Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
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