s share in bank i in country k, and FBik is the total share in bank i in country
k that is owned by foreign blockholders.49 The shares of a bank that are widely-held are
computed as follows:
WIDEik= 1- DBik-GBik-FBik
(17)
To obtain the total domestic, government, and foreign blockholders’ stake in the
top ten banks in country k, each of the individual bank ownership measures (DBik, GBik,
FBik) is multiplied by the bank’s total assets (TA); the resulting number is summed across
49
The BankScope database provides ownership information for financial institutions only. If a bank’s
shareholder was a nonfinancial institution, various sources were used to determine that shareholder’s
ownership structure. These sources included Mergent Online, as well as company websites.
65
all banks in a country and divided by the total assets of the top 10 banks. The country
level ownership measures are thus:
10
DB k =
∑ DB
i =1
10
ik TA ik
; GB k =
10
∑ TA
∑ GB
i =1
10
ik TA ik
∑ TA
ik
i =1
; FB k =
10
i =1
ik
∑ FB
i =1
ik TA ik
(18)
10
∑ TA
ik
i =1
All of these ownership measures are constructed with data from 1995, 2000, and
2005. Table 2 shows descriptive statistics of these measures for the 90 countries in my
sample. The ownership data does not allow me to identify the ultimate owners, given
that I do not have data on multiple classes of shares or any pyramidal structures.
However, the data does allow me to determine whether a bank’s largest shareholder is
either: (1) the state, (2) foreigners, (3) or a large domestic blockholder, where a large
domestic blockholder could be an individual, a family, or a local company. Appendix C
provides a detailed example of the calculation of these ownership measures for three
banks. To construct the individual bank’s ownership measures, each of the bank’s
shareholders is classified by type (e.g. individual, government). As an example, for the
Slovenian bank, Abanka Vipa, there is only one large blockholder, Zavarovalnica
Triglav, dd, a Slovenian insurance company, which owns 32.92% of its shares. As such,
DBAbanka is 32.92%. There are no foreign or government shareholders for Abanka, so
GBAbanka and FBAbanka equal zero.
As a proxy for the banking sector performance measures, I compute average
annual performance measures for all banks covered by BankScope for each country.50 As
50
All banks that are primarily in the business of making commercial loans are included. This includes
commercial banks, cooperative banks, and specialized government credit institutions, but excludes
investment banks, and non-bank credit institutions.
66
robustness tests, in unreported results, data on banking system performance from 19952005 obtained from IMF Global Financial Stability Reports (2006, 2005, and 2003) are
used and produce similar results.
3.3.1. Bank Performance Measures
I will explore how bank ownership structure affects bank’s asset quality,
profitability, and operational efficiency, using various widely-used measures of bank
performance. The asset quality measures are the ratio of non-performing loans to gross
loans (NPL-GL), which reveals the extent of problem loans in the bank’s portfolio, and
the ratio of loan-loss reserves-to-non-performing loans (LLR-NPL), which measures how
well the bank is provisioning to cover potential losses stemming from bad loans. The
return on average equity (ROAE) and the return on average assets (ROAA) are used to
determine banks’ profitability. Banks’ operational efficiency is measured using the net
interest margin (NIM, or, the interest income minus interest expense as a fraction of
average earning assets) and the cost-to-income ratio.
To further examine the relation between performance and ownership, I will also
use data from DataStream and WorldScope to construct Tobin’s Q (the ratio of market
value of a firm divided by the replacement value of its assets) measures for the subset of
banks that are publicly traded. The proxy for Tobin’s Q measure will be the market-tobook value ratio.
67
3.4.
Bank Ownership and Bank Performance
Table 14 shows descriptive statistics of the sample of banks with available data by
year. Coverage varies widely by year, depending on the performance measure used.51
The BankScope database provides eight years of historical financial data. There are thus
a very limited number of banks with available data prior to 1998 as well as for 2005.
Because of these data constraints, the sample period used in the following analyses runs
from 1998 through 2004.
Panel B of Table 14 shows some unconditional average
performance measures for the full sample. In terms of asset quality, over the 1998-2004
period, the banks in the sample have an average non-performing loans-to-gross loans
ratio of 7.9%, while on average, their loan loss reserves cover 97.0% of non-performing
loans. The average ROA and ROE are 0.8% and 9.6%, respectively. Net interest margin
and the cost-to-income ratios are 3.62% and 60.1%, respectively, while the average
Tobin’s Q measure is 1.68. Panel C splits the sample into emerging and developed
countries. Not surprisingly, banks in emerging markets have poorer asset quality (higher
nonperforming loans and lower coverage of problem loans), are less efficient (higher
operational costs), and have lower Tobin’s Qs than banks in developed countries.
However, banks in emerging markets appear to be more profitable than their counterparts
in developed countries. This latter result could be explained by the fact that banks in
emerging markets may have to charge higher rates on their loans due to the higher
riskiness of such loans in those countries; the higher net interest margins for banks in
emerging markets add support to this argument.
51
There is a more drastic reduction in sample size when the asset quality measures are used. In addition,
because many of the banks are private, Tobin’s Q measures are only available for a small number of banks
with publicly traded shares that are covered by DataStream and WorldScope.
68
3.4.1. Changes in Bank Ownership and Bank Performance
Given the vast changes in bank ownership structure that occurred primarily over
the second half of the 1990s, I will explore how these changes affect future performance.
Given the well documented poor performance of government-owned banks, coupled with
the decline in government ownership of banks over the period, the question of interest is
how bank performance is affected by changes in control (e.g. from government control to
foreign control). To explore this issue, I identify banks that were under government
control as of the beginning of the period (1995) and changed to domestic or foreign
blockholder control as of 2000, using the 20% threshold to identify control.52 I will run
pooled regressions, with standard errors clustered by bank to address any bias in the
standard error estimates due to autocorrelation in the various performance measures. In
addition, to correct for cross correlation in the performance measures across banks in a
given year, year dummy variables are included in all regressions. Given the limited
number of years in the analysis (4), this procedure yields virtually identical results as
clustering the standard errors by both firm and time (Petersen, 2007). The following
regression framework will be employed to identify significant changes in bank
performance associated with changes in control.
PERFict = α + β1DOMi + β 2GOVi + β3 FORi + β 4 ∆D + δ c Controls+ γ t + ε ic t
(19)
where PERFi,t refers to the various annual performance measures from 2001 through
2004; DOM, GOV, FOR are dummies for domestic, government, and foreign ownership
of banks, using the 20% threshold; ∆D refers to dummies for changes in control
(specifically, GD and GF are dummy variables which equal one if a bank changed from
52
In unreported results, I use other thresholds (10%, 50%) to determine control.
69
government control as of 1995 to domestic blockholder or foreign control as of 2000,
respectively, using the 20% threshold to define control), and the controls include the lag
of assets (log), the lag of non-interest-income-to-total assets, the lag of market share
(MS), a dummy variable for countries experiencing a systemic banking crisis during the
year, a dummy variable for restrictions on foreign ownership of banks, and a dummy
identifying a country with common law origin of their commercial laws.
53
Finally, γt
represents year fixed effects. Regional dummies are included in all regressions to control
for other factors that may affect bank performance.
There are obvious selection bias and endogeneity concerns associated with the
regressions in Equation 19.
For example, government banks sold to domestic
blockholders (captured by the GD dummy) may differ substantially from other banks.
These differences may not be captured in the limited set of controls included in the
regressions. The coefficients from the OLS regressions will be biased if the different
characteristics of these banks are related to bank performance (LHS variable).
In an attempt to address these concerns, I will use the Heckman model (1979) to
correct the potential bias created by the sample selection problems. The first stage
equation (selection equation) is a probit model to estimate the probability that a bank
changes from government to domestic (or foreign) blockholder control; the dependent
variables in these selection equations are the GD, and GF dummy variables, which equal
one if the bank changed from government to domestic or foreign blockholder control,
respectively, and 0 otherwise. The second-stage equation is similar to Equation 19, but it
53
Ownership data is available for 1995, 2000, and 2005. In these regressions, the 1995 ownership dummy
variable is used for the years 1995-2000, and the ownership dummy variable for 2000 is used in subsequent
years.
70
also includes the variable λ (the inverse Mills ratio), which is generated from the firststage equation, and attempts to correct for the selection bias.
An important condition for the use of the Heckman procedure is that the selection
equation must contain at least one instrument that is not related to the dependent variable
in the second stage (in this case, the bank performance measures). I will use the level of
government ownership of banks as of 1995, Kaufmann et al. (2006) corruption index as
of the beginning of the period (1996), and a proxy for the size of the productive
population (population ages 15-64 as a fraction of the total population) as instruments. In
theory, corruption level in a country, the level of government ownership of banks, and the
size of the productive population as of the beginning of the period should not affect
individual bank performance.
Yet, domestic blockholders may be more inclined to
acquire stakes in banks in countries where there is more corruption, particularly if these
blockholders are acquiring stakes in banks to extract private benefits of control. In
addition, foreigners may be more inclined to acquire stakes in banks in countries with a
larger customer base (larger productive population). Finally, privatizations are more
likely to occur in countries with high government ownership of banks as of the beginning
of the period.
The results from the first-stage probit model are shown in Panel A of Table 15.
The results show that bank privatizations involving domestic blockholders are more
likely in more corrupt countries, while privatizations involving foreign blockholders are
more likely in countries with a larger productive population and in those with a large
government presence in the banking sector as of the beginning of the period. Panels B
and C show the results from the second stage regressions. Even after controlling for
71
potential selection bias, albeit not perfectly, the results indicate that changes in control
from government to domestic blockholders are associated with deterioration in asset
quality and profitability.
These results are not only statistically, but also economically
significant. All else equal, the results show that banks changing from government to
domestic blockholder control over the 1995-2000 period have higher (by 0.21) nonperforming loans-to-gross loans (NPL-to-GL) ratio. The average NPL-to-GL ratio over
the period is 9.1%; this translates into a 30.1% ratio for banks changing to domestic
blockholder control over the period. The average loan portfolio of the banks in the
sample is $25.3 million, so on average, banks changing to domestic blockholder control
over the period have an extra $5.3 million in nonperforming loans.
The effect on
profitability is also economically significant. The average ROAE, 8.9%, would drop to 20.9% for banks in which domestic blockholders took control. The results in Panel C
reveal that banks in which domestic blockholders assumed control operate with higher
loan loss reserves in emerging markets. This could be a result of the poor asset quality of
banks operating in emerging markets; more reserves are needed to cover the potentially
higher losses from lending activity in emerging markets. Other than the difference in
loan loss reserves, there is no additional difference in performance between banks in
emerging and developed countries.
The results in Table 15 also add support to prior findings. Government ownership
of banks is associated with poor performance across the board. In particular, government
banks have poor asset quality (higher non-performing loans), lower profitability (ROAE),
and had higher costs than their peers, although these banks also operate with higher levels
of reserves in emerging markets, as Panel C shows. Finally, the results also present some
72
evidence pointing towards benefits of involving foreigners in the privatization process.
Banks changing from government to foreign control improve profitability (ROAE) and
operational efficiency (NIM). These findings are in line with the well documented
superior performance of foreign banks relative to their peers. Lower funding costs for
foreign banks, through their ability to tap into external liquidity through their parent
banks, can help explain the improvements in net interest margins experienced by
government banks acquired by foreigners. The results show that on average, for banks
privatized and later sold to foreigners the return on equity would increase by 8.7
percentage points (to 16.8%), while the net interest margin would increase to 5.5%.54
Previous studies document significant differences in the performance of foreign banks
between emerging and developed markets. Consistent with these findings, the results in
Panel C show that the improvements in profitability (ROAE), asset quality (higher loan
loss reserves), and operational efficiency (lowers costs) associated with foreign
ownership of banks occurs primarily in emerging markets.
The results in this section provide evidence consistent with a negative (positive)
impact on subsequent bank performance for banks changing from government to
domestic (foreign) blockholders. The negative impact associated with increases in
domestic blockholder ownership of banks is consistent with the entrenchment-based
view. The evidence shows that large domestic blockholders may be pursuing selfish
interests at the expense of minority shareholders, resulting in the substantial deterioration
in bank performance. In addition, these results are also in line with the looting view. The
54
The average ROAE and NIM for the banks in the sample are 8.1%, and 4%, respectively.
73
negative impact on asset quality and profitability could be a result of detrimental related
lending practices of these banks.
The results presented here rely on an arbitrarily chosen benchmark (20%) to
measure changes control. As robustness tests, in unreported results, I have also run the
regressions using both a 10% and a 50% threshold. Using the 10% threshold, the results
are fairly similar, although the statistical significance of the results does drop. Domestic
blockholders continue to have a negative and significant impact on bank’s profitability
(ROAE), but the impact on asset quality, while still of the same sign, is no longer
significant. Arguably, an explanation for the weaker results obtained using this alternate
threshold is the fact that 10% might be too low of a threshold to establish control. The
results using the 50% threshold are similar to the ones presented here. The statistical
significance of the results does decline, however. Using the 50% threshold creates a
problem, however, because there are only a handful of banks changing control over the
period using that benchmark.
Although the sample selection bias is an issue, in unreported results I also
replicate the results in this section using conventional OLS. Those results also confirm
the negative (positive) impact of increases in domestic (foreign) ownership of banks on
bank performance. Next, I examine whether these changes in control affect bank value.
3.4.2. Bank Ownership and Bank Performance – Q measure of performance
I now turn to analyze how changes in ownership affect the value of the banks,
using market-to-book value as a proxy for Tobin’s Q. For this analysis, the sample size is
reduced significantly (see Table 14) given that the banks need to be publicly traded and
74
covered by DataStream and WorldScope. The impact of changes in bank ownership on
bank value will be explored using the following regression framework:
Qit = α1 DOM + α 2 GOV + α 3 FOR + α 5 ∆D + α 6 Assets i ,t −1 + α 7
NII i ,t −1
TAi ,t −1
α 8 Qi ,t −1 + α 9 CRISIS + α 10 FR + α 11COM + α 12 GDPGrowth t −1 + γ t + ε it
+
(20)
where Qi,t refers to bank i’s Tobin’s Q measure; DOM, GOV, and FOR are dummies for
domestic, government, and foreign ownership of banks, using the 20% threshold; ∆D
refers to dummies for changes in control (from government to domestic or foreign
control, respectively), and the controls include the lag of assets (log) and the lag of noninterest-income-to-total assets (to differentiate retail commercial banks from those large
banks deriving their income primarily from investment banking activities), and the
growth in real GDP per capita. CRISIS is a dummy variable which equals one if the
country experienced a systemic banking crisis during the year; FR is a dummy variable
which equals one if the country has any restrictions on foreign ownership of banks; COM
is a dummy for countries with common law origin of their commercial laws, and γt
represents year fixed effects.
To address sample selection bias issues, I will again use the Heckman (1979) twostage procedure, in which I will include the variable λ, estimated from a first-stage
selection equation (see Panel A of Table 15) to the regressions in Equation 20. The
results from the second stage regressions are shown in Table 16. Consistent with the
results presented earlier, Panel A shows that changes in control to domestic blockholders
have a negative impact on bank value. In particular, banks changing from government to
domestic blockholder control have significantly lower Tobin’s Qs.
75
In fact, for the
average bank, whose Tobin’s Q is 1.63, its Q would become negative (-0.69) if domestic
blockholders assumed control over the period.
This result is both statistically and
economically significant. Panel B shows that this negative impact on bank value brought
about by changes to domestic blockholders comes primarily from banks in developed
markets.
In contrast, changes in control to foreign banks appear to have a positive impact
on bank value, although the statistical significance of these results is lower. Panel B also
reveals that the positive impact of foreign ownership on bank value is much lower in
emerging markets. This finding contradicts some of the previous findings documenting
improved performance of foreign banks relative to their peers in emerging markets
(Claessens, et al., 2001; Micco, et al., 2004). It should be noted, however, that those
studies do not examine Tobin’s Q as a measure of bank performance.
As robustness tests, to verify the somewhat arbitrary selection of the 20%
threshold as signifying change in control, I run the regressions using both the 10% and
the 50% thresholds, and the results are similar, although the statistical significance of the
results diminishes.
In addition, I also run regular pooled OLS regressions, without
controlling for sample selection bias, and obtain similar results.
While the results in this section cover only a small subsample of banks, they
confirm prior results. Bank privatizations involving domestic (foreign) blockholders
appear to be associated with a negative (positive) impact on bank performance and bank
value. Having analyzed the impact of bank ownership on individual bank performance,
the next section examines how the broad changes in bank ownership structure of the
largest banks in a country affect the performance of the rest of the banking sector.
76
3.5.
Spillover Effects of Changes in Bank Ownership Structure
This section explores how the rest of the banking sector is affected by the changes
in the bank ownership structure of the top ten banks in a country. Prior studies have
shown that increased foreign presence improves domestic banks’ efficiency and
competitiveness (Claessens, et al., 2001; Micco, et al., 2004). Given the vast changes in
bank ownership structure over the past ten years, it is interesting to explore how these
changes have affected the banking sector’s performance. This is especially interesting
given the fact that these changes in bank ownership structure have not been
homogeneous.
For example, foreign ownership has virtually replaced government
ownership of banks in several former socialist countries including Bulgaria, Romania,
and Poland, while domestic blockholders have increased their ownership stakes in
countries such as Belgium, Norway, and Germany (Figure 1).
To explore the impact of changes in bank ownership structure on the banking
sector’s performance, average banking sector performance measures are calculated for
each country, using financial information for all banks covered by the BankScope
database each year, excluding the top ten banks in each country (i.e. those banks whose
ownership information was used to construct the aggregate country-level bank ownership
structure measures).55 The following pooled regressions are used:
PERFk ,t = α + β1∆OWNERSHIP
k + β 2 ASSETSt −1 + β 3 INFt −1 + β 4 GDPt −1 +
β 5CRISIS+ β 6 FR + β 7 COMMON+ γ t + ε t
55
(21)
This includes all banks whose primary activity is to provide credit. It includes all commercial,
cooperative banks, and specialized government credit institutions, but excludes investment banks, and nonbank credit institutions.
77
where PERFk,t represents the various banking system performance measures for country
k; ∆OWNERSHIP represents the changes in domestic, government, and foreign
blockholder ownership of the top 10 banks in country k between 1995 and 2000; ASSETS
is the log of banking system assets (log); INF is the inflation rate; GDP is the real GDP per
capita; CRISIS is a dummy variable which equals one if the country experienced a
systemic banking crisis during the year; FR is a dummy for foreign restrictions on bank
ownership, and COMMON is a dummy which equals one if the country is a common law
country.56 In addition, regional dummies (e.g. North America, Western Europe) are used
to control for other factors that may explain bank performance. Finally, γt represents year
fixed effects, and standard errors are clustered by country to address potential
autocorrelation of banking system performance measures. As robustness tests, in
unreported results the Fama-MacBeth approach is used and yields similar results.
There are obvious endogeneity concerns associated with the regressions in
Equation 21. For example, improving asset quality or profitability of the banking system
could be attractive to domestic or foreign blockholders, and this in turn could result in the
observed increased domestic (or foreign) blockholder ownership of banks. Thus, the
improvements in asset quality for the banking system may not be caused by changes in
the ownership structure of banks.
To address these issues, in this section I will use a Two-Stage Least Squares
regression approach, using instruments as of the beginning of the period (1995) to
forecast changes in domestic and foreign blockholder ownership of banks.
56
Bank
The data used to construct the foreign restriction dummy was obtained from Ross Levine’s website. The
data comes from Barth, Caprio and Levine (2005). The foreign restriction dummy equals one if a country
restricts foreign entry through an acquisition, a subsidiary, or a branch.
78
concentration (assets of the top 3 banks divided by the assets of all commercial banks), a
measure of dispersed ownership of banks (WIDE), and Kaufmann et al. (2006) regulatory
quality index as of 1996 are used to forecast changes in DB. Domestic blockholders may
be reluctant to acquire stakes in banks in countries with a high concentration of banking
system assets, and it should be easier for them to increase their stakes in banks in
countries where dispersed ownership of banks is prevalent. Further, if these domestic
blokcholders are indeed pursuing ways to extract private benefits of control, domestic
blockholders are more likely to increase their stakes in banks in countries where the rule
of law is not well established. However, these measures should not directly affect the
performance of the banking sector. The instruments used to forecast changes in FB
include a proxy for the size of the productive population (population ages 15-64 as a
percent of the total population) and the Kaufmann et al. (2006) political stability index as
of the beginning of the period. Countries that are politically stable and have potentially
larger work forces could attract foreign banks (e.g. larger potential customer base); yet,
these measures should not directly affect banking sector performance.
Table 17 shows the results of the regressions from Equation 21 using instruments
for changes in DB and FB. The total sample size is reduced to 70 countries because of
data availability.57 Panel A shows the results from the first-stage regressions. Following
Larcker and Rusticus (2007), I report the results of several tests that address the validity
of the instruments and that examine whether the IV methods are indeed preferred to
traditional ordinary least squares procedures.
57
The results show that the respective
For some countries, the top ten banks in the country represent 100 percent of the commercial banks that
are covered by BankScope. Such countries include Israel, Saudi Arabia, South Africa, Sri Lanka.
Tanzania, Trinidad and Tobago, United Arab Emirates, Zimbabwe, Algeria, Honduras, Iran, Iraq, Ivory
Coast, Kuwait, Libya, Macau, Oman, Qatar, Senegal, and Venezuela.
79
instruments work well in forecasting changes in domestic (DB) and foreign (FB)
ownership of banks (the partial R2 is 0.20 and 0.21 for the regressions of changes in DB
and FB, respectively). The partial F-test rejects the hypothesis that the instruments are
jointly zero. In addition, the test of over-identifying restrictions fails to reject the null
that at least one of the instruments is valid. Finally, the Hausman test fails to reject the
null of no endogeneity problems for the changes in domestic and foreign blockholder
ownership of banks. The latter results alleviate some of the concerns about potential
endogeneity issues.
Panels B and C of Table 17 show the results from the second stage regressions.
The results show that, even after addressing endogeneity concerns using the IV approach,
increases in DB are associated with improvements in asset quality (lower non-performing
loans, and higher loan loss reserves), and improved profitability (higher ROA) of the
banking sector. A country experiencing a 1% increase in DB will have a 5.4 percentage
point decrease in non-performing loans-to-gross loans ratio (from 9.2% to 3.8%), a
significant reduction. These results add further support to the looting view. Given that
domestic blockholders engage in less than optimal banking practices, increases in DB
will not improve the competitiveness of other banks.
These banks will still find
profitable opportunities when domestic blockholder presence increases. The results from
Panel C show that the positive impact on banking sector asset quality associated with
increases in DB is significantly lower in emerging markets, but the impact on profitability
comes primarily from emerging markets.
The results also show that increased foreign presence does not appear to improve
the competitiveness of the banking sector. Increases in FB have a positive impact on
80
banking sector net interest margins.
These results are consistent with the cream
skimming argument. If foreign banks lend only to the best hard information borrowers,
the remaining banks are left with a worse (riskier) pool of borrowers. To compensate for
the increased riskiness of the borrower pool, the remaining banks may be forced to
charge higher interest rates, which have a positive impact on net interest margins.
Finally, increases in government ownership of banks are associated with improvements in
asset quality (higher loan loss reserves) and operational efficiency (lower costs) of the
banking sector. The latter results should be interpreted with caution. One way in which
governments can increase their stakes in banks is by taking over problem banks. In doing
so, the performance of the rest of the banking sector could be artificially improved with
increased government presence in the banking sector (e.g. the problem loans are taken
out). As such, the remaining banks face stiffer competition (i.e. the weaker banks are
taken out), which could result in the improvements in efficiency.
Acknowledging that the use of the IV approach is not a perfect solution to the
endogeneity concerns, as robustness tests, in unreported results, I also run the regressions
in Equation 21 using the conventional OLS. The results tend to confirm the results
presented here. There is positive spillover effect on banking sector asset quality and
profitability associated with increases in DB, and a negative impact on the
competitiveness of the banking sector associated with increases in FB.
As has been shown in this section, there is a positive spillover effect associated
with increases in large domestic blockholder ownership of banks. Other banks continue
to enjoy higher profits and better asset quality when there is an increase in domestic
blockholder ownership of banks. In contrast, increases in foreign bank presence affect
81
the competitiveness of the banking sector adversely. Finally, increased government bank
presence appears to improve banking sector’s asset quality, and competitiveness.
This
could signify that governments may be able to play a positive role in the banking sector,
at least by bailing out failing banks.
3.6.
Conclusion
Large scale bank privatizations over the past ten years have vastly changed bank
ownership structure around the world. This essay examines whether these changes in
ownership affect individual bank performance, incorporating the role of large domestic
shareholders into the analysis. In addition, the spillover effects of these changes on the
banking sector’s performance are analyzed.
Large domestic blockholder ownership of banks has a negative impact on bank
performance. Banks changing from government to domestic blockholder control tend to
have poor asset quality, lower profitability, and lower bank value. These findings are
consistent with the entrenchment-based view, and with the looting view, which argue that
large blockholders affect firm value adversely. In addition, increases in large domestic
blockholder presence in the banking sector have positive ramifications for the banking
sector (i.e. improved asset quality and profitability). These results are consistent with the
view that these domestic blockholder-owned banks operate in a niche market, and do not
necessarily compete with other small banks in their respective countries.
In line with prior findings, foreign banks exhibit superior performance relative to
their peers. In particular, bank privatizations involving foreigners are associated with
82
improved profitability and bank value. Contrary to prior findings, increases in foreign
bank presence are not associated with improvement in the competitiveness of the banking
sector. Increases in foreign presence actually improve the banking sector margins. This
adds support to the cream-skimming argument, which argues that foreign banks tend to
target only the best hard information borrowers.
Government banks continue to perform poorly relative to their peers. These
results apply to both emerging and developed markets. Finally, increases in government
ownership of banks do appear to improve the asset quality and competitiveness of the
banking sector. These results could be driven by the fact that governments usually take
over problem banks, and by doing so, they may artificially improve the asset quality of
the banking sector (e.g. by removing a large percentage of problem loans from the
banking sector).
Although bank privatizations are almost over in several countries, in others,
including China, the process is in its early stages. An interesting area for future research
is to analyze the long-term impact of these bank privatizations on the banking sector and
on the overall economy of these countries. As foreign banks become more familiar with
the domestic markets, their soft information disadvantages may become less severe,
which may lead to an increase in their lending activities (i.e. they will no longer cream
skim). Increased foreign bank lending activity could potentially jump-start economic and
financial development. It remains to be seen whether this will actually occur.
83
CHAPTER 4:
CONCLUSION
This dissertation documents the vast changes in bank ownership structure that
have occurred throughout the world over the last decade, and examines the overall impact
of these changes. The first essay analyzes whether the efficiency of capital allocation has
been affected by these changes in bank ownership structure throughout world. I find that
the decline in government ownership of banks, by itself, has not had much of an impact
on capital allocation efficiency; what matters is whether foreigners or large domestic
blockholders take over the stakes that are relinquished by the government.
Increases in domestic blockholder ownership of banks are associated with
deterioration in capital allocation efficiency.
Countries experiencing an increase in
domestic blockholder ownership of banks tend to allocate more credit to industries that
are less productive and less dependent on external financing, which suggests a
misallocation of funds by domestic blockholders. This result is consistent with the
looting view (La Porta, et al., 2003), which argues that banks controlled by large
84
domestic blockholders, who usually have substantial interests in other nonfinancial firms,
direct a significant portion of their lending activities to related, yet inefficient companies.
By pursuing such activities, other firms in need of financing may be neglected and unable
to obtain funding.
In contrast, increases in foreign presence in the banking sector appear to improve
capital allocation efficiency. Countries in which foreign banking presence increased
allocate more credit to more productive industries (those that generate more value added),
and to industries that rely more on external financing. This finding adds support to the
findings of Giannetti and Ongena (2007), and is consistent with the view that foreign
banks improve capital allocation efficiency by mitigating the related lending problems
associated with government or domestic blockholder ownership of banks.
The second essay examines whether these changes in ownership have affected
individual bank performance, incorporating the role of large domestic shareholders into
the analysis. In addition, the spillover effects of these changes on the banking sector’s
performance are analyzed.
In general, large domestic blockholders have a negative impact on bank
performance. In particular, banks changing from government to domestic blockholder
control perform poorly in terms of asset quality and profitability. In addition, these
changes in control have a detrimental impact on bank value. This evidence suggests that
domestic blockholders may be acquiring stakes in banks to extract private benefits of
control, consistent with the entrenchment-based view of blockholder ownership. These
findings are also consistent with the looting view, which argues that large domestic
blockholders may engage in detrimental related lending practices. In addition, increases
85
in large domestic blockholder presence in the banking sector have positive ramifications
for the banking sector (i.e. improved asset quality and profitability). These results are
consistent with the view that these domestic blockholder-owned banks operate in a niche
market, and do not necessarily compete with other small banks in their respective
countries.
In line with prior findings, foreign banks exhibit superior performance
(operational efficiency) relative to their peers. Banks in which foreign blockholders
acquire control improve their profitability, operational efficiency, and valuation.
Contrary to prior findings, I find no evidence that increased foreign bank presence
improves the competitiveness of the banking sector.
In addition, in line with prior findings, government banks continue to perform
poorly relative to their peers. These results apply to both emerging and developed
markets. Finally, increases in government ownership of banks do appear to improve the
asset quality, profitability, and competitiveness of the banking sector. These results
could be driven by the fact that governments usually take over problem banks, and by
doing so, they may artificially improve the asset quality of the banking sector (e.g. by
removing a large percentage of problem loans from the banking sector).
86
LIST OF REFERENCES
Barajas, Adolfo, Roberto Steiner, and Natalia Salazar, 2000, The impact of liberalization
and foreign investment in Colombia's financial sector, Journal of Development
Economics 63, 157-196.
Barth, James, Gerard Caprio, and Ross Levine, 1999, Banking systems around the globe:
Do regulation and ownership affect performance and stability?, World Bank,
Working Paper.
---, 2001, The regulation and supervision of banks around the world: A new database,
World Bank, Development Economics Department, Policy Research Working
Paper.
---, 2004, Bank regulation and supervision: What works best?, Journal of Financial
Intermediation 13, 205-248.
---, 2005. Rethinking Bank Supervision and Regulation: Until Angels Govern (Cambridge
University Press, Cambridge, UK).
Beck, Thorsten, Asli Demirgüç-Kunt, and Ross Levine, 2000a, A New Database on
Financial Development and Structure, World Bank Economic Review 14, 597-605.
Beck, Thorsten, and Ross Levine, 2002, Industry growth and capital allocation: does
having a market- or bank-based system matter?, Journal of Financial Economics
64, 147-180.
Beck, Thorsten, Ross Levine, and Norman Loayza, 2000b, Finance and the sources of
growth, Journal of Financial Economics 58, 261-300.
Berger, Allen N., Seth Bonime, Lawrence G. Goldberg, and Lawrence White, 2004, The
dynamics of market entry: The effects of mergers and acquisitions on entry in the
banking
system,
Journal
of
Business
77,
797-834.
87
Berger, Allen N., George R.G. Clarke, Robert Cull, Leora Klapper, and Gregory F. Udell,
2005, Corporate governance and bank performance: A joint analysis of the static,
selection, and dynamic effects of domestic, foreign, and state-ownership, World
Bank, Working Paper.
Berger, Allen N., Leora Klapper, and Gregory F. Udell, 2001, The ability of banks to
lend to informationally opaque small businesses, Journal of Banking & Finance
25, 2127-2167.
Bonin, John P., Iftekhar Hasan, and Paul Wachtel, 2005, Bank performance, efficiency
and ownership in transition countries, Journal of Banking & Finance 29, 31-53.
Booth, L.V., A. Aivazian, Asli Demirgüç-Kunt, and V. Maksimovic, 2001, Capital
Structures in Developing Countries, Journal of Finance 56, 87-130.
Caprio, Gerard, and Daniela Klingebiel, 2003, Episodes of systemic and borderline
financial crises, World Bank, Working Paper.
Claessens, Stijn, Asli Demirgüç-Kunt, and Harry Huizinga, 2001, How does foreign
banking entry affect domestic banking markets?, Journal of Banking and Finance
25, 891-911.
Claessens, Stijn, Simeon Djankov, Joseph P.H. Fan, and Larry H.P. Lang, 2002,
Disentangling the Incentive and Entrenchment Effects of Large Shareholdings,
Journal of Finance 57, 2741-2771.
Clarke, George , Robert Cull, Maria Soledad Martínez Peria, and Susana M. Sánchez,
2003, Foreign Bank Entry: Experience, Implications for Developing Economies,
and Agenda for Future Research The World Bank Research Observer 18, 25-59.
Clarke, George, Robert Cull, and Maria Soledad Martínez Peria, 2001, Does foreign bank
penetration reduce access to credit in developing countries? Evidence from
asking borrowers, World Bank, Development Economics Department, Policy
Research Working Paper 2716.
Clarke, George R. G., Robert Cull, and Mary M. Shirley, 2005, Bank privatization in
developing countries: A summary of lessons and findings, Journal of Banking &
Finance 29, 1905-1930.
Cornett, M.M., L. Guo, S. Khaksari, and H. Tehranian, 2003, The impact of corporate
governance on performance differences in privately-owned versus state-owned
banks: An international comparison, Boston College, Working Paper.
Demirgüç-Kunt, Asli, and Enrica Detragiache, 2005, Cross-Country Empirical Studies of
Systemic Bank Distress: A Survey, National Institute Economic Review 68-83.
88
Demirgüç-Kunt, Asli, and Harry Huizinga, 1999, Determinants of commercial bank
interest margins and profitability: some international evidence, The World Bank
Economic Review 13, 379-408.
Detragiache, Enrica, Thierry Tressel, and Poonam Gupta, 2006, Foreign Banks in Poor
Countries: Theory and Evidence, IMF Working Paper,
Dinç, Serdar, 2005, Politicians and banks: Political influences on government-owned
banks in emerging markets, Journal of Financial Economics 77, 453-479.
Djankov, Simeon, Rafael La Porta, Florencio Lopez De Silanes, and Andrei Shleifer,
2007, The Law and Economics of Self-Dealing, Journal of Financial Economics,
forthcoming.
Echikson, William, 2007, Tapping into Turkey, The Wall Street Journal Online.
Focarelli, Dario, and Alberto Pozzolo, 2000, The determinants of cross-border bank
shareholdings: An analysis with bank-level data from OECD countries, Bank of
Italy, Research Department, Working paper.
Galindo, Arturo, and Alejandro Micco, 2004, Do state-owned banks promote growth?
Cross-country evidence from manufacturing industries, Economic Letters 84, 371376.
Gerschenkron, Alexander, 1962. Economic Backwardness in Historical Perspective
(Harvard University Press, Cambridge, MA).
Giannetti, Mariassunta, 2003, Do Better Institutions Mitigate Agency Problems?
Evidence from Corporate Finance Choices, Journal of Financial and Quantitaive
Analysis 38, 185-212.
Giannetti, Mariassunta , and Steven Ongena, 2007, Financial Integration and Firm
Performance: Evidence from Foreign Bank Entry in Emerging Markets,
Stockholm School of Economics, Working Paper.
Goldsmith, R., 1969. Financial Structure and Development (Yale University Press, New
Haven, CT).
Greenwood, J., and B. Jovanovic, 1990, Financial Development, Growth, and the
Distribution of Income, Journal of Political Economy 98, 1076-1107.
Grosse, Robert, and Lawrence G. Goldberg, 1991, Foreign Bank Activity in the United
States: An Analysis by Country of Origin, Journal of Banking and Finance 15,
1092-1112.
89
Heckman, J., 1979, Sample selection bias as a specification error, Econometrica 47, 153161.
Jensen, Michael C., and William H. Meckling, 1976, Theory of the Firm: Managerial
Behavior, Agency Costs and Ownership Structure, Journal of Financial
Economics 3, 305-360.
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi, 2006, Governance Matters V:
Governance Indicators for 1996-2005, World Bank Policy Research, Working
Paper.
Kornai, Janos, 1979, Resource-constrained versus demand-constrained systems,
Econometrica 47.
La Porta, Rafael, Florencio Lopez-de-Silanes, and Andrei Shleifer, 1999, Corporate
ownership around the world, Journal of Finance 54, 471-517.
---, 2002a, Government ownership of banks, Journal of Finance 57, 265-301.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny,
2002b, Investor Protection and Corporate Valuation, Journal of Finance 57,
1147-1170.
La Porta, Rafael, Florencio Lopez-De-Silanes, and Guillermo Zamarripa, 2003, Related
Lending, Quarterly Journal of Economics 118, 231.
Laeven, Luc, 2001, Insider Lending and Bank Ownership: The Case of Russia, Journal of
Comparative Economics 29, 207-229.
Larcker, David F., and Tjomme O. Rusticus, 2007, On the Use of Instrumental Variables
in Accounting Research, SSRN, Working Paper.
Majnoni, Giovanni, Rashmi Shankar, and Eva Varhegyi, 2003, The dynamics of foreign
bank ownership: Evidence from Hungary, World Bank, Policy Research Working
Paper.
McKinnon, R., 1973. Money and Capital in Economic Development (Brookings,
Washington, D.C.).
Megginson, William, 2005, The economics of bank privatization, Journal of Banking and
Finance 29, 1931-1980.
Mian, Atif, 2006a, Distance Constraints: The Limits of Foreign Lending in Poor
Economies, Journal of Finance 61, 1465-1505.
90
---, 2006b, Foreign, Private Domestic, and Government Banks: New Evidence from
Emerging Markets, Journal of Banking & Finance.
Micco, Alejandro, Ugo Panizza, and Monica Yañez, 2004, Bank ownership and
performance, Inter-American Development Bank, Working Paper.
---, 2006, Bank ownership and lending behavior, Inter-American Development Bank,
Working Paper.
Morse, Andrew, and Tom Wright, 2007, Reliance on Bank Loans Shields Asian Firms
The Wall Street Journal Online.
Petersen, Mitchell A., 2007, Estimating Standard Errors in Finance Panel Data Sets:
Comparing Approaches, forthcoming in The Review of Financial Studies.
Rajan, Raghuram G., and Luigi Zingales, 1998, Financial Dependence and Growth, The
American Economic Review 88, 559-586.
---, 2003, The great reversals: the politics of financial development in the twentieth
century, Journal of Financial Economics 69, 5-50.
Sapienza, Paola, 2004, The effects of government ownership on bank lending, Journal of
Financial Economics 72, 357-384.
Shaw, E., 1973. Financial Deepening in Economic Development (Brookings,
Washington, D.C.).
Shleifer, Andrei, and Robert Vishny, 1994, Politicians and firms, Quarterly Journal of
Economics 109, 995-1025.
Shleifer, Andrei, and Robert W. Vishny, 1997, A Survey of Corporate Governance,
Journal of Finance 52, 737-783.
Stulz, Rene 1988, Managerial control of voting rights: Financing policies and the market
for corporate control, Journal of Financial Economics 20, 25-54.
Tschoegl, Adrian E., 1983, Size, growth, and transnationality of among the world's
largest banks, Journal of Business 56, 187-201.
Unite, Angelo A., and Michael J. Sullivan, 2003, The effect of foreign entry and
ownership structure on the Philippine domestic banking market, Journal of
Banking and Finance 27, 2323-2345.
91
Wurgler, Jeffrey, 2000, Financial markets and the allocation of capital, Journal of
Financial Economics 58, 187-214.
92
APPENDIX A
CREDIT DATA SOURCES
Country
Website
Source
Argentina
Institution Compiling Credit
Data
Banco Central de Argentina
www.bcra.gov.ar
Estadisticas e indicadores –
monetarias y financierasPréstamos por actividades (detalle
trimestral)
Australia
Reserve Bank of Australia
www.rba.gov.au/
Austria
www.oenb.at
Bahrain
Statistik Hotline
Oesterreichische Nationalbank
Central Bank of Bahrain
Statistics- Bank Lending to
Business- Total Credit Outstanding
by Size and Sector
Data provided via email
Bangladesh
Bangladesh Bank
Belgium
National Bank of Belgium –
Belgostat Online Financial
Statistics and Markets
Superintendencia de Bancos y
Entidades Financieras de
Bolivia;
Banco Central de Bolivia
Banco Central do Brasil
www.bangladesh
-bank.org
www.nbb.be/bel
gostat;
www.nbb.be
www.sbef.gov.b
o;
www.bcb.gov.bo
Bolivia
Brazil
www.cbb.gov.bh
www.bcb.gov.br
Bulgaria
Bulgarian National Bank –
Statistics (Monetary)
www.bnb.bg
Canada
Government of Canada SME
Financing Data Initiative
sme-fdi.gc.ca
China
The People’s Bank of China
www.pbc.gov.cn
Colombia
Superintendencia Financiera
de Colombia –
Establecimientos de Credito
www.superfinan
ciera.gov.co/
93
Statistical Bulletin- Table 18Outstanding Loans and Advances
by Economic Sector
Bangladesh Bank Quarterly – Bank
Advances by Economic Purpose
Central Credit Register –
Credit to Enterprises Quarterly Sectoral statistics
Boletin Informativo (Super.)
Financiamiento concedido por
Sector Bancario (Central Bank)
Time Series Management System –
Credit Indicators
Statistics on deposits and loans by
amount categories and economic
activities
Survey of Suppliers of Business
Financing 2000–2006
Survey & Statistics – Statistics –
Sources and Uses of Credit Funds
of Financial Institutions
Establecimientos de credito Cifras Económicas y Financieras –
Operaciones Activas de Crédito
Country
Costa Rica
Institution Compiling Data
Banco Central de Costa Rica
Website
www.bccr.fi.cr
Cyprus
Central Bank of Cyprus –
Money & banking statistics
www.centralban
k.gov.cy
Czech
Republic
Czech National Bank –
Banking Statistics
www.cnb.cz/en/i
ndex.html
Denmark
Danmarks Nationalbank
Egypt
Central Bank of Egypt
El Salvador
Superintendencia del Sistema
Financiero
Información Financiera
Statistics Finland
Deutsche Bundesbank –
Statistics- Banks-Time Series
www.nationalba
nken.dk
www.cbe.org.eg/
timeSeries.htm
www.ssf.gob.sv
Finland
Germany
www.bundesban
k.de
Greece
Bank of Greece
www.bankofgree
ce.gr
Hong Kong
Hong Kong Monetary
Authority
www.info.gov.h
k/hkma
Hungary
Magyar Nemzeti Bank
India
Reserve Bank of India
www.mnb.hu/En
gine.aspx
www.rbi.org.in
Indonesia
Bank Indonesia
www.bi.go.id
Ireland
Central Bank & Financial
Services Authority of Ireland
www.centralban
k.ie/
Israel
Bank of Israel
www.bankisrael.
gov.il/
94
Source
Banking System Credit to Private
Sector by Economic Activity
Monetary Survey All banks - Sectoral distribution of
bank credit
Statistics- Banking Statistics –
ARAD Time series database ARAD -Monetary and financial
statistics - Banking statistics
(commercial banks) - Loans Client - Economic activity
breakdown
Statistics Database- B/S and flows
of MFI Sector
Time Series – Bank Lending to
Private Business Sector
Información Financiera –Cartera
de Préstamos – Montos otorgados
por categoría de riesgo
Data provided via email
Statistics-Banks- Tables- Lending Total All Banks by economic
sector
Annual Report – Credit
Developments – Credit to the
Private Sector by Branch of
Economic Activity
Statistics- Monthly Statistical
Bulletin – Domestic Loans by
economic sector
Statistics-Statistical Time Series –
Credit to nonfinancial corporations
Database- Basic Statistical Returns
Table No. 5.5
Population Group and Bank
Group-Wise Classification of
Outstanding Credit of Scheduled
Commercial Banks According to
Occupation
Statistics-Money and Banking –
Outstanding of Credit in local &
foreign currency of commercial
banks by group of banks and
economic sector
Data provided via email by the
Statistics Department. Data
compiled in Table C8 of the
Quarterly Bulletin.
Information and Data- Data on
Israel’s banking system - Credit
Risk by Industry
Country
Italy
Institution Compiling Data
Bank of Italy – Statistics
Website
www.bancaditali
a.it
Source
Statistical Database - Banks: Loans
by Branch of Economic Activity
Japan
Bank of Japan
www.boj.or.jp
Jordan
Central Bank of Jordan
www.cbj.gov.jo
Kenya
Central Bank of KenyaStatistical Bulletin
www.centralban
k.go.ke
Kuwait
Central Bank of Kuwait
www.cbk.gov.k
w
Statistics- Deposits and Loan
Markets – Loans Outstanding by
Sector
Table 12: Sectoral Distribution of
Licensed Banks’ Credit
Table 1.7 – Statistical BulletinSectoral Distribution of Credit
Facilities – Banking System
Monthly Stats – Sectoral
Distribution of Used Credit
Facilities by residents
Malaysia
Bank Negara Malaysia
Mexico
Banco de Mexico
Financiamiento e información
financiera de intermediarios
financieros
New Zealand
Reserve Bank of New Zealand
Nicaragua
Banco Central de Nicaragua
www.bnm.gov.m
y
http://www.banx
ico.org.mx/polm
oneinflacion/esta
disticas/financBa
lanIntermFinan/f
inancBalanInter
mFinan.html
www.
rbnz.govt.nz/
www.bcn.gob.ni
Oman
Central Bank of Oman
www.cbooman.org
Central Bank of Oman- Quarterly
Statistical Bulletin - Bank Credit
by Sectors
Pakistan
State Bank of Pakistan
www.sbp.org.pk/
ecodata/index2.a
sp
Publications- Banking Statistics of
Pakistan- Classification of
Schedules Banks’ Advances
Panama
Peru
Superintendencia de Bancos
de Panama
Central Bank of Peru
Estadísticas Carta Bancaria
Reporte de Estabilidad Financiera Colocaciones por Sector
Económico
Philippines
Central Bank of Philippines
Portugal
Banco de Portugal
www.superbanco
s.gob.pa
http://www.bcrp.
gob.pe/bcr/Publi
caciones/Reporte
-de-EstabilidadFinanciera.html
http://www.bsp.g
ov.ph/banking/bs
psup_pbs_01.asp
www.bportugal.p
t
95
Monthly Statistical Bulletin –
Loans by Sector
Estadísticas-sistema financieroFinanciamiento e información
financiera de intermediarios
financieros –Banca commercialcrédito por actividad de
prestatarios
Money, credit & financial statistics
– Sector Credit
Annual Economic Statistics
Bulletin – Loans by Economic
Sector- Deposit Taking Institutions
Banking Statistics – Outstanding
loans (by sector)
Central Credit Register Statistical
Information -Breakdown of loans
extended to non-financial
corporations by branch of activity
Country
Romania
Institution Compiling Data
National Bank of Romania
Website
www.bnro.ro/def
_en.htm
Source
Monthly Bulletin 12/200-12/2006
Table 17b- Loans granted &
commitments assumed by banks
Russia
Bank of Russia
Saudi Arabia
Saudi Arabian Monetary
Agency
Monetary Authority of
Singapore-
www.cbr.ru/eng/
publ/print.asp?fil
e=BBS/bank_bul
letin_reg.htm
www.sama.gov.s
a
www.mas.gov.sg
Bulletin of Banking StatisticsRegional Supplement – Credits
Extended to Legal Entities by
Economic Activities
Table 12 (d) Bank Credit
Classified by Economic Activity
Monthly Statistical Bulletin
database
Singapore
Spain
Banco de España
www.bde.es;
www.bde.es/info
est/htmls/capit04
.htm
Boletín Sadistic – (Chapter 4) –
Crédito y dudosos del crédito para
financiar actividades productivas
de las empresas y los empresarios
individuales que reciben los
créditos. Detalle por la actividad
principal
Sri Lanka
Central Bank of Sri Lanka
www.cbsl.gov.lk
Switzerland
Swiss National Bank
www.snb.ch
Tanzania
Bank of Tanzania
www.bot-tz.org
Thailand
Bank of Thailand
Trinidad &
Tobago
Central Bank of Trinidad &
Tobago
http://www.bot.o
r.th/bothomepag
e/databank/Finan
cial_Institutions/
New_Fin_Data/
CB_Menu_E.ht
m
www.centralbank.org.tt
Annual Report- Money, Credit &
Interest Rate- Ch. 7 - Sectoral
Distribution of Loans and
Advances by Commercial Banks
Monthly Bulletin of Banking
Statistics- Credit Volume
Statistics- Credit by Economic
Activities- All Banks
Bank of Tanzania Economic
Bulletin- Table 3.11 – Commercial
Banks- Domestic Lending by
Activities
Data Bank- Financial Institutions
Data- Commercial Banks - Credits
Classified by Types of Businesses
(Table 7)
Tunisia
Central Bank of Tunisia
www.bct.gov.tn
Turkey
Central Bank of the Republic
of Turkey
Central Bank of the UAE
www.tcmb.gov.t
r
www.centralban
k.ae
UAE
96
Economic Bulletin – Commercial
Banks- Distribution of Loans
Monetary Statistics- Trend in the
financial system loans to the
economy by sector activity
Loans Extended to firms by sector
Data- Table 6 : Banks' Credit
(Gross) to Residents by Economic
Activity
Country
UK
Institution Compiling Data
Bank of England
Venezuela
Superintendencia de Bancos e
Instituciones Financieras
Website
www.bankofengl
and.co.uk
www.sudeban.go
b.ve/inf_estadisti
ca.php
97
Source
Statistics- Interactive DatabaseTables- c Deposits and Lending
Series Anuales – Cartera de
Crédito por destino
APPENDIX B
DESCRIPTION OF INDUSTRIES FOR CREDIT DATA
Industry
Subsectors Included
Agriculture, fishing, hunting, and forestry
AGRICULTURE
CONSTRUCTION
FINANCE AND
INSURANCE
HEALTHCARE
HOTELS AND
RESTAURANTS
MANUFACTURING
METALS AND
NONPRECIOUS
METALS
MINING
OIL AND GAS
TOURISM
TRANSPORTATION
AND STORAGE
UTILITIES
All construction companies
Financial institutions, real estate firms,
insurance companies
Health and social work organizations
Accommodation and food service activities
All manufacturers
All basic metals and metal products
Mining and quarrying
Petroleum, coal products, nuclear fuel, natural
gas
All tourism related activities excluding hotels &
restaurants
Transportation, storage, and communications
Electricity, gas, and water
All trade related activities
WHOLESALE AND
RETAIL TRADE
SERVICES
OTHER
All other services and service organizations,
including education providers, recreational,
cultural, and sporting organizations; public
administration and defense
All unclassified sectors
98
Corresponding ISIC
Industry (UN data)
Agriculture, hunting,
forestry, and fishing
Construction
Other Activities
Other Activities
Wholesale, retail
trade, restaurants and
hotels
Manufacturing
Mining and utilities
Mining and utilities
Mining and utilities
Wholesale, retail
trade, restaurants and
hotels
Transport, storage,
and communications
Mining and utilities
Wholesale, retail
trade, restaurants and
hotels
Other Activities
Other Activities
99
Country
Slovakia
Bank Name
Slovenska Pol'nohospodarska
Banka (Agricultural Bank)
Shareholders
Slovak Insurance Co.
EBRD
Ministry of Agriculture
Agrobanka Praha
Other minority shareholders
Shareholders
ZAVAROVALNICA TRIGLAV dd
FMR dd
HIT dd
Stajerski Avto
Other minority shareholders
Barclays Bank PLC
General Public
Country of
Shareholder
Slovenia
Slovenia
Slovenia
Slovenia
Slovenia
UK
Kenya
Country of
Foreign
Shareholder
N
Slovakia
Y
SUPRANATIONAL
N
Slovakia
Y
Czech Rep.
N
Slovakia
Foreign
N
N
N
N
N
Y
N
Ownership %age
31.90%
20.00%
8.30%
3.80%
36.00%
Ownership %age
32.92%
9.84%
6.10%
5.00%
46.14%
68.50%
31.50%
Type
Government
Other
Government
Foreign
Widely-held
Type
Local company
Local company
Local company
Local company
Widely held
Foreign
Widely held
Government's Stake in
Shareholder
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Government's Stake in
Shareholder
100.00%
2.49%
100.00%
0.00%
0.00%
Foreign Stake in
shareholder
0.00%
0.00%
0.00%
0.00%
0.00%
100.00%
0.00%
Foreign Stake in
shareholder
0.00%
97.51%
0.00%
100.00%
0.00%
The above table presents ownership data for a Slovenian bank, a Kenyan bank, and a Slovakian bank as of 2005, 2000, and 1995, respectively. The data
was obtained from BankScope and The Bankers’ Almanac. Using the BankScope data, I am usually able to identify whether a shareholder is a
government entity, a foreign entity, or a private domestic shareholder. Identifying local, government, and foreign shareholders was a more involved
process when calculating the 2000 and 1995 ownership measures. The Bankers’ Almanac, Mergent Online, and company websites were used to
determine the shareholder type in those cases. When the shareholder was a foreign bank, the location of the foreign bank’s headquarters was used to
determine the country of the foreign shareholder.
When regional development banks (owned by various governments) or other supranational entities are shareholders in a bank, I follow La Porta et al.’s
(2002) strategy. The equity ownership in the regional bank by the local government is estimated as the proportion of the banks’ assets that are in the
country. The calculation of the government’s ownership stakes in Slovenska Pol'nohospodarska Banka illustrates this point.
Kenya
Country
Slovenia
Barclays Bank of Kenya Ltd.
Bank Name
Abanka Vipa dd
EXAMPLE OF OWNERSHIP VARIABLE CONSTRUCTION
APPENDIX C
100
DBBBK=0%
GBBBK=0%.
The foreign ownership variable is computed in a similar manner. There is only one foreign shareholder owning more than
10% of the shares in SDB, the EBRD. Accounting for the fact that the Slovakian government has a 2.49% stake in EBRD,
the remaining 97.51% is owned by foreign governments. Thus, the total foreign blockholder ownership of SPB is
comprised of the 19.502% owned indirectly by various foreign governments through their stakes in EBRD. Thus, foreign
blockholders’ share in SPB is: FB SPB = ( 20 % × 0 .9751 ) = 19 .502 % ; there is no large domestic blockholder, so
DBSPB=0%
GB SPB = 31 . 9 % + ( 20 % × 0 . 0249 ) + 8 . 3 % = 40.698% .
Slovenska’s Ownership Variables:
The computation of the ownership variables is not as straight forward for Slovenska Pol'nohospodarska Banka (SDB). The
total government’s share in SPB is comprised of three elements. First, 31.9% is owned by the government through Slovak
Insurance Co. (a state-owned enterprise). Second, the Slovakian government has an additional 8.3% stake in SPB through
the Ministry of Agriculture. Finally, the Slovakian government has a 2.49% stake in the European Bank for Reconstruction
and Development (EBRD) as of 1995 (2.49% of EBRD’s total operating assets are in Slovakia - this information was
obtained directly from EBRD’s annual report). EBRD holds a 20% stake in SPB. Thus, the Slovakian government
indirectly owns an additional 0.498% stake in SPB. Thus, the total government’s share is computed as follows:
FBBBK=68.5%
Barclays Bank of Kenya Ownership Variables:
Barclays Bank of Kenya (BBK) has only one large blockholder, Barclays Bank PLC, UK. The total share of BBK owned
by foreigners is thus the 68.5% owned by Barclays Bank PLC. So,
Abanka Vipa does not have any government or foreign shareholders. Thus, GBAbanka=0%, and FBAbanka =0%.
Abanka Vipa’s Ownership Variables:
Abanka Vipa has only one domestic shareholder (Zavarovalnica Triglav dd, an insurance company) owning more than 10%
of its shares. Thus, the total share of Abanka Vipa, dd owned by large domestic blockholders (DBi) corresponds to the
32.92% owned by Zavarovalnica Triglav, dd. DBAbanka=32.92%.
APPENDIX D
TABLES
This table shows descriptive statistics for the sample of countries with data available on credit
outstanding to industry and industry value added, respectively. Data on outstanding credit by
industry was obtained from various Central Banks and Financial Institutions Regulatory
Authorities. In general, the data represents the total credit provided to specific industries by the
total banking system. Data on value added by industry was obtained from the United Nations
Statistics Division National Accounts. Gross value added represents the value of output less the
value of intermediate goods. Industry value added describes the generation of gross value added
by industrial classification of economic activities according to the International Standard
Industrial Classification (ISIC). It measures the contribution to GDP made by a particular
industry.
Table 1: Summary statistics (continued)
101
Table 1 (continued)
Country
Argentina
Australia
Austria
Bahrain
Bangladesh
Belgium
Bolivia
Brazil
Bulgaria
Canada
China
Colombia
Costa Rica
Cyprus
Czech Republic
Denmark
Egypt
El Salvador
Finland
Germany
Greece
Hong Kong
Hungary
India
Indonesia
Ireland
Israel
Italy
Japan
Jordan
Kenya
Kuwait
Malaysia
Mexico
New Zealand
Nicaragua
Oman
Pakistan
Panama
Peru
Philippines
Portugal
Romania
Russia
Saudi Arabia
Singapore
Spain
Sri Lanka
Switzerland
Tanzania
Thailand
Trinidad
Tunisia
Turkey
UAE
UK
Venezuela
TOTAL
Anuual Industry Credit Growth
Mean
Std. dev.
10.40%
0.432
9.22%
0.121
4.49%
0.175
26.33%
0.543
14.22%
0.083
1.42%
0.110
0.13%
0.140
13.06%
0.092
44.76%
0.384
4.15%
0.100
12.11%
0.111
15.52%
0.214
21.51%
0.176
2.02%
0.106
2.84%
0.153
18.23%
0.215
4.65%
0.074
3.16%
0.178
3.82%
0.068
-1.21%
0.043
7.33%
0.107
3.77%
0.104
13.77%
0.176
25.52%
0.242
21.16%
0.175
14.49%
0.187
1.67%
0.120
6.63%
0.060
-5.02%
0.061
8.43%
0.165
8.52%
0.187
17.01%
0.347
1.32%
0.091
-3.79%
0.183
12.74%
0.049
24.14%
0.095
19.19%
0.152
20.33%
0.310
14.17%
0.334
4.00%
0.199
1.79%
0.099
3.45%
0.075
44.37%
0.180
43.69%
0.214
17.65%
0.255
4.97%
0.237
12.85%
0.113
16.90%
0.107
0.29%
0.101
59.92%
0.929
2.98%
0.107
19.41%
0.615
5.03%
0.050
42.31%
0.345
23.75%
0.325
5.38%
0.164
88.34%
0.383
13.27%
0.292
Industry Value Added
Mean
Std. dev.
20.31%
0.322
7.52%
0.094
3.11%
0.038
12.09%
0.140
12.48%
0.029
2.93%
0.037
6.43%
0.087
13.82%
0.045
8.70%
0.062
4.96%
0.034
13.43%
0.087
10.98%
0.079
15.74%
0.065
5.28%
0.046
6.30%
0.060
3.33%
0.084
8.95%
0.060
6.38%
0.056
4.35%
0.051
1.18%
0.056
5.33%
0.065
0.55%
0.076
8.73%
0.057
12.23%
0.051
16.51%
0.083
8.44%
0.089
4.27%
0.062
3.62%
0.037
-0.14%
0.032
10.28%
0.060
11.58%
0.076
18.86%
0.183
9.42%
0.075
9.02%
0.044
5.57%
0.083
10.54%
0.065
18.47%
0.134
13.95%
0.101
5.44%
0.095
8.26%
0.071
10.84%
0.050
2.05%
0.042
29.80%
0.188
21.45%
0.128
10.23%
0.108
3.57%
0.092
6.99%
0.043
14.42%
0.070
2.38%
0.017
14.44%
0.043
8.80%
0.041
15.01%
0.190
6.02%
0.072
29.08%
0.180
17.94%
0.154
4.56%
0.053
37.24%
0.135
9.89%
0.117
102
ρCREDIT, VA
0.856***
0.969***
0.906***
-0.018
0.122
0.983***
0.043
0.956***
0.872***
0.951***
0.386**
0.511***
0.831***
0.467**
0.875***
0.980***
0.746***
0.817***
0.968***
0.961***
0.896***
0.937***
0.901***
0.806***
0.863***
0.816***
0.761***
0.805***
0.982***
0.755***
0.678***
0.300*
-0.198
0.940***
0.968***
0.183
0.484**
0.483***
0.977***
0.933***
0.924***
0.901***
0.944***
0.936***
0.244
0.682***
0.911***
0.421*
0.075
0.165
0.833***
0.240
0.982***
0.631***
0.210
0.968***
0.177
0.679
Average #
Industry-Year
Industries per year Observations
6
36
6
36
7
42
6
30
6
18
7
42
6
30
3
18
6
24
7
42
5
30
7
42
7
42
7
21
7
42
7
42
4
24
7
21
7
42
7
42
4
20
3
18
7
42
7
42
7
42
7
42
7
35
7
42
6
36
7
42
7
42
5
32
7
21
7
42
3
18
3
15
7
14
7
42
6
36
7
42
7
35
7
28
4
24
6
30
7
35
6
36
7
42
4
16
5
30
6
38
6
36
7
42
3
18
7
42
7
42
7
42
3
9
6
1876
103
DB95
0.12%
2.51%
0.00%
2.65%
48.32%
9.13%
0.00%
2.79%
36.10%
5.95%
55.10%
15.46%
6.15%
1.63%
28.07%
20.11%
99.42%
10.85%
3.82%
13.26%
10.12%
7.23%
81.12%
100.00%
2.19%
22.48%
9.13%
DB00
0.92%
24.66%
0.73%
1.04%
1.18%
19.59%
0.00%
0.00%
24.48%
1.22%
58.17%
48.56%
5.05%
9.96%
14.42%
51.22%
99.66%
4.31%
0.00%
4.37%
17.62%
6.53%
86.08%
100.00%
26.86%
24.26%
9.96%
DB05
0.00%
7.97%
0.00%
0.69%
9.12%
1.84%
0.57%
0.00%
21.22%
12.35%
43.83%
1.75%
22.68%
8.61%
7.13%
10.17%
72.49%
9.82%
38.82%
13.79%
15.75%
14.33%
89.53%
93.03%
20.12%
20.62%
10.17%
GB95
3.39%
7.83%
88.35%
0.00%
1.01%
1.75%
88.20%
2.31%
48.24%
23.50%
18.41%
0.00%
13.64%
91.80%
6.13%
15.05%
0.00%
66.74%
85.35%
16.23%
12.87%
49.34%
0.00%
0.00%
29.79%
26.80%
13.64%
GB00
0.18%
6.78%
80.55%
0.00%
1.42%
0.00%
75.78%
0.00%
25.46%
14.62%
0.00%
0.00%
5.87%
77.87%
16.82%
0.90%
0.00%
58.66%
21.43%
39.51%
13.59%
42.34%
0.00%
0.00%
6.04%
19.51%
6.04%
GB05
1.98%
9.25%
71.66%
0.77%
0.00%
22.15%
54.95%
0.00%
13.24%
11.35%
0.00%
0.00%
8.48%
37.53%
20.22%
10.51%
2.80%
44.70%
1.46%
22.13%
17.50%
44.96%
0.00%
0.00%
35.39%
17.24%
10.51%
FB95
1.97%
24.85%
0.00%
3.22%
10.32%
77.93%
0.00%
11.43%
0.00%
37.53%
0.29%
84.54%
14.74%
0.45%
16.59%
0.00%
0.32%
2.40%
8.50%
4.62%
13.46%
2.86%
0.00%
0.00%
40.23%
14.25%
3.22%
FB00
0.95%
64.13%
0.00%
3.54%
11.18%
61.49%
6.00%
28.43%
0.00%
34.75%
11.44%
51.44%
3.78%
2.94%
15.48%
0.00%
0.31%
0.30%
69.29%
0.00%
18.05%
4.13%
0.00%
0.00%
49.42%
17.48%
4.13%
FB05
3.59%
60.53%
0.00%
2.64%
24.47%
63.10%
12.82%
17.04%
0.00%
25.83%
15.85%
98.25%
0.00%
19.61%
13.61%
0.07%
24.02%
1.45%
44.61%
10.23%
15.06%
0.93%
10.47%
0.00%
35.13%
19.97%
13.61%
Coverage2
98.97%
98.62%
71.21%
94.54%
99.23%
87.46%
65.70%
95.33%
100.00%
92.40%
77.73%
100.00%
84.81%
97.48%
100.00%
80.37%
100.00%
100.00%
100.00%
94.20%
100.00%
100.00%
82.77%
61.84%
100.00%
Table 2: Ownership of banks around the world: 1995, 2000, and 2005 (continued)
Panel A shows the percentage of assets of the top 10 banks in each country that as of 1995, 2000, & 2005 are owned by domestic
(DB), government (GB), or foreign blockholders (FB). Countries are classified by the legal origin of their commercial laws. Panels
B, C, and D show the results of tests of means across legal origin (common vs. civil law), financial development, and whether or not
the countries experienced a banking crisis over the period, respectively.
Country
Australia
Bahrain
Bangladesh
Canada
Cyprus
Hong Kong
India
Ireland
Israel
Kenya
Malaysia
New Zealand
Nigeria
Pakistan
Saudi Arabia
Singapore
South Africa
Sri Lanka
Tanzania
Thailand
Trinidad & Tobago
United Arab Emirates
UK
USA
Zimbabwe
English Origin Average
English Origin Median
Panel A- Ownership of the top 10 banks by type of owner1 as of 1995, 2000 & 2005
104
Country
Algeria
Argentina
Belgium
Bolivia
Brazil
Chile
Colombia
Costa Rica
Dominican Republic
Egypt
El Salvador
France
Greece
Guatemala
Honduras
Indonesia
Iran
Iraq
Italy
Ivory Coast
Jordan
Kuwait
Lebanon
Libya
Macau
Mexico
Morocco
Netherlands
Nicaragua
Oman
Panama
Paraguay
Peru
Philippines
Portugal
Qatar
Senegal
Spain
Tunisia
Turkey
Uruguay
Venezuela
French origin average
French origin median
Table 2 (continued)
DB95
31.57%
20.99%
27.77%
37.98%
26.91%
13.78%
18.93%
9.53%
31.01%
0.71%
0.00%
24.67%
2.61%
22.81%
38.03%
18.23%
0.00%
0.00%
33.18%
4.73%
31.78%
1.51%
33.34%
0.00%
43.68%
28.38%
7.74%
47.05%
0.53%
25.49%
8.38%
17.49%
60.97%
25.72%
54.78%
25.30%
1.90%
6.56%
2.83%
30.26%
4.58%
11.86%
19.85%
19.96%
DB00
27.69%
15.11%
83.99%
43.90%
39.40%
43.40%
50.48%
13.87%
50.25%
3.07%
40.95%
28.88%
0.00%
37.55%
59.59%
4.60%
0.00%
0.00%
33.04%
1.11%
33.88%
22.63%
42.47%
0.00%
38.66%
33.27%
17.58%
39.74%
18.70%
30.00%
36.49%
4.33%
45.92%
32.13%
32.51%
20.62%
14.85%
6.96%
4.42%
37.76%
16.73%
29.07%
27.04%
29.53%
DB05
20.69%
17.65%
85.14%
53.89%
30.07%
33.86%
36.31%
11.31%
44.10%
3.33%
20.14%
47.86%
1.40%
44.16%
32.69%
13.88%
2.27%
0.00%
18.12%
1.52%
15.85%
11.51%
44.27%
0.00%
53.01%
30.42%
19.68%
79.75%
10.88%
29.06%
67.98%
0.00%
30.14%
23.84%
11.04%
7.77%
23.21%
16.41%
9.74%
32.67%
0.00%
40.18%
25.61%
20.41%
GB95
68.43%
59.90%
20.19%
0.02%
42.13%
21.75%
39.49%
89.02%
38.44%
84.93%
24.93%
18.49%
71.87%
15.79%
12.66%
71.65%
100.00%
100.00%
34.54%
3.54%
9.17%
17.43%
6.34%
96.78%
38.69%
33.20%
28.98%
8.56%
32.19%
5.98%
0.00%
30.12%
27.02%
24.25%
34.73%
31.21%
22.53%
0.00%
25.94%
59.75%
68.59%
43.50%
37.21%
30.66%
GB00
71.95%
33.70%
0.00%
0.01%
41.48%
18.88%
3.39%
71.22%
25.12%
80.60%
2.88%
12.01%
23.96%
21.97%
5.13%
87.87%
129.62%
100.00%
0.00%
2.85%
1.79%
32.77%
0.01%
95.72%
0.00%
27.13%
11.14%
4.35%
8.46%
11.82%
0.00%
14.58%
11.32%
24.20%
27.00%
28.45%
3.74%
0.00%
21.49%
40.01%
42.80%
7.44%
27.31%
16.73%
GB05
76.70%
52.69%
0.25%
0.01%
43.37%
19.24%
13.37%
67.91%
33.03%
78.87%
6.81%
0.00%
14.66%
16.47%
6.31%
44.52%
97.17%
87.74%
5.17%
2.87%
0.00%
24.28%
0.00%
94.39%
13.59%
17.22%
30.65%
2.82%
0.00%
2.51%
0.00%
11.99%
11.85%
27.86%
28.40%
23.39%
6.69%
0.00%
19.74%
39.48%
55.06%
13.14%
25.96%
15.56%
FB95
0.00%
9.19%
13.90%
19.83%
2.83%
19.75%
0.00%
0.00%
0.00%
6.22%
0.00%
9.24%
0.00%
0.00%
14.20%
0.00%
0.00%
0.00%
2.19%
83.04%
8.51%
0.00%
15.31%
0.00%
0.00%
2.64%
12.18%
2.25%
0.35%
16.31%
49.57%
36.00%
3.70%
5.99%
0.00%
0.00%
67.13%
1.99%
6.97%
0.00%
22.92%
6.89%
10.45%
2.74%
FB00
0.35%
43.15%
15.73%
27.06%
6.69%
20.50%
21.38%
3.00%
13.99%
6.33%
6.58%
0.82%
16.61%
0.00%
3.71%
0.00%
0.00%
0.00%
0.00%
92.81%
4.03%
0.07%
21.39%
0.00%
47.07%
29.81%
20.78%
0.25%
12.88%
7.58%
25.79%
79.18%
37.91%
8.53%
11.24%
0.00%
40.28%
2.66%
13.77%
3.12%
36.28%
50.90%
17.43%
9.88%
Coverage2
100.00%
74.51%
92.84%
98.61%
87.56%
85.83%
98.57%
83.56%
96.17%
97.21%
85.42%
59.39%
96.68%
66.73%
100.00%
83.82%
100.00%
100.00%
79.58%
100.00%
75.03%
100.00%
85.60%
100.00%
100.00%
91.92%
77.63%
70.64%
95.95%
100.00%
74.68%
76.06%
97.43%
74.88%
88.79%
100.00%
100.00%
73.81%
91.02%
99.81%
61.60%
100.00%
(continued)
FB05
2.23%
21.11%
14.40%
21.46%
15.38%
31.13%
18.49%
20.54%
6.32%
8.94%
16.60%
5.94%
28.69%
0.00%
39.84%
12.94%
0.00%
0.00%
8.77%
70.62%
11.68%
4.70%
14.22%
0.00%
19.52%
44.04%
15.97%
0.00%
0.00%
10.55%
23.19%
67.88%
49.64%
5.37%
18.08%
2.28%
36.04%
2.89%
16.16%
3.25%
44.30%
26.81%
18.09%
14.89%
105
Country
Austria
Germany
Japan
Korea
Switzerland
Taiwan
German origin average
German origin median
Denmark
Finland
Norway
Sweden
Scandinavian origin average
Scandinavian origin median
Bulgaria
China
Croatia
Czech Republic
Hungary
Kazakhstan
Macedonia
Poland
Romania
Russia
Slovakia
Slovenia
Vietnam
Socialist origin average
Socialist origin median
WORLD MEAN
WORLD MEDIAN
Table 2 (continued)
DB95
45.19%
7.17%
0.00%
0.00%
4.12%
4.40%
10.15%
4.26%
32.90%
66.60%
3.74%
9.98%
28.30%
21.44%
3.02%
0.36%
22.78%
1.03%
3.99%
12.30%
14.39%
0.00%
2.57%
3.98%
0.80%
2.62%
0.00%
5.22%
2.62%
18.20%
9.75%
DB00
27.19%
11.17%
59.14%
1.71%
58.42%
7.76%
27.56%
19.18%
27.56%
33.62%
26.30%
28.86%
29.09%
28.21%
15.13%
2.37%
0.00%
0.50%
0.00%
28.38%
7.69%
0.84%
0.00%
10.01%
28.51%
13.48%
0.22%
8.24%
2.37%
23.68%
19.15%
DB05
43.81%
33.21%
69.04%
27.87%
1.93%
46.10%
36.99%
38.51%
34.67%
26.98%
61.51%
13.48%
34.16%
30.82%
21.79%
1.53%
4.47%
3.56%
0.00%
44.42%
11.55%
1.25%
7.11%
24.95%
0.00%
9.61%
0.92%
10.09%
4.47%
23.12%
16.13%
GB95
24.45%
32.65%
0.00%
25.25%
15.15%
76.43%
28.99%
24.85%
0.00%
22.48%
47.91%
14.00%
21.10%
18.24%
76.17%
99.64%
0.00%
52.00%
33.59%
34.26%
0.00%
78.03%
67.42%
35.06%
64.26%
53.58%
99.19%
53.32%
53.58%
35.38%
28.00%
GB00
0.00%
27.07%
5.02%
38.00%
4.35%
57.39%
21.97%
16.04%
0.00%
1.00%
0.00%
3.45%
1.11%
0.50%
2.36%
96.08%
18.53%
17.26%
8.72%
0.78%
0.40%
7.49%
42.64%
52.06%
26.69%
43.75%
94.87%
31.66%
18.53%
24.25%
11.92%
GB05
0.00%
16.10%
0.00%
20.34%
6.68%
30.43%
12.26%
11.39%
0.00%
0.00%
11.42%
10.06%
5.37%
5.03%
0.04%
84.93%
2.18%
3.67%
4.55%
0.03%
2.37%
14.99%
0.23%
35.80%
0.00%
26.87%
90.28%
20.46%
3.67%
20.91%
11.92%
FB95
3.56%
0.00%
0.00%
0.00%
0.00%
0.00%
0.59%
0.00%
0.00%
6.57%
7.24%
0.24%
3.51%
3.40%
0.00%
0.00%
0.00%
12.64%
34.02%
1.30%
0.98%
13.21%
0.98%
2.00%
5.51%
4.90%
0.81%
5.87%
1.30%
9.88%
2.32%
FB05
29.36%
11.41%
0.00%
18.24%
2.17%
0.00%
10.20%
6.79%
25.66%
69.61%
21.65%
9.69%
31.65%
23.66%
71.18%
0.00%
87.41%
49.33%
88.06%
12.48%
21.25%
51.46%
59.13%
28.67%
91.27%
36.03%
0.36%
45.89%
49.33%
22.71%
15.91%
(continued)
FB00
49.83%
0.00%
0.00%
9.22%
0.68%
0.00%
9.95%
0.34%
0.55%
53.55%
31.10%
19.71%
26.23%
25.41%
70.08%
0.00%
61.40%
60.08%
68.26%
20.32%
47.09%
58.67%
21.75%
4.83%
36.89%
15.78%
0.65%
35.83%
36.89%
20.00%
11.21%
76.41%
77.03%
91.04%
83.37%
95.33%
100.00%
100.00%
99.66%
80.96%
56.94%
90.12%
94.21%
93.57%
96.29%
81.71%
87.34%
81.59%
2
Coverage
98.61%
56.72%
50.99%
80.98%
72.19%
62.70%
106
DB95
16.55%
22.48%
-5.94%
(-1.17)
DB95
26.92%
15.02%
11.89%
(2.37**)
DB95
21.92%
11.77%
10.15%
(2.20**)
798
860
873
Civil Law (n=65)
Common Law (n=25)
Difference
t-stats (Common vs. Civil)
Developed Markets
Emerging Markets
Difference
Emerging vs. Developed
No Crisis
Crisis
Difference
Crisis vs. No Crisis
Total # of banks (1995):
Total # of banks (2000):
Total # of banks (2005):
Table 2 (continued)
Panel B - Means by legal origin (t-statistics in Italics are for differences in means)
DB00
DB05
GB95
GB00
GB05
FB95
23.45%
24.09%
38.68%
26.07%
22.33%
8.20%
24.27%
20.62%
26.80%
19.51%
17.24%
14.25%
-0.81%
3.46%
11.88%
6.56%
5.09%
-6.05%
(-0.15)
(0.64)
(1.66*)
(0.93)
(0.84)
(-1.44)
Panel C - Means by country characteristics
DB00
DB05
GB95
GB00
GB05
FB95
33.58%
31.27%
15.36%
4.61%
5.46%
9.86%
20.08%
20.16%
42.66%
31.39%
26.53%
9.89%
13.51%
11.11%
-27.30%
-26.78%
-21.08%
-0.03%
(2.54**)
(2.09**)
(-4.05***)
(-4.06***)
(-3.68***)
(-0.01)
Panel C - Means - countries experiencing a banking crisis between 1995 and 2005
DB00
DB05
GB95
GB00
GB05
FB95
28.00%
25.26%
29.40%
21.02%
19.55%
10.50%
16.21%
19.44%
45.70%
29.83%
23.28%
8.81%
11.79%
5.82%
-16.30%
-8.80%
-3.73%
1.68%
(2.41**)
(1.17)
(-2.50**)
(-1.35)
(-0.66)
(0.43)
FB00
16.28%
26.42%
-10.14%
(-2.03**)
FB00
14.99%
21.82%
-6.83%
(-1.24)
FB00
20.96%
17.48%
3.48%
(0.63)
FB05
18.94%
29.22%
-10.28%
(-1.98*)
FB05
19.50%
23.88%
-4.38%
(-0.76)
FB05
23.76%
19.97%
3.79%
(0.67)
n
57
33
n
24
66
n
65
25
This table shows the changes in bank ownership structure over the 1995-2005 period.
Countries with good (poor) shareholder protection are those countries with an Anti-SelfDealing Index (ASDI) greater (lower) than the median ASDI for all countries in the sample.
The ASDI was obtained from Djankov, et al. (2007). * Significant at the 10% level; **
significant at the 5% level; *** significant t the 1% level.
Table 3: Changes in bank ownership structure between 1995 and 2005 (continued)
107
Table 3 (continued)
Year
1995
2005
Change
Difference (t-stat)
Year
1995
2005
Change
Difference (t-stat)
Year
1995
2005
Change
Difference (t-stat)
Year
1995
2005
Change
Difference (t-stat)
Year
1995
2005
Change
Difference (t-stat)
Panel A- Average Ownership of Banks by Type
DB
GB
FB
18.20%
35.38%
9.88%
23.12%
20.91%
22.71%
4.93%
-14.47%
12.83%
(-3.43***)
(4.05***)
(1.49)
Panel B- Average Ownership of Banks by Type - Emerging & Developed Markets
Emerging Markets
DB
GB
FB
15.02%
42.66%
9.89%
20.16%
26.53%
23.88%
5.14%
-16.13%
13.99%
(1.65)
(-3.16***)
(3.93***)
Developed Markets
DB
GB
FB
26.92%
15.36%
9.86%
31.27%
5.46%
19.50%
4.35%
-9.90%
9.64%
(-2.41**)
(0.50)
(1.42)
Panel C- Differences in Ownership of Banks by Legal Origin
Civil Law Countries
DB
GB
FB
16.55%
38.68%
8.20%
24.09%
22.33%
23.76%
7.54%
-16.35%
15.56%
(2.25**)
(-3.27***)
(4.38***)
Common Law Countries
22.48%
26.80%
14.25%
20.62%
17.24%
19.97%
-1.86%
-9.56%
5.72%
(-0.23)
(-1.25)
(0.85)
WIDELY
36.55%
33.26%
-3.29%
(-0.83)
n
90
WIDELY
32.43%
29.43%
-3.00%
(-0.75)
n
66
WIDELY
47.87%
43.78%
-4.09%
(-0.43)
n
24
WIDELY
36.57%
29.83%
-6.74%
(-1.54)
n
65
36.47%
42.17%
5.70%
(0.67)
25
WIDELY
37.26%
32.79%
-4.47%
(-0.76)
n
32
WIDELY
39.65%
34.08%
-5.56%
(-0.72)
n
31
1
Year
1995
2005
Change
Difference (t-stat)
Year
1995
2005
Change
Difference (t-stat)
∆DB
∆GB
∆FB
∆WIDE
∆DB
∆GB
∆FB
∆WIDE
Panel D- Average Ownership of Banks by Protection of Minority Shareholders
Poor Shareholder Protection Markets
DB
GB
FB
16.46%
36.51%
9.77%
23.04%
15.69%
28.49%
6.58%
-20.82%
18.72%
(1.38)
(-3.84***)
(3.68***)
Good Shareholder Protection Markets
DB
GB
FB
22.80%
29.13%
8.42%
28.78%
17.12%
20.02%
5.98%
-12.02%
11.60%
(-1.77*)
(2.06**)
(0.82)
Panel E- Descriptive Statistics of Changes in Ownership Measures
Mean
Max
Min
4.93%
69.04%
-43.74%
-14.47%
20.40%
-83.89%
12.83%
87.41%
-31.09%
-3.29%
49.88%
-71.29%
Panel F- Correlation Matrix for Changes in Ownership Measures
∆DB
∆GB
∆FB
-0.25**
-0.17
1.00
-0.17
-0.51***
1.00
-0.25**
-0.51***
1.00
-0.52***
-0.28***
-0.27**
108
Std. Dev.
19.87%
20.20%
21.03%
21.65%
∆WIDE
-0.52***
-0.28***
-0.27**
1.00
109
87
87
87
87
87
87
87
87
71
85
71
82
87
N
Table 4: Changes in bank ownership and country characteristics
The first column shows the correlation between the changes in the measures of bank ownership between 1995 and 2005 (∆GB, ∆DB, and ∆FB) and
various country characteristics related to the initial level of financial development and governance. The second column shows the coefficients from
the following regression: ∆OWNERSHIP = α + βX + δGDP per capita in 1995, where X represents the vector of independent variables. Robust
(White) t-statistics are shown in parentheses. ***Significant at 1 percent level; **significant at 5 percent level; *significant at 10 percent level.
Common Law Dummy
Banking Crisis Dummy
Control of Corruption
Rule of Law
Regulatory Quality
Government Effectiveness
Political Stability
Voice and Accountability
Stock Market Capitalization/GDP in 1995
Commercial Bank Assets/Total Bank assets 1995
Liquid Liabilities/GDP in 1995
Private Credit/GDP in 1995
Log of GDP per Capita 1995
Independent Variables
Dependent Variable: ∆DB
Dependent Variable: ∆GB
Dependent Variable: ∆FB
Raw
Reg. Coefficients
Raw
Reg. Coefficients
Raw
Reg. Coefficients
Correlations (Robust t-stats) Correlations (Robust t-stats) Correlations (Robust t-stats)
PANEL A: Initial Level of Development
0.006
0.001
0.041
0.006
0.211**
0.031*
(0.05)
(1.94)
(0.44)
0.076
0.067
0.301***
0.146***
-0.192*
-0.203***
(0.72)
(3.45)
(-3.75)
0.060
0.060
0.187
0.044
-0.158
-0.163**
(0.52)
(0.65)
(-2.47)
-0.114
0.166
0.092
0.122
0.172
-0.194**
(-2.00)
(0.81)
(1.07)
-0.113
-0.069*
0.304***
0.129**
-0.216*
-0.105**
(-1.92)
(2.57)
(-2.09)
PANEL B: Governance Measures
0.048
0.014
-0.048
-0.078***
0.115
0.035
(0.54)
(-3.59)
(1.13)
0.036
0.015
-0.037
0.127
-0.057**
0.042*
(0.62)
(-2.00)
(1.67)
-0.037
-0.028
0.020
-0.084
0.181*
-0.075*
(-0.84)
(0.63)
(-1.96)
-0.032
-0.028
0.086
-0.026
0.038
0.006
(-0.85)
(-0.87)
(0.21)
-0.010
-0.007
0.127
-0.022
-0.046
-0.062
(-0.22)
(-0.72)
(-1.39)
-0.092
-0.049
0.137
-0.003
-0.008
-0.026
(-1.61)
(-0.14)
(-0.77)
PANEL C: Crisis and Stability & Legal Origin
0.106
0.038
-0.301***
-0.097*
0.276***
0.151***
(0.78)
(-1.97)
(2.76)
-0.213**
-0.090**
0.152
0.066
-0.211**
-0.106***
(-2.23)
(1.40)
(-2.65)
110
Table 5: Which countries experienced more changes in bank ownership structure? (continued)
This table shows results from the following OLS regressions:
∆OWNERSHIP= α + β1PRIVCREDIT95 + β2COMMON + β3FR + β4CRISIS + ΛGOVERNANCE
where ∆OWNERSHIP is the change in the ownership variables between 1995 and 2005; PRIVCREDIT95 is the private credit/GDP as
of 1995; COMMON is a dummy variable which equals 1 if the country is a common law country; FR is a dummy which equals one if
the country has any restrictions on foreign ownership of banks; CRISIS is a dummy for countries suffering a systemic banking crisis
during the period, and GOVERNANCE refers to each of six governance indicators: voice and accountability; political stability;
government effectiveness; regulatory quality; rule of law, and control of corruption. Robust (White) t-statistics are shown in
parentheses. ***Significant at 1 percent level; **significant at 5 percent level; *significant at 10 percent level. The governance
variables were obtained from Kaufmann et al. (2006), and they are measured as of 1996. Higher values correspond to better
governance.
111
R
N
2
Intercept
Crisis Dummy
Foreign Restriction
Control of corruption (1996)
Rule of law (1996)
Regulatory quality (1996)
Government effectiveness (1996)
Political Stability (1996)
Voice & accountability (1996)
Common Law dummy
Independent Variables
Private credit/GDP in 1995
Table 5 (continued)
0.193
85
0.086
85
0.093
85
0.106
85
0.103
85
PANEL A - Changes in Private Blockholder Ownership of Banks
Dependent Variable: ∆DB
0.143
0.092
0.114
0.162
0.137
(1.38)
(1.07)
(1.29)
(1.52)
(1.49)
-0.107**
-0.113**
-0.122**
-0.117**
-0.117**
(-2.10)
(-2.24)
(-2.31)
(-2.35)
(-2.34)
0.093**
-0.005
(2.23)
(-0.19)
-0.027
-0.024
(-0.71)
(-0.79)
-0.051
-0.045
(-0.60)
(-1.09)
-0.070
-0.049
(-1.36)
(-1.29)
0.152*
(1.69)
-0.136**
(-2.11)
0.050
0.067
0.063
0.053
0.054
(0.99)
(1.48)
(1.37)
(1.14)
(1.21)
0.011
0.031
0.031
0.009
0.022
(0.23)
(0.72)
(0.70)
(0.20)
(0.53)
0.012
0.008
-0.002
0.002
0.011
(0.24)
(0.16)
(-0.04)
(0.05)
(0.24)
0.117
85
-0.047
(-1.55)
0.046
(0.97)
0.005
(0.13)
0.004
(0.08)
0.158*
(1.80)
-0.114**
(-2.30)
(continued)
0.091
85
0.057
(1.18)
0.020
(0.45)
0.007
(0.14)
-0.024
(-0.60)
0.122
(1.21)
-0.113**
(-2.24)
112
R
N
2
Intercept
Crisis Dummy
Foreign Restriction
Control of corruption (1996)
Rule of law (1996)
Regulatory quality (1996)
Government effectiveness (1996)
Political Stability (1996)
Voice & accountability (1996)
Common Law dummy
Independent Variables
Private credit/GDP in 1995
Table 5 (continued)
0.315
85
0.263
85
0.233
85
0.200
85
0.206
85
PANEL B - Changes in Government Ownership of Banks
Dependent Variable: ∆GB
0.180***
0.230***
0.208***
0.181***
0.183***
(3.01)
(4.30)
(3.62)
(3.05)
(3.77)
-0.019
-0.004
-0.007
0.015
0.013
(-0.34)
(-0.07)
(-0.13)
(0.29)
(0.27)
-0.119***
-0.072***
(-3.03)
(-3.00)
-0.030
-0.056**
(-0.75)
(-2.22)
0.092
-0.024
(1.16)
(-0.82)
0.029
-0.041
(0.67)
(-1.35)
-0.090
(-1.31)
0.061
(1.01)
0.123*
0.113*
0.115**
0.121**
0.117**
(1.93)
(1.92)
(2.07)
(2.03)
(2.00)
-0.095**
-0.126**
-0.112**
-0.121**
-0.118**
(-2.05)
(-2.63)
(-2.36)
(-2.43)
(-2.42)
-0.232***
-0.219***
-0.228***
-0.205***
-0.200***
(-5.83)
(-5.87)
(-5.74)
(-5.30)
(-5.21)
0.205
85
-0.028
(-1.10)
0.115*
(1.94)
-0.125**
(-2.46)
-0.205***
(-5.31)
0.184***
(3.37)
0.016
(0.31)
(continued)
0.216
85
0.107*
(1.84)
-0.134***
(-2.65)
-0.206***
(-5.40)
-0.049*
(-1.77)
0.213***
(3.62)
0.014
(0.28)
113
R
N
2
Intercept
Crisis Dummy
Foreign Restriction
Control of corruption (1996)
Rule of law (1996)
Regulatory quality (1996)
Government effectiveness (1996)
Political Stability (1996)
Voice & accountability (1996)
Common Law dummy
Independent Variables
Private credit/GDP in 1995
Table 5 (continued)
0.239
85
0.203
85
0.199
85
0.164
85
0.188
85
PANEL C - Changes in Foreign Ownership of Banks
Dependent Variable: ∆FB
-0.125**
-0.150***
-0.146***
-0.140**
-0.147***
(-2.05)
(-3.72)
(-2.94)
(-2.31)
(-3.55)
-0.072*
-0.069*
-0.051
-0.057
-0.048
(-1.22)
(-1.41)
(-1.22)
(-1.77)
(-1.76)
0.068***
0.035
(0.72)
(2.95)
0.047
0.069***
(1.17)
(2.64)
-0.107
0.045
(-1.45)
(1.49)
0.080**
0.048
(0.92)
(2.62)
-0.005
(-0.05)
0.066
(0.98)
-0.049
-0.059
-0.058
-0.059
-0.051
(-0.82)
(-1.12)
(-1.12)
(-1.06)
(-0.93)
0.118**
0.135***
0.122**
0.143**
0.136***
(2.22)
(2.68)
(2.48)
(2.63)
(2.70)
0.179***
0.175***
0.192***
0.166***
0.157***
(4.01)
(4.55)
(4.52)
(4.12)
(4.08)
0.178
85
-0.046
(-0.82)
0.151***
(2.83)
0.165***
(4.14)
0.063*
(1.90)
-0.157***
(-2.66)
-0.074*
(-1.80)
0.192
85
0.061**
(2.28)
-0.046
(-0.83)
0.154***
(2.84)
0.166***
(4.23)
-0.157***
(-2.96)
-0.074*
(-1.77)
Dependent Variable
Independent Variables
∆DB (1995-2000)
Industry Credit Growth1
Industry Credit Growth1
0.059*
(1.86)
∆GB (1995-2000)
Industry Credit Growth1
-0.241***
(-3.72)
∆FB (1995-2000)
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Intercept
R2
Year Fixed Effects
N
Industry Credit Growth1
0.253**
(2.52)
0.750***
(3.41)
3.099***
(8.01)
0.031*
(1.72)
-0.029
(-1.17)
0.272***
(2.66)
0.783***
(3.52)
3.151***
(7.96)
0.037*
(1.95)
-0.042
(-1.49)
0.193**
(2.13)
0.805***
(3.75)
3.000***
(7.95)
0.041**
(2.19)
-0.052**
(-2.06)
0.150**
(2.45)
0.242**
(2.51)
0.685***
(3.04)
2.882***
(7.63)
0.043**
(2.13)
-0.035
(-1.39)
0.137
Yes
1031
0.138
Yes
1031
0.163
Yes
1031
0.148
Yes
1031
Pooled regressions. The dependent variable is the annual growth in industry credit for each
of 41 countries in the sample from 2001-2006. ∆DB, ∆GB, ∆FB refer to changes in
domestic, government, and foreign ownership of banks between 1995 and 2000. Lending rate
is the average lending rate for a country; inflation is the growth rate in CPI for each country;
the growth in GDP represents the growth in real GDP per capita. The latter variables were
obtained from the World Bank's World Development Indicators database. Common law
dummy equals one if the country has common law origin of its commercial laws. Year fixed
effects are included in all regressions. Standard errors are clustered by country. Robust
(White) t-statistics are shown in parentheses. *, **, and *** indicate significance at the 10%,
5%, and 1% level, respectively. Data on credit provided to industry segments was obtained
from various Central Banks and Financial Institutions Regulatory Authorities. In general, the
data captures the total credit provided to specific industries by the banking system. In some
countries, total credit includes credit by banks and nonbank financial institutions (e.g. credit
unions, finance companies).
Table 6: Effect of changes in bank ownership structure on industry credit growth
114
Pooled regressions. The dependent variable is the annual growth in industry credit for each
of 41 countries in the sample from 2001-2006. The HIVA dummy equals one if industry i's
average value added is above the median valued added for all industries in its country.
Average industry value added is obtained from 1995-2000 in Panel A, and from 1990-1995 in
Panel B. ∆DB, ∆GB, and ∆FB are the changes in domestic, government, and foreign
ownership of banks between 1995 and 2000. Lending rate is the average lending rate for a
country; inflation is the growth rate in CPI for each country; the growth in GDP represents
the growth in real GDP per capita. The latter variables were obtained from the World Bank's
World Development Indicators database. Common law dummy equals one if the country has
common law origin of its commercial laws. Year fixed effects are included in all regressions.
Standard errors are clustered by country. Robust (White) t-statistics are shown in
parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.
Table 7: Effect of changes in bank ownership structure on the allocation of credit
(continued)
115
Table 7 (continued)
Independent Variables
HIVA DUMMY
HIVA*∆DB
HIVA*∆GB
Panel A - Industries are classified based on value added between 1995 and 2000
Dependent Variable
Industry Credit Growth1 Industry Credit Growth1 Industry Credit Growth1
0.020
0.027
0.012
(1.07)
(0.83)
(0.54)
-0.107*
(-2.07)
-0.077
(-0.67)
HIVA*∆FB
Industry Credit Growth1
-0.005
(-0.24)
0.238**
(2.19)
∆DB (1995-2000)
∆GB (1995-2000)
∆FB (1995-2000)
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Intercept
R2
Year Fixed Effects
N
Independent Variables
HIVA DUMMY
-0.002
(-0.04)
-0.162
(-1.29)
-0.084
(-0.78)
0.231
(0.66)
0.626
(1.29)
2.690***
(4.13)
0.044
(1.50)
-0.041
(-0.75)
0.075
(1.22)
-0.066
(-0.65)
0.351*
(2.18)
0.502**
(2.87)
2.612***
(5.37)
0.048**
(2.68)
-0.041
(-0.95)
-0.087
(-0.86)
0.314
(0.90)
0.492
(1.04)
2.573***
(3.98)
0.046
(1.57)
-0.021
(-0.46)
0.246
(0.71)
0.568
(1.17)
2.606***
(4.17)
0.047*
(1.85)
-0.036
(-0.81)
0.118
0.114
0.116
Yes
Yes
Yes
1031
1031
1031
Panel B - Industries are classified based on value added between 1990 and 1995
Dependent Variable
1
Industry Credit Growth
0.020
(1.10)
HIVA*∆DB
1
Industry Credit Growth
0.027
(0.85)
-0.108*
(-2.09)
HIVA*∆GB
0.119
Yes
1031
1
Industry Credit Growth
0.013
(0.56)
1
Industry Credit Growth
-0.004
(-0.21)
-0.076
(-0.66)
HIVA*∆FB
0.236**
(2.18)
∆DB (1995-2000)
∆GB (1995-2000)
∆FB (1995-2000)
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Intercept
2
R
Year Fixed Effects
N
-0.002
(-0.04)
-0.162
(-1.29)
-0.084
(-0.78)
0.231
(0.66)
0.626
(1.29)
2.690***
(4.13)
0.044
(1.50)
-0.041
(-0.75)
0.075
(1.23)
0.351*
(2.18)
0.502**
(2.85)
2.612***
(5.38)
0.048**
(2.68)
-0.041
(-0.96)
0.246
(0.71)
0.568
(1.17)
2.606***
(4.17)
0.047*
(1.85)
-0.036
(-0.81)
-0.087
(-0.85)
0.314
(0.90)
0.491
(1.04)
2.573***
(3.98)
0.046
(1.57)
-0.021
(-0.47)
0.118
Yes
1031
0.114
Yes
1031
0.117
Yes
1031
0.119
Yes
1031
-0.066
(-0.65)
116
Panel A- First Stage Regression Results
Dependent Variable:
∆DB
-0.236
(-1.40)
0.281***
(2.84)
0.010
(0.29)
Independent Variables:
Bank Concentration
WIDE95
Rule of Law
Productive Population (% total)
∆FB
1.175**
(2.24)
-0.023
(-0.82)
Political Stability
Transition Economy
Lending Rate
Inflation
Growth in GDP
Common Law Dummy
Intercept
2
R
N
Partial F-statistic
2
Partial R
2
Overidentifying restrictions test (c )
p-value
Hausman test (Wu's F-statistic)
p-value
-0.118
(-0.75)
0.195
(0.66)
-0.984
(-1.05)
-0.041
(-0.70)
0.117
(1.13)
-0.004
(-0.04)
0.965
(1.34)
-0.093*
(-1.85)
-0.617*
(-1.92)
0.239
52
3.40**
0.243
52
2.75*
0.185
0.193
1.03
(0.309)
0.07
(0.976)
1.81
(0.179)
0.42
(0.659)
Two-Stage Least Squares Regressions. In the first-stage regressions a proxy for the size of
the productive population as a percent of total population, and Kaufmann et al. (2006)
political stability index as of the beginning of the period (1996) are used as instruments for
changes in foreign ownership of banks. A bank concentration proxy (the assets of the top 3
banks as a percent of all commercial banks’ assets), a measure of dispersed ownership of
banks as of 1995 and Kaufman et al. (2006) rule of law index as of 1996 are used to forecast
changes in domestic blockholder ownership of banks. The control variables are measured as
of 1995. Following Larcker & Rusticus (2007) the partial R2, which shows the explanatory
power of the instruments that are unique to the first-stage regression, is computed as R2p=
(R2y, z - R2y,z1)/(1-R2y,z1), where z is the combined set of control variables (z1) and instrumental
variables. The overidentifying restriction test statistic is obtained from a regression of the
second-stage residuals on all exogenous variables. The nR2 from this model is distributed
χ2K-L, where K is the number of instruments (2 and 3 for the changes in FB and DB,
respectively) and L is the number of endogenous explanatory variables (1). Panel A shows
results from the first-stage regressions, while Panel B shows second-stage regression results.
Robust (White) t-statistics are shown in parentheses. *, **, *** indicate significance at the
10%, 5%, and 1% level, respectively.
Table 8: Regressions using instrumental variables for changes in bank ownership
(continued)
117
Table 8 (continued)
Panel B - Second Stage Regressions: Industries are classified based on value added between 1990 and 1995
Dependent Variable
Independent Variables
HIVA DUMMY
1
Industry Credit Growth
0.021
(1.31)
HIVA*∆DBIV
1
Industry Credit Growth
0.037*
(1.82)
-0.247*
(-1.74)
HIVA*∆GB
1
Industry Credit Growth
0.013
(0.56)
1
Industry Credit Growth
0.066*
(1.94)
-0.076
(-0.66)
HIVA*∆FBIV
0.349*
(1.88)
∆DBIV
-0.013
(-0.11)
∆GB
-0.066
(-0.65)
∆FBIV
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Intercept
2
R
Year Fixed Effects
N
0.335**
(1.99)
0.491**
(2.15)
2.589
(6.55)
0.046**
(2.51)
-0.033
(-1.05)
0.237
(1.26)
0.534**
(2.31)
2.419***
(5.59)
0.041**
(2.19)
-0.017
(-0.44)
0.112
Yes
1031
0.116
Yes
1031
118
0.246
(0.71)
0.568
(1.17)
2.606***
(4.17)
0.047*
(1.85)
-0.036
(-0.81)
0.117
Yes
1031
0.028
(0.16)
0.417**
(2.49)
0.394*
(1.72)
2.431***
(5.48)
0.071***
(3.12)
-0.088**
(-2.26)
0.118
Yes
1031
Pooled regressions. The dependent variable is the annual growth in industry credit for each
of 41 countries in the sample from 2001-2006. In Panel A, the HIVA dummy equals one if
industry i's average value added (1990-1995) is above the median valued added for all
industries in its country. In Panel B, HIVA dummy equals one if the slope of industry i's
value added time trend (1970-1995) is above the median for all industries in its country.
Gov’t-to-Domestic dummy equals one if the increase in DB and the decline in GB exceed
10%. ∆DB, ∆GB, and ∆FB are the changes in domestic, government, and foreign ownership
of banks between 1995 and 2000. Lending rate is the average lending rate for a country;
inflation is the growth rate in CPI for each country; the growth in GDP represents the growth
in real GDP per capita. The latter variables were obtained from the World Bank's World
Development Indicators database. Common law dummy equals one if the country has
common law origin of its commercial laws. Year fixed effects are included in all regressions.
Standard errors are clustered by country. Robust (White) t-statistics are shown in
parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.
Table 9: Testing the impact of domestic blockholder ownership of banks (continued)
119
Table 9 (continued)
Panel A - Industries are classified based on value added between 1990 and 1995
Dependent Variable
Independent Variables
HIVA DUMMY
HIVA*Gov't to Domestic Dummy
Industry Credit Growth1
0.032
(1.57)
-0.094***
(-3.77)
HIVA*∆DB
Industry Credit Growth1
0.032
(0.95)
-0.092*
(-1.99)
-0.006
(-0.11)
-0.126
(-1.30)
HIVA*∆GB
∆DB (1995-2000)
0.045
(0.70)
∆GB (1995-2000)
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Gov't to Domestic Dummy
Intercept
Industry Credit Growth1
0.021
(0.94)
-0.108***
(-3.96)
0.330
(0.92)
0.490
(1.05)
2.581***
(3.77)
0.046
(1.60)
0.046*
(1.88)
-0.037
(-0.75)
0.356*
(2.12)
0.501**
(2.79)
2.599***
(5.48)
0.049**
(2.60)
0.029
(0.95)
-0.043
(-1.00)
-0.048
(-0.44)
0.228
(0.65)
0.562
(1.15)
2.564***
(4.02)
0.046*
(1.82)
0.038
(1.53)
-0.035
(-0.76)
R2
0.115
0.116
0.121
Year Fixed Effects
Yes
Yes
Yes
N
1031
1031
1031
Panel B - Industries are classified based on the slope of the value added time trend between 1970 and 1995
Dependent Variable
Independent Variables
HIVA DUMMY
Industry Credit Growth1
0.018
(1.13)
HIVA*∆DB
Industry Credit Growth1
0.024
(1.54)
-0.107**
(-1.99)
HIVA*∆GB
Industry Credit Growth1
0.009
(0.47)
-0.083
(-0.67)
HIVA*∆FB
∆DB (1995-2000)
∆GB (1995-2000)
∆FB (1995-2000)
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Intercept
R2
Year Fixed Effects
N
Industry Credit Growth1
-0.008
(-0.47)
0.243**
(2.14)
0.335**
(1.99)
0.492**
(2.15)
2.590***
(6.55)
0.046**
(2.51)
-0.026
(-1.06)
-0.008
(-0.34)
-0.054**
(-2.43)
-0.031
(-1.00)
0.075
(1.54)
0.351*
(1.96)
0.502**
(2.19)
2.612***
(6.28)
0.048**
(2.37)
-0.040
(-1.11)
0.245
(1.45)
0.568**
(2.39)
2.606***
(6.62)
0.047**
(2.56)
-0.034
(-1.09)
-0.090
(-1.44)
0.313*
(1.89)
0.492**
(2.15)
2.573***
(6.56)
0.046**
(2.45)
-0.020
(-0.64)
0.111
Yes
1031
0.113
Yes
1031
0.116
Yes
1031
0.119
Yes
1031
-0.063
(-0.94)
120
Dependent Variable
Independent Variables
HIFINDEP DUMMY
1
Industry Credit Growth
3.008**
(2.27)
HIFINDEP*∆DB
Industry Credit Growth
3.172**
(2.26)
-3.282*
(-1.86)
HIFINDEP*∆GB
1
1
Industry Credit Growth
1.804**
(2.26)
1
Industry Credit Growth
2.344**
(2.20)
-13.815**
(-1.97)
HIFINDEP*∆FB
10.758**
(2.13)
∆DB
2.586**
(2.12)
∆GB
-5.692***
(-2.59)
∆FB
Lending Rate
Inflation
Growth in GDP
COMMON LAW DUMMY
Intercept
2
R
Year Fixed Effects
N
-9.237***
(-2.74)
37.433***
(3.18)
0.435
(0.06)
3.722***
(3.58)
2.020
(0.99)
0.032
Yes
812
-8.970***
(-2.72)
38.665***
(3.14)
1.620
(0.21)
3.819***
(3.62)
1.716
(0.82)
0.032
Yes
812
-7.723**
(-2.46)
33.371***
(3.02)
9.103
(1.17)
3.723***
(3.65)
1.259
(0.64)
0.042
Yes
812
5.386***
(3.05)
-6.992**
(-2.37)
26.474***
(2.83)
1.337
(0.18)
4.424***
(3.58)
1.425
(0.74)
0.037
Yes
812
Pooled regressions. The dependent variable is the annual growth in industry credit for each
of 32 countries in the sample from 2001-2006. The HIFINDEP dummy equals one if
industry i’s median external finance dependence ratio (1995-2000) is above the median for all
industries in its country, and 0 otherwise. ∆DB, ∆GB, ∆FB represent the changes in
domestic, government, and foreign ownership of banks between 1995 and 2000. Lending
rate is the average lending rate for a country; inflation is the growth rate in CPI for each
country; the growth in GDP represents the growth in real GDP per capita. The latter
variables were obtained from the World Bank's World Development Indicators database.
Common law dummy equals one if the country has common law origin of its commercial
laws. Year fixed effects are included in all regressions. Standard errors are clustered by
country. Robust (White) t-statistics are shown in parentheses. *, **, and *** indicate
significance at the 10%, 5%, and 1% level, respectively.
Table 10: Credit growth in industries by extent of dependence on external finance
121
The table shows descriptive statistics for two measures of capital allocation efficiency. The
measures, ηc (original) and ηc (new), are estimated from the following regressions, following
Wurgler (2000):
I
VAict
ln ict = α c + η c ln
+ ε ict
I ict −1
VAict −1
where I is real gross fixed capital formation, VA is real value added, i indexes firms and/or
industries, c indexes countries, and t indexes years. The original measure (first column) is
constructed using industry level data for gross fixed capital formation and value added
obtained from UNIDO's INDSTAT3 database using data from 1995 through 2004 or the
latest available date. The new measure ηc(new) is estimated using firm-level data from 2001
through 2005 for real fixed assets as a proxy for gross fixed capital formation, and for sales
growth as a proxy for value added. Data was obtained from Thomson Financial's
WorldScope database. The correlation between the two measures, ηc (original) and ηc (new)
is 0.36 and is significant at the 5% level. When the new measure is estimated using the 19952005 time period, the correlation increases to 0.45 and is significant at the 1% level.
Table 11: Descriptive statistics - measures of capital allocation efficiency (continued)
122
Table 11 (continued)
Country
Argentina
Australia
Austria
Bangladesh
Belgium
Bolivia
Brazil
Bulgaria
Canada
Chile
China
Colombia
Cyprus
CzechRepublic
Denmark
Egypt
ElSalvador
Finland
France
Germany
Greece
HongKong
Hungary
India
Indonesia
Iran
Ireland
Israel
Italy
Japan
Jordan
Kenya
Korea
Kuwait
Macau
Malaysia
Mexico
Morocco
Netherlands
NewZealand
Norway
Oman
Pakistan
Panama
Peru
Philippines
Poland
Portugal
Romania
Russia
Singapore
Slovenia
SouthAfrica
Spain
SriLanka
Sweden
Taiwan
Tanzania
Thailand
Tunisia
Turkey
UK
USA
Uruguay
Venezuela
Mean
Std. Dev.
Correlation (h, hnew)
ηc (original)
.
0.640
0.694
0.265
0.103
0.280
.
0.568
.
0.178
.
0.203
0.404
0.909
.
.
-0.029
0.750
0.516
0.185
0.072
0.560
0.543
0.204
-0.124
0.639
0.290
0.659
1.138
0.411
0.202
0.100
0.806
0.451
-0.397
0.488
0.272
0.576
0.807
.
0.352
0.299
.
-0.158
.
.
0.288
0.288
0.410
.
0.545
0.638
.
0.753
0.189
.
.
1.007
.
0.338
0.439
0.873
0.883
0.480
0.263
0.425
0.311
0.363***
123
ηc (new)
0.282
0.184
0.289
.
0.220
.
0.373
.
0.272
0.054
0.118
-0.132
.
0.429
0.428
0.076
.
0.702
0.278
0.220
0.110
0.141
0.195
0.202
0.160
.
0.344
0.035
0.220
0.309
.
.
0.075
.
.
0.266
0.137
0.128
0.349
0.182
0.170
.
0.135
.
0.275
0.113
0.230
0.196
.
0.209
0.164
.
0.367
0.172
0.025
0.195
0.099
.
0.094
.
0.176
0.316
0.150
.
0.314
0.209
0.132
Two-Stage Least Squares Regressions. In the first-stage regressions a proxy for the size of
the productive population as a percent of total population, and Kaufmann et al. (2006)
political stability index as of the beginning of the period are used as instruments for changes
in foreign ownership of banks. A bank concentration proxy (the assets of the top 3 banks as a
percent of all commercial banks’ assets) and a measure of dispersed ownership of banks as of
1995 are used to forecast changes in domestic blockholder ownership of banks. The control
variables are a summary measure of financial development, FD, the log of one plus the
average sum of stock market capitalization and private credit to GDP; dummies for common
law origin of commercial laws and systemic banking crisis. Following Larcker & Rusticus
(2007) the partial R2, which shows the explanatory power of the instruments that are unique
to the first-stage regression, is computed as R2p= (R2y, z – R2y,z1)/(1-R2y,z1), where z is the
combined set of control variables (z1) and instrumental variables. The overidentifying
restriction test statistic is obtained from a regression of the second-stage residuals on all
exogenous variables. The nR2 from this model is distributed χ2K-L, where K is the number of
instruments (2) and L is the number of endogenous explanatory variables (1). Panels B & C
show results for OLS regressions for the cross-section of countries. The dependent variable is
the estimate of the elasticity of industry investment to industry value added, ηc, estimated
from the following regressions, following Wurgler (2000):
ln
I ict
VA ict
= α c + η c ln
+ ε ict
I ict −1
VA ict −1
where i indexes ISIC-3 manufacturing industries, c indexes countries, and t indexes years. I is
real gross fixed capital formation, and VA is real value added. These estimate were
constructed using data from 1995-2004. The independent variables are a summary measure of
financial development, FD, the log of one plus the average sum of stock market capitalization
and private credit to GDP; dummies for common law origin of commercial laws and systemic
banking crisis, and the changes in domestic, government, and foreign blockholder ownership
of banks between 1995 and 2000 (these variables are standardized to have unit variance).
Panel B shows the results using the actual values for the changes in bank ownership structure,
while Panel C uses the instrumental variables for changes in domestic and foreign ownership
of banks. Finally, Panel D shows differences in the impact of changes in bank ownership
structure on capital allocation efficiency between countries with good and poor governance,
using Transparency International’s Corruption Perception Index (CPI). Countries with a CPI
below the median for all countries in the sample are considered poor governance countries.
The changes in domestic blockholder and foreign ownership of banks are instrumented to
mitigate endogeneity concerns.
Table 12: Impact of changes in bank ownership structure on capital allocation –
instrumental variables approach (continued)
124
Table 12 (continued)
Panel A- First-Stage Regression Results
Independent Variables:
Dependent Variable:
∆DB
∆FB
Bank Concentration
-0.369**
(-2.43)
WIDE95
0.276***
(2.74)
Productive Population (% total)
1.243**
(2.26)
Political Stability
0.080**
(2.17)
FD
-0.030
-0.138***
(-0.9)
(-3.9)
Common Law Dummy
-0.073
-0.018
(-1.3)
(-0.34)
Crisis Dummy
-0.036
0.081
(-0.62)
(1.54)
Intercept
0.219*
-0.592*
(1.89)
(-1.71)
R2
N
Partial F-statistic
Partial R2
2
Overidentifying restrictions test (c )
p-value
Hausman test (Wu's F-statistic)
p-value
0.293
49
7.00***
0.440
49
7.47***
0.246
0.258
1.78
(0.182)
0.62
2.42
(0.120)
2.99*
(0.434)
(0.091)
(continued)
125
Table 12 (continued)
Panel B- Actual Changes in Bank Ownership Structure
(1)
(2)
(3)
Dependent Variable: ηc
∆DB
-0.070**
(-2.22)
∆GB
-0.052
(-1.12)
∆FB
0.112**
(2.58)
FDc
0.078
0.094*
0.107**
(1.46)
(1.69)
(2.28)
Common Law Dummy
0.001
0.045
0.069
(0.01)
(0.48)
(0.79)
Crisis Dummy
0.124
0.110
0.069
(1.32)
(1.12)
(0.67)
Intercept
0.347***
0.265***
0.241***
(5.26)
(2.79)
(3.08)
R2
N
Dependent Variable: ηc
∆DBIV
∆GB
∆FBIV
FD
Common Law Dummy
Crisis Dummy
Intercept
R2
N
Panel C - Instrumental Variables for ∆DB & ∆FB
(1)
(2)
(3)
Dependent Variable: ηc
∆DBIV
-0.073*
(-1.93)
∆GB
-0.371
(-1.41)
∆FBIV
0.152***
(2.84)
FDc
0.082
0.123**
0.143**
(1.56)
(2.07)
(2.26)
Common Law Dummy
-0.031
0.041
0.145
(-0.3)
(0.46)
(1.43)
Crisis Dummy
0.123
0.104
0.030
(1.26)
(1.08)
(0.34)
Intercept
0.370***
0.217**
0.114
(5.13)
(2.13)
(0.96)
0.132
0.098
0.183
R2
0.113
0.110
0.188
49
49
49
N
49
49
49
Panel D - By Quality of Governance (Transparency International's Corruption Perceptions Index)
Poor Governance Countries
Good Governance Countries
(1)
(2)
(3)
(1)
(2)
(3)
Dependent Variable: ηc
-0.033
∆DBIV
-0.157*
(-1.84)
(-0.75)
-0.287
∆GB
-0.364
(-0.89)
(-0.95)
0.072
∆FBIV
0.133*
(0.61)
(1.73)
-0.073
-0.042
-0.025
FD
0.027
0.054
0.115
(-0.7)
(-0.31)
(-0.13)
(0.34)
(0.66)
(1.04)
-0.187
-0.002
0.070
Common Law Dummy
(-1.38)
(-0.01)
(0.38)
-0.003
0.067
0.034
Crisis Dummy
0.018
0.059
0.114
(-0.03)
(0.46)
(0.27)
(0.14)
(0.51)
(1.03)
0.470***
0.209
0.202
Intercept
0.449***
0.349**
0.192
(4.52)
(1.21)
(0.97)
(4.31)
(2.68)
(1.06)
0.225
22
0.153
22
0.132
22
R2
N
126
0.109
27
0.119
27
0.155
27
Panel A
Dependent Variable: ηcNEW
∆DB
(1)
-0.044*
(-1.80)
(2)
∆GB
(3)
-0.012
(-0.65)
∆FB
FD
Common Law Dummy
Crisis Dummy
Intercept
0.046***
(3.15)
-0.020
(-0.50)
0.042
(1.07)
-0.186
(-1.44)
0.060**
(2.17)
-0.051
(-1.17)
0.032
(0.76)
0.154***
(3.59)
2
0.2609
0.1003
46
46
Panel B - By Protection of Minority Shareholders
(1)
(2)
Dependent Variable: ηcNEW
∆DB
-0.010
(-0.60)
∆DB*POOR PROTECTION DUMMY
-0.089*
(-1.80)
∆GB
0.053
(1.34)
∆GB*POOR PROTECTION DUMMY
-0.077
(-1.44)
∆FB
R
N
Common Law Dummy
Crisis Dummy
Poor Protection Dummy
Intercept
2
R
N
0.2562
46
(3)
0.063**
(2.09)
-0.012
(-0.27)
0.045
(1.15)
0.079*
(1.95)
0.105**
(2.19)
0.064**
(2.26)
-0.020
(-0.47)
-0.001
(-0.02)
0.051
(1.08)
0.126***
(2.95)
0.016
(0.34)
0.056
(1.04)
0.077**
(2.69)
0.001
(0.03)
0.013
(0.32)
0.038
(0.90)
0.082
(1.59)
0.3078
46
0.2195
46
0.3036
46
∆FB*POOR PROTECTION DUMMY
FD
0.069**
(2.40)
0.076***
(2.69)
-0.013
(-0.35)
0.020
(0.46)
0.104**
(2.24)
OLS regressions for the cross-section of countries. The dependent variable is the estimate of the
elasticity of investment to value added estimated from the following firm-level regression, similar to
I ict
VA ict
Wurgler (2000):
ln
= α + η ln
+ ε
I ict
−1
c
c
VA
ict − 1
ict
where i indexes firms, c indexes countries, and t indexes years. I is growth in real tangible fixed assets
from Worldscope (code 339), which is used as a proxy for gross fixed capital formation, and VA is the
firm's annual real sales growth used as a proxy for value added These estimates were constructed
using data obtained from DataStream and WorldScope for the years 2001 through 2005. The
independent variables are a summary measure of financial development, FD, the log of one plus the
average sum of stock market capitalization and private credit to GDP; a common law dummy, a
dummy for systemic banking crisis, and the changes in domestic, government, and foreign blockholder
ownership of banks between 1995 and 2000 (these variables were standardized to have unit variance).
In Panel B, a poor protection dummy is introduced which equals one if the country's Anti-Self Dealing
Index (Djankov et al. 2007) falls below the median for all countries in the sample. Robust t-statistics
are shown in parentheses.
Table 13: Regressions using new measure of capital allocation efficiency
127
Panel A - Data Availabity for the sample of banks
Asset Quality Measures:
Profitability Measures:
NPL_GL
LLR_NPL
ROAA
ROAE
# of banks
# of
# of banks
# of
# of banks
# of
# of banks
# of
Year
countries
countries
countries
countries
1998
325
76
352
75
495
90
535
90
1999
346
79
377
80
499
90
542
90
2000
344
81
380
83
499
90
543
90
2001
348
79
381
81
499
90
543
90
2002
216
67
249
74
326
82
369
87
2003
213
69
244
75
320
83
362
87
2004
201
65
232
72
304
80
345
86
Operational Measures:
Valuation Measure:
NIM
Cost_I
Tobin's Q Measure
# of banks
# of
# of banks
# of
# of banks
# of
Year
countries
countries
countries
1998
488
90
479
89
156
45
1999
493
90
484
90
160
45
2000
494
90
489
90
161
45
2001
496
90
490
90
163
46
2002
323
82
317
82
162
46
2003
317
83
312
83
163
46
2004
301
80
299
79
163
46
Panel B - Descriptive statistics of sample (1998-2004)
Mean
Std. Dev.
Max
Min
Performance Measure:
NPL_GL
7.92%
0.076
39.72%
0.00%
LLR_NPL
96.98%
0.764
617.56%
0.00%
0.77%
0.023
10.96%
-38.66%
ROA
9.61%
0.220
84.65%
-316.78%
ROE
3.62%
0.034
48.18%
-23.02%
NIM
Cost-to-Income
60.10%
0.223
433.08%
8.73%
Q
1.675
1.176
14.574
0.102
Panel C - Average Performance Measures (1998-2004) - Emerging vs. Developed
Emerging
Developed
Performance Measure:
9.19%
2.76%
NPL_GL
85.16%
145.37%
LLR_NPL
ROA
0.84%
0.52%
ROE
10.00%
8.30%
4.15%
1.75%
NIM
Cost-to-Income
60.66%
58.18%
Q
1.396
1.983
The asset quality measures used are the non-performing loans-to-gross loans ratio (NPL_GL) and the
loan loss reserves-to-non-performing loans ratio (LLR_NPL). The return on average assets (ROAA)
and the return on average equity (ROAE) are the measures of profitability. The net interest margin
(NIM) and the cost-to-income ratio are the measures of operational efficiency. Finally, the bank’s Q
(proxied by the market-to-book value) is used as a valuation measure. All measures were obtained
from Bureau Van Dijk's BankScope database, with the exception of Q, which was obtained from
DataStream/WorldScope database. While the final sample of banks consists of 628 banks from 90
countries, data availability limits the number of banks per year as described above. Panel B shows
descriptive statistics for the full sample of banks, while Panel C splits the sample between emerging
and developed markets.
Table 14: Descriptive statistics - banks with available data by year
128
Independent Variables
Corruption Index
GB95
Assets (1995)
Market Share (1995)
NII-to-Assets (1995)
COMMON LAW DUMMY
CRISIS
FOREIGN RESTRICTIONS
Intercept
Observations
Pseudo R2
Panel A- First-Stage Probit Regressions
Dependent Variable:
Gov't-to-Domestic Dummy Independent Variables
Productive Population (1995)
0.497***
(2.59)
0.418
GB95
(1.01)
0.022
Assets (1995)
(0.33)
-2.395
Market Share (1995)
(-1.44)
-0.690
NII-to-Assets (1995)
(-0.23)
1.435***
COMMON LAW DUMMY
(2.61)
0.520
CRISIS
(1.53)
0.270
FOREIGN RESTRICTIONS
(0.86)
-3.575***
Intercept
(-4.39)
619
Observations
Pseudo R2
0.121
Dependent Variable:
Gov't-to-Foreign Dummy
2.973*
(1.74)
0.711***
(2.59)
-0.080*
(-1.81)
0.957*
(1.86)
2.015
(1.30)
-0.574**
(-2.45)
0.141
(0.54)
-0.226
(-0.83)
-3.034***
(-2.92)
609
0.165
The dependent variables in the first-stage probit regressions (Panel A) are the Gov't-toDomestic and Gov't-to-Foreign dummies, which equal one if a bank was under government
control (at the 20% threshold) as of 1995 and changed to domestic or foreign blockholder
control as of 2000, respectively, and 0 otherwise. The corruption index is the Kaufmann et
al. (2006) corruption index as of 1996; GB95 is the percentage of the assets of the top 10
banks in a country owned by the government as of 1995; productive population is the fraction
of people ages 15 through 64 as a fraction of the total population as of 1995; the other
controls are the lag of bank's assets (log), the ratio of non-interest income-to-assets and the
market share. Crisis is a dummy which equals one if the country experienced a banking crisis
between 2000 and 2005; foreign restriction is a dummy which equals one if the country has
any restrictions on foreign ownership of banks. From the first-stage regressions, I obtain the
lambda (the inverse Mills ratio from the Heckman equation) which is used as an explanatory
variable in the second stage regressions (Panel B), in which the dependent variables are the
various bank performance measures defined in Table 14 from 2001-2004. Robust t-statistics
are shown in parentheses; standard errors are clustered by bank. *, **, and *** indicate
significance at the 10%, 5%, and 1% level.
Table 15: Impact of changes in bank control on bank performance – Heckman Model
(continued)
129
Table 15 (continued)
Panel B - Changes in Control and Subsequent Bank Performance - Heckman Model
IMPACT OF CHANGES FROM GOVERNMENT TO DOMESTIC BLOCKHOLDER CONTROL
Dependent Variable
Asset Quality
Profitability
Operations
Independent Variables
NPL-to-GL LLR-to-NPL
ROAA
ROAE
NIM
Cost-to-Income
λ (Inverse Mills Ratio - Heckman)
0.108***
0.314
-0.004
-0.143**
-0.019
0.164
(3.15)
(1.11)
(-0.33)
(-2.56)
(-1.04)
(1.16)
DOMESTIC
-0.010
0.051
-0.003
-0.010
0.002
0.121
(-0.67)
(0.38)
(-0.44)
(-0.54)
(0.37)
(1.19)
GOVERNMENT
-0.123
-0.006
-0.006
0.038**
-0.080***
0.139***
(2.02)
(-1.18)
(-1.31)
(-3.63)
(-0.84)
(3.08)
Gov't-to-Domestic Dummy
0.209***
0.341
-0.014
-0.298**
-0.026
0.323
(2.75)
(0.54)
(-0.52)
(-2.50)
(-0.69)
(1.35)
Lag of Assets
-0.015***
0.032
0.001
0.013***
-0.002
-0.011
(-2.99)
(1.10)
(0.99)
(2.71)
(-1.29)
(-0.43)
Lag of Market Share
-0.039
-0.298
0.003
0.092
0.022
-0.119
(-0.74)
(-0.66)
(0.26)
(1.00)
(0.81)
(-1.05)
Lag of NII-to-Assets
0.267
-1.173*
0.103
-0.217
-0.010
0.253
(1.41)
(-1.77)
(1.29)
(-0.65)
(-0.19)
(0.63)
CRISIS
0.007
0.045
-0.005
0.024
0.046**
0.009
(0.34)
(0.36)
(-0.46)
(0.57)
(2.58)
(0.17)
Foreign Restrictions
0.026
0.083
0.025***
0.107***
0.036***
-0.054
(1.58)
(0.61)
(3.63)
(3.38)
(2.69)
(-1.36)
COMMON LAW DUMMY
-0.019
-0.043
0.002
0.070***
0.023**
-0.009
(-1.37)
(-0.37)
(0.44)
(3.02)
(2.56)
(-0.13)
Intercept
-0.031
1.024
0.014
0.281**
0.071**
0.265
(-0.37)
(1.54)
(0.67)
(2.52)
(2.24)
(0.59)
2
0.186
0.168
0.049
0.123
0.206
0.013
R
Regional Dummies
Yes
Yes
Yes
Yes
Yes
Yes
Year fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
N
952
947
1374
1135
1366
1358
IMPACT OF CHANGES FROM GOVERNMENT TO FOREIGN BLOCKHOLDER CONTROL
Dependent Variable
Asset Quality
Profitability
Operations
Independent Variables
NPL-to-GL LLR-to-NPL
ROAA
ROAE
NIM
Cost-to-Income
λ (Inverse Mills Ratio - Heckman)
0.006
-0.188***
-0.001
-0.003
-0.001
-0.039**
(1.01)
(-3.14)
(-0.49)
(-0.36)
(-0.58)
(-2.20)
GOVERNMENT
0.058***
-0.284**
-0.006
-0.065***
0.000
0.127*
(2.95)
(-2.51)
(-0.96)
(-3.81)
(-0.09)
(1.85)
FOREIGN
-0.003
-0.001
-0.037
-0.034
-0.275**
-0.011***
(-0.16)
(-2.20)
(-0.11)
(-1.48)
(-2.73)
(-0.80)
Gov't-to-Foreign Dummy
0.013
0.274
0.006
0.042
0.087**
0.015**
(0.33)
(1.03)
(0.97)
(2.26)
(2.29)
(0.42)
Lag of Assets
-0.015***
0.010
0.001
0.008*
-0.005***
-0.015
(-2.79)
(0.31)
(0.58)
(1.70)
(-4.17)
(-0.56)
Lag of Market Share
-0.061
0.020
0.008
0.139
0.006
-0.129
(-1.21)
(0.04)
(0.59)
(1.60)
(0.33)
(-1.19)
Lag of NII-to-Assets
0.269
-0.819
0.110
-0.246
0.040
0.307
(1.37)
(-1.52)
(1.37)
(-0.73)
(1.32)
(0.82)
CRISIS
-0.017
-0.163
-0.017
-0.021
-0.004
-0.030
(-0.55)
(-1.39)
(-1.24)
(-0.68)
(-0.58)
(-0.37)
Foreign Restrictions
0.044
-1.967***
0.010
0.030
-0.011
-0.425**
(0.52)
(-3.25)
(0.45)
(0.35)
(-0.56)
(-2.45)
COMMON LAW DUMMY
-0.007
-0.161
-0.003
0.019
0.000
-0.028
(-0.46)
(-1.22)
(-0.79)
(1.00)
(0.03)
(-0.41)
Intercept
0.207***
1.068***
0.004
0.046
0.073***
0.517**
(3.81)
(2.77)
(0.36)
(0.93)
(6.79)
(2.43)
2
R
Regional Dummies
Year fixed effects
N
0.184
Yes
Yes
858
0.194
Yes
Yes
854
0.043
Yes
Yes
1276
0.090
Yes
Yes
1053
0.107
Yes
Yes
1268
0.012
Yes
Yes
1260
(continued)
130
Table 15 (continued)
Panel C - Changes in Control and Subsequent Bank Performance - Heckman Model - Emerging vs. Developed
IMPACT OF CHANGES FROM GOVERNMENT TO DOMESTIC BLOCKHOLDER CONTROL
Dependent Variable
Asset Quality
Profitability
Operations
Independent Variables
NPL-to-GL LLR-to-NPL
ROAA
ROAE
NIM
Cost-to-Income
0.106***
0.412
-0.004
-0.149***
-0.019
0.201
λ (Inverse Mills Ratio - Heckman)
(3.01)
(1.40)
(-0.34)
(-2.66)
(-1.01)
(1.26)
DOMESTIC
0.014
0.013
0.029
0.019
0.471
-0.021**
(-2.04)
(0.04)
(1.15)
(1.05)
(1.38)
(1.07)
GOVERNMENT
0.051
-1.164***
-0.008*
-0.086*
-0.001
-0.039
(1.14)
(-3.27)
(-1.89)
(-1.73)
(-0.22)
(-0.27)
Gov't-to-Domestic Dummy (GD)
0.241***
-0.597
-0.022
-0.295**
-0.053
-0.091
(2.81)
(-0.81)
(-0.95)
(-2.43)
(-1.39)
(-0.32)
GD×EMERGING
-0.042
1.252***
0.010
-0.020
0.033
0.619
(-0.87)
(3.21)
(0.59)
(-0.36)
(1.22)
(1.17)
DOMESTIC×EMERGING
0.014
0.037
-0.021
-0.047
-0.021
-0.467
(0.74)
(0.09)
(-1.52)
(-1.28)
(-1.31)
(-1.03)
GOVERNMENT×EMERGING
-0.016
0.000
-0.001
-0.006
0.210
1.232***
(-0.33)
(3.22)
(0.06)
(-0.02)
(-0.57)
(1.32)
Lag of Assets
-0.013**
-0.023
0.002
0.019***
-0.001
-0.016
(-2.35)
(-0.72)
(1.63)
(3.44)
(-0.60)
(-0.59)
Lag of Market Share
-0.047
0.017
-0.001
0.058
0.019
-0.074
(-0.90)
(0.04)
(-0.06)
(0.64)
(0.64)
(-0.59)
Lag of NII-to-Assets
0.268
-1.104*
0.102
-0.230
-0.011
0.251
(1.42)
(-1.79)
(1.28)
(-0.70)
(-0.20)
(0.67)
CRISIS
0.008
-0.007
-0.005
0.021
0.046**
0.017
(0.37)
(-0.05)
(-0.47)
(0.52)
(2.56)
(0.29)
Foreign Restrictions
0.022
0.240*
0.024***
0.098***
0.035**
-0.012
(1.29)
(1.75)
(3.40)
(3.07)
(2.58)
(-0.23)
COMMON LAW DUMMY
-0.015
-0.223*
0.003
0.082***
0.025***
-0.056
(-0.97)
(-1.84)
(0.73)
(3.42)
(2.61)
(-0.62)
EMERGING
0.014
-0.881***
0.016***
0.089***
0.016**
-0.036
(0.64)
(-3.02)
(3.18)
(2.94)
(2.29)
(-0.40)
Intercept
-0.044
1.891**
-0.003
0.191
0.051
0.259
(-0.50)
(2.37)
(-0.13)
(1.63)
(1.56)
(0.48)
R2
Regional Dummies
Year fixed effects
N
0.188
Yes
Yes
952
0.213
Yes
Yes
947
0.054
Yes
Yes
1374
0.132
Yes
Yes
1135
0.209
Yes
Yes
1366
0.022
Yes
Yes
1358
(continued)
131
Table 15 (continued)
IMPACT OF CHANGES FROM GOVERNMENT TO FOREIGN BLOCKHOLDER CONTROL
Dependent Variable
Asset Quality
Profitability
Operations
Independent Variables
NPL-to-GL LLR-to-NPL
ROAA
ROAE
NIM
Cost-to-Income
λ (Inverse Mills Ratio - Heckman)
0.006
-0.180***
-0.001
-0.005
-0.002
-0.039**
(1.01)
(-2.87)
(-0.58)
(-0.55)
(-0.74)
(-2.23)
GOVERNMENT
0.045
-1.192***
-0.012
-0.076
-0.015*
-0.253
(0.96)
(-3.10)
(-1.51)
(-1.52)
(-1.84)
(-0.93)
FOREIGN
-0.044***
-0.483**
0.003
0.079***
-0.013**
-0.363**
(-2.79)
(-2.03)
(0.58)
(3.36)
(-2.10)
(-2.17)
Gov't-to-Foreign Dummy (GF)
0.017
-0.132
0.000
-0.081**
0.020**
0.375***
(0.65)
(-0.45)
(0.00)
(-2.10)
(2.34)
(3.47)
GF×EMERGING
-0.007
0.005
-0.009
0.632*
0.175***
-0.333**
(-0.18)
(1.77)
(0.59)
(3.43)
(-0.98)
(-2.56)
FOREIGN×EMERGING
0.278
0.006
0.003
0.051**
-0.143***
0.409**
(2.00)
(1.07)
(0.53)
(-3.71)
(0.39)
(2.26)
GOVERNMENT×EMERGING
0.014
-0.005
0.000
0.016
1.136***
0.488*
(0.26)
(2.69)
(-0.63)
(0.00)
(1.51)
(1.74)
Lag of Assets
-0.014**
-0.051
0.002
0.015***
-0.004***
-0.029
(-2.26)
(-1.37)
(1.28)
(2.85)
(-2.87)
(-1.16)
Lag of Market Share
-0.067
0.380
0.002
0.098
-0.004
-0.140
(-1.31)
(0.80)
(0.14)
(1.15)
(-0.20)
(-1.32)
Lag of NII-to-Assets
0.263
-0.831
0.112
-0.235
0.044
0.283
(1.32)
(-1.60)
(1.41)
(-0.71)
(1.40)
(0.78)
CRISIS
-0.019
-0.158
-0.018
-0.028
-0.006
-0.031
(-0.58)
(-1.30)
(-1.31)
(-0.90)
(-0.89)
(-0.38)
Foreign Restrictions
0.042
-1.751***
0.005
-0.007
-0.018
-0.403**
(0.49)
(-2.77)
(0.20)
(-0.07)
(-0.83)
(-2.40)
COMMON LAW DUMMY
-0.004
-0.306**
-0.002
0.024
0.001
-0.074
(-0.25)
(-2.23)
(-0.46)
(1.22)
(0.17)
(-0.81)
EMERGING
0.001
-0.847***
0.011*
0.101***
0.010
-0.303*
(0.05)
(-2.89)
(1.68)
(3.37)
(1.30)
(-1.81)
Intercept
0.203***
2.208***
-0.010
-0.078
0.057***
0.856***
(2.73)
(3.46)
(-0.68)
(-1.06)
(3.87)
(4.31)
R2
Regional Dummies
Year fixed effects
N
0.189
Yes
Yes
858
0.227
Yes
Yes
854
0.045
Yes
Yes
1276
132
0.111
Yes
Yes
1053
0.114
Yes
Yes
1268
0.018
Yes
Yes
1260
Panel A: Valuation - Heckman Model
CHANGES FROM GOVERNMENT TO DOMESTIC BLOCKHOLDER CONTROL
Dependent Variable
Independent Variables :
Q
Q
Q
-1.259***
-1.194***
-1.194***
λ (Inverse Mills Ratio - Heckman)
(-2.93)
(-2.69)
(-2.72)
DOMESTIC
-0.157
-0.209
(-1.07)
(-1.51)
GOVERNMENT
-0.192
-0.253
(-0.88)
(-1.14)
Government-to-Domestic
-2.320**
-2.219**
-2.281**
(-2.42)
(-2.29)
(-2.39)
Lag of Assets
0.022
0.016
0.027
(0.51)
(0.34)
(0.64)
Lag of NII-to-Assets
-4.844
-5.755
-4.838
(-1.46)
(-1.64)
(-1.46)
CRISIS
-0.096
-0.120
-0.106
(-0.43)
(-0.51)
(-0.46)
Foreign Restrictions
0.087
-0.013
0.017
(0.46)
(-0.06)
(0.08)
Common Law Dummy
0.151
0.162
0.224
(1.00)
(1.06)
(1.46)
Lag Q
1.155**
1.141**
1.190**
(2.17)
(2.22)
(2.33)
Growth in GDP (Lag)
5.342
5.868*
5.142
(1.58)
(1.75)
(1.50)
Intercept
-1.355
-1.124
-1.333
(-1.36)
(-1.05)
(-1.26)
R2
Regional Dummies
Year Dummies
N
0.161
Yes
Yes
692
0.161
Yes
Yes
692
0.182
Yes
Yes
692
Pooled regressions. The dependent variables are the banks' Tobin’s Q measures from 20012005, proxied by the market-to-book value. The independent variables are dummy variables
for domestic, government, and foreign ownership of banks using the 20% threshold. The lag
of bank's assets (log) and the ratio of non-interest income-to-assets are included as controls.
Crisis is a dummy which equals one if the country experienced a banking crisis between 2000
and 2005; foreign restriction is a dummy which equals one if the country has any restrictions
on foreign ownership of banks. Lambda (Inverse Mills ratio) is obtained from first-stage
probit regressions where the Gov’t-to-Domestic and Govt’-to-Foreign dummies are the
dependent variables, respectively. Finally, regional dummies and year fixed effects are
included in all regressions. Standard errors are clustered by bank. Robust (White) t-statistics
are reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1%
level, respectively.
Table 16: Changes in control and bank value (continued)
133
Table 16 (continued)
CHANGES FROM GOVERNMENT TO FOREIGN BLOCKHOLDER CONTROL
Dependent Variable
Independent Variables :
Q
Q
Q
λ (Inverse Mills Ratio - Heckman)
0.749
0.653
0.537
(1.50)
(1.34)
(1.17)
FOREIGN
0.383
0.363
(1.55)
(1.49)
GOVERNMENT
-0.226
-0.194
(-1.17)
(-1.02)
Government-to-Foreign
1.800
1.393
1.958*
(1.60)
(1.76)
(1.34)
Lag of Assets
0.011
-0.009
0.014
(0.20)
(-0.15)
(0.26)
Lag of NII-to-Assets
-6.234*
-6.654*
-6.312*
(-1.90)
(-1.91)
(-1.86)
CRISIS
-0.247
-0.234
-0.264
(-1.11)
(-1.06)
(-1.20)
Foreign Restrictions
-0.206
-0.300
-0.207
(-0.77)
(-1.10)
(-0.78)
Common Law Dummy
-0.243
-0.187
-0.109
(-0.78)
(-0.57)
(-0.35)
Lag Q
1.358***
1.306***
1.341***
(2.76)
(2.71)
(2.75)
Growth in GDP (Lag)
5.589*
5.807*
5.562*
(1.66)
(1.73)
(1.66)
Intercept
-0.291
0.384
0.120
(-0.41)
(0.55)
(0.18)
2
R
0.155
0.149
0.157
Regional Dummies
Yes
Yes
Yes
Year Dummies
Yes
Yes
Yes
N
692
692
692
(continued)
134
Table 16 (continued)
Panel B: Valuation - EMERGING vs. DEVELOPED - Heckman Model
GOVERNMENT TO DOMESTIC BLOCKHOLDER CONTROL
Dependent Variable
Independent Variables :
Q
Q
Q
λ (Inverse Mills Ratio - Heckman)
0.015
-0.002
0.051
(0.15)
(-0.01)
(0.40)
DOMESTIC
-0.318
-0.336*
(-1.58)
(-1.73)
GOVERNMENT
0.041
-0.057
(0.08)
(-0.10)
Government-to-Domestic
-0.558
-0.762**
-0.551
(-1.57)
(-2.29)
(-1.55)
Gov-to-Domestic * EMERGING
0.251
0.625
0.683*
(0.59)
(1.67)
(1.32)
DOMESTIC * EMERGING
0.340
0.094
(1.21)
(0.26)
GOVERNMENT * EMERGING
-0.395
-0.452
(-0.68)
(-0.67)
Lag of Assets
-0.004
0.003
0.002
(-0.09)
(0.08)
(0.04)
Lag of NII-to-Assets
-3.589
-4.853
-4.167
(-1.11)
(-1.37)
(-1.24)
CRISIS
-0.208
-0.237
-0.213
(-1.00)
(-1.10)
(-1.01)
Foreign Restrictions
0.146
0.021
0.009
(0.74)
(0.09)
(0.04)
EMERGING
-0.530*
-0.256
-0.213
(-1.82)
(-1.06)
(-0.52)
Lag Q
1.105*
1.086**
1.107**
(1.96)
(2.01)
(2.11)
Growth in GDP (Lag)
6.174*
6.726**
6.438*
(1.85)
(2.05)
(1.94)
Intercept
1.936**
1.735**
1.774**
(2.18)
(2.17)
(2.24)
2
R
0.151
0.152
0.160
Regional Dummies
Yes
Yes
Yes
Year Dummies
Yes
Yes
Yes
N
692
692
692
GOVERNMENT TO FOREIGN BLOCKHOLDER CONTROL
Dependent Variable
Independent Variables :
Q
Q
Q
λ (Inverse Mills Ratio - Heckman)
0.476**
0.419*
0.462**
(2.27)
(1.66)
(2.16)
GOVERNMENT
0.147
0.197
(0.27)
(0.36)
FOREIGN
0.448
0.454
(0.82)
(0.83)
Government-to-Foreign (GF)
2.044***
1.545*
1.577*
(3.05)
(1.77)
(1.85)
Gov't-to-Foreign * EMERGING
-0.841
-0.692
-0.779*
(-1.78)
(-1.22)
(-0.92)
FOREIGN*EMERGING
0.012
-0.057
(0.02)
(-0.09)
GOVERNMENT*EMERGING
-0.527
-0.500
(-0.91)
(-0.87)
Lag of Assets
-0.025
-0.020
-0.013
(-0.58)
(-0.48)
(-0.33)
Lag of NII-to-Assets
-4.740
-3.653
-4.162
(-1.41)
(-1.28)
(-1.31)
CRISIS
-0.142
-0.126
-0.146
(-0.63)
(-0.57)
(-0.65)
Foreign Restrictions
-0.186
0.042
-0.097
(-0.71)
(0.17)
(-0.36)
EMERGING
-0.155
-0.409*
-0.220
(-0.66)
(-1.83)
(-1.03)
Lag Q
1.234**
1.208**
1.229**
(2.31)
(2.12)
(2.32)
Growth in GDP (Lag)
6.116*
5.884*
6.094*
(1.89)
(1.88)
(1.91)
Intercept
1.053
1.075
0.827
(1.31)
(1.08)
(1.04)
2
R
Regional Dummies
Year Dummies
N
0.164
Yes
Yes
692
135
0.168
Yes
Yes
692
0.174
Yes
Yes
692
Panel A- First Stage Regression Results
Independent Variables:
Dependent Variable:
∆DB
∆FB
-0.070
Bank Concentration
(-0.41)
WIDE95
0.207**
(2.08)
-0.072
Regulatory Quality
(-0.99)
Productive Population (% total)
1.788**
(2.53)
Political Stability
-0.029
(-0.65)
Lag of Banking System Assets
0.008
-0.032**
(0.58)
(-2.45)
Log of GDP per Capita
0.014
-0.004
(0.33)
(-0.13)
Lag of Inflation
0.065
-0.001
(0.65)
(-0.01)
CRISIS
-0.031
0.116**
(-0.55)
(2.04)
Foreign Restrictions
0.000
-0.082
(0.01)
(-1.13)
Common Law Dummy
-0.062
-0.109*
(-1.16)
(-1.99)
Intercept
-0.096
-0.683*
(-0.35)
(-1.74)
2
R
N
Partial F-statistic
2
Partial R
2
Overidentifying restrictions test (χ )
pvalue
Hausman test (Wu's F-statistic)
pvalue
0.160
69
2.21*
0.196
0.284
69
3.31**
0.213
0.43
(0.513)
0.07
(0.974)
0.45
(0.503)
0.11
(0.895)
Two-Stage Least Squares Regressions. In the first-stage regressions a proxy for the size of
the productive population as a percent of total population, and Kaufmann et al. (2006)
political stability index as of the beginning of the period (1996) are used as instruments for
changes in foreign ownership of banks. A bank concentration proxy (the assets of the top 3
banks as a percent of all commercial banks’ assets), a measure of dispersed ownership of
banks as of 1995, and Kaufman et al. (2006) regulatory quality index as of 1996 are used to
forecast changes in domestic blockholder ownership of banks. The control variables are
measured as of 1995. Following Larcker & Rusticus (2007) the partial R2, which shows the
explanatory power of the instruments that are unique to the first-stage regression, is computed
as R2p= (R2y, z - R2y,z1)/(1-R2y,z1), where z is the combined set of control variables (z1) and
instrumental variables. The overidentifying restriction test statistic is obtained from a
regression of the second-stage residuals on all exogenous variables. The nR2 from this model
is distributed χ2K-L, where K is the number of instruments (2 and 3 for the changes in FB and
DB, respectively) and L is the number of endogenous explanatory variables (1). Panel A
shows results from the first-stage regressions, while Panels B-D show second-stage
regression results. Robust (White) t-statistics are shown in parentheses. *, **, *** indicate
significance at the 10%, 5%, and 1% level, respectively.
Table 17: Spillover effects of changes in bank ownership structure – Instrumental
Variable (IV) approach (continued)
136
Table 17 (continued)
Dependent Variables
∆DBIV
∆GB
∆FBIV
Lag of banking system assets
Lag of GDP per capita
Lag of Inflation
CRISIS
Foreign Restrictions
COMMON LAW DUMMY
Intercept
R2
Regional Dummies
Year Dummies
N
Dependent Variables
∆DBIV
∆GB
∆FBIV
Lag of banking system assets
Lag of GDP per capita
Lag of Inflation
CRISIS
Foreign Restrictions
COMMON LAW DUMMY
Intercept
R2
Regional Dummies
Year Dummies
N
Panel B: Second-Stage Regressions
ASSET QUALITY
Dependent Variable
Dependent Variable
NPL-to-GL
NPL-to-GL
NPL-to-GL LLR-to-NPL LLR-to-NPL LLR-to-NPL
-0.054*
0.960*
(-1.74)
(1.98)
-0.001
0.439**
(-0.03)
(2.23)
-0.018
0.312
(-0.50)
(0.65)
0.001
0.000
0.000
-0.025
-0.021
-0.005
(0.76)
(0.24)
(-0.06)
(-1.08)
(-0.96)
(-0.21)
0.003
0.003
0.003
0.032
0.028
0.030
(0.89)
(0.93)
(0.91)
(0.74)
(0.65)
(0.71)
0.000
0.000
0.000
0.001
0.002
0.005
(0.23)
(0.15)
(0.12)
(0.12)
(0.19)
(0.59)
0.006
0.006
0.007
-0.117*
-0.141*
-0.129*
(0.82)
(0.90)
(0.90)
(-1.71)
(-1.98)
(-1.78)
0.008
0.006
0.005
0.009
0.047
0.065
(1.15)
(0.88)
(0.67)
(0.10)
(0.56)
(0.57)
-0.011
-0.006
-0.009
0.095
0.002
0.049
(-1.33)
(-0.76)
(-0.96)
(1.17)
(0.02)
(0.53)
-0.032
-0.034*
-0.027
0.201
0.389
0.089
(-1.38)
(-1.69)
(-1.07)
(0.72)
(1.27)
(0.31)
0.805
Yes
Yes
246
0.803
0.804
Yes
Yes
Yes
Yes
246
246
PROFITABILITY
Dependent Variable
ROAA
ROAA
ROAA
0.050**
(2.13)
0.005
(0.59)
0.003
(0.21)
-0.001
0.000
0.000
(-1.16)
(-0.61)
(-0.24)
0.000
0.000
0.000
(0.23)
(0.25)
(0.29)
-0.001
-0.001
-0.001
(-0.88)
(-0.74)
(-0.72)
-0.003
-0.003
-0.003
(-0.52)
(-0.57)
(-0.56)
0.004
0.006
0.006*
(1.07)
(1.57)
(1.87)
0.010***
0.005**
0.006*
(3.02)
(2.03)
(1.86)
0.001
0.005
0.002
(0.08)
(0.50)
(0.18)
0.351
Yes
Yes
275
0.340
Yes
Yes
275
0.338
Yes
Yes
275
0.682
Yes
Yes
240
ROAE
0.118
(0.73)
0.684
Yes
Yes
240
0.678
Yes
Yes
240
Dependent Variable
ROAE
ROAE
0.003
(0.05)
0.007
(1.59)
-0.021***
(-2.66)
0.000
(0.19)
-0.001
(-0.02)
0.040
(0.89)
0.082***
(2.72)
0.152*
(1.80)
0.007
(1.48)
-0.019**
(-2.52)
0.001
(0.38)
-0.001
(-0.02)
0.045
(1.02)
0.082***
(2.74)
0.161*
(1.77)
-0.014
(-0.13)
0.008
(1.50)
-0.021***
(-2.69)
0.001
(0.30)
-0.001
(-0.02)
0.044
(1.09)
0.069**
(2.55)
0.164**
(2.05)
0.228
Yes
Yes
275
0.231
Yes
Yes
275
0.226
Yes
Yes
275
(continued)
137
Table 17 (continued)
Dependent Variables
∆DBIV
NIM
-0.014
(-0.46)
∆GB
OPERATIONS
Dependent Variable
NIM
NIM
Cost-to-Income
-0.167
(-1.06)
-0.011
(-1.15)
∆FBIV
Lag of banking system assets
Lag of GDP per capita
Lag of Inflation
CRISIS
Foreign Restrictions
COMMON LAW DUMMY
Intercept
2
R
Regional Dummies
Year Dummies
N
Dependent Variable
Cost-to-Income Cost-to-Income
-0.153**
(-2.20)
0.001
(1.17)
-0.004**
(-2.12)
0.002**
(2.06)
0.012*
(1.84)
0.006
(0.84)
0.000
(0.00)
0.039***
(3.36)
0.001
(1.13)
-0.003*
(-1.98)
0.002**
(2.06)
0.013*
(1.90)
0.005
(0.81)
0.003
(0.45)
0.034**
(2.61)
0.054*
(1.93)
0.003
(1.65)
-0.004**
(-2.32)
0.002**
(2.30)
0.011
(1.56)
0.010
(1.41)
0.010
(1.33)
0.015
(0.81)
0.521
Yes
Yes
275
0.520
Yes
Yes
275
0.524
Yes
Yes
275
-0.006
(-0.91)
0.011
(1.13)
-0.012
(-1.50)
0.041
(1.39)
-0.035
(-1.32)
-0.088***
(-2.69)
0.384***
(3.97)
-0.004
(-0.83)
0.012
(1.30)
-0.012
(-1.49)
0.048*
(1.69)
-0.040
(-1.49)
-0.075***
(-2.76)
0.332***
(3.59)
-0.011
(-0.10)
-0.008
(-1.13)
0.011
(1.09)
-0.012
(-1.56)
0.043
(1.45)
-0.043
(-1.56)
-0.074**
(-2.19)
0.379***
(3.42)
0.465
Yes
Yes
275
0.479
Yes
Yes
275
0.464
Yes
Yes
275
(continued)
138
Table 17 (continued)
Panel C Second-Stage Regressions - Emerging vs. Developed
ASSET QUALITY
Dependent Variable
Dependent Variable
Dependent Variables
NPL-to-GL
NPL-to-GL
LLR-to-NPL
LLR-to-NPL
∆DBIV
-0.007
0.139
(-0.23)
(0.17)
∆FBIV
-0.078
0.333
(-1.40)
(0.40)
∆DBIV x EMERGING
-0.080*
1.474
(-1.80)
(1.46)
∆FBIV x EMERGING
0.049
0.358
(0.75)
(0.47)
Lag of banking system assets
0.001
0.000
-0.029
-0.006
(0.79)
(-0.29)
(-1.18)
(-0.26)
Lag of GDP per capita
0.004
0.005
0.011
-0.004
(1.24)
(1.33)
(0.31)
(-0.11)
Lag of Inflation
0.000
0.000
0.001
0.005
(0.22)
(0.10)
(0.14)
(0.65)
CRISIS
0.004
0.006
-0.076
-0.139*
(0.49)
(0.69)
(-1.21)
(-1.89)
Foreign Restrictions
0.010
0.004
-0.009
0.111
(1.28)
(0.44)
(-0.12)
(0.87)
COMMON LAW DUMMY
-0.011
-0.013
0.101
0.080
(-1.34)
(-1.19)
(1.29)
(0.72)
EMERGING
0.012
0.005
-0.224
-0.232
(0.89)
(0.29)
(-1.51)
(-1.30)
Intercept
-0.051
-0.041
0.588*
0.497
(-1.53)
(-1.13)
(1.90)
(1.53)
2
R
Regional Dummies
Year Dummies
N
Dependent Variables
∆DBIV
∆FBIV
∆DBIV x EMERGING
∆FBIV x EMERGING
Lag of banking system assets
Lag of GDP per capita
Lag of Inflation
CRISIS
Foreign Restrictions
COMMON LAW DUMMY
EMERGING
Intercept
2
R
Regional Dummies
Year Dummies
N
0.806
0.804
Yes
Yes
Yes
Yes
246
246
PROFITABILITY
Dependent Variable
ROAA
ROAA
0.020
(1.05)
0.011
(0.53)
0.050**
(2.03)
-0.008
(-0.36)
-0.001
0.000
(-1.01)
(-0.18)
0.000
0.000
(-0.09)
(0.04)
-0.001
-0.001
(-0.81)
(-0.72)
-0.002
-0.003
(-0.32)
(-0.51)
0.003
0.006*
(0.81)
(1.94)
0.010***
0.006*
(2.93)
(1.92)
-0.005
0.000
(-1.25)
(0.01)
0.006
0.003
(0.42)
(0.16)
0.356
Yes
Yes
275
0.338
Yes
Yes
275
0.686
Yes
Yes
240
0.681
Yes
Yes
240
Dependent Variable
ROAE
ROAE
-0.032
(-0.24)
-0.013
(-0.10)
0.249
(1.20)
-0.018
(-0.11)
0.007
0.008
(1.59)
(1.49)
-0.021**
-0.019*
(-2.12)
(-1.72)
0.001
0.001
(0.23)
(0.29)
0.005
-0.001
(0.08)
(-0.01)
0.035
0.042
(0.78)
(1.07)
0.084***
0.068**
(2.69)
(2.26)
-0.018
0.011
(-0.59)
(0.25)
0.161
0.147
(1.47)
(1.16)
0.231
Yes
Yes
275
0.226
Yes
Yes
275
(continued)
139
Table 17 (continued)
Dependent Variables
∆DBIV
∆FBIV
∆DBIV x EMERGING
∆FBIV x EMERGING
Lag of banking system assets
Lag of GDP per capita
Lag of Inflation
CRISIS
Foreign Restrictions
COMMON LAW DUMMY
EMERGING
Intercept
2
R
Regional Dummies
Year Dummies
N
OPERATIONS
Dependent Variable
NIM
NIM
-0.044
(-1.43)
0.100**
(2.00)
0.049
(1.15)
-0.051
(-1.32)
0.002
0.003*
(1.26)
(1.71)
-0.002
-0.004*
(-1.08)
(-1.80)
0.002**
0.002**
(2.07)
(2.21)
0.013*
0.012*
(1.97)
(1.71)
0.004
0.009
(0.65)
(1.14)
0.001
0.012
(0.13)
(1.37)
0.003
0.006
(0.36)
(0.65)
0.023
0.010
(0.98)
(0.46)
0.522
Yes
Yes
275
0.525
Yes
Yes
275
140
Dependent Variable
Cost-to-Income Cost-to-Income
-0.005
(-0.02)
0.545
(1.43)
-0.260
(-1.09)
-0.538
(-1.38)
-0.007
-0.005
(-1.01)
(-0.81)
0.004
-0.001
(0.30)
(-0.04)
-0.012
-0.013
(-1.51)
(-1.59)
0.037
0.055*
(1.14)
(1.81)
-0.029
-0.044
(-1.02)
(-1.52)
-0.092***
-0.049
(-2.66)
(-1.58)
-0.020
0.019
(-0.36)
(0.29)
0.481**
0.426**
(2.64)
(2.51)
0.469
Yes
Yes
275
0.474
Yes
Yes
275
APPENDIX E
FIGURES
Changes in Bank Ownership Structure between 1995 and 2005
1
0.8
0.6
0.4
0.2
Change in GB
Change in DB
0
Change in FB
Sl
ov
ak
i
Bu a
lg
ar
R o ia
m
an
ia
Po
Cz
la
ec
nd
hR
ep
ub
T a l ic
nz
an
ia
G
re
ec
Ve
e
ne
zu
e
Pa la
kis
ta
n
No
rw
ay
G
er
m
Ka
an
y
za
kh
st
a
Be n
lg
iu
m
Ta
iw
an
Fr
an
ce
-0.2
-0.4
-0.6
-0.8
-1
This graph shows the changes in government (GB), domestic (DB), and foreign (FB)
blockholder ownership of banks between 1995 and 2005 for a sample of countries. These
measures represent the average government, domestic, and foreign blockholder ownership of
the top ten banks in a country.
Figure 1: Changes in bank ownership structure
141