Volume One A Reader in International Corporate Finance Edited by Stijn Claessens and Luc Laeven A Reader in International Corporate Finance Volume One A Reader in International Corporate Finance Edited by Stijn Claessens and Luc Laeven Volume One ©2006 The International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org E-mail: feedback@worldbank.org All rights reserved. 1 2 3 4 5 09 08 07 06 This volume is a product of the staff of the International Bank for Reconstruction and Develop- ment / The World Bank. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The bound- aries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone: 978-750-8400; fax: 978-750-4470; Internet: www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. ISBN-10: 0-8213-6698-X ISBN-13: 978-0-8213-6698-1 eISBN: 0-8213-6699-8 eISBN-13: 978-0-8213-6699-8 DOI: 10.1596/978-0-8213-6698-1 Library of Congress Cataloging-in-Publication data has been applied for. Contents FOREWORD vii ACKNOWLEDGMENTS ix INTRODUCTION xi VOLUME I. PART I. LAW AND FINANCE 1 Law, Endowments, and Finance 1 Thorsten Beck, Asli Demirgüç-Kunt, and Ross Levine 2 Financial Development, Property Rights, and Growth 47 Stijn Claessens and Luc Laeven 3 Does Legal Enforcement Affect Financial Transactions? 83 The Contractual Channel in Private Equity Josh Lerner and Antoinette Schoar VOLUME I. PART II. CORPORATE GOVERNANCE 4 Disentangling the Incentive and Entrenchment Effects 107 of Large Shareholdings Stijn Claessens, Simeon Djankov, Joseph P. H. Fan, and Larry H. P. Lang 5 Private Benefits of Control: An International Comparison 139 Alexander Dyck and Luigi Zingales 6 Ferreting Out Tunneling: An Application to Indian Business Groups 203 Marianne Bertrand, Paras Mehta, and Sendhil Mullainathan 7 Cross-country Determinants of Mergers and Acquisitions 231 Stefano Rossi and Paolo F. Volpin VOLUME I. PART III. BANKING 8 The Effects of Government Ownership on Bank Lending 259 Paola Sapienza 9 Related Lending 287 Rafael La Porta, Florencio López-de-Silanes, and Guillermo Zamarripa v vi Contents 10 The Value of Durable Bank Relationships: 325 Evidence from Korean Banking Shocks Kee-Hong Bae, Jun-Koo Kang, and Chan-Woo Lim 11 Do Depositors Punish Banks for Bad Behavior? 359 Market Discipline, Deposit Insurance, and Banking Crises Maria Soledad Martinez Peria and Sergio L. Schmukler INDEX 383 Foreword This two-volume set reprints more than twenty of what we think are the most in- fluential articles on international corporate finance published over the course of the past six years. The book covers a range of topics covering the following six areas: law and finance, corporate governance, banking, capital markets, capital structure and financing constraints, and political economy of finance. All papers have ap- peared in top academic journals and have been widely cited in other work. The purpose of the book is to make available to researchers and students, in an easy way and at an affordable price, a collection of articles offering a review of the present thinking on topics in international corporate finance. The book is ideally suited as an accompaniment to existing textbooks for courses on corporate finance and emerging market finance at the graduate economics, law, and MBA levels. The articles selected reflect two major trends in the corporate finance literature that are significant departures from prior work: One is the increased interest in international aspects of corporate finance, particularly topics specific to emerging markets. The other is the increased awareness of the importance of institutions in explaining differences in corporate finance patterns--at the country and firm levels--around the world. The latter has culminated in a new literature known as the "law and finance literature," which focuses on the legal underpinnings of finance. It has also been accompanied by a greater understanding of the importance of political economy factors in countries' economic development and has led to the increased application of a political economy framework to the study of corporate finance. This collection offers an overview of the present thinking on topics in interna- tional corporate finance. We hope that the papers in this book will serve the role of gathering in one place the background reading most often used for an advanced course in corporate finance. We also think that researchers will appreciate the ben- efit of having all these articles in one place, and we hope that the book will stimu- late new research and thinking in this exciting new field. We trust the students and their instructors will deepen their understanding of international corporate finance by reading the papers. Of course, any of the remaining errors in the papers included in this book are entirely those of the authors and not of the editors. vii Acknowledgments The editors wish to thank the following authors and publishers who have kindly given permission for the use of copyright material. Blackwell Publishing for the following articles: Stijn Claessens and Luc Laeven (2003), "Financial Development, Property Rights, and Growth," Journal of Finance, Vol. 58 (6), pp. 2401­36; Stijn Claessens, Simeon Djankov, Joseph Fan, and Larry Lang (2002), "Disentangling the Incen- tive and Entrenchment Effects of Large Shareholdings," Journal of Finance, Vol. 57 (6), pp. 2741­71; Alexander Dyck and Luigi Zingales (2004), "Private Benefits of Control: An International Comparison," Journal of Finance, Vol. 59 (2), pp. 537­600; Maria Soledad Martinez Peria and Sergio L. Schmukler (2001), "Do Depositors Punish Banks for Bad Behavior? Market Discipline, Deposit Insurance, and Banking Crises," Journal of Finance, Vol. 56 (3), pp. 1029­51; Peter Blair Henry (2000), "Stock Market Liberalization, Economic Reform, and Emerging Market Equity Prices," Journal of Finance, Vol. 55 (2), pp. 529­64; Utpal Bhat- tacharya and Hazem Daouk (2002), "The World Price of Insider Trading," Journal of Finance, Vol. 57 (1), pp. 75­108; Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer (2006), "What Works in Securities Laws?" Journal of Finance, Vol. 61 (1), pp. 1­32; Art Durnev, Randall Morck, and Bernard Yeung (2004), "Value-Enhancing Capital Budgeting and Firm-Specific Stock Return Variation," Journal of Finance, Vol. 59 (1), pp. 65­105; Laurence Booth, Varouj Aivazian, Asli Demirgüç-Kunt, and Vojislav Maksimovic (2001), "Capital Structures in Develop- ing Countries," Journal of Finance, Vol. 56 (1), pp. 87­130; Mihir Desai, Fritz Foley, and James Hines (2004), "A Multinational Perspective on Capital Structure Choice and Internal Capital Markets," Journal of Finance, Vol. 59 (6), pp. 2451­ 87; Thorsten Beck, Asli Demirgüç-Kunt, and Vojislav Maksimovic (2005), "Finan- cial and Legal Constraints to Growth: Does Firm Size Matter?" Journal of Finance, Vol. 60 (1), pp. 137­77. Elsevier for the following articles: Thorsten Beck, Asli Demirgüç-Kunt, and Ross Levine (2003), "Law, Endowments, and Finance," Journal of Financial Economics, Vol. 70 (2), pp. 137­81; Stefano Rossi and Paolo F. Volpin (2004), "Cross-Country Determinants of Mergers and Acquisitions," Journal of Financial Economics, Vol. 74 (2), pp. 277­304; Paola Sa- pienza (2004), "The Effects of Government Ownership on Bank Lending," Journal of Financial Economics, Vol. 72 (2), pp. 357­84; Kee-Hong Bae, Jun-Koo Kang, and Chan-Woo Lim (2002), "The Value of Durable Bank Relationships: Evidence from Korean Banking Shocks," Journal of Financial Economics, Vol. 64 (2), pp. ix x Acknowledgments 147­80; Geert Bekaert, Campbell R. Harvey, and Christian Lundblad (2005), "Does Financial Liberalization Spur Growth?" Journal of Financial Economics, Vol. 77 (1), pp. 3­55; Raghuram G. Rajan and Luigi Zingales (2003), "The Great Reversals: The Politics of Financial Development in the 20th Century," Journal of Financial Economics, Vol. 69 (1), pp. 5­50; Simon Johnson and Todd Mitton (2003), "Cronyism and Capital Controls: Evidence from Malaysia," Journal of Financial Economics, Vol. 67 (2), pp. 351­82. Oxford University Press for the following article: Inessa Love (2003), "Financial Development and Financing Constraints: Interna- tional Evidence from the Structural Investment Model," Review of Financial Stud- ies, Vol. 16 (3), pp. 765­91. American Economic Association for the following article: Raymond Fisman (2001), "Estimating the Value of Political Connections," Ameri- can Economic Review, Vol. 91 (4), pp. 1095­1102. MIT Press for the following articles: Josh Lerner and Antoinette Schoar (2005), "Does Legal Enforcement Affect Finan- cial Transactions? The Contractual Channel in Private Equity," Quarterly Journal of Economics, Vol. 120 (1), pp. 223­46; Marianne Bertrand, Paras Mehta, and Sendhil Mullainathan (2002), "Ferreting Out Tunneling: An Application to Indian Business Groups," Quarterly Journal of Economics, Vol. 117 (1), pp. 121­48; Rafael La Porta, Florencio Lopez-de-Silanes, and Guillermo Zamarripa (2003), "Related Lending," Quarterly Journal of Economics, Vol. 118 (1), pp. 231­68. We would like to thank Rose Vo for her assistance in obtaining the copyrights of the articles from the authors and publishers, Joaquin Lopez for his technical assis- tance in reproducing the papers, Stephen McGroarty of the Office of the Publisher of the World Bank for his assistance and guidance in publishing the book, and the World Bank for financial support. The views presented in these published papers are those of the authors and should not be attributed to, or reported as reflecting, the position of the World Bank, the International Monetary Fund, the executive directors of both organizations, or any other organization mentioned therein. The book was largely completed when the second editor was at the World Bank. Introduction Volume I. Part I. Law and Finance Volume I begins with an examination of the legal and financial aspects of inter- national capital markets. In recent years, there has been an increased interest in international aspects of corporate finance. There are stark differences in financial structures and financing patterns of corporations around the world, particularly as they relate to emerging markets. Recent work has suggested that most of these differences can be explained by differences in laws and institutions of countries and in countries' economic and other endowments. These relationships have been the focus of a new literature on law and finance. La Porta et al. (1997, 1998) were the first to show that the legal traditions of a country determine to a large extent the financial development of a country. They started a large literature investigating the determinants and effects of legal systems across countries. In chapter 1, "Law, Endowments, and Finance," Thorsten Beck, Asli Demirguc- Kunt, and Ross Levine contribute to this literature by assessing the importance of both legal traditions and property rights institutions. The law and finance theory suggests that legal traditions brought by colonizers differ in protecting the rights of private investors in relation to the state, with important implications for financial markets. The endowments theory argues that initial conditionsas proxied by natural endowments, including the disease environmentinfluence the formation of long-lasting property rights institutions that shape financial development, even decades or centuries later. Using information on the origin of the law and on the disease environment encountered by colonizers centuries ago, the authors extract the independent effects of both law and endowments on financial development. They find evidence supporting both theories, although the initial endowments theory explains more of the cross-country variation in financial development than the legal traditions theory does. This suggests that there are economic and other forces at play that make certain initial conditions translate into the institutional environments of today. In chapter 2, "Financial Development, Property Rights, and Growth," Stijn Claessens and Luc Laeven add to this literature by showing that better legal and property rights institutions affect economic growth through two equally impor- tant channels: one is improved access to finance resulting from greater financial development, the channel already highlighted in the law and finance literature; the other is improved investment allocation resulting from more secure property rights, as firms and other investors allocate resources raised in a more efficient manner. Quantitatively, the effects of these two channels on economic growth are similar. This suggests that the legal system is important not only for financial sector devel- xi xii Introduction opment but also for an efficient operation of the real sectors. Better property rights, for example, can stimulate investment in sectors that are more intangibles-intensive or that heavily depend on intellectual property rights, such as the services, soft- ware, and telecommunications industries. As these industries have become drivers of growth in many countries, the second channel has become more important. In chapter 3, "Does Legal Enforcement Affect Financial Transactions? The Contractual Channel in Private Equity," Josh Lerner and Antoinette Schoar show that legal tradition and law enforcement have direct implications for how finan- cial contracts are shaped. Taking a much more micro approach and using data on private equity investments in developing countries, they show that investments in high-enforcement and common law nations often use convertible preferred stock with covenants, while investments in low-enforcement and civil law nations tend to use common stock and debt and rely on equity and board control. While relying on ownership rather than contractual provisions may help to alleviate legal enforce- ment problems, there appears to be a real cost to operating in a low-enforcement environment because transactions in low-enforcement countries have lower valua- tions and returns. In other words, the low-enforcement environments force inves- tors to use less-than-optimal contracts to assure their ownership and control rights, which in turn makes the operations of the businesses less efficient. Volume I. Part II. Corporate Governance Corporate governance is another field that has gained increased interest from aca- demics and policy makers around the world in the past decade, spurred by major corporate scandals and governance problems in a host of countries, including the corporate scandals of Enron in the United States and Parmalat in Italy and the expropriation of minority shareholders in the East Asian crisis countries and other emerging countries. Governance problems are particularly pronounced in many emerging countries where family control is the predominant form of corporate ownership and where minority shareholder rights are often not enforced. In chapter 4, "Disentangling the Incentive and Entrenchment Effects of Large Shareholdings," Stijn Claessens, Simeon Djankov, Joseph Fan, and Larry Lang show that ownership of firms in East Asian countries is highly concentrated and that there is often a large difference between the control rights and the cash-flow rights of the principal shareholder of the firm. They argue that the larger the cash-flow rights of the shareholder, the more his or her incentives are aligned with those of the minority shareholder because the investor has his or her own money at stake. On the other hand, control rights give the principal owner the ability to direct the firm's resources. The larger the difference between control and cash-flow rights, the more likely that the principal shareholder is entrenched and that the minority shareholders are expropriated as the controlling owner directs resources to his or her own advantages. Using data on a large number of listed companies in eight East Asian countries, the authors find that firm value increases with the cash- flow rights of the largest shareholder, consistent with a positive incentive effect; however, firm value falls when the control rights of the largest shareholder exceed Introduction xiii its cash-flow ownership, consistent with an entrenchment effect. This suggests expropriation, which may have further economic costs as resources are poorly invested. The private benefits of control for the controlling shareholder are often substan- tial, particularly in environments where shareholder rights are low. This explains why concentrated ownership is the predominant form of ownership around the world, particularly in developing economies, but also in continental Europe, where property rights are weaker and often poorly enforced. In chapter 5, "Private Ben- efits of Control: An International Comparison," Alexander Dyck and Luigi Zin- gales propose a method that estimates the private benefits of control. For a sample of 39 countries and using individual transactions, they find that private benefits of control vary widely across countries, from a low of -4 percent to a high of +65 percent. Across countries, higher private benefits of control are associated with less developed capital markets, more concentrated ownership, and more privately nego- tiated privatizations. Legal institutions plus enforcement and pressure by the media appear to be important factors in curbing private benefits of control. Because private benefits are associated with inefficient investment, their findings confirm the importance of establishing strong property rights and enforcing these to increase growth. Controlling shareholders often devise complex ownership structures of firms (for example, through pyramidal structures) to create a gap between voting rights and cash-flow rights and to be able to direct resources through internal markets to affiliated firms. This is particularly the case for business groups in emerging mar- kets. Owners of such business groups are often accused of expropriating minority shareholders by tunneling resources from firms where they have low cash-flow rightswith little costs of taking away moneyto firms where they have high cash-flow rightswith large gains of bringing in money. In chapter 6, "Ferreting Out Tunneling: An Application to Indian Business Groups," Marianne Bertrand, Paras Mehta, and Sendhil Mullainathan propose a methodology to measure the extent of tunneling activities in business groups. This methodology rests on isolat- ing and then testing the distinctive implications of the tunneling hypothesis for the propagation of earnings shocks across firms within a group. Using data on Indian business groups, the authors find a significant amount of tunneling, much of it occurring via nonoperating components of profit. This suggests a cost-of- business group that may have to be mitigated by some other measures, such as better property rights, increased disclosure, and specific restrictions (such as pre- venting or limiting intragroup ownership structures). The threat of takeover can play a potentially important disciplining role for poorly governed firms because management risks being removed; however, in practice, the market for corporate control is generally inactive in countries where it is most needed: where shareholder protection is weak. The rules limiting takeovers are often more restricted in these environments, making domestic takeovers more difficult. Still, there is evidence that foreign takeovers can have important positive implications for the governance of local target firms, particularly in countries with poor investor protection. This is the theme of chapter 7, "Cross-Country Deter- xiv Introduction minants of Mergers and Acquisitions," by Stefano Rossi and Paolo Volpin. They study the determinants of mergers and acquisitions (M&As) around the world by focusing on differences in laws and regulations across countries. They find that M&A activity is significantly larger in countries with better accounting standards and stronger shareholder protection. In cross-border deals, targets are typically from countries with poorer investor protection than their acquirers' countries, suggesting that cross-border transactions play a governance role by improving the degree of investor protection within target firms. As such, globalization and internationalization of financial services can help countries improve their corporate governance arrangements. Volume I. Part III. Banking Another common feature of developing countries is the predominance of state banks. State banks also played an important role in many industrial countries, at least until recently, but many governments have privatized in the past decade. In 1995, government ownership of banks around the world averaged around 42 per- cent (La Porta et al. 2002). In chapter 8, "The Effects of Government Ownership on Bank Lending," Paola Sapienza uses information on individual loan contracts in Italy, where lending by state-owned banks represents more than half of total lending, to study the effects of government ownership on bank lending behavior. She finds that lending by state banks is inefficient. State-owned banks charge lower interest rates than do privately owned banks to similar or identical firms, even if firms are able to borrow more from privately owned banks. State-owned banks also favor large firms and firms located in depressed areas, again in contrast to the choices of private banks. Finally, the lending behavior of state-owned banks is af- fected by the electoral results of the party affiliated with the bank: the stronger the political party in the area where the firm is borrowing, the lower the interest rates charged. This suggests that the political forces affect the lending behavior of state- owned banks in an adverse manner and offers an argument for the privatization of state-owned banks. Private banks can, however, also have problems when not properly governed and monitored. When banks are privately owned in emerging economies, they are often part of business groups. This can create incentive problems that result in lending on preferential terms. More generally, banks in many countries lend to firms controlled by the bank's owners. This type of lending is known as "insider lending" or "related lending." In chapter 9, "Related Lending," Rafael La Porta, Florencio Lopez-de-Silanes, and Guillermo Zamarripa examine the benefits of related lending, using data on bank-borrower relationships in Mexico. The authors show that related lending in Mexico is prevalent and takes place on better terms than arm's-length lending. This could still be consistent with an efficient allocation of resources, but the authors show that related loans are significantly more likely to default and that when they default, they have lower recovery rates than unrelated loans. Their evidence for Mexico supports the view that related lending is often a manifestation of looting, particularly in weak institutional environments. The costs Introduction xv of this are often incurred by the government and taxpayers, as happened in Mexico when many of the private banks experienced financial distress and had to be res- cued by the government, which provided fiscal resources for their recapitalization. However, close ties between banks and industrial groups need not be inefficient; they can create valuable relationships, particularly in environments where hard in- formation on borrowers is sparse. As such, relationships can substitute for a weak- er institutional environment. In chapter 10, "The Value of Durable Bank Relation- ships: Evidence from Korean Banking Shocks," Kee-Hong Bae, Jun-Koo Kang, and Chan-Woo Lim examine the value of durable bank relationships in the Republic of Korea, using a sample of exogenous events that negatively affected Korean banks during the financial crisis of 1997­98. The authors show that adverse shocks to banks have a negative effect not only on the value of the banks themselves but also on the value of their client firms. They also show that this adverse effect on firm value is a decreasing function of the financial health of both the banks and their client firms. These results indicate that bank relationships were valuable to this group of firms; however, whether the relationship supported an efficient allocation of resources is not clear. Given the importance of banks in developing countries' financial intermediation, it is essential that banks be properly supervised and monitored, a task most often assigned to the bank supervisory agency. When bank supervisors fail to discipline banks, however, it is up to the depositors to monitor banks and punish banks for bad behavior by withdrawing deposits. In chapter 11, "Do Depositors Punish Banks for Bad Behavior? Market Discipline, Deposit Insurance, and Banking Cri- ses," Maria Soledad Martinez Peria and Sergio Schmukler study whether this form of market discipline is effective and whether it is affected by the presence of deposit insurance. They focus on the experiences of Argentina, Chile, and Mexico during the 1980s and 1990s. They find that depositors discipline banks by withdrawing deposits and by requiring higher interest rates, and their responsiveness to bank risk taking increases in the aftermath of crises. Deposit insurance does not appear to diminish the extent of market discipline. This suggests that in a weak institu- tional environment, where bank supervision fails to mitigate excessive risks taking by banks, depositors and other bank claimholders can play an important role in the monitoring of financial institutions. Volume II. Part I. Capital Markets Volume II opens with a selection of articles on capital markets. Equity and bond finance raised in capital markets (as an alternative to bank finance) has become increasingly important for corporations around the world. The increase in the use of markets for raising capital are in part resulting from rising equity prices that have triggered new issuance. Lower interest rates have also caused many firms to opt for corporate bonds. Also important, especially in developing countries, as institutional fundamentals are improving substantially, there has been an improved willingness on the part of international investors to invest and provide funds. As xvi Introduction emerging stock markets have been liberalized, global investors have been increas- ingly seeking to diversify assets in these markets. The effects of these measures have been researched in a number of papers. Stock market liberalization (that is, the decision by a country's government to allow foreigners to purchase shares in that country's stock market) has been found to have real effects on the economic performance of a country. In chapter 1, "Stock Market Liberalization, Economic Reform, and Emerging Market Equity Prices," Peter Blair Henry shows that a country's aggregate equity price index experiences substantial abnormal returns during the period leading up to the implementation of its initial stock market liberalization. This result is consistent with the prediction of standard international asset-pricing models that stock market liberalization reduces a country's cost of equity capital by allowing for risk sharing between domestic and foreign agents. This reduced cost of capital in turn can be expected to lead to greater investment and growth. Stock market liberalization has indeed been found to have positive ramifications for overall investment and economic growth. In chapter 2, "Does Financial Liber- alization Spur Growth?" Geert Bekaert, Campbell Harvey, and Christian Lundblad show that equity market liberalizations, on average, lead to a 1 percent increase in annual real economic growth. This effect appears to have been most pronounced in countries with a strong institutional environment, suggesting that liberalization must be accompanied by a strengthening of the institutional environment to reap all of the benefits. Other evidence confirms the need for additional policy measures besides liber- alization. Not all stock markets work as efficiently as they should. In particular, insider trading is a common feature of many stock markets. Although most stock markets have established laws to prevent insider trading, enforcement is poor in many countries, and investors get worse prices and rates of return. In chapter 3, "The World Price of Insider Trading," Utpal Bhattacharya and Hazem Daouk analyze the quality of enforcement of insider trading laws. They show that while insider trading laws exist in the majority of countries with stock markets, enforce- ment--as evidenced by actual prosecutions of people engaging in insider trading-- has taken place in only about one-third of these countries. Their empirical analysis shows that the cost of equity in a country does not change after the introduction of insider trading laws, but only decreases significantly after the first prosecution, suggesting that enforcement of the law is critical, rather than just the adoption of the insider trading law. The question remains, however, whether stock markets should be regulated by relying mostly on the government using public enforcement by securities commis- sions and the like or whether the emphasis should be on self-regulation, relying on private enforcement by giving individuals the legal tools to litigate in case of abuses. In chapter 4, "What Works in Securities Laws?" Rafael La Porta, Florencio Lopez-De-Silanes, and Andrei Shleifer tackle this complex matter by examining the effect of different designs of securities laws on stock market development in 49 countries. The authors find little evidence that public enforcement benefits stock markets, but strong evidence that laws mandating disclosure and facilitating pri- Introduction xvii vate enforcement through liability rules benefit stock markets' developmentwith regard to the size of the market, the number of firms listed, and the new issuance. Their results echo those analyzing the banking system, where it has been found that supervision by government authorities often does not deliver the results de- sired, but that private sector oversight can be effective, especially in weak institu- tional environments. A well-functioning stock market should allow firms not only to raise financing but also to produce more informative stock prices. Where stock prices are more informative, this induces better governance and more efficient capital investment decisions. However, in many developing countries, the cost of collecting informa- tion on firms is high, resulting in less trading by investors with private information, leading to less informative stock prices. In chapter 5, "Value-Enhancing Capital Budgeting and Firm-Specific Stock Return Variation," Art Durnev, Randall Morck, and Bernard Yeung introduce a method to gauge the informativeness of a compa- ny's stock price. They base their measure of informativeness on the magnitude of firm-specific return variation. The idea is that a more informative stock displays a higher stock variation because stock variation occurs because of trading by inves- tors with private information. The authors document this measure of stock price informativeness for a large number of countries. They then go on to show that the economic efficiency of corporate investment, as measured by Tobin's Q (the ratio of the market value of a firm's assets to the replacement value of its assets--a mea- sure of firm efficiency and growth prospects), is positively related to the magnitude of firm-specific variation in stock returns, suggesting that more informative stock prices facilitate more efficient corporate investment. Volume II. Part II. Capital Structure and Financial Constraints Because of large institutional differences and differences in the relative importance of the banking system and the equity and bond markets, it will come as no surprise that capital structures of firms vary widely across countries. In chapter 6, "Capi- tal Structures in Developing Countries," Laurence Booth, Varouj Aivazian, Asli Demirguc-Kunt, and Vojislav Maksimovic document capital structure choices of firms in 10 developing countries and then analyze the determinants of these struc- tures. They find that although some of the factors that are important in explaining capital structure in developed countries (such as profitability and asset tangibil- ity of the firm) carry over to developing countries, there are persistent differences across countries, indicating that specific country factors are at work. The authors explore obvious candidates such as the institutional framework governing bank- ruptcy, accounting standards, and the availability of alternative forms of financing, but their smaller set of countries does not allow them to explain in a definite way which of these may be more important. More generally, it is difficult to disentangle the impact of different institutional features on capital structure choices in a cross-country setting because there are so many country-specific factors to control for. In chapter 7, "A Multinational Per- xviii Introduction spective on Capital Structure Choice and Internal Capital Markets," Mihir Desai, Fritz Foley, and James Hines therefore take advantage of a unique dataset on the capital structure of foreign affiliates of U.S. multinationals to further our under- standing of the institutional determinants of capital structure. The authors find that capital structure choice is significantly affected by three institutional factors: tax environment, capital market development, and creditor rights. They show that financial leverage of subsidiaries is positively affected by local tax rates. They also find that multinational affiliates are financed with less external debt in countries with underdeveloped capital markets or weak creditor rights, likely reflecting the disadvantages of higher local borrowing costs. Instrumental variable analysisto control for other factors driving these resultsindicates that greater borrowing from parent companies substitutes for three-quarters of reduced external borrow- ing induced by weak local capital market conditions. Multinational firms therefore appear to employ internal capital markets opportunistically to overcome imperfec- tions in external capital markets. As such, globalization and internationalization of financial services can offer some benefits for countries with weak institutional environments. Besides a limited way to control for cross-country differences, another compli- cation of studying the determinants of capital structure is that not all firms de- mand external finance. Many successful firms finance their investments internally and do not need to access outside finance. For these firms, financial sector devel- opment thus matters less. The important question is whether those firms that are financially constrained are better able to obtain external finance in more developed financial systems, with positive ramifications for firm growth. Here the difficulty arises in how to measure which firms are financially constrained. In chapter 8, "Financial Development and Financing Constraints: International Evidence from the Structural Investment Model," Inessa Love addresses this question by using an investment Euler equation to infer the degree of financing constraints of individual firms. She provides evidence that financial development affects growth by reducing the financing constraints of firms and in that way improving the efficient allocation of investment. The magnitude of the changes, which run through changes in the cost of capital, is large: in a country with a low level of financial development, the cost of capital is twice as large as in a country with an average level of financial development. In chapter 9, "Financial and Legal Constraints to Growth: Does Firm Size Mat- ter?" Thorsten Beck, Asli Demirguc-Kunt, and Vojislav Maksimovic expand on the analysis of what financial sector development means for the growth prospects of individual firms. They use firm-level survey data covering 54 countries to construct a self-reported measure of financing constraints to address the question of how much faster firms might grow if they had more access to financing. The authors find that financial and institutional development weakens the constraining effects of financing constraints on firm growth in an economically and statistically signifi- cant way and that it is the smallest firms that benefit most from greater financial sector development. Introduction xix Volume II. Part III. Political Economy of Finance Politics plays an important role in finance. Financial development and financial reform are often driven by political economy considerations, and where finance is a scarce commodity, political connections are often especially valuable for firms in need of external finance. Whether these connections are good, in the sense that they support an efficient allocation of resources, is one question that has been more closely analyzed recently. Also, a number of papers have also researched from various angles how political economy factors affect the institutions necessary for financial sector development. In chapter 10, "The Great Reversals: The Politics of Financial Development in the 20th Century," Raghuram Rajan and Luigi Zingales show that financial de- velopment does not change monotonically over time. By most measures, countries were more financially developed in 1913 than in 1980 and only recently have many countries surpassed their 1913 levels. To explain these changes, they propose an interest group theory of financial development wherein incumbents oppose finan- cial development because it fosters greater competition through lowering entry barriers for newcomers. The theory predicts that incumbents' opposition will be weaker when an economy allows both cross-border trade and capital flows because then their hold on the allocation of rents is less. Consistent with this theory, they find that trade and capital flows can explain some of the cross-country and time- series variations in financial development. This in turn suggests that liberalization of trade and capital flows can be an important means of fostering greater financial sector development because they weaken the political economy factors holding back an economy. The last two chapters in Volume II provide further empirical evidence of the value of political connections in developing countries, but now using firm-level data for particular countries. In chapter 11, "Estimating the Value of Political Con- nections," Raymond Fisman shows that the market value of politically connected firms in Indonesia under President Suharto declined more when adverse rumors cir- culated about the health of the president. Because the same firms did not perform better than other firms, this suggests that these connected firms obtained favors, yet allocated resources less efficiently. In chapter 12, "Cronyism and Capital Controls: Evidence from Malaysia," Simon Johnson and Todd Mitton provide empirical evidence for Malaysia that the imposition of capital controls during the Asian financial crises benefited primarily firms with strong connections to Prime Minister Mahathir, again without an improved performance when compared with other firms. These chapters indicate that the operation of corporations in developing countries, including their financing and financial structure, importantly depends on their relationships with politicians. As such, financial sector reform cannot avoid considering how to address political economy issues. Chapter One 1 Journal of Financial Economics 70 (2003) 137­181 Law, endowments, and finance$ Thorsten Becka, Asli Demirgu.@-Kunta, Ross Levineb,c,* aThe World Bank, Washington, DC 20433, USA bDepartment of Finance, Carlson School of Management, University of Minnesota, Minneapolis, MN 55455, USA cNational Bureau of Economic Research, Inc., Cambridge, MA 02138-5398, USA Received 5 October 2001; accepted 4 September 2002 Abstract Using a sample of 70 former colonies, this paper assesses two theories regarding the historical determinants of financial development. The law and finance theory holds that legal traditions, brought by colonizers, differ in terms of protecting the rights of private investors vis-a-vis the state, with important implications for financial markets. The endowment theory " argues that the disease environment encountered by colonizers influences the formation of long-lasting institutions that shape financial development. The empirical results provide evidence for both theories. However, initial endowments explain more of the cross-country variation in financial intermediary and stock market development. r 2003 Elsevier B.V. All rights reserved. JEL classification: G2; K2; O11; P51 $ We thank David Arseneau, Pam Gill, and Tolga Sobaci for excellent research assistance, and Agnes Yaptenco and Kari Labrie for assistance with the manuscript. We thank without implicating Daron Acemoglu, John Boyd, Maria Carkovic, Tim Guinnane, Patrick Honohan, Phil Keefer, Paul Mahoney, Alexander Pivovarsky, Andrei Shleifer, Oren Sussman, an anonymous referee, seminar participants at the Banco Central de Chile, the University of Minnesota, Harvard University, the World Bank, the University of Maryland, and UCLA, and conference participants at the Fedesarrollo conference on Financial Crisis and Policy Responses in Cartagena, the Crenos conference on Finance, Institutions, Technology, and Growth in Alghero, and the CEPR Summer Finance Conference in Gerzensee. We give special thanks to Simon Johnson. His guidance led us to focus and thereby improve the paper. Parts of this paper were originally part of a working paper titled ``Law, Politics, and Finance,'' which was a background paper for the 2002 World Development Report. This paper's findings, interpretations, and conclusions are entirely those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. *Corresponding author. Department of Finance, Carlson School of Management, University of Minnesota, Minneapolis, MN 55455, USA. Tel.: +1-612-624-9551; fax: +1-612-626-1335. E-mail address: rlevine@csom.umn.edu (R. Levine). 0304-405X/03/$- see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0304-405X(03)00144-2 2 A Reader in International Corporate Finance 138 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 Keywords: Law; Endowments; Financial development; Economic development; Property rights 1. Introduction A substantial body of work suggests that well-functioning financial intermediaries and markets promote economic growth (see, e.g., Levine, 1997). The view that financial systems exert a first-order impact on economic growth raises critical questions: How have some countries developed well-functioning financial systems, while others have not? Why do some countries have strong laws and property rights protection that support private contracting and financial systems, while others do not? While considerable research examines the finance-growth relationship, much less work examines the fundamental sources of differences among nations in financial development. This paper empirically evaluates two theories concerning the historical determi- nants of financial systems. First, the law and finance theory holds that: (a) legal traditions differ in terms of the priority they attach to protecting the rights of private investors vis-a-vis the state; (b) private property rights protection forms the basis of " financial contracting and overall financial development; and, (c) the major legal traditions were formed in Europe centuries ago and were then spread through conquest, colonization, and imitation (see La Porta et al., 1998, henceforth LLSV). Thus, the law and finance theory predicts that historically determined differences in legal traditions help explain international differences in financial systems today. The law and finance theory focuses on the differences between the two most influential legal traditions, the British Common law and the French Civil law (see, e.g., Hayek, 1960; LLSV, 1998). According to this theory, the British Common law evolved to protect private property owners against the crown (Merryman, 1985).1 This facilitated the ability of private property owners to transact confidently, with positive repercussions on financial development (North and Weingast, 1989). In contrast, the French Civil law was constructed to eliminate the role of a corrupt judiciary, solidify state power, and restrain the courts from interfering with state policy.2 Over time, state dominance produced a legal tradition that focuses more on 1While landholding rights in England were originally based on King William I's feudal system, the courts developed legal rules that treated large estate holders as private property owners and not as tenants of the king. Indeed, the common law at the dawn of the 17th century was principally a law of private property (e.g., Littleton, 1481; Coke, 1628). During the great conflict between Parliament and the English kings in the 16th and 17th centuries, the crown attempted to reassert feudal prerogatives and sell monopoly rights to cope with budgetary shortfalls. Parliament (composed mostly of landowners and wealthy merchants) along with the courts took the side of the property owners against the crown. While King James I argued that royal prerogative superseded the common law, the courts asserted that the law is king, Lex, Rex. The Stuarts were thrown out in 1688. 2By the 18th century, there was a notable deterioration in the integrity and prestige of the judiciary. The crown sold judgeships to rich families and the judges unabashedly promoted the interests of the elite. [Refer to Dawson, 1968, p. 373]. Unsurprisingly, the French Revolution strove to eliminate the role of the Chapter One 3 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 139 the rights of the state and less on the rights of individual investors than the British Common law (Hayek, 1960; Mahoney, 2001). According to the law and finance theory, a powerful state with a responsive legal system will have the incentives and capabilities to divert the flow of society's resources from optimal toward favored ends, and therefore this power will hinder the development of free, competitive financial systems. Thus, the law and finance theory predicts that countries that have adopted a French Civil law tradition will tend to place less emphasis on private property rights protection and will enjoy correspondingly lower levels of financial development than countries with a British Common law tradition. The law and finance theory focuses on the origin of a country's legal tradition. The French imposed the Napoleonic Code in all conquered lands and colonies. Furthermore, the Code shaped the Spanish and Portuguese legal systems, which further spread the French Civil law to Spanish and Portuguese colonies. Similarly, the British instituted the Common law in its colonies. According to the law and finance theory, the spread of legal traditions had enduring influences on national approaches to private property rights and financial development--British colonizers advanced a legal tradition that stresses private property rights and fosters financial development, whereas in contrast colonizers that spread the French Civil law implanted a legal tradition that is less conducive to financial development. The endowment theory, on the other hand, emphasizes the roles of geography and the disease environment in shaping institutional development; we apply this theory to the development of private property rights and financial institutions. Acemoglu et al. (2001, henceforth AJR) base their theory on three premises. First, AJR note that Europeans adopted different types of colonization strategies. At one end of the spectrum, the Europeans settled and created institutions to support private property and check the power of the state. These settler colonies include the United States, Australia, and New Zealand. At the other end of the spectrum, Europeans did not aim to settle but rather to extract as much from the colony as possible. In these ``extractive states,'' Europeans did not create institutions to support private property rights; instead, they established institutions that empowered the elite to extract gold, silver, etc. (e.g., Congo, Ivory Coast, and much of Latin America). The second component of AJR's theory holds that the type of colonization strategy was heavily influenced by the feasibility of settlement. Mortality rates were startlingly high in some places. In the first year of the Sierra Leone Company, 72 percent of the Europeans died. In the 1805 Mungo park expedition in Gambia and Niger, all of the Europeans died before completing the trip. In these inhospitable environments, Europeans tended to create extractive states (AJR, 2001). In areas where endowments favored settlement, Europeans tended to form settler colonies. (footnote continued) judiciary in making and interpreting the law. Robespierre even argued that, ``the word jurisprudence... must be effaced from our language.'' [Quoted from Dawson, 1968, p. 426] Glaeser and Shleifer (2002) explain how antagonism toward jurisprudence and the exaltation of the role of the state encouraged the development of easily verifiable ``bright-line-rules'' that do not rely on the discretion of judges. Thus, codification supported the strengthening of the government and relegated judges to a relatively minor, bureaucratic role. 4 A Reader in International Corporate Finance 140 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 For instance, AJR note that the Pilgrims decided to settle in the American colonies instead of Guyana partially because of the high mortality rates in Guyana. Moreover, Curtin (1964, 1998) documents that European newspapers published colonial mortality rates widely, so that potential settlers would have information about colonial endowments. Thus, according to the endowment theory, the disease environment shaped colonization strategy and the types of institutions established by European colonizers. The final piece of the AJR theory of institutional development stresses that the institutions created by European colonizers endured after independence. Settler colonies tended to produce post-colonial governments that were more democratic and more devoted to defending private property rights than extractive colonies. In contrast, since extractive colonies had institutions for effectively extracting resources, the post-colonial elite frequently assumed power and readily exploited the pre-existing extractive institutions. Young (1994) presents historical evidence that once authoritarian institutions are efficiently extracting resources from the bulk of society, post-independence rulers tend to use these institutions to their own advantage and profit. This was the case in Sierra Leone, Senegal, and Congo. Latin America was similar. For instance, while Mexicans gained independence from European colonialists, the elite that assumed power took advantage of the existing institutions to extract resources rather than create institutions to protect private property contracts, and foster broad-based economic development. Furthermore, Engerman et al. (1998) demonstrate the long-lasting impact of initial institutions on voting rights: once regimes restrict voting rights to protect the elite from the masses, the government tends to resist changes in suffrage policies for long periods. While AJR (2001) focus on institutional development in general, their theory is applicable to the financial sector. In an extractive environment, colonizers will not construct institutions that favor the development of free, competitive financial markets because competitive markets may threaten the position of the extractors. In settler colonies, however, colonizers will be much more likely to construct institutions that protect private property rights and hence foster financial development. Thus, according to the endowment theory, differences in endowments shaped initial institutions and these initial institutions have had long-lasting repercussions on private property rights protection and financial development.3 Although the law and endowment theories both stress the importance of initial institutions in shaping the financial systems we observe today, they highlight very different causal mechanisms. The law and finance theory focuses on the legal 3 Engerman and Sokoloff (1997) note another channel through which geographical endowments shape initial institutions with enduring effects on economic development. Namely, they show that agriculture in southern North America and much of South America is conducive to large plantations. Thus, colonists developed long-lasting institutions to protect the few landowners against the many peasants. In contrast, northern North America's agriculture is conducive to small farms, so more egalitarian institutions emerged. Thus, again, endowments influence the formation of institutions associated with openness and competition. Our primary reason for focusing on the AJR (2001) measure of settler mortality and not also examining agricultural endowments is that AJR (2001) have assembled data for a broad cross-section of countries. Chapter One 5 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 141 tradition brought by the colonizer. The endowment theory focuses on the disease and geography endowments encountered by the colonizer and how these endowments shaped both colonization strategy and the construction of long-lasting institutions. In the law and finance theory, the identity of the colonizer is crucial, but the identity of the colonizer is irrelevant according to the endowment theory. Similarly, in the endowment theory, the endowments of the lands where Europeans arrived are crucial, but the law and finance theory gives no weight to the mortality rates of European colonizers in explaining the development of today's private property rights and financial systems. This is admittedly overstated. Proponents of the law and finance theory do not argue that endowments are irrelevant. Similarly, proponents of the endowment theory do not contend that legal origin is irrelevant. Rather, each theory articulates very distinct mechanisms about how the colonization period shaped national views toward private property rights and financial development. We stress--and empirically evaluate--these distinct predictions. While these two explanations of financial development offer very different causal mechanisms, they are not necessarily mutually exclusive. To evaluate empirically the law and endowment theories of financial development, we use cross-country regressions on a sample of 70 former colonies, for reasons described below. We examine whether cross-country differences in financial institutions are accounted for by cross-country differences in legal tradition and/or initial endowments, while controlling for other possible determinants. To measure financial development, we use measures of: (i) financial intermediary development; (ii) equity market development; and, (iii) private property rights protection. For simplicity, we use the term ``financial development'' to refer to each of these three measures. We measure financial development over the period 1990­1995. To measure legal tradition, we use the LLSV (1999) indicators specifying whether the country has a British or French legal tradition, as determined by the origin of each country's Company/Commercial law. To measure initial endowments, we primarily use the AJR measure of settler mortality rates as European settlers arrived in various parts of the globe. For robustness, we also use the absolute value of the latitude of each country as an alternative, albeit less precise, indicator of initial endowments, since many authors argue that tropical climates are not conducive to institutional and economic development. In conducting the cross-country comparisons, we control for other potential determinants of financial development. Specifically, we include measures of ethnic diversity, religious composition, years of independence since 1776, and continent dummy variables. Further, we also assess whether the political structure of a country is the only mechanism through which the legal tradition and initial endowments influence current financial development. We focus on a sample of 70 former colonies for two reasons. First, we have the AJR (2001) data on settler mortality, which is a key building block of AJR's (2001) empirical assessment of the endowment theory. Second, some observers stress that European colonization offers a unique break, i.e., a natural identifying condition (AJR, 2001, 2002; Engerman and Sokoloff, 1997). As European conquerors and colonizers landed, they brought different legal traditions. Colonization represents a period during which legal traditions were exogenously established around the globe 6 A Reader in International Corporate Finance 142 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 and thus provides a natural starting point for examining the law and endowment theories of financial development. For these reasons, we use a sample of 70 former colonies with data on settler mortality. This sample only includes countries with British and French legal origins. This paper makes four contributions.4 First, this paper applies AJR's (2001) endowment theory of institutions directly to the study of financial development. Although AJR (2001) carefully document the connections running from endow- ments to institutions to the level of economic development today, we examine whether initial colonial endowments explain a wide array of current measures of financial development. Since financial development helps explain technological innovation, the efficiency of capital allocation across industries and firms, output volatility, the likelihood of a systemic banking crisis, and economic growth, even when controlling for the levels of economic and institutional development, it is important to assess whether endowments influence financial development.5 Second, this is the first paper to consider simultaneously the legal and endowment views of financial development. This is crucial to assessing two very different visions of how the institutions founded by Europeans continue to shape national approaches to private property and financial systems in former colonies. Third, although others have shown that legal tradition shapes financial development (LLSV, 1997, 1998, 2000), this paper goes much further in evaluating the robustness of the law and finance view by controlling for endowments, religion, ethnic diversity, length of independence, etc. This assessment is critical if we are to have much confidence in legal theories of financial development. Fourth, while some analysts argue that the structure and competitiveness of the political system shapes institutions and policies, this is the first paper to examine whether legal origin and both disease and geographical endowments explain cross-country differences in financial development beyond their ability to account for differences in national political systems. The paper is organized as follows. Section 2 describes the data and presents figures that motivate the analysis. Section 3 discusses the regression results, and a series of robustness tests are presented in Section 4. Section 5 concludes. 2. Data and initial assessments This section describes the data and presents figures that document: (1) British Common law countries tend to have higher levels of financial development than 4Pivovarsky (2001) also examines the relationship between institutions and financial development. He analyzes the impact of current institutions, instrumented by settler mortality and legal origin, on financial development and finds a strong effect of the exogenous component of institutions on financial development. Our contribution is distinct, however, in that we compare the direct effects of endowments and legal origin on financial system development. 5In particular, see Beck et al. (2000) on the finance and productivity growth relationship, Wurgler (2000) on the finance and industry allocation of capital relationship, Demirgu.@-Kunt and Maksimovic (1998) on the finance and firm growth link, Demirgu.@-Kunt and Detragiache (2002) on the finance and crisis relationship, Easterly et al. (2000) on the finance and output volatility link, and Levine and Zervos (1998), Rajan and Zingales (1998), and Beck and Levine (2002, 2003) on the finance­growth relationship. Chapter One 7 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 143 French Civil law countries; and, (2) countries with high levels of European mortality during the initial stages of colonization tend to have lower levels of financial development than those countries with initially low settler-mortality rates. 2.1. Financial development To measure financial development, we use indicators of financial intermediary development, stock market development, and property rights protection. The goal is to proxy for the degree to which national financial systems facilitate the acquisition of firm information, ease corporate governance, help agents manage risk, and mobilize savings effectively. Unfortunately, we do not have direct and comparable measures of the ability of national financial systems to provide these benefits for a broad cross-section of countries. Thus, we use a variety of indicators of financial development to assess the connections between law, endowments, and finance. PRIVATE CREDIT equals financial intermediary credits to the private sector divided by gross domestic product (GDP) and is measured over 1990­1995. PRIVATE CREDIT excludes credit to the public sector and cross-claims between financial intermediaries, and thus measures the amount of savings that is channeled through debt-issuing financial intermediaries to private borrowers. For most countries, PRIVATE CREDIT is obtained from data available from the Interna- tional Monetary Fund (IMF). To maximize the size of the sample, however, we also use World Bank data sources for a few countries that lack IMF data; the countries and sources are specified in the data appendix. Past work shows a strong connection between PRIVATE CREDIT and economic growth (see Levine et al., 2000). PRIVATE CREDIT ranges from values above 0.9 in the United States, Hong Kong, Singapore, South Africa, and Malaysia, to values less than 0.03 in Sierra Leone, Uganda, Angola, and Zaire. STOCK MARKET DEVELOPMENT equals the total value of outstanding equity shares as a fraction of GDP and is averaged over the period 1990­1995.6 This measures the overall size of the equity market relative to the size of the economy.7 The data are primarily collected from the World Bank's International Finance Corporation. However, we use additional data sources to complete the dataset, as specified in the appendix. There are large cross-country differences as shown in 6For both STOCK MARKET DEVELOPMENT and PRIVATE CREDIT, we have conducted the analyses using data averaged over the 1975­1995 period instead of the 1990­1995 period. We get the same results. Since there are fewer countries with data over the 1975­1995 period, we present the results with the 1990­1995 averages. 7Since there are differences in ownership concentration across countries, LLSV (1998) suggest using an adjustment whereby STOCK MARKET DEVELOPMENT is multiplied by one minus the median ownership share of the three largest shareholders in the ten largest non-financial, privately-owned domestic firms in the country. This paper obtains the same conclusions using this adjusted measure. Since we only have these ownership share figures for a sub-sample of countries, however, making this adjustment substantially reduces our dataset. Thus, we report the results using the standard STOCK MARKET DEVELOPMENT indicator for market size. 8 A Reader in International Corporate Finance 144 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 Table 1, Panel A. STOCK MARKET DEVELOPMENT is greater than 0.65 in the United States, Chile, Singapore, South Africa, Hong Kong, and Malaysia, and is indistinguishable from zero in 29 countries. PROPERTY RIGHTS is an index of the degree to which the government enforces laws that protect private property. The data are for 1997 and were obtained from LLSV (1999) and the Index of Economic Freedom. While PRIVATE CREDIT and STOCK MARKET DEVELOPMENT are direct measures of the size of financial intermediaries and equity markets respectively, PROPERTY RIGHTS does not directly measure the size of a component of the financial sector. Rather, PROPERTY RIGHTS measures a key input into the efficient operation of financial contracts and the development of formal financial institutions: the degree of protection of private property rights. The law and endowment theories stress the degree to which national institutions emphasize private property rights versus the rights of the state. This difference in emphasis may influence a variety of indicators of financial development. While PROPERTY RIGHTS as defined is one attempt to measure this difference, there may be measurement problems or other differences in emphasis on state versus private rights that affect financial contracting beyond narrow indicators of property rights protection. Hence, we examine a variety of financial development indicators. The maximum value of PROPERTY RIGHTS is five, while a value of one indicates the weakest property rights protection. Nine former colonies have the maximum value of five. Only Haiti and Rwanda have the minimum value of one, while 15 countries have a value of two for PROPERTY RIGHTS. We do not have data on PROPERTY RIGHTS for the Central African Republic, so there are only 69 countries in the PROPERTY RIGHTS regressions. 2.2. Legal origin LLSV (1998, 1999) identify the legal origin of each country's company or commercial law as French, British, German, Scandinavian, or Socialist.8 Given we are examining former colonies with data on settler mortality from AJR (2001), we 8 One may further refine the categorization of legal traditions, as described by the following examples. First, Franks and Sussman (1999) and Coffee (2000) describe differences in two Common law countries: the United Kingdom and the United States. While in the U.K. there is freedom of contracting (Glendon et al., 1982), in the U.S. the judiciary has a more important role to play in developing law. In both systems, however, the legislature does not have a monopoly on creating law, as in the original French legal system, as designed by Napoleon. In both the U.K. and the U.S., case law is a source of law, while not in France. Second, different colonization strategies may have intensified differences across legal traditions. England did not try to replace Islamic, Hindu, or African law. English courts in the colonies, therefore, used local laws and customs in deciding cases. This quickly produced an Indian Common law distinct from English Common law. While perhaps chaotic, this allowed for the integration of common law with local circumstances. In contrast, the French imposed the Code although serious conflicts frequently existed with local customs. Also, legal scholars study differences across the French Civil law countries of Latin America. While recognizing that each country's legal system is special, the comparative law literature clearly emphasizes that there are key differences across the major legal families Chapter One 9 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 145 have data for only French and British legal-origin countries.9 Thus, we do not include many of the most developed countries in the LLSV (1998, 1999) sample. The FRENCH LEGAL ORIGIN dummy variable equals one if the country adopted its company/commercial law from the French Civil law and zero otherwise. In the regressions, British legal origin is captured in the constant. Fig. 1 clearly shows that financial development is substantially higher in countries with a British Common law tradition than in countries with a French Civil law tradition. French Civil law countries have, on average, lower levels of PRIVATE CREDIT, STOCK MARKET DEVELOPMENT, and PROPERTY RIGHTS than British Common law countries. There are 45 French Civil law countries and 25 British Common law countries. Table 1, Panel B correlations confirm Fig. 1: the FRENCH LEGAL ORIGIN dummy variable is significantly, negatively correlated with each of the three financial development indicators. Furthermore, Fig. 2 illustrates that in Common law countries, eight countries have PRIVATE CREDIT greater than 0.6 (Australia, Canada, New Zealand, Malaysia, Singapore, South Africa, Hong Kong, and the United States), while among French Civil law countries, only Malta has PRIVATE CREDIT greater than 0.6. Fig. 2 also demonstrates clearly that legal origin does not completely explain the cross-country variation observed in financial systems today. Fig. 2 documents that there are many Common law countries with poorly developed financial inter- mediaries, and a few French legal origin countries that have well-developed financial intermediaries. For instance, many Common law countries have PRIVATE CREDIT less than 0.3, with countries such as Uganda, Sierra Leone, Ghana, Sudan, and Tanzania registering extremely low PRIVATE CREDIT levels. Thus, we need to know more than legal origin to account for cross-country differences in financial systems. 2.3. Endowments As Europeans arrived around the world, they encountered very different environments. In some lands, Europeans found hospitable environments. In others, conditions were less hospitable and Europeans died in large numbers. According to AJR (2001), these location specific endowments fundamentally influenced the types of long-lasting institutions created by European colonists. To measure endowments, we use the AJR (2001) measure of SETTLER MORTALITY. AJR (2001) compile data on the death rates faced by settlers. Curtin (1989) constructs data on the mortality and disease rates of European soldiers in colonies during the early nineteenth century. The raw data come from the British, 9Although we have data on settler mortality for Vietnam and Myanmar (which are classified as socialist legal origin countries by LLSV, 1999), we do not include these two countries because we do not have comparable information on financial development for these economies. Also, there are 70 countries in our sample of former colonies with settler mortality data. We also constructed a larger sample of 95 non- European countries. This 95-country sample, however, does not have settler mortality data. For the 95- country sample, we conducted the analyses using latitude instead of settler mortality and obtained the same results reported below. 10 146 T. Beck Table 1 Summary statistics and correlations A et Summary statistics are presented in Panel A and correlations are presented in Panel B, respectively. Private Credit is the value of credits by financial Reader al. intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property Rights reflects the degree to which government enforces laws that protect private property, with higher numbers indicating better enforcement. /Journal French Legal Origin is a dummy variable that takes on the value one for countries with French Civil law tradition, and zero otherwise. Settler Mortality is the in log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. Latin America and Africa are dummy variables International that take the value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Catholic, Muslim, and Other Religion indicate the of percentage of the population that follows a particular religion (Catholic, Muslim, or religions other than Catholic, Muslim, or Protestant, respectively). Financial Independence is the percentage of years since 1776 that a country has been independent. Ethnic Fractionalization is the probability that two randomly selected individuals in a country will not speak the same language. Legislative Competition is an indicator of competition in the last legislative election. Checks measures the number of veto-players in the political decision making process. These last two measures are averaged over 1990­1995. Detailed variable Economic Corporate definitions and sources are given in the data appendix. Panel A: Summary statistics: s N Mean Std. dev Min Max 70 Finance Private Credit 70 0.32 0.30 0.01 1.48 (2003) Stock Market Development 70 0.19 0.40 0.00 1.89 Property Rights 69 3.12 0.99 1.00 5.00 French Legal Origin 70 0.64 0.48 0.00 1.00 137 Settler Mortality 70 4.67 1.24 2.15 7.99 ­181 Africa 70 0.40 0.49 0.00 1.00 Latin America 70 0.36 0.48 0.00 1.00 Catholic 70 39.44 36.89 0.10 97.3 Muslim 70 23.90 33.87 0.00 99.4 Other Religion 70 25.79 23.58 0.30 86.0 Independence 70 0.32 0.32 0.00 1.00 Ethnic Fractionalization 70 0.42 0.31 0.00 0.89 Legislative Competition 68 5.81 1.62 1.00 7.00 Checks 68 2.68 1.40 1.00 6.00 Panel B: Correlation matrix of variables Stock French Private Market Property Legal Settler Latin Other Ethnic Legislative Credit Develop- Rights Origin Mortality Africa America Catholic Muslim Religion Indepen- Fractiona- Competi- ment dence lization tion Stock Market Development T. Beck Property Rights 0.618*** 0.487*** French Legal 0.370*** 0.430*** 0.461*** et Origin al. Settler Mortality 0.669*** 0.528*** 0.438*** 0.238** Africa 0.408*** 0.228* 0.426*** 0.061 0.651*** /Journal Latin America 0.105 0.140 0.064 0.244** 0.178 0.609*** Catholic 0.133 0.194 0.114 0.479*** 0.118 0.356*** 0.706*** Muslim 0.157 0.141 0.103 0.006 0.271** 0.240** 0.500*** 0.652*** of Other Religion 0.283** 0.421*** 0.187 0.552*** 0.137 0.166 0.379*** 0.548*** 0.175 Financial Chapter Independence 0.057 0.016 0.041 0.330*** 0.323*** 0.475*** 0.630*** 0.700*** 0.421*** 0.384*** Ethnic Fractionalization 0.269** 0.062 0.213* 0.076 0.433*** 0.718*** 0.551*** 0.370*** 0.229* 0.229* 0.437*** Economic One Legislative Competition 0.408*** 0.271** 0.401*** 0.032 0.601*** 0.699*** 0.513*** 0.425*** 0.387*** 0.143 0.392*** 0.506*** Checks 0.378*** 0.323** 0.373*** 0.202* 0.497*** 0.543*** 0.383*** 0.248** 0.285** 0.010 0.317*** 0.306** 0.664*** s *, * *, * * * indicate significance levels of 10%, 5%, and 1%, respectively. 70 (2003) 137 ­181 147 11 12 A Reader in International Corporate Finance 148 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 0.5 0.45 Civil Law 0.4 Common Law 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 Private Credit Stock Market Development Property Rights (divided by ten) Fig. 1. Financial development across Common and Civil law countries. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property Rights reflects the degree to which government enforces laws that protect private property, with higher numbers indicating better enforcement. Civil law countries are countries whose legal system is of French Civil law origin, whereas Common law countries are countries whose legal system is of British Common law origin. French, and United States governments during the period 1817­1848. The standard measure is annualized deaths per thousand soldiers, with each death replaced by a new soldier. Curtin (1998) adds similar data on soldier mortality during the second half of the nineteenth century. Finally, Gutierrez (1986) uses Vatican records to construct estimates of the mortality rates of bishops in Latin America from 1604 to 1876. Since some of these data overlap with Curtin's separate estimates, AJR confirm the compatibility of the two data series before constructing an overall measure of the logarithm of annualized deaths per thousand Europeans, SETTLER MORTALITY, for a large group of former colonies. As in AJR (2001), we use the logarithm to diminish the impact of outliers. The AJR (2001) measure forms the core of our analysis of the relation between endowments and finance. This measure ranges from 2.15 (Australia and New Zealand) to 7.99 (Mali). Fig. 3 shows a generally negative, though certainly not linear, relation between SETTLER MORTALITY and financial development.10 The absence of a linear relationship is especially pronounced for STOCK MARKET DEVELOPMENT since many countries have stock market capitalization ratios of zero. Consequently, we use a Tobit estimator to check our results. Table 1, Panel B shows that there is a 10When we experimented with a non-linear transformation (e.g., the inverse of the log settler mortality rate), we obtain the same conclusions discussed below. Furthermore, we re-ran the analyses using the logarithm of PRIVATE CREDIT. Again, we confirm the conclusions discussed below. Chapter One 13 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 149 0.9 MLT 0.8 0.7 0.6 TUN CHL 0.5 PAN Credit IDN 0.4 SUR MRT MUS Private BOL MAR 0.3 COL CIV EGY BRA MEX HND NIC SEN TGO URY DOM SLV 0.2 DZA PRY VEN CMR ETH ECU ARG CRI MDG BFA COG GTM GAB HTI MLI 0.1 NER GIN TCD PER RWA CAF AGO 0 ZAR (a) 1.6 USA 1.4 HKG 1.2 1 SGP MYS ZAF Credit 0.8 AUS CAN NZL Private0.6 BHS TTO 0.4 BRB KEN JAM IND PAK 0.2 BGD NGA GUY LKA GMB GHA SDN TZA SLE UGA 0 (b) Fig. 2. (a) Private credit in Civil law countries: Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Civil law countries are countries whose legal system is of French Civil law origin, whereas Common law countries are countries whose legal system is of British Common law origin. There are 45 Civil law and 25 Common law countries in the sample. (b) Private Credit in Common law countries: Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Civil law countries are countries whose legal system is of French Civil law origin, whereas Common law countries are countries whose legal system is of British Common law origin. There are 45 Civil law and 25 Common law countries in the sample. significant, negative correlation between SETTLER MORTALITY and each of the three financial development indicators at the one-percent significance level. The data indicate that in colonies where early settlers found very inhospitable environments, we do not observe well-developed financial systems today. 14 A Reader in International Corporate Finance 14 A Reader in International Corporate Finance ARTICLE IN PRESS 150 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 Chapter One 15 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 151 2.4. Other possible determinants of financial development To assess the robustness of our results, we include several other potential determinants of financial development in our empirical analysis. ETHNIC FRACTIONALIZATION measures the probability that two randomly selected individuals from a country are from different ethnolinguistic groups. LSSV (1999, p. 231) argue, ``...political theories predict that, as ethnic heterogeneity increases, governments become more interventionist.'' Recent studies show that in highly ethnically diverse economies, the group that comes to power tends to implement policies that: (a) expropriate as many resources as possible from the ethnic losers; (b) restrict the rights of other groups; and, (c) prohibit the growth of industries or sectors that threaten the ruling group (see, e.g., Alesina et al., 1999; Easterly and Levine, 1997). When this view is applied to the financial sector, the implication is clear: greater ethnic diversity implies the adoption of policies and institutions that are focused on maintaining power and control, rather than on creating an open and competitive financial system. Table 1, Panel B indicates that there is a significant, negative correlation between ETHNIC FRACTIONALIZATION and PRIVATE CREDIT. Thus we include ETHNIC FRACTIONALIZATION to examine the independent impacts of law and endowments on financial development. INDEPENDENCE equals the fraction of years since 1776 that a country has been independent. We include this measure because a longer period of independence may provide greater opportunities for countries to develop institutions, policies, and regulations independent of their colonial heritage. In the simple correlations, however, we do not find a significant link between INDEPENDENCE and financial development. We also examine religious composition. Many scholars argue that religion shapes national views regarding property rights, competition, and the role of the state (LLSV, 1999; Stulz and Williamson, 2003). Putnam (1993, p. 107), for instance, contends that the Catholic Church fosters ``vertical bonds of authority'' rather than ``horizontal bonds of fellowship.'' Similarly, Landes (1998) argues that Catholic and Muslim countries tend to develop xenophobic cultures and powerful bonds between church and state to maintain control, bonds which limit competition and private property rights protection. Fig. 3. (a) Settler Mortality and Private Credit: Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Settler Mortality is the log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. The sample comprises 70 countries of Common law and French Civil law origin. (b) Settler Mortality and Stock Market Development: Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP Settler Mortality is the log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. The sample comprises 70 countries of Common law and French Civil law origin. (c) Settler Mortality and Property Rights: Property rights reflects the degree to which government enforces laws that protect private property, with higher numbers indicating better enforcement. Settler Mortality is the log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. The sample comprises 70 countries of Common law and French Civil law origin. 16 A Reader in International Corporate Finance 152 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 CATHOLIC, MUSLIM, and OTHER RELIGION equal the fraction of the population that is Catholic, Muslim, or of another (non-Protestant) religion. The Protestant share of the population is omitted (and therefore captured in the regression constant). The data are from LLSV (1999). Table 1, Panel B shows that countries with a higher population proportion that is neither Catholic, nor Muslim, nor Protestant, have higher levels of financial development than countries where a higher fraction of the country is either Catholic or Muslim. Thus, we control for religious composition in examining the independent relations between financial development and both legal origin and endowments. We note there is a very large, positive, and significant correlation between CATHOLIC and FRENCH LEGAL ORIGIN (0.48). Thus, it may be particularly difficult to distinguish fully between CATHOLIC and the Civil law tradition. Finally, we include one dummy variable for countries in LATIN AMERICA and another for countries in Sub-Saharan AFRICA. A large number of studies find that countries in Sub-Saharan Africa and Latin America perform more poorly than countries in other regions of the world even after controlling for economic policies, institutional development, and other factors. Easterly and Levine (1997) provide related analyses and citations. There are important problems with including continent dummies. First, continent dummies do not proxy for a clear explanation of why countries in these regions have worse institutions or perform more poorly. Second, Latin America is primarily a French legal-origin continent; the correlation between Catholic and Latin America is 0.71 and is significant at the one-percent level. Thus, including continent dummies may weaken our ability to identify linkages between financial development and legal origin without offering a clear, alternative explanation. Third, many Sub-Saharan African countries have high settler mortality rates. The correlation between AFRICA and SETTLER MORTALITY is 0.65 and is significant at the one-percent level. Thus, including the AFRICA dummy may decrease the ability to find a link between financial development and endowments without offering an alternative theory. Including these continent dummies, however, may control for region-specific characteristics that are not captured by any of the other explanatory variables. Therefore, while recognizing the problems associated with interpreting continent dummies, we include them in assessing the relations between law, endowments, and finance.11 3. Regression results This section presents regressions on the relationship between financial develop- ment and both law and endowments while controlling for other possible 11In a previous version, we also included GDP per capita as a control variable. However, institutional development also influences economic development (as shown by AJR, 2001), so including GDP per capita together with initial endowments may bias the coefficient on legal origin and settler mortality/latitude toward zero. Further, unlike the other regressors, GDP per capita is endogenous, which causes estimation problems as shown by AJR (2001). Chapter One 17 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 153 determinants of financial development. The dependent variable is one of the three measures of financial development, PRIVATE CREDIT, STOCK MARKET DEVELOPMENT, or PROPERTY RIGHTS. We use the dummy variable FRENCH LEGAL ORIGIN to assess the links between law and finance. We use SETTLER MORTALITY to assess the relationship between endowments and finance. As control variables, we use continent dummy variables (for Latin American and Africa), measures of religious composition, the percentage of years the country has been independent since 1776, and ethnic diversity. We also include a regression where we control concurrently for continent dummies, time since independence, and ethnic fractionalization. We do not include religious composition dummies in this regression since they never enter significantly at the five-percent significance level. The reasons for including these particular controls were discussed above. 3.1. Law and finance Table 2 presents regressions of financial development on French legal origin and various combinations of the control variables. Table 2 does not include measures of endowments. The results indicate a strong, negative relation between French legal origin and financial development. When controlling for continent, religious composition, ethnic diversity, and independence, French legal origin enters negatively and significantly at the five-percent level in all of the financial development regressions. The results suggest an economically large impact. For instance, the smallest coefficient (in absolute value) on FRENCH LEGAL ORIGIN in the STOCK MARKET DEVELOPMENT regressions is 0:27; and the mean and standard deviation values of STOCK MARKET DEVELOPMENT are 0.19 and 0.40, respectively. For illustrative purposes, the coefficient suggests that if Argentina had a British Common law tradition, its low level of stock market capitalization (0.10) would be substantially larger and closer to that of New Zealand (0.37). In sum, French Civil law countries tend to have lower levels of financial development than British Common law countries after controlling for many national characteristics. This result is consistent with the LLSV (1998) view that the identity of the colonizer matters because of the legal traditions the colonizers brought. 3.2. Endowments and finance Table 3 indicates a robust, negative association between SETTLER MORTAL- ITY and financial development. SETTLER MORTALITY enters with a negative coefficient and is significant at the five percent level in all of the PRIVATE CREDIT and STOCK MARKET DEVELOPMENT regressions. The coefficient sizes are economically large. According to the smallest coefficient (in the absolute sense) in the PRIVATE CREDIT regression in Table 3 ð 0:14Þ; a one standard deviation reduction in the logarithm of mortality rates (1.24) would increase PRIVATE CREDIT by 0.17, and the mean and standard deviation of PRIVATE CREDIT are 0.32 and 0.30, respectively. Thus, the estimates in Table 3 can account for why 18 154 Table 2 Law and finance The regression estimated is: Financial Sector Development ¼ a þ b1 French Legal Origin + b2X; where Financial Sector Development is either Private Credit, Stock Market Development, or Property Rights. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property Rights reflects the degree to which government T. enforces laws that protect private property, with higher numbers indicating better enforcement. French Legal Origin is a dummy variable that takes on the Beck value one for countries with French Civil law tradition, and zero otherwise. The regressions also include a vector of control variables, X: Latin America and Africa are dummy variables that take the value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Catholic, Muslim, and A et Reader Other Religion indicate the percentage of the population that follows a particular religion (Catholic, Muslim, or religions other than Catholic, Muslim, or al. Protestant, respectively). Independence is the percentage of years since 1776 that a country has been independent. Ethnic Fractionalization is the probability /Journal that two randomly selected individuals in a country will not speak the same language. Regressions are estimated using Ordinary Least Squares. Robust in standard errors are given in parentheses. *, **, *** indicate significance at the 10% 5%, and 1% levels, respectively. Detailed variable definitions and sources International are given in the data appendix. of Financial Ethnic French Legal Latin Africa Catholic Muslim Other Independence Fractionalization Adjusted-R2 Obs. Origin America Religion Economic Private 0.233*** 0.124 70 Corporate Credit (0.088) 0.136** 0.292*** 0.417*** 0.378 70 s (0.067) (0.092) (0.100) 70 0.181** 0.002 0.003 0.001 0.121 70 (2003) Finance (0.086) (0.003) (0.003) (0.005) 0.275*** 0.191 0.148 70 (0.097) (0.136) 137 0.247*** 0.289*** 0.203 70 ­181 (0.084) (0.095) 0.168** 0.352*** 0.348*** 0.170 0.109 0.384 70 (0.080) (0.112) (0.107) (0.179) (0.133) Stock 0.356*** 0.173 70 Market (0.118) Development 0.278*** 0.242* 0.312** 0.240 70 (0.101) (0.128) (0.143) 0.265** 0.002 0.002 0.006 0.199 70 (0.107) (0.004) (0.004) (0.005) 0.395*** 0.176** 0.179 70 (0.111) (0.082) 0.362*** 0.121 0.170 70 (0.117) (0.122) 0.308*** 0.299*** 0.315* 0.224 0.087 0.237 70 T. (0.102) (0.104) (0.177) (0.150) (0.176) Beck Property 0.947*** 0.198 69 et al. Rights (0.241) 0.836*** 0.250 0.969*** 0.351 69 /Journal (0.206) (0.265) (0.243) 1.065*** 0.002 0.005 0.007 0.182 69 (0.291) (0.009) (0.009) (0.011) of 1.103*** 0.692** 0.232 69 Financial Chapter (0.235) (0.346) 0.995*** 0.813** 0.253 69 (0.232) (0.339) Economic One 0.856*** 0.286 1.014*** 0.182 0.178 0.334 69 (0.203) (0.297) (0.293) (0.393) (0.477) s 70 (2003) 137 ­181 155 19 20 156 T. Beck Table 3 Endowments and finance A et The regression estimated is: Financial Sector Development ¼ a þ b1 Settler Mortality +b2X; where Financial Sector Development is either Private Credit, Reader al. Stock Market Development, or Property Rights. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property rights reflects the degree to which government /Journal enforces laws that protect private property, with higher numbers indicating better enforcement. Settler Mortality is the log of the annualized deaths per in thousand European soldiers in European colonies in the early 19th century. The regressions also include a vector of control variables, X: Latin America and International of Africa are dummy variables that take the value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Catholic, Muslim, and Financial Other Religion indicate the percentage of the population that follows a particular religion (Catholic, Muslim, or religions other than Catholic, Muslim, or Protestant, respectively). Independence is the percentage of years since 1776 that a country has been independent. Ethnic Fractionalization is the probability that two randomly selected individuals in a country will not speak the same language. Regressions are estimated using Ordinary Least Squares. Robust standard errors are given in parentheses. The symbols *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively. Detailed variable definitions Economic Corporate and sources are given in the data appendix. Ethnic s Settler Mortality Latin America Africa Catholic Muslim Other Religion Independence Fractionalization Adjusted-R2 Obs. 70 (2003) Finance Private 0.164*** 0.440 70 Credit (0.030) 0.137*** 0.230*** 0.163 0.500 70 137 (0.038) (0.086) (0.113) ­181 0.161*** 0.004 0.003 0.002 0.490 70 (0.028) (0.003) (0.210) (0.004) 0.178*** 0.168 0.460 70 (0.031) (0.138) 0.166*** 0.025 0.432 70 (0.033) (0.076) 0.140*** 0.224* 0.131 0.038 0.080 0.489 70 (0.038) (0.128) (0.121) (0.176) (0.103) Stock 0.170*** 0.267 70 Market (0.047) Development 0.182** 0.204 0.008 0.305 70 (0.071) (0.132) (0.199) 0.159*** 0.001 0.001 0.004 0.372 70 (0.042) (0.003) (0.003) (0.005) 0.191*** 0.260 0.297 70 T. (0.056) (0.158) Beck 0.198*** 0.261 0.292 70 (0.059) (0.167) et 0.189** 0.145 0.057 0.099 0.141 0.294 70 al. (0.073) (0.127) (0.198) (0.180) (0.183) /Journal Property 0.349*** 0.177 69 Rights (0.099) of 0.151 0.489* 0.903** 0.220 69 Chapter Financial (0.117) (0.290) (0.352) 0.339*** 0.015* 0.012 0.010 0.194 69 (0.092) (0.009) (0.008) (0.011) One Economic 0.377*** 0.336 0.175 69 (0.104) (0.387) 0.338*** 0.102 0.166 69 (0.113) (0.415) s 0.180 0.271 1.010** 0.418 0.345 0.214 69 70 (2003) (0.125) (0.407) (0.392) (0.550) (0.514) 137 ­181 157 21 22 A Reader in International Corporate Finance 158 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 countries such as Nicaragua and Jamaica with bad endowments (log settler mortality rates of 5.1 and 4.9, respectively) have lower levels of financial intermediary development (0.25 and 0.27, respectively) than Chile (0.54), which had a log settler mortality rate of 4.2. Furthermore, SETTLER MORTALITY enters all of the PROPERTY RIGHTS regressions negatively and significantly, except those including continent dummies. As noted, there is an extremely high correlation between AFRICA and SETTLER MORTALITY. Also, as we report below, when we use an alternative measure of property rights protection, settler mortality continues to enter significantly even when controlling for AFRICA. These results support the view that high settler mortality rates are negatively associated with the level of financial development today, and are robust to an assortment of control variables. Such findings are fully consistent with the AJR (2001, 2002) assertion that a colony's environmental endowments influenced how it was colonized--whether it was an extractive colony or a settler colony--with long- lasting implications for institutional development. 3.3. Law, endowments, and finance Table 4 presents regression results on the relation between financial development and both law and endowments while controlling for other exogenous determinants of financial development. Table 4 regressions provide strong support for the endowment view of financial development. SETTLER MORTALITY enters all of the PRIVATE CREDIT and STOCK MARKET DEVELOPMENT regressions significantly at the five-percent level even when controlling for legal origin, continent, religious composition, the length of time the country has been independent, and ethnic diversity. The sizes of the coefficients on SETTLER MORTALITY in the PRIVATE CREDIT and STOCK MARKET DEVELOPMENT regressions are very similar to those in Table 3, in which the regressions do not also control for legal origin. Also similar to Table 3, the Table 4 regressions indicate that SETTLER MORTALITY exerts a statistically significant impact on PROPERTY RIGHTS except when controlling for the AFRICA dummy variable (because of the very high correlation between the rate of settler mortality and countries in Sub-Saharan Africa). As discussed below, however, when we use an alternative measure of property rights protection, settler mortality enters significantly even when controlling for the AFRICA dummy variable. In sum, poor endowments--as measured by settler mortality--are negatively associated with financial development today. Even when controlling for the legal tradition of the colonizers and other possible determinants of financial development, initial endowments of the colonies help explain cross-country variation in financial development today, which is strongly supportive of the AJR (2001, 2002) endowment view. Table 4 regressions also provide support for the law and finance view, though some qualifications are necessary. When controlling for SETTLER MORTALITY, the relationship between financial intermediary development (PRIVATE CREDIT) Chapter One 23 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 159 and legal origin is not robust to the inclusion of various control variables. However, FRENCH LEGAL ORIGIN is negatively and significantly associated with PROPERTY RIGHTS in all of the regressions when controlling for SETTLER MORTALITY. Putting aside regressions that include CATHOLIC (which is extremely positively correlated with French Civil law), FRENCH LEGAL ORIGIN is also negatively and significantly linked with STOCK MARKET DEVELOPMENT. To the extent that equity markets rely more than banking institutions on well-functioning legal systems to defend the rights of indi- vidual investors, these findings are consistent with the thrust of the law and finance view. Subject to the qualifications discussed above, we interpret the results as generally consistent with the LLSV (1998) theory that the French Civil law tends to place greater emphasis on the rights of the state versus the rights of individuals, with negative repercussions on financial contracting. In contrast, the British Common law tends to place greater emphasis on the contractual rights of individual investors, with positive implications for financial development. While LLSV (1998) document the link between financial development and legal origin, this paper goes much further in controlling for alternative explanations. Our results demonstrate a strong connection between legal origin and both stock market development and private property rights protection, but we also show that the link between legal origin and financial intermediary development is not robust to the inclusion of numerous control variables. In comparing the independent explanatory power between law and endowments, Tables 2­4 indicate that endowments explain a greater amount of the cross-country variation in financial intermediary and stock market development than legal origin. Consider, for instance, the regressions in Tables 2­4 that do not include any regressors beyond FRENCH LEGAL ORIGIN and SETTLER MORTALITY. The adjusted R-square in the PRIVATE CREDIT­FRENCH LEGAL ORIGIN regression is 0.12 (Table 2), while it is 0.44 in the PRIVATE CREDIT­SETTLER MORTALITY regression (Table 3). Furthermore, when adding FRENCH LEGAL ORIGIN to the SETTLER MORTALITY regression, the adjusted R-square only rises from 0.44 to 0.48 (Table 4). As also indicated above, legal origin does not enter the PRIVATE CREDIT regression robustly when including various control variables, but endowments remain negatively and significantly linked with financial intermediary development across various control variables. Turning to private property rights protection, the explanatory power of law and endowments in the PROPERTY RIGHTS regressions is very similar. However, the STOCK MARKET DEVELOPMENT regressions again illustrate the greater explanatory power of endowments. The adjusted R-square in the STOCK MARKET DEVELOPMENT- FRENCH LEGAL ORIGIN regression is 0.17 (Table 2), and is 0.27 in the SETTLER MORTALITY regression (Table 3). Furthermore, when adding FRENCH LEGAL ORIGIN to the SETTLER MORTALITY regression, the adjusted R-square only rises from 0.27 to 0.36 (Table 4). Thus, while legal origin significantly enters all of the stock market development regressions that do not control for religious composition (Table 4), endowments explain a greater 24 160 Table 4 Law, endowments, and finance. The regression estimated is: Financial Sector Development ¼ a þ b1 French Legal Origin+b2 Settler Mortality+b3X; where Financial Sector Development is either Private Credit, Stock Market Development, or Property Rights. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property rights reflects the degree to which government enforces laws that protect private property, with higher numbers indicating better enforcement. French Legal Origin is a dummy variable T. that takes on the value one for countries with French Civil law tradition, and zero otherwise. Settler Mortality is the log of the annualized deaths per thousand Beck European soldiers in European colonies in the early 19th century. The regressions also include a vector of control variables, X: Latin America and Africa are dummy variables that take the value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Catholic, Muslim, and Other Religion A et indicate the percentage of the population that follows a particular religion (Catholic, Muslim, or religions other than Catholic, Muslim, or Protestant, Reader al. respectively). Independence is the percentage of years since 1776 that a country has been independent. Ethnic Fractionalization is the probability that two randomly selected individuals in a country will not speak the same language. Regressions are estimated using Ordinary Least Squares. Robust standard errors /Journal are given in parentheses. The symbols *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively. Detailed variable definitions and sources are in given in the data appendix. International of Ethnic Financial Settler French Legal Latin Africa Catholic Muslim Other Independence Fractionalization Adjusted-R2 Obs. Mortality Origin America Religion Economic Private 0.151*** 0.141** 0.480 70 Corporate Credit (0.026) (0.059) 0.130*** 0.097* 0.194* 0.148 0.514 70 (0.034) (0.055) (0.082) (0.108) s 0.157*** 0.054 0.004 0.003 0.002 0.486 70 70 (2003) Finance (0.028) (0.074) (0.003) (0.002) (0.003) 0.160*** 0.115 0.090 0.478 70 (0.028) (0.077) (0.134) 137 0.148*** 0.144** 0.024 0.472 70 (0.028) (0.059) (0.073) ­181 0.127*** 0.108 0.0214* 0.110 0.029 0.100 0.505 70 (0.035) (0.069) (0.117) (0.121) (0.185) (0.110) Stock 0.145*** 0.268*** 0.358 70 Market (0.038) (0.085) Development 0.164*** 0.229*** 0.118 0.028 0.363 70 (0.061) (0.079) (0.123) (0.181) 0.147*** 0.146 0.000 0.001 0.005 0.380 70 (0.040) (0.090) (0.003) (0.003) (0.005) 0.156*** 0.240*** 0.095 0.353 70 (0.049) (0.072) (0.135) 0.167*** 0.246*** 0.178 0.364 70 (0.049) (0.080) (0.150) 0.161** 0.232*** 0.123 0.012 0.044 0.098 0.346 70 T. (0.063) (0.071) (0.115) (0.190) (0.163) (0.171) Beck Property 0.279*** 0.781*** 0.304 69 et al. Rights (0.080) (0.223) 0.088 0.810*** 0.183 0.786** 0.348 69 /Journal (0.101) (0.216) (0.274) (0.334) 0.277*** 0.853*** 0.004 0.003 0.008 0.281 69 (0.082) (0.310) (0.008) (0.008) (0.010) of 0.251*** 0.856*** 0.256 0.299 69 Financial Chapter (0.087) (0.227) (0.371) 0.232** 0.833*** 0.398 0.307 69 (0.095) (0.231) (0.401) Economic One 0.082 0.816*** 0.197 0.860** 0.091 0.184 0.328 69 (0.110) (0.216) (0.328) (0371) (0.434) (0.480) s 70 (2003) 137 ­181 161 25 26 A Reader in International Corporate Finance 162 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 proportion of the cross-country variation in stock market development than legal origin. It is difficult to compare the sizes of the coefficients on SETTLER MORTALITY and FRENCH LEGAL ORIGIN because a change in legal origin is obviously large and discrete. Nevertheless, we compare a change in legal origin with a change in SETTLER MORTALITY from the second quintile to the fourth quintile (i.e., a change of 2.1), which is less than a two standard deviation change in SETTLER MORTALITY (2.5). Using, for instance, the coefficients in the last row of the stock market development indicators in Table 4, this implies a change in STOCK MARKET DEVELOPMENT of 0.23 from a legal origin change and 0.34 from the endowment change. The effect of the endowment change is approximately 50% larger. Turning to the control variables, the regression analyses do not indicate a robust, consistent relationship between the continent dummy variables, the religious composition measures, the length of national independence, nor the level of ethnic diversity, on the one hand, and financial development, on the other hand, when controlling for legal origin and national endowments. The Table 4 regressions--as well those in Tables 2 and 3 --do not demonstrate a significant, robust relation between any of these control variables and any of the measures of financial development when controlling for legal origin and endowments. As emphasized above, French Civil law countries also tend to be predominantly Catholic, much of Latin America adopted the French Civil law tradition, and Sub-Saharan Africa had very high rates of settler mortality. Nevertheless, while a consistent pattern of results emerges for law and endowments, we do not observe a robust set of results on the continent dummies, religious composition variables, independence indicator, or ethnic diversity measure. 4. Robustness test 4.1. Political structure As a robustness check, we control for political structure. North (1990) argues that once groups gain power, they shape policies and institutions to their own advantages. The work of Finer (1997) and Damaska (1986) further suggests that centralized or otherwise powerful states will be more responsive to and efficient at implementing the interests of the elite than a decentralized or more competitive political system endowed with checks and balances. LLSV (1998) do not control for political structure in their examination of the law and finance view. In a different approach, Rajan and Zingales (2003) argue that financial systems do not develop monotonically over time. This observation is not fully consistent with the law and endowment theories, which are based on time invariant factors. Rajan and Zingales (2003) instead propose a theory of financial development based on controlling interest groups. In our sensitivity analyses, we focus on the political structure view because we encounter data limitations concerning interest groups for our broad cross-section of countries. Chapter One 27 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 163 To assess whether law and endowments continue to explain cross-country differences in financial development after controlling for the structure of the political environment, we use two measures of political openness. LEGISLATIVE COMPETITION is an index of the degree of competitiveness of the last legislative election, ranging from 1 (non-competitive) to 7 (most competitive). CHECKS measures the number of influential veto players in legislative and executive initiatives. These data are from Beck et al. (2001a). The politics and finance view predicts that greater competition and more checks and balances will limit the ability of the elite to dictate policy and institutional development. To control for the endogenous determination of political structures, we use instrumental variables.12 As instruments, we include the religious composition variables, independence, and ethnic diversity. We include the religious variables since Landes (1998) and others argue that the Catholic and Muslim religions tend to produce hierarchical political systems. We include independence since more years of independence may permit greater latitude to shape domestic political institutions. We include ethnic diversity since some theories suggest that ethnic diversity will tend to create political systems that stymie competition and permit greater discretion on the part of the controlling party (see, e.g., Alesina et al., 1999). The instrumental variables significantly explain cross-country variation in the political structure indexes at the one-percent significance level. Nevertheless, given the valid skepticism associated with obtaining fully acceptable instrumental variables for political structure, we note that: (i) we present these exploratory results as a robustness check on the endowment and law theories and not as a strong test of the political channel; and, (ii) we are particularly circumspect in interpreting these instrumental variable regressions. Table 5 instrumental variable results are consistent with the law and endowment theories while controlling for the structure of the political system, and suggest that the politics mechanism is not the only channel through which legal origin and endowments influence financial development. As shown, legal origin and endow- ments continue to enter the financial development regressions significantly even when controlling for the exogenous component of political structure except for SETTLER MORTALITY in the PROPERTY RIGHTS regressions. The political structure variables do not enter any of the financial development regressions significantly. Thus, there is no evidence in Table 5 that political structure explains cross-country variation in financial development beyond the explanatory power of legal origin and environmental endowments. Furthermore, the results do not suggest that political structure is the only channel through which legal origin and initial endowments influence financial development. If political structure were the only channel through which law and initial endowments influence financial development, we would have found significant coefficients on the political structure indicators and insignificant coefficients on the legal origin and endowment indicators. We find the opposite. Moreover, we run two-stage least squares regressions with financial 12We find the same results hold when using ordinary least squares and not instrumenting for political structure. 28 A Reader in International Corporate Finance 164 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 Table 5 Law, endowments, politics, and finance The regression estimated in is: Financial Sector Development ¼ a þ b1 French Legal Origin+b2 Settler Mortality+b3: Political Structure, where Financial Sector Development is either Private Credit, Stock Market Development, or Property Rights and Political Structure is either Legislative Competition or Checks. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property Rights reflects the degree to which government enforces laws that protect private property, with higher numbers indicating better enforcement. French Legal Origin is a dummy variable that takes on the value one for countries with French Civil law tradition, and zero otherwise. Settler Mortality is the log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. Legislative Competition is an indicator of competition in the last legislative election. Checks measures the number of veto-players in the political decision process. These last two measures are averaged over 1990­ 1995. Detailed variable definitions and sources are given in the data appendix. All regressions are estimated using Instrumental Variables, two-stage least squares. In the first-stage regressions the Political Structure indicators are regressed on Legal Origin, Settler Mortality, Catholic, Muslim, Other Religion, Independence and Ethnic Fractionalization. Robust standard errors are given in parentheses. The symbols *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively. P-Values are given in parentheses for the test of the over-identifying restrictions (OIR). OIR w2-test Settler French Legislative Checks Adjusted-R2 Obs. (p-value) Mortality Legal Competition Origin Private 3.693 0.169*** 0.123** 0.037 0.429 68 Credit (0.449) (0.051) (0.059) (0.048) 2.405 0.184*** 0.160** 0.083 0.317 68 (0.662) (0.044) (0.064) (0.060) Stock 1.232 0.199** 0.215** 0.090 0.224 68 Market (0.873) (0.090) (0.083) (0.079) Development 2.445 0.177** 0.274** 0.095 0.192 68 (0.655) (0.074) (0.105) (0.086) Property 3.214 0.186 0.858*** 0.093 0.348 67 Rights (0.523) (0.154) (0.223) (0.154) 3.055 0.177 0.780*** 0.160 0.323 67 (0.549) (0.159) (0.225) (0.243) development as the dependent variable and political structure as the only explanatory variable in the second stage. The instruments are legal origin and settler mortality. While political structure enters the financial development regression significantly and with the predicted sign, the instruments do not pass the test of over- identifying restrictions. These results do not reject the importance of political factors in shaping finance. Rather, the evidence in this paper suggests that legal origin and endowments influence financial development beyond the structure of the political system.13 13Beck et al. (2003) examine the different channels through which legal origin affects financial development. Chapter One 29 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 165 4.2. Alternative samples To assess the robustness of the results, we examine different subsamples of countries. In these robustness checks, we only include two regressions to keep the table to a manageable length. We include one regression with only the legal origin and endowment variables as regressors and a second regression that also includes continent dummy variables, years of independence, and ethnic diversity. We do not include the religious indicators because they do not enter any of the Tables 2­4 regressions significantly at the five percent level. Table 6 presents regression results on five different sub-samples of countries. Panel A excludes Australia, Canada, New Zealand, and the United States from the regression. After omitting these countries, the data continue to support both the law and endowment views of financial development. The results are fully consistent with the full-sample results in Table 4. FRENCH LEGAL ORIGIN enters all of the STOCK MARKET DEVELOPMENT and PROPERTY RIGHTS regressions significantly, but does not enter the PRIVATE CREDIT regression significantly when controlling for other determinants. SETTLER MORTALITY enters all of the PRIVATE CREDIT and STOCK MARKET DEVELOPMENT regressions significantly, but does not enter significantly in the PROPERTY RIGHTS regression when controlling for AFRICA. In Panels B and C, we examine French legal origin and British legal origin countries separately to test whether settler mortality accounts for cross-country variation in financial development within each group. Again, the results support the view that the disease environment encountered by European settlers shaped the formation of long-lasting financial institutions. The results do suggest, however, that the SETTLER MORTALITY-finance relationship is stronger for the British legal origin sample of countries than for the French legal origin sample. SETTLER MORTALITY enters negatively and significantly in all the regressions in Panel C (British-only legal origin countries), except for the PROPERTY RIGHTS regression in which we include the African dummy variable (which we discuss above). SETTLER MORTALITY is not as robustly related to equity market development and property rights in the French legal origin subsample--it does not enter significantly once we control for AFRICA. Further, SETTLER MORTALITY explains less than half of the cross-country variation in financial development among French Civil law countries than among British Common law countries, as can be seen from comparing the adjusted R2 statistics in Panels B and C. Finally, we also examine high and low settler mortality countries. Here, we assess whether legal origin explains financial development within the high (above the median) settler mortality countries and within the low (below the median) settler mortality countries. Note there are more countries in Panel E than Panel D because Algeria and Morocco have exactly the median level of SETTLER MORTALITY and are allocated to the below-median group. When we allocate them to the above-median group, or split them between the two groups, we obtain the same results. The results are broadly consistent with earlier findings. FRENCH LEGAL ORIGIN is not strongly associated with financial intermediary develop- ment (PRIVATE CREDIT) in the high-mortality countries. Nevertheless, legal 30 166 Table 6 Law, endowments, and finance: alternative samples. The regressions estimated in Panel A are: Financial Sector Development ¼ a þ b1 French Legal Origin + b2 Settler Mortality+b3X; where Financial Sector Development is either Private Credit, Stock Market Development, or Property Rights. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property rights reflects the degree to which government enforces laws that protect private property, with higher numbers indicating better enforcement. French Legal Origin is a dummy variable that takes on the value one for countries with French Civil law tradition, and zero otherwise. Settler Mortality is the log of the annualized T. deaths per thousand European soldiers in European colonies in the early 19th century. The regressions also include a vector of control variables, X: Latin Beck America and Africa are dummy variables that take the value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Independence A is the percentage of years since 1776 that a country has been independent. Ethnic Fractionalization is the probability that two randomly selected individuals in et Reader a country will not speak the same language. The regressions in Panel A exclude Australia, Canada, New Zealand and the U.S., the regressions in Panels B­C al. are: Financial Sector Development ¼ a þ b1 Settler Mortality + b2X: The regressions in Panel B include only French Legal Origin and in Panel C only British Legal Origin countries. The regressions estimated in Panels D­E are: Financial Sector Development ¼ a þ b1 French Legal Origin + b2X: The regressions in /Journal in Panel D include countries with Settler Mortality above the median and the regressions in Panel E countries with Settler Mortality below the median. There are International more countries in Panel E than in Panel D because Algeria and Morocco have exactly the median level of Settler Mortality and are allocated to the below- of median group. Regressions are estimated using Ordinary Least Squares. Robust standard errors are given in parentheses. The symbols *, **, *** indicate Financial significance at the 10%, 5%, and 1% levels, respectively. Detailed variable definitions and sources are given in the data appendix. Panel A: Excluding Australia, Canada, New Zealand, and the United States Ethnic Economic Corporate Settler Mortality French Legal Latin America Africa Independence Fractionalization Adjusted-R2 Obs. Origin s Private 0.129*** 0.102* 0.379 66 70 Finance Credit (0.030) (0.061) (2003) 0.127*** 0.031 0.072 0.088 0.216** 0.063 0.419 66 (0.041) (0.064) (0.095) (0.114) (0.100) (0.100) 137 Stock 0.161*** 0.291*** 0.342 66 ­181 Market (0.051) (0.106) Development 0.180** 0.281*** 0.212 0.009 0.147 0.046 0.342 66 (0.069) (0.100) (0.158) (0.192) (0.212) (0.166) Property 0.200** 0.654*** 0.173 65 Rights (0.084) (0.233) 0.025 0.571** 0.243 0.832** 0.517 0.380 0.238 65 (0.101) (0.230) (0.369) (0.323) (0.484) (0.478) Panel B: French Legal Origin countries Ethnic Settler Mortality Latin America Africa Independence Fractionalization Adjusted-R2 Obs. Private 0.080*** 0.217 45 Credit (0.029) 0.066** 0.044 0.161* 0.243** 0.082 0.390 45 T. (0.029) (0.088) (0.086) (0.095) (0.086) Beck Stock 0.037** 0.057 45 et Market (0.016) al. Development 0.018 0.023 0.001 0.034 0.054 0.018 45 (0.024) (0.059) (0.065) (0.077) (0.065) /Journal Property 0.204* 0.047 44 Rights (0.112) of Chapter 0.015 0.073 0.937** 0.141 0.352 0.087 44 Financial (0.120) (0.269) (0.392) (0.389) (0.509) One Panel C: British legal origin countries Economic Ethnic Settler Mortality Latin America Africa Independence Fractionalization Adjusted-R2 Obs. s Private 0.204*** 0.532 25 70 Credit (0.042) (2003) 0.158** 0.074 0.017 0.561 0.136 0.526 25 (0.066) (0.217) (0.261) (0.444) (0.387) 137 Stock 0.227*** 0.330 25 ­181 Market (0.064) Development 0.313** 0.176 0.007 0.547 0.477 0.329 25 (0.113) (0.313) (0.478) (0.573) (0.687) Property 0.335*** 0.205 25 Rights (0.108) 0.086 0.226 0.816 1.339 0.131 0.184 25 167 (0.193) (0.909) (0.750) (0.870) (1.471) 31 32 Table 6 (continued) 168 Panel D: Countries above median for settler mortality Ethnic French Legal Origin Latin America Africa Independence Fractionalization Adjusted-R2 Obs. Private 0.039 0.014 34 Credit (0.060) T. 0.025 0.055 0.331*** 0.356*** 0.040 0.538 34 Beck (0.040) (0.046) (0.024) (0.082) (0.083) A et Reader Stock 0.082** 0.178 34 al. Market (0.037) Development 0.062** 0.078 0.152*** 0.136 0.050 0.342 34 /Journal in (0.027) (0.047) (0.013) (0.097) (0.057) International Property 1.036*** 0.249 33 of Rights (0.327) Financial 0.654** 0.374 0.783*** 2.458*** 0.346 0.400 33 (0.309) (0.535) (0.181) (0.740) (0.723) Economic Corporate Panel E: Countries below median for settler mortality Ethnic French Legal Origin Latin America Africa Independence Fractionalization Adjusted-R2 Obs. s 70 Finance Private 0.414*** 0.297 36 (2003) Credit (0.128) 0.303** 0.305* 0.012 0.197 0.150 0.314 36 (0.142) (0.170) (0.235) (0.285) (0.294) 137 Stock 0.611*** 0.313 36 ­181 Market (0.190) Development 0.613*** 0.001 0.290 0.011 0.037 0.255 36 (0.217) (0.255) (0.399) (0.290) (0.429) Property 0.870** 0.194 36 Rights (0.324) 0.824** 0.569* 1.424*** 0.968** 0.120 0.358 36 (0.318) (0.284) (0.473) (0.420) (0.775) Chapter One 33 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 169 origin is strongly and negatively associated with STOCK MARKET DEVELOP- MENT and PROPERTY RIGHTS in both subsamples and PRIVATE CREDIT in the low-mortality sample. While one notes some differences when looking across different subsamples, the same basic pattern emerges as in the full sample: law and endowments explain financial development, though the endowment-intermediary (PRIVATE CREDIT) relationship is more robust than the law-intermediary (PRIVATE CREDIT) relationship. 4.3. Alternative indicators of financial development Next, we examine alternative measures of financial development. Specifically, instead of examining financial intermediary credit to the private sector (PRIVATE CREDIT), we use the demand and interest-bearing liabilities of financial intermediaries (LIQUID LIABILITIES). Also, instead of using market capitaliza- tion to measure stock market development, we examine the total value of stock transactions in the economy as a share of GDP (TOTAL VALUE TRADED). Finally, instead of utilizing the private property rights protection index as used by LLSV (1999), we examine: (a) the International Country Risk Guide (ICRG) measure of the degree to which a country adheres to the rule of law (RULE OF LAW); and, (b) the Kaufmann et al. (1999) AGGREGATE RULE OF LAW index. However, the RULE OF LAW and AGGREGATE RULE OF LAW indicators are available for fewer countries, 63 and 68, respectively, than the PROPERTY RIGHTS measure used throughout the paper thus far. Table 7 indicates that these alternative indicators produce results that are consistent with those discussed above. Settler mortality is significantly, negatively associated with the new measures of financial intermediary development, stock market development, and property rights protection. Although the RULE OF LAW­SETTLER MORTALITY relationship weakens when including continent dummy variables, years of independence, and ethnic diversity, the AGGREGATE RULE OF LAW­SETTLER MORTALITY relationship remains significant when controlling for these country traits. Since SETTLER MORTALITY loses its significant relationship with two of our three measures of private property rights protection, only when including a dummy variable for AFRICA (where settler mortality rates were very high), we interpret these findings as broadly consistent with the view that the initial endowments in the various colonies helped shape institutional approaches to the protection of private property rights. FRENCH LEGAL ORIGIN is negatively associated with all the alternative financial development indicators except financial intermediary development. As noted above, the relationship between law and financial intermediary development is more fragile than the endowment­intermediary relationship. Unlike in the PROPERTY RIGHTS regressions of Tables 2­4, SETTLER MORTALITY explains a larger share of the variation in the RULE OF LAW and AGGREGATE RULE OF LAW regressions than FRENCH LEGAL ORIGIN. As discussed in Section 3.3, we draw this conclusion by comparing adjusted-R2 statistics across regressions with only legal origin, with only SETTLER MORTALITY, and then 34 170 Table 7 Law, endowments, and finance alternative finance indicators. The regression estimated is: Financial Sector Development ¼ a þ b1 French Legal Origin + b2 Settler Mortality + b3X; where Financial Sector Development is either Liquid Liabilities, Total Value Traded, Rule of Law, or Aggregate Rule of Law. Liquid Liabilities is currency plus demand and interest-bearing liabilities of banks and nonbank financial intermediaries, divided by GDP. Total value traded is the total value of shares traded as a share of GDP. Rule of law (ICRG) accounts for the degree to which a country adheres to the rule of law. Aggregate Rule of Law is an aggregate indicator estimated with an unobserved- T. components model using a large number of individual indicators from different sources (Kaufmann et al., 1999). French Legal Origin is a dummy variable that Beck takes on the value one for countries with French Civil law tradition, and zero otherwise. Settler Mortality is the log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. The regressions also include a vector of control variables, X: Latin America and Africa are A et dummy variables that take the value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Independence is the percentage of years Reader al. since 1776 that a country has been independent. Ethnic Fractionalization is the probability that two randomly selected individuals in a country will not speak the same language. Regressions are estimated using Ordinary Least Squares. Robust standard errors are given in parentheses. The symbols *, **, *** indicate /Journal significance at the 10%, 5%, and 1% levels, respectively. Detailed variable definitions and sources are given in the data appendix. in International Ethnic of Settler Mortality French Legal Origin Latin America Africa Independence Fractionalization Adjusted-R2 Obs. Financial Liquid 0.150*** 0.073 0.433 70 Liabilities (0.02958) (0.05731) 0.148*** 0.054 0.085 0.210** 0.439*** 0.015 0.604 70 Economic Corporate (0.034) (0.058) (0.079) (0.083) (0.117) (0.107) Total 0.058*** 0.105** 0.274 70 s Value (0.018) (0.041) 70 Finance Traded 0.043** 0.081*** 0.129** 0.109 0.035 0.049 0.292 70 (2003) (0.020) (0.030) (0.050) (0.074) (0.070) (0.087) 137 Rule of Law 0.285** 0.553* 0.141 63 (0.133) (0.314) ­181 0.041 0.668** 1.246*** 0.764 0.881 1.109* 0.238 63 (0.180) (0.334) (0.448) (0.592) (0.555) (0.625) Aggregate Rule 0.362*** 0.395* 0.349 68 of Law (0.076) (0.190) 0.292** 0.373* 0.494* 0.169 0.187 0.441 0.348 68 (0.129) (0.216) (0.262) (0.407) (0.303) (0.355) Chapter One 35 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 171 with SETTLER MORTALITY and legal origin dummies included simultaneously. The regressions with only SETTLER MORTALITY and only the legal origin dummy variable for this sample of countries are not reported. 4.4. Alternative endowment indicator Next, we use an alternative measure of endowments, LATITUDE, which equals the absolute value of the latitude of each country normalized to lie between zero and one. We take the data from LLSV (1999). Countries that are closer to the equator will tend to have a more tropical climate that is inhospitable to European settlers and therefore will more likely foster extractive institutions.14 However, LATITUDE is not as precise an indicator of the conditions facing European settlers as SETTLER MORTALITY and thus LATITUDE is not as precise an empirical proxy for the AJR (2001) endowment theory as SETTLER MORTALITY. LATITUDE directly measures geographic location, not climatic conditions. Accordingly, we have focused our analyses on SETTLER MORTALITY, and only include LATITUDE in our robustness checks. Table 8 regressions with LATITUDE indicate, albeit less robustly than those with SETTLER MORTALITY, that countries closer to the equator have lower levels of financial development than countries in more temperate climates. LATITUDE is positively associated with PROPERTY RIGHTS after using the array of control variables discussed above. LATITUDE is also significantly and positively linked with PRIVATE CREDIT in all of the regressions that do not include AFRICA, which is very highly correlated with LATITUDE. There is not a strong link between LATITUDE and stock market development. Using LATITUDE, we do find a strong link between legal origin and financial development. FRENCH LEGAL ORIGIN enters significantly in all regressions and its inclusion substantially increases the adjusted R2 over those regressions that only include LATITUDE. Especially given the imprecise nature of LATITUDE as proxy for the AJR (2001) endowment theory, we view Table 8 as confirmation of our earlier findings. 4.5. Tobit estimation Finally, we estimate the stock market development equations using a Tobit estimator. Both STOCK MARKET DEVELOPMENT (market capitalization divided by GDP) and TOTAL VALUE TRADED (stock market trading divided by GDP) have many countries with zero values. Thus, we re-estimate the equation using a Tobit estimator. As shown in Table 9, we find that both legal origin and endowments enter significantly in all of the regressions when using the Tobit estimator, confirming earlier results. 14While some authors stress the direct impact of tropical environments on production (Kamarck, 1976; Crosby, 1989; and Gallup et al., 1998), AJR (2002) and Easterly and Levine (2003) show that the environment tends to influence economic development primarily through its impact on institutions. 36 172 Table 8 Law, endowments, and finance: alternative endowment indicator T. The regression estimated in Panel A is: Financial Sector Development ¼ a þ b1 Latitude + b3X; where Financial Sector Development is either Private Credit, Beck Stock Market Development, or Property Rights. Private Credit is the value of credits by financial intermediaries to the private sector as a share of GDP. Stock Market Development measures the value of shares listed on the stock exchange as a share of GDP. Property Rights reflects the degree to which government A et Reader enforces laws that protect private property, with higher numbers indicating better enforcement. Latitude is the absolute value of the latitude of a country, al. scaled between zero and one. The regressions also include a vector of control variables, X: Latin America and Africa are dummy variables that take the value /Journal one if the country is located in Latin America or Sub-Saharan Africa, respectively. Independence is the percentage of years since 1776 that a country has been in independent. Ethnic Fractionalization is the probability that two randomly selected individuals in a country will not speak the same language. The regression estimated in Panel B is: Financial Sector Development ¼ a þ b1 French Legal Origin + b2 Latitude + b3X: French Legal Origin is a dummy variable that International of takes on the value one for countries with French Civil law tradition, and zero otherwise. Regressions are estimated using Ordinary Least Squares. Robust Financial standard errors are given in parentheses. The symbols *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively. Detailed variable definitions and sources are given in the data appendix. Panel A: Latitude and finance Economic Corporate Ethnic Latitude Latin America Africa Independence Fractionalization Adjusted-R2 Obs. s Private 1.048*** 0.189 70 70 Credit (0.300) (2003) Finance 0.423 0.319** 0.380*** 0.034 0.018 0.346 70 (0.327) (0.147) (0.125) (0.168) (0.135) 137 Stock 0.491 0.012 70 ­181 Market (0.386) Development 0.171 0.402** 0.470* 0.085 0.121 0.120 70 (0.680) (0.198) (0.244) (0.126) (0.198) Property 3.232*** 0.165 69 Rights (0.784) 2.600*** 0.040 1.122*** 0.569 0.708 0.267 69 (0.952) (0.429) (0.339) (0.474) (0.462) Panel B: Latitude, law and finance Ethnic Latitude French Legal Origin Latin America Africa Independence Fractionalization Adjusted-R2 Obs. Private 0.970*** 0.206*** 0.286 70 Credit (0.276) (0.079) 0.381 0.162** 0.288** 0.312** 0.122 0.055 0.392 70 T. (0.301) (0.078) (0.127) (0.122) (0.171) (0.141) Beck Stock 0.360 0.346*** 0.175 70 et Market (0.355) (0.122) al. Development 0.251 0.312*** 0.341* 0.339 0.256** 0.051 0.229 70 /Journal (0.613) (0.104) (0.179) (0.134) (0.127) (0.173) Property 2.924*** 0.873*** 0.335 69 of Rights (0.659) (0.224) Chapter Financial 2.398*** 0.821*** 0.120 0.783** 0.120 0.517 0.392 69 (0.843) (0.201) (0.341) (0.308) (0.353) (0.453) Economic One s 70 (2003) 137 ­181 173 37 38 174 Table 9 T. Law, endowments, and stock market development: Tobit regressions The regression estimated is: Financial Sector Development ¼ a þ b1 French Legal Origin + b2 Settler Mortality + b3X; where Financial Sector Development Beck is either Stock Market Development or Total Value Traded. Stock Market Development measures the value of shares listed on the stock exchange as a share of A et GDP. Total value traded is the total value of shares traded as a share of GDP. French Legal Origin is a dummy variable mat takes on the value one for Reader al. countries with French Civil law tradition, and zero otherwise. Settler Mortality is the log of the annualized deaths per thousand European soldiers in European colonies in the early 19th century. The regressions also include a vector of control variables, X: Latin America and Africa are dummy variables that take the /Journal value one if the country is located in Latin America or Sub-Saharan Africa, respectively. Independence is the percentage of years since 1776 that a country has in been independent. Ethnic Fractionalization is the probability that two randomly selected individuals in a country will not speak the same language. International Regressions are estimated using Tobit, censored-normal. Standard errors are given in parentheses. The symbols *, **, *** indicate significance at the 10%, 5%, of and 1% levels, respectively. Detailed variable definitions and sources are given in the data appendix. Financial Ethnic Settler Mortality French Legal Origin Latin America Africa Independence Fractionalization Adjusted-R2 Obs. Economic Corporate Stock 0.269*** 0.353*** 0.337 70 Market (0.051) (0.116) Development 0.207*** 0.413*** 0.087 0.347 0.246 0.342 0.329 70 s (0.069) (0.140) (0.177) (0.234) (0.244) (0.291) 70 (2003) Finance Total Value 0.117*** 0.144*** 0.792 70 Traded (0.024) (0.055) 137 0.059* 0.170*** 0.121 0.301*** 0.142 0.176 1.014 70 (0.031) (0.064) (0.080) (0.108) (0.111) (0.134) ­181 Chapter One 39 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 175 5. Conclusions This paper assesses two theories regarding the historical determinants of financial development. The law and finance theory predicts that historically determined differences in legal origin can explain cross-country differences in financial development observed today. Specifically, the law and finance theory predicts that countries that inherited the British Common law tradition obtained a legal tradition that tends to both emphasize private property rights and support financial development to a much greater degree than countries that obtained the French Civil law tradition. The endowment theory, on the other hand, predicts that the initial environmental endowments encountered by European colonizers shaped the types of long-lasting institutions created by those colonizers. Specifically, hospitable endowments favored the construction of settler colonies, where Europeans established secure property rights. In contrast, colonies with high settler mortality rates fostered the construction of extractive colonies, where Europeans established institutions that facilitated state control and resource extraction. According to the endowment theory, the long-lasting institutions created by colonizers continue to influence financial development today. Although both the law and endowment theories stress the importance of how initial conditions influence institutions today, there are crucial differences. The law and finance theory focuses on the legal tradition spread by the colonizer. Thus, the identity of the colonizer is key. The endowment theory focuses on how the colony's endowments shaped the construction of long-lasting institutions. Thus, the endowment theory focuses on the conditions of the colony, not the identity of the colonizer. The paper provides qualified support for the law and finance theory (Hayek, 1960; LLSV, 1998). One important qualification is that the connection between legal origin and financial intermediary development is not robust to controlling for endowments and other country characteristics. Legal origin, however, explains cross-country differences in private property rights protection even after controlling for initial endowment indicators, religious composition, ethnic diversity, and the fraction of years the country has been independent since 1776. Furthermore, except when controlling for religious composition (there is a strong correlation between French legal heritage and the Catholic religion), there is a robust link between legal origin and stock market development--French Civil law countries have significantly lower levels of stock market development than British Common law countries after controlling for other country characteristics. The data provide strong support for the endowment view. Countries with poor geographical endowments, as measured by the log of settler mortality, tend to have less developed financial intermediaries, less developed stock markets, and weaker property rights protection. These results hold after controlling for legal origin, the percentage of years since 1776 the country has been independent, the religious composition of the country, and the degree of ethnic diversity. In terms of comparing the law and endowment theories, the empirical results indicate that both the legal systems brought by colonizers and the initial endowments in the colonies are 40 A Reader in International Corporate Finance 176 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 important determinants of stock market development and private property rights protection. However, initial endowments are more robustly associated with financial intermediary development than legal origin. Moreover, initial endowments explain more of the cross-country variation in financial intermediary and stock market development than legal origin. In sum, and consistent with AJR's (2001) endowment theory, we find a robust link between initial endowments and current levels of financial development. Appendix A Table 10 The financial development and institutions across countries are presented in Table 10. In Table 11 a description of the variables is presented. Financial development and institutions across countries Stock Country Private market Property Legal Settler Legislative Country name code credit development rights origin mortality competition Checks Algeria DZA 0.19 0.00 3 F 78.2 3.50 1.00 Angola AGO 0.03 0.00 2 F 280 4.83 2.00 Argentina ARG 0.15 0.10 4 F 68.9 7.00 4.00 Australia AUS 0.81 0.54 5 B 8.55 7.00 4.33 Bahamas BHS 0.55 0.00 5 B 85 7.00 4.00 Bangladesh BGD 0.21 0.02 2 B 71.41 6.67 3.17 Barbados BRB 0.39 0.21 3 B 85 6.67 3.67 Bolivia BOL 0.34 0.01 3 F 71 7.00 5.33 Brazil BRA 0.27 0.16 3 F 71 7.00 4.17 Burkina Faso BFA 0.12 0.00 3 F 280 4.00 1.00 Cameroon CMR 0.18 0.00 2 F 280 5.75 2.00 Canada CAN 0.80 0.51 5 B 16.1 7.00 4.00 Central African CAF 0.06 0.00 F 280 5.17 1.67 Republic Chad TCD 0.08 0.00 2 F 280 2.50 1.00 Chile CHL 0.54 0.79 5 F 68.9 7.00 4.00 Colombia COL 0.30 0.12 3 F 71 7.00 2.00 Congo COG 0.13 0.00 2 F 240 5.00 2.00 Costa Rica CRI 0.15 0.06 3 F 78.1 7.00 2.33 Cote d'Ivoire CIV 0.31 0.05 2 F 668 5.67 1.83 Dominican DOM 0.22 0.00 2 F 130 7.00 5.00 Republic Ecuador ECU 0.18 0.10 3 F 71 7.00 3.67 Egypt EGY 0.28 0.07 3 F 67.8 6.00 2.00 El Salvador SLV 0.23 0.06 3 F 78.1 7.00 3.33 Ethiopia ETH 0.19 0.00 2 F 26 2.67 1.00 Gabon GAB 0.11 0.00 3 F 280 6.50 1.67 Gambia GMB 0.11 0.00 4 B 1470 5.50 2.67 Ghana GHA 0.05 0.12 3 B 668 3.00 2.00 Guatemala GTM 0.13 0.01 3 F 71 7.00 3.17 Guinea GIN 0.09 0.00 2 F 483 1.00 1.00 Guyana GUY 0.20 0.00 3 B 32.18 6.50 1.50 Chapter One 41 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 177 Table 10 (continued) Stock Country Private market Property Legal Settler Legislative Country name code credit development rights origin mortality competition Checks Haiti HTI 0.12 0.00 1 F 130 6.00 1.83 Honduras HND 0.26 0.05 3 F 78.1 7.00 2.00 Hong Kong HKG 1.36 1.79 5 B 14.9 N/A N/A India IND 0.24 0.27 3 B 48.63 7.00 5.83 Indonesia IDN 0.44 0.14 3 F 170 6.00 1.00 Jamaica JAM 0.21 0.42 4 B 130 6.67 3.67 Kenya KEN 0.31 0.15 3 B 145 5.50 2.00 Madagascar MDG 0.14 0.00 3 F 536.04 6.33 2.83 Malaysia MYS 0.93 1.89 4 B 17.7 7.00 6.00 Mali MLI 0.12 0.00 3 F 2940 5.00 2.00 Malta MLT 0.84 0.12 3 F 16.3 7.00 3.00 Mauritania MRT 0.37 0.00 2 F 280 3.50 2.50 Mauritius MUS 0.37 0.22 2 F 30.5 7.00 5.00 Mexico MEX 0.27 0.32 3 F 71 6.83 2.00 Morocco MAR 0.34 0.08 4 F 78.2 7.00 1.00 New Zealand NZL 0.81 0.40 5 B 8.55 7.00 2.83 Nicaragua NIC 0.25 0.00 2 F 163.3 7.00 2.25 Niger NER 0.11 0.00 3 F 400 3.67 1.67 Nigeria NGA 0.22 0.05 3 B 2004 1.00 1.00 Pakistan PAK 0.23 0.16 4 B 36.99 7.00 5.50 Panama PAN 0.50 0.07 3 F 163.3 7.00 3.17 Paraguay PRY 0.20 0.01 3 F 78.1 7.00 3.00 Peru PER 0.08 0.08 3 F 71 7.00 3.67 Rwanda RWA 0.07 0.00 1 F 280 4.17 1.00 Senegal SEN 0.24 0.00 4 F 164.66 6.50 2.00 Sierra Leone SLE 0.03 0.00 2 B 483 2.67 1.00 Singapore SGP 0.96 1.33 5 B 17.7 6.00 2.00 South Africa ZAF 0.94 1.56 3 B 15.5 7.00 2.00 Sri Lanka LKA 0.20 0.17 3 B 69.8 7.00 3.17 Sudan SDN 0.05 0.00 2 B 88.2 N/A N/A Surinam SUR 0.41 0.00 3 F 32.18 7.00 4.33 Tanzania TZA 0.05 0.00 3 B 145 4.50 1.00 Togo TGO 0.24 0.00 3 F 668 4.33 1.50 Trinidad and TTO 0.48 0.12 5 B 85 6.67 3.67 Tobago Tunisia TUN 0.58 0.08 3 F 63 5.17 1.00 Uganda UGA 0.03 0.00 4 B 280 4.00 1.00 Uruguay URY 0.23 0.01 4 F 71 7.00 4.00 USA USA 1.48 0.69 5 B 15 7.00 4.67 Venezuela VEN 0.19 0.12 3 F 78.1 7.00 4.67 Zaire ZAR 0.00 0.00 2 F 240 2.83 1.00 42 A Reader in International Corporate Finance 178 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 Table 11 Variable descriptions and sources Variable Description Sources Private Credit {ð0:5Þn½FðtÞ=P eðtÞ þ Fðt 1Þ=P eðt 1Þ }/½GDPðtÞ=P aðtÞ ; Beck et al. (2001b), where F is credit by deposit money banks and other IFS, IFC, and own financial institutions to the private sector (lines 22d and 42d calculations in International Financial Statistics, IFS), GDP is line 99b, P e is end-of-period CPI (line 64), and P a is the average CPI for the year. Average for 1990­1995. Data for Angola, Guinea, and Tanzania are calculated using data from IFS and World Development Indicators (WDI); for Angola, IFS data for 1996­1998 are used and GDP data are from WDI; for Guinea, GDP data from WDI are used and given the lack of CPI indicators, the ratio of line 22d plus 42d divided by GDP is calculated. Stock Market {ð0:5Þn½FðtÞ=P eðtÞ þ Fðt 1Þ=P eðt 1Þ }/½GDPðtÞ=P aðtÞ ; Beck et al. (2001b), Development where F is the total value of outstanding shares, GDP is line IFC, IFS, WDI and 99b (IFS), P e is end-of, period CPI (line 64, IFS) and P a is own calculations the average CPI for the year. Average for 1990­1995. For Guatemala and El Salvador, IFC data from 1996 and 1997 are used to calculate the variables. For Malta, data for 1994 and 1995 are taken from the stock exchange's web-page. For all countries that do not have stock markets or that introduced stock markets after 1995, a zero was entered. Also, for Nicaragua, a zero was entered since no data is found, the exchange was founded in 1993, and it is reported to be very small. Property Rights An index of the degree to which government protects and La Porta et al. enforces laws that protect private property. Measured in (1999), Heritage 1997 and ranges from 1 to 5. Foundation Liquid Liabilities {ð0:5Þn½FðtÞ=P eðtÞ þ Fðt 1Þ=P eðt 1Þ }/½GDPðtÞ=P aðtÞ ; where F is currency plus demand and interest-bearing liabilities of banks and nonbank financial intermediaries (line 55l in IFS), GDP is line 99b, P e is end-of period CPI (line 64) and P a is the average CPI for the year. Average for 1990­1995. Data for Angola, Guinea, and Tanzania are calculated using data from IFS and World Development Indicators (WDI); for Angola, IFS data for 1996­1998 are used and GDP data are from WDI; for Guinea, GDP data from WDI are used and given the lack of CPI indicators, the ratio of line 551 divided by GDP is calculated Total Value The total value of shares traded as a ratio of GDP. Average Beck et al. (2001b), Traded for 1990­1995. For Guatemala and El Salvador IFC data IFC, IFS and own from 1996 and 1997 are used to calculate the variable. For calculations Malta, data for 1994 and 1995 are taken from the stock exchange's web-page. For all countries that do not have stock markets or that introduced stock markets after 1995, a Chapter One 43 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 179 Table 11 (continued) Variable Description Sources zero was entered. Also, for Nicaragua, a zero is entered, since no data is found, the exchange was founded in 1993, and it is reported to be very small. Rule of Law An indicator of the degree to which the country adheres to International the rule of law (ranging from 0 to 6). Average for 1990­ Country Risk Guide 1995. (ICRG) Aggregate Rule An indicator of the strength and impartiality of the legal Kaufmann et al. of Law system. An aggregate indicator that is estimated with an (1999) unobserved-component model from individual indicators of the efficiency of the legal system from 11 sources. Measured in 1998. French Legal Dummy variable that takes on value one if a country legal La Porta et al. (1999) Origin system is of French Civil law origin. Settler Mortality Log of the annualized deaths per thousand European Acemoglu et al. soldiers in European colonies in the early 19th century. (2001) Latitude Absolute value of the latitude of a country, scaled between La Porta et al. (1999) zero and one. Africa Dummy variable that takes on value one if country is in Easterly and Levine Sub-Saharan Africa. (1997) Latin America Dummy variable that takes on value one if country is in Easterly and Levine Latin America. (1997) Catholic Percentage of population that follows Catholic religion, in La Porta et al. (1999) 1980. Ranges from 0 to 100. Muslim Percentage of population that follows Muslim religion, in La Porta et al. (1999) 1980. Ranges from 0 to 100. Other Religion Percentage of population that follows religion other than La Porta et al. (1999) Catholic, Muslim, or Protestant, in 1980. Ranges from 0 to 100. Independence Percentage of years since 1776 that a country has been Easterly and Levine independent. (1997) Ethnic Probability that two randomly selected individuals in a Easterly and Levine Fractionalization country will not speak the same language. (1997) Legislative Index of the number of parties competing in the last Beck et al. (2001b) Competition legislative election, ranging from 1 (non-competitive) to 7 (competitive). Average for 1990­1995. Checks Measure of the number of veto-players in the political Beck et al. (2001b) decision-making process, both in the executive and the legislature. Average for 1990­1995. 44 A Reader in International Corporate Finance 180 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 References Acemoglu, D., Johnson, S., Robinson, J.A., 2001. The colonial origins of comparative development: an empirical investigation. American Economic Review 91, 1369­1401. Acemoglu, D., Johnson, S., Robinson, J.A., 2002. Reversal of fortunes: geography and institutions in the making of the modern world income distribution. Quarterly Journal of Economics 117, 1231­1294. Alesina, A., Baqir, R., Easterly, W., 1999. Public goods and ethic divisions. Quarterly Journal of Economics 114, 1243­1284. Beck, T., Levine, R., 2002. Industry growth and capital allocation: does having a market- or bank-based system matter? Journal of Financial Economics 64, 147­180. Beck, T., Levine, R., 2003. Stockmarkets, banks, and growth: panel evidence. Journal of Banking and Finance, forthcoming. Beck, T., Levine, R., Loayza, N., 2000. Finance and the sources of growth. Journal of Financial Economics 58, 261­300. Beck, T., Clarke, G., Groff, A., Keefer, P., Walsh, P., 2001a. New tools and new tests in comparative political economy: the database of political institutions. World Bank Economic Review 15, 165­176. Beck, T., Demirgu.@-Kunt, A., Levine, R., 2001b. The financial structure database. In: Demirguc-Kunt, A., Levine, R. (Eds.), Financial Structure and Economic Growth: A Cross-Country Comparison of Banks, Markets, and Development. MIT Press, Cambridge, MA, pp. 17­80. Beck, T., Demirgu.@-Kunt, A., Levine, R., 2003. Law and finance. Why does legal origin matter? Journal of Comparative Economics, forthcoming. Coke, E., 1628. [1979 version]. The First Part of the Institutes of the Laws of England. Garland Publishing, Inc., New York, NY. Coffee, J., 2000. Privatization and corporate governance: the lessons from the securities market failure. Unpublished working paper 158, Columbia Law School, New York. Crosby, A.W., 1989. Ecological Imperialism: The Biological Expansion of Europe, 900­1900. Cambridge University Press, Cambridge. Curtin, P.D., 1964. The Image of Africa. University of Wisconsin Press, Madison, WI. Curtin, P.D., 1989. Death by Migration: Europe's Encounter with the Tropical World in the Nineteenth Century. Cambridge University Press, New York, NY. Curtin, P.D., 1998. Disease and Empire: The Health of European Troops in the Conquest of Africa. Cambridge University Press, New York, NY. Damaska, M.R., 1986. The Faces of Justice and State Authority: A Comparative Approach to the Legal Process. Yale University Press, New Haven, CT. Dawson, J.P., 1968. The Oracles of the Law. University of Michigan Law School, Ann Arbor, MI (Reprinted in 1986 by William S. Hein & Co., Inc., Buffalo, New York). Demirgu.@-Kunt, A., Detragiache, E., 2002. Does deposit insurance increase banking system stability? An empirical investigation. Journal of Monetary Economics 49, 1373­1406. Demirgu.@-Kunt, A., Maksimovic, V., 1998. Law, finance, and firm growth. Journal of Finance 53, 2107­2137. Easterly, W., Levine, R., 1997. Africa's growth tragedy: policies and ethnic divisions. Quarterly Journal of Economics 112, 1203­1250. Easterly, W., Levine, R., 2003. Tropics, germs, and crops: how endowments influence economic development. Journal of Monetary Economics 50, 3­40. Easterly, W., Islam, R., Stiglitz, J., 2000. Shaken and stirred: explaining growth volatility. World Bank Conference on Development Economics. Engerman, S., Sokoloff, K., 1997. Factor endowments, institutions, and differential paths of growth among new world economies. In: Haber, S.H. (Ed.), How Latin America Fell Behind. Stanford University Press, Stanford, CA, pp. 260­304. Engerman, S., Mariscal, E., Sokoloff, K., 1998. Schooling, suffrage, and the persistence of inequality in the Americas, 1800­1945. Unpublished working paper, Department of Economics, UCLA. Finer, S., 1997. The History of Government, Vols. I­III. Cambridge University Press, Cambridge, UK. Chapter One 45 T. Beck et al. / Journal of Financial Economics 70 (2003) 137­181 181 Franks, J., Sussman, O., 1999. Financial innovations and corporate insolvency. Unpublished working paper, London Business School. Gallup, J.L., Sachs, J.D., Mellinger, A.D., 1998. Geography and economic development. National Bureau of Economic Research Working Paper No. 6849. Glaeser, E., Shleifer, A., 2002. Legal origins. Quarterly Journal of Economics 117, 1193­1230. Glendon, M.A., Gordon, M.W., Osakwe, C., 1982. Comparative Legal Tradition in a Nutshell. Wes Publishing Co., St. Paul, MN. Gutierrez, H., 1986. La mortalite des eveques latino-americains aux XVIIe et XVII siecles. Annales de Demographie Historique 29­39. Hayek, F., 1960. The Constitution of Liberty. The University of Chicago Press, Chicago, IL. Kamarck, A., 1976. The Tropics and Economic Development: A Provocative Inquiry into the Poverty of Nations. The Johns Hopkins University Press, Baltimore, MD. Kaufmann, D., Kraay, A., Zoido-Lobaton, P., 1999. Aggregating Governance Indicators. World Bank ! Research Working Paper 2195. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1997. Legal determinants of external finance. Journal of Finance 52, 1131­1150. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and finance. Journal of Political Economy 106, 1113­1155. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1999. The quality of government. Journal of Law, Economics, and Organization 15, 222­279. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 2000. Investor protection and corporate governance. Journal of Financial Economics 58, 3­27. Landes, D., 1998. The Wealth and Poverty of Nations. W.W. Norton, New York, NY. Levine, R., 1997. Financial development and economic growth: views and agenda. Journal of Economic Literature 35, 688­726. Levine, R., Zervos, S., 1998. Stock markets, banks, and economic growth. American Economic Review 88, 537­538. Levine, R., Loayza, N., Beck, T., 2000. Financial intermediation and growth: causality and causes. Journal of Monetary Economics 46, 31­77. Littleton, T., 1481. [1903 version] In: Wambaugh, E. (Ed.), Littleton's Tenures in English. John Bryne & Co., Washington, D.C. Mahoney, P., 2001. The common law and economic growth: Hayek might be right. Journal of Legal Studies 30, 503­525. Merryman, J.H., 1985. The Civil Law Tradition: An Introduction to the Legal Systems of Western Europe and Latin America. Stanford University Press, Stanford, CA. North, D., 1990. Institutions, Institutional Change, and Economic Performance. Cambridge University Press, Cambridge, UK. North, D., Weingast, B.R., 1989. Constitutions and commitment: the evolution of institutions governing public choice in seventeenth-century England. Journal of Economic History 49, 803­832. Pivovarsky, A., 2001. Essays on institutions and finance. Ph.D. Dissertation, John F. Kennedy School, Harvard University. Putnam, R., 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton University Press, Princeton, NJ. Rajan, R.G., Zingales, L., 1998. Financial dependence and growth. American Economic Review 88, 559­586. Rajan, R.G., Zingales, L., 2003. The great reversals: the politics of financial development in the 20th century. Journal of Financial Economics 69, 5­50. Stulz, R., Williamson, R., 2003. Culture, openness, and finance. Journal of Financial Economics, forthcoming. Wurgler, J., 2000. Financial markets and the allocation of capital. Journal of Financial Economics 58, 187­214. Young, C., 1994. The African Colonial State in Comparative Perspective. Yale University Press, New Haven, CT. Chapter Two 47 THE JOURNAL OF FINANCE VOL. LVIII, NO. 6 DECEMBER 2003 Financial Development, Property Rights, and Growth STIJN CLAESSENS and LUC LAEVENn ABSTRACT In countries with more secure property rights, ¢rms might allocate resources better and consequentially grow faster as the returns on di¡erent types of as- sets are more protected against competitors'actions. Using data on sectoral value added for a large number of countries, we ¢nd evidence consistent with better property rights leading to higher growth through improved asset allo- cation. Quantitatively, the growth e¡ect is as large as that of improved access to ¢nancing due to greater ¢nancial development. Our results are robust using various samples and speci¢cations, including controlling for growth opportunities. RECENTLY, NUMEROUS PAPERS HAVE ESTABLISHED that ¢nancial development fosters growth and that a country's ¢nancial development is related to its institutional characteristics, including its legal framework. The ¢nancial development and growth literature has established that ¢nance matters for growth both at the macroeconomic and microeconomic level (King and Levine (1993), Levine (1997)). The law and ¢nance literature has found that ¢nancial markets are better devel- oped incountrieswithstronglegal frameworks (La Portaet al. (1998), Beck, Demir- gc-Kunt, and Levine (2003)). These well-developed ¢nancial markets make it easier for ¢rms to attract ¢nancing for their investment needs (Demirgc-Kunt and Maksimovic (1998), Rajan and Zingales (1998)). Related work has established that debt structures of ¢rms di¡er across institutional frameworks (Rajan and Zingales (1995), Demirgc-Kunt and Maksimovic (1999), and Booth et al. (2000)).1 n Claessens is from the University of Amsterdam and CEPR and Laeven is from the World Bank.We are grateful to Richard Green (the editor) and an anonymous referee who helped us to substantially improve the paper. We thank Thorsten Beck, Sudipto Dasgupta, Charles Goodhart, Simon Johnson, Ross Levine, Inessa Love, Enrico Perotti, Sheridan Titman, and Chris Woodru¡ for helpful suggestions, and Ying Lin for excellent research assistance. We received helpful comments from seminar participants at Korea University, Universidad Ar- gentina de la Empresa, London School of Economics, Stockholm School of Economics, Univer- sity of Amsterdam, the 17th Annual Congress of the European Economic Association in Venice, and the Third Annual Conference on Financial Market Development in Emerging and Transition Economies at Hong Kong University of Science and Technology.We thank Ra- ghu Rajan and Luigi Zingales for the use of their data, Ray Fisman and Inessa Love for pro- viding their data on U.S. sectoral sales growth, and Walter Park for providing his index data on patent rights.The views expressed in this paper are those of the authors and do not neces- sarily represent those of the World Bank. 1In particular, it has been established that ¢rms in developing countries have a smaller fraction of their total debt in the form of long-term debt. 2401 48 A Reader in International Corporate Finance 2402 The Journal of Finance Thus far the literature has not paid much attention to di¡erences across coun- tries in terms of ¢rms' asset structure, that is, to di¡erences in the allocation of investable funds by ¢rms across various types of assets. However, these di¡er- ences are large as well. Demirgc-Kunt and Maksimovic (1999) ¢nd that ¢rms in developing countries have higher proportions of ¢xed assets to total assets and less intangible assets than ¢rms in developed countries. This is sur- prising since the literature on ¢rms' optimal capital structure (Harris and Raviv (1991)) suggests that a lackof long-term ¢nancingFtypical in a developing countryFwould make it more di/cult to ¢nance ¢xed assets.Why is it that ¢rms in developing countries have more ¢xed assets? Is it that they need more ¢xed collateral to attract external ¢nancing? Or does the preference for ¢xed assets and a corresponding lower share of intangible assets arise in countries with worse property rights because the returns on ¢xed assets are easier to se- cure from the ¢rm's point of view than the returns on intangible assets? More generally, what is the role of property rights in terms of a¡ecting investment pat- terns of ¢rms? In this paper, we empirically explore the role of property rights in in£uenc- ing the allocation of investable resources. We start from the well-established proposition that greater ¢nancial sector development increases the availability of external resources and thereby enhances ¢rm investment. We also acknowl- edge the literature demonstrating the importance of a good legal frame- work and well-established property rights for overall economic growth. In terms of channels through which property rights a¡ect ¢rm growth, we focus on the allocation of investable resources by a ¢rm. At the ¢rm level, our idea of property rights is the degree of protection of the return on assets against power- ful competitors.This notion of property rights is di¡erent from what is common in the literature where it is typically regarded as the protection of assets against actions by government. By focusing on the asset side of a ¢rm's balance sheet, we instead use the term property rights as referring to the protection of entre- preneurial and other investment in ¢rm assets against actions of other ¢rms. We argue that a ¢rm operating in a market with weaker property rights may be led to invest more in ¢xed assets relative to intangible assets because it ¢nds it relatively more di/cult to secure returns from intangible assets than from ¢xed assets. The argument goes as follows. A ¢rm is always at riskof not getting the returns from its assets (tangible or intangible) due to actions by the government, its own employees, or other ¢rms. Since our notion of property rights is protection against powerful competitors, rather than against the government, we assume no risk of expropriation by the government (or equivalently, we assume the risk to be identical for tangible and intangible assets). For the ¢rm's employees and other ¢rms, in particular powerful competitors, it is relatively easy to steal the intangible assets of a ¢rm if property rights are not secure. In a narrow sense, this is because the value of many intangible assetsFpatents (property rights to inventions and other technical improvements), copyrights (property rights to authors, artists, and composers), and trademarks (property rights for distinctive commercial marks or symbols)Fpurely derive from the existence of (intellec- Chapter Two 49 Financial Development, Property Rights, and Growth 2403 tual) property rights.Without property rights protection, employees can simply walk away with many of a ¢rm's intangible assets and competitors can easily copy them. As such, property rights in a narrow sense are very important for securing returns on intangible assets. In contrast, stealing physical property such as buildings and machinery is more di/cult, particularly for competing ¢rms, evenwhen general property rights are not secure. In abroader sense there- fore, property rights matter more for securing returns from intangible assets than from tangible assets. It follows that property rights matter more for intan- gible assets than for tangible assets. More generally, we argue that the degree to which ¢rms allocate resources in an optimal way will depend on the strength of a country's property rights, with the allocation e¡ect being important for conse- quent ¢rm growth. As noted, the literature has already shown that across countries, ¢rm growth is a¡ected by the development of ¢nancial markets. As such, there are two e¡ects to consider in a cross-country study, a ¢nance e¡ect and an asset allocation ef- fect. The ¢nance e¡ect determines the available resources for investment and thus a¡ects ¢rm growth.The asset allocation e¡ect determines the e/ciency of ¢rm investment and thus also a¡ects growth.We empirically investigate the im- portance of the ¢nance and asset allocation e¡ects for di¡erent industries in a large number of countries.We ¢nd less growth in countries with a lower level of ¢nancial development, consistent with the hypothesis that ¢rms lack access to ¢nance and thus underinvest. And in countries with less secure property rights, there is less growth, consistent with the hypothesis that the allocation of ¢rms' investment is ine/cient as ¢rms underinvest in intangible assets. Our results are robust to using di¡erent country samples and estimation techniques, including instrumental variables and variations in country controls. Empirically, the two e¡ects appear to be equally important drivers of growth in sectoral value added. Our estimates predict that the di¡erence in growth rates between the 75th and 25th percentile intangible-intensive industry will be 1.4% per year higher in a country with a property rights index of ¢ve, the 75th percentile country, com- pared to an index of three, the 25th percentile country. For comparison, the aver- age growth rate in our sample is 3.4% per year. Therefore, a di¡erential rate of 1.4% due to an improvement in the property rights index from three to ¢ve repre- sents a large increase. Althoughwe do an arrayof robustness tests, our results do come with provisos. Apart from the usual caveats related to possible weaknesses in the data and the choice of a particular time period and country sample, there are methodological issues. Most important may be the fact that to test fully for the role of the asset allocation mechanism, we need both an instrument for the mechanism and an instrument for property rights.While instruments for the property rights have been developed, instruments for the actual asset allocation do not (yet) exist. When and if appropriate instruments are found, the asset allocation mechanism needs to be tested further. The paper is structured as follows. Section I reviews the related literature, de- velops the ¢nance and asset allocation e¡ects, and presents our methodology to separate the two e¡ects empirically. Section II presents the data used in our 50 A Reader in International Corporate Finance 2404 The Journal of Finance empirical application. Section III presents the empirical results concerning the relationships between growth in value added and the ¢nance and asset allocation e¡ects. Section IV presents a number of robustness tests, and SectionVconcludes. I. Related Literature and Hypothesis Our work is related to several strands of literature. The starting point is the work by King and Levine (1993), Levine and Zervos (1998), Beck, Levine, and Loayza (2000), and others that has established an empirical link between ¢nan- cial development and economic growth. Also related is the law and ¢nancelitera- ture initiated by La Porta et al. (1997).This literature focuses on the relationship between the institutional framework of a country and its ¢nancial development (see also La Porta et al. (1998), Rajan and Zingales (1998), and Demirgc-Kunt and Maksimovic (1998)).The literature has established that ¢nancial sector develop- ment is higher in countries with betterlegal systems and stronger creditor rights since such environments increase the ability of lenders to collateralize their loans and ¢nance ¢rms. In an extension, Beck et al. (2003) show that both legal systems and a country's initial endowments are important determinants of ¢nan- cial development and private property rights protection, with initial endow- ments explaining relatively more of the cross-country variation in ¢nancial development than legal origin. The second strand we draw on is the capital structure literature (Myers (1977), Titman and Wessels (1988), and Harris and Raviv (1991)). This literature relates ¢rms' liability structure to ¢rm asset choices, among others. It has established that real, tangible assets, such as plant and equipment, can support more debt than intangible assets. In particular, ¢xed assets can support more long-term debt because they have greater liquidation and collateralizable value. Holding other factors constant, debt ratios will be lower the larger the proportion of ¢rm values represented by intangible assets (Myers (1977)). Bradley, Jarrell, and Kim (1984) provide empirical support for the argument that a larger amount of intan- gible assets reduces the borrowing capacity of a ¢rm.2 The third strand of literature relates to the role of property rights in a¡ecting overall investment and investment patterns. Besley (1995) shows the role of prop- erty rights for investment incentives and provides evidence for the importance of property rights in the context of land ownership by farmers in Ghana. Johnson, McMillan, and Woodru¡ (2002) show for a sample of ¢rms in post-communist countries that weaker property rights discourage the reinvestment of ¢rm earn- ings, even when bank loans are available, suggesting that secure property rights are both a necessaryand su/cient condition for entrepreneurial investment.The role of property rights in a¡ecting investment patterns has also been acknowl- edged, although less explicitly studied. Mans¢eld (1995) hints that there may be a relationship between the protection of property rights and the allocation of in- vestable resources between ¢xed and intangible assets. Using a survey of ¢rm 2Work by Rajan and Zingales (1995) and Demirgc-Kunt and Maksimovic (1999) con¢rms that debt maturity and asset structures for cross sections of countries are related in this way, with ¢rms with more ¢xed assets being able to support a greater amount of long-term debt. Chapter Two 51 Financial Development, Property Rights, and Growth 2405 managers, he states that ``most of the ¢rms we contacted seemed to regard intel- lectual property rights protection to be an important factor . . . [in£uencing] investment decisions'' (p. 24). Stern, Porter, and Furman (2000) show that the strength of acountry's intellectual property rights a¡ects its innovative capacity, as measured by the degree of international patenting. In developing countries, the lower degree of investment in intangible assets may relate to the weaker pro- tection of property rights. More generally, the institutional economics literature (North (1990)) suggests that investment in particular types of assets will be high- er the more protected the property rights of the assets are. These three strands have not yet merged in investigating empirically the ef- fects of institutions on both ¢rm ¢nancing and asset allocation, and conse- quently on growth. Here we want to test two hypotheses: whether ¢rms in countries with better developed ¢nancial systems have more access to ¢nance and are therefore able to invest more overall, and whether ¢rms in countries with better property rights invest more e/ciently across types of assets. In turn, both aspects will be re£ected in higher growth rates. The law and ¢nance literature has already established that ¢rms in a country with a better legal framework and more developed ¢nancial markets ¢nd it easier to attract external ¢nancing. Empirical investigation of how a country's property rights protection a¡ects ¢rms'asset allocation has not yet occurred. For our empirical tests, we use the setup of Rajan and Zingales (1998, RZ here- after) to assess the relationship between ¢nancial development, property rights, and growth.3 The RZ model relates the growth in realvalue added in a sector in a particular country to a number of country and industry-speci¢c variables. In the case of RZ, the speci¢c test focuses on ¢nancial development and the argument of RZ is that ¢nancially dependent ¢rms can be expected to grow more in countries with a higher level of ¢nancial development. In addition to including country in- dicators and industry indicators, they overcome some of the identi¢cation pro- blems encountered in standard cross-country growth regressions by interacting a country characteristic (¢nancial development of a particular country) with an industry characteristic (external ¢nancial dependence of a particular industry). This approach is less subject to criticism regarding an omitted variable bias or model speci¢cation than traditional approaches and allows them to isolate the impact of ¢nancial development on growth. In the regression results explaining sectoral growth, RZ ¢nd a positive sign for the interaction between the external ¢nancial dependence ratio and thelevel of ¢nancial development.Theyalso ¢nd a similar e¡ect when including an interaction term between the typical external dependence variable for the particular sector and the quality of a country's legal framework. Their results provide support for the ¢nance e¡ect.We expand the RZ model to test for the asset allocation e¡ect.We add to the basic model in RZ avariable that is the interaction of the typical ratio for each industrial sector of intangible-to- 3Other papers that use this approach include Cetorelli and Gambera (2001), which investi- gates the e¡ects of bank concentration on sectoral growth, and Fisman and Love (2003), which investigates the e¡ects of trade credit usage on sectoral growth. 52 A Reader in International Corporate Finance 2406 The Journal of Finance ¢xed assets and an index of the strengthofcountries'property rights.We thentest whether industrial sectors that typically use many intangible assets grow faster (slower) in countries with more (less) secure property rights. If intangible-inten- sive sectors grow faster in countries with better property rights, then we have indirect evidence that property rights a¡ect ¢rms' asset choices and conse- quently (through that channel) growth.We also perform a number of robustness tests on the importance of controlling for country-speci¢c factors and using in- strumental variables to control for the possible (residual) endogeneity of some variables. In line with RZ, we use U.S. ¢rm data to construct proxies at the industry level for the typical external ¢nancial dependence for a particular industrial sector and the typical ratio of intangible to ¢xed assets for a particular industry. The presumption here is that the well-developed ¢nancial markets and the well-pro- tected property rights in the United States should allow U.S. ¢rms to achieve the desired ¢nancial and asset structures for their respective industrial sector.This approach o¡ers a way to identify the desired extent of external ¢nancial depen- dence and the optimal asset mix of an industry anywhere in the world.4 It as- sumes that there are technological and economic reasons why some industries depend more on external ¢nance and intangible assets than others do, and that these di¡erences, to a large degree, prevail across countries.This does not mean that we assume a sector in two countries with the same degree of property pro- tection to have exactly the same optimal mix of intangibles and tangible assets. Local conditions such as growth opportunities are allowed to di¡er between countries.We onlyassume the rankorder ofoptimal asset mixes across industries to be similar across countries. Furthermore, we explicitly conduct tests for the importance of this assumption. Following RZ, the regressions include the industry's market sharein total man- ufacturing in the speci¢c country to control for di¡erences in growth potential across industries. Industries with large market shares may have less growth potential than industries with small initial market shares when there is an industry-speci¢c convergence. The initial share may also help to control for other variations between countries, such as in their initial comparative advan- tage among certain industries based on factors other than ¢nancial development and property rights protection. Finally, in line with RZ, we use country and industry dummies to control for country-speci¢c and industry-speci¢c factors. II. Data We use industry-speci¢c and country-speci¢c data from a variety of sources. Table I presents an overview of the variables used in our empirical analysis and their sources. Most of the variables are self-explanatory and have been used in other cross-country studies of ¢rm ¢nancing structures and ¢rm growth. 4The advantage of this approach is that we do not need information on the actual asset mix for industries in di¡erent countries.The comparability of such data would be limited because accounting practices, particularly with respect to intangible assets, di¡er greatly around the world. Chapter Two 53 Financial Development, Property Rights, and Growth 2407 In line with RZ, we use the ratio of private credit to GDP as a proxy for ¢nan- cial development. As proxies for the level of protection of property rights, we use three broad indexes of property rights and two indexes of intellectual property rights, as well as a speci¢c index of patent rights.These indexes of property rights come from di¡erent sources, each having some advantages and disadvantages. Our main property rights index is the rating of protection of property rights from the Index of Economic Freedom constructed by the Heritage Foundation. This relatively broad index of property rights is available for a large set of countries and has been used by other researchers (e.g., Johnson, Kaufmann, and Zoido-Lo- baton (1998) and La Porta et al. (1999, 2002)). A second index of property rights rates the protection of intellectual property rights in particular by using data from the ``Special 301''placements of the O/ce of the U.S.Trade Representative (USTR).``Special 301'' requires the USTR to identify those countries that deny adequate and e¡ective protection for intellectual property rights or deny fair and equitable market access for persons that rely on intellectual property protec- tion. Countries can be placedon di¡erentlists depending on their relative protec- tion of intellectual property. For example, countries which have the most onerous or egregious acts, policies, or practices and which have the greatest adverse im- pact on relevant U.S. products are designated ``priority foreign countries.'' As such, the index weights the degree of property rights protectionwiththe econom- ic impact that protection de¢ciencies have on U.S. trade.We use these quali¢ca- tions to construct an index of intellectual property rights protection.The third index is the patent rights index constructed by Ginarte and Park (1997).This in- dex focuses more speci¢cally on the protection of patents. A fourth index is the property rights index of the World Economic Forum (2002), which measures the general legal protection of private property in a country. The ¢fth index is the intellectual property rights index of theWorld Economic Forum, which measures the protection of intellectual property in a country.The twoWorld Economic For- um indexes are available only for the year 2001. The sixth index is the property rights index constructed by Knack and Keefer (1995) using data from the Inter- national Country Risk Guide (ICRG). This index measures property rights in a broad sense and includes ¢ve measures: quality of the bureaucracy, corruption in government, rule of law, expropriation risk, and repudiation of contracts by the government. Table I presents more details on these six indexes of property protection. Our main index of protection of property rights covers the period 1995 to 1999; the Special 301 index of protection of intellectual property rights covers the per- iod 1990 to 1999; theWorld Economic Forum indexes refer to 2001; and the Knack and Keefer index covers the period1982 to1995.The growth regressions, however, include data for the period 1980 to 1989, as in RZ. Ideally, one would want to use property rights indexes for the period 1980 to 1989 as well; however, this is not possible for the property rights indexes available to us due to data limitations. The one exception is the Ginarte and Park patent rights index, for which we do have data for the period 1980 to 1989. Therefore, this index does not su¡er from the nonoverlapping time period problem and we can use the patent rights index for the year 1980Fthe beginning of the period 1980 to 1989Fin the regressions. 54 A Reader in International Corporate Finance 2408 The Journal of Finance TableI De¢nition and Source of theVariables This table describes the variables collected for our study. The ¢rst column gives the names of the variable as we use it.The second column describes the variable and provides the source from which it was collected. Variable Description Property A rating of property rights in eachcountry (on a scale from1to 5).The more (Freedom) protection private property receives, the higher the score. The score is based, broadly, on the degree of legal protection of private property, the probability that the government will expropriate private property, and the country's legal protection of private property. The index equals the median rating for the period 1995 to 1999. Source: The Index of Economic Freedom from the Heritage Foundation. We reversed the original order of the index. Intellectual An index of intellectual property rights (on a scale from 1 to 5).The more Property (301) protection private property receives, the higher the score. The index is calculated using the ``Special 301'' placements of the O/ce of the U.S. Trade Representative (USTR). Special 301 requires the USTR to identify those countries that deny adequate and e¡ective protection for intellectual property rights or deny fair and equitable market access for persons that rely on intellectual property protection. Countries that have the most onerous or egregious acts, policies, or practices and that have the greatest adverse impact on relevant U.S. products are designated ``Priority foreign countries.'' Countries can also be placed on other lists.We assign the following ratings:1¼Priority foreign countries; 2¼306 Monitoring; 3¼Priority watch list; 4¼Watch list; 5¼Not listed. The index equals the median rating for the period 1990 to 1999. Source: International Intellectual PropertyAlliance. Original source: USTR. Patent rights An index of patent rights (on a scale from 0 to 5) in 1980. The more (GP) protection patents receive, the higher the score.The index criteria are: coverage, membership, duration, enforcement, and loss of rights. Source: Ginarte and Park (1997). Property An index of property rights (on a scale from 1 to 7) in 2001. The more (WEF) protection private property receives, the higher the score. A 1 indicates that assets are poorly delineated and not protected by law, while 7 indicates that assets are clearly delineated and protected by law. Source: Global Competitiveness Report,World Economic Forum (2002). Intellectual An index of intellectual property rights (on a scale from 1 to 7) in 2001. property (WEF) The more protection intellectual property receives, the higher the score. A 1 indicates that intellectual property protection is weak or nonexistent, while 7 indicates that intellectual property protection is equal to the world's most stringent. Source: Global Competitiveness Report,World Economic Forum (2002). Property A measure of property rights in each country (on a scale from 0 to 10).The (ICRG) index equals the average rating between 1982 and 1995. The more protection private property receives, the higher the score. The score is based on the average of ¢ve measures: quality of the bureaucracy, corruption in government, rule of law, expropriation risk, and repudiation of contracts by the government. Source: International Country Risk Guide and Knack and Keefer (1995). Private credit Private credit dividedby GDP in1980. Source: Rajan and Zingales (1998) and the International FinancialStatisticsof the InternationalMonetary Fund. Chapter Two 55 Financial Development, Property Rights, and Growth 2409 Variable Description Market cap Stock market capitalization divided by GDP in 1980. Source: Rajan and Zingales (1998). Accounting Accounting standards in 1983 (on a scale from 0 to 90). Higher scores indicate more disclosure. Source: Center for International Financial Analysis and Research and Rajan and Zingales (1998). Human capital Human capital is the average for 1980 of the years of schooling attained by the population over 25 years of age. Source: Barro and Lee (1993). Rule of Law Assessment of the law and order tradition in the country (on a scale from 0 to 10). Average of the months of April and October of the monthly index between 1982 and 1995. Lower scores indicate less tradition for law and order. Source: International Country Risk Guide and La Porta et al. (1997). Legal origin Identi¢es the legal origin of the Company Law or Commercial Code of each country. There are four possible origins: (1) English Common law, (2) French Commercial Code, (3) German Commercial Code, and (4) Scandinavian Commercial Code. Source: La Porta et al. (1999). European settler European settler mortality rate, measured in terms of deaths per annum mortality per 1000 ``mean strength.'' Source: Acemoglu et al. (2001). GDP per capita The logarithm of GDP per capita in 1980. Source:World Development Indicators of theWorld Bank. Growth in Average annual real growth rate of value added in a particular sector in a value added particular countryover the period1980 to1989.The sectors are classi¢ed on the basis of ISIC. Source: United Nations Database on Industrial Statistics and Rajan and Zingales (1998). Growth in Average growth in average size by ISIC sector over the period 1980 to 1989. average size Source: United Nations Database on Industrial Statistics and Rajan and Zingales (1998). Growth in Average growth in numberofestablishments by ISIC sector over the period number 1980 to 1989. Source: United Nations Database on Industrial Statistics and Rajan and Zingales (1998). Fraction of sector Fraction of ISIC sector in value added of total manufacturing sector in in value added 1980. Source: Rajan and Zingales (1998). Financial External ¢nancial dependence of U.S. ¢rms by ISIC sector averaged over dependence the period 1980 to 1989. Source: Rajan and Zingales (1998). Sales growth Real annual growth in sales of U.S. ¢rms by ISIC sector averaged over the period 1980 to 1989. Source: Fisman and Love (2002). Tobin's Q Tobin's Q of U.S. ¢rms by ISIC sector averaged over the period 1980 to 1989. Tobin's Q is de¢ned as the sum of the market value of equity plus the book value ofliabilitiesover thebook value of totalassets.Source:COMPUSTAT. Intangible Ratio of intangible assets-to-net ¢xed assets of U.S. ¢rms by ISIC sector intensity over the period 1980 to 1989. Source: COMPUSTAT. Intangibles is COMPUSTAT item 33 and represents the net value of intangible assets. Intangibles are assets that have no physical existence in themselves, but represent rights to enjoy some privilege. In COMPUSTAT, this item includes blueprints or building designs, patents, copyrights, trademarks, franchises, organizational costs, client lists, computer software patent costs, licenses, and goodwill (except on unconsolidated subsidiaries). Intangibles excludes goodwill on unconsolidated subsidiaries, which are included in Investments and Advances under the Equity Method (COMPUSTAT item 31). Net ¢xed assets is COMPUSTAT item 8 and represents net property, plant and equipment, which equals gross property, plant and equipment (COMPUSTAT item 7) less accumulated depreciation, depletion and amortization (COMPUSTAT item 196). 56 A Reader in International Corporate Finance 2410 The Journal of Finance For the other property rights indexes, we use index values as of their ¢rst avail- able date. Although the indexes of property protection are from di¡erent sources and for di¡erent time periods, they appear quite related and are highly positively corre- lated.The correlation between our main property rights index and the other ¢ve indexes of protection of (intellectual) property rights ranges, for example, from 0.49 to 0.78.The fact that the property rights indexes relate to di¡erent time peri- ods could nevertheless raise concerns in our speci¢cation, in part because prop- erty rights may have evolved in response to economic performance. We believe these concerns to be small, mostly because measures of institutional frameworks have been found to be stable over long periods of time (Acemoglu, Johnson, and Robinson (2001, 2002)). Also, RZ show that the sample means of the accounting standards variable they use do not di¡er signi¢cantly between 1983 and 1990. This stability also applies to our property rights indexes, which do not change much over the time for which they are available. Table II shows that the mean property rights index for countries sampled in the ¢rst and last available year is not statistically signi¢cantly di¡erent for any of the three indexes. Note that the sample mean of the Ginarte and Park patents rights indexFthe only index for which we have data for the period 1980 to 1989Ffor countries sampled in 1980 does not signi¢cantly di¡er statistically from the sample mean in 1990 for the same set of countries. In addition, we ¢nd that the relative ordering of the di¡er- ent property rights indexes does not change much over time, as the Spearman rank order correlations of the respective indexes are high. A t-test of di¡erences further con¢rms that the property rights indexes in the ¢rst and last available year are not statistically di¡erent. As a further robustness check, we also per- form our regressions instrumenting the property rights indexes with variables that predate the period 1980 to 1989, using the methodology used by Beck et al. (2000) and byAcemoglu et al. (2001). Table III presents the summary statistics of the country-speci¢c variables grouped by developing and developed countries (Table AI in the Appendix pre- sents the same summary statistics, but by individual country). We only use the classi¢cation developing versus developed countries to illustrate the di¡erences in the various variables by institutional settings.The country summary statistics show that, as a group, developingcountries haveless developed ¢nancial systems, weaker law and order systems, worse protection of (intellectual) property rights, and fewer patents per capita. Allvariables except for the stock market capitaliza- tion-to-GDP ratio and the accounting standards show a statistically signi¢cant di¡erence between the two groups of countries. Other work has documented ex- tensively the di¡erences in the degree of law and order between developed and developing countries. This di¡erence in legal frameworks partly relates to the di¡erence in the private credit-to-GDP ratio between these two groups of coun- tries, where low contract enforcement environments have hindered the develop- ment of ¢nancial systems in developing countries. The degree of ¢nancial development and the protection of property rights tend to go together and are both related to the overall level of development of a coun- try. As such, it could be di/cult to analyze the di¡erential e¡ects of ¢nancial F TableII Stability of Property Rights Measures overTime inanc This table reports for each of the three property rights indexes the sample mean and standard deviation for the ¢rst year and the last year of the ia sample period across all sampled countries, the t-statistic for a test of di¡erence in the sample means assuming unequalvariances, the rank order lD correlation coe/cient, and a test of independence of the property rights indexes in the ¢rst year and the last year of the sample period.The null hypothesis of the test of independence is that the property rights indexes are independent.The sources and de¢nitions of the data are reported in Table I. Signi¢cance level corresponds to 1%. a evelopm Statistics across Countries Test of Di¡erence Rank Order Test of ent,P Chapter Property rights Number of in Means Correlation Independence rope index Year Mean Std. Dev. Observations t-statistic Spearman's r p-value rt Two y Property (Freedom) 1995 3.93 0.96 44 R Property (Freedom) 2000 3.89 0.97 44 0.22 0.90 0.000a Intellectual property (301) 1990 4.29 0.60 28 Intellectual property (301) 2000 4.03 0.81 28 1.36 0.76 0.000a ights,and G Patents (GP) 1980 2.69 0.91 44 Patents (GP) 1990 2.74 1.00 44 0.29 0.97 0.000a row th 24 11 57 58 A Reader in International Corporate Finance 2412 The Journal of Finance TableIII Descriptive Statistics of InstitutionalVariables This table reports summary statistics of the variables used in our study. For each variable, we report the mean across all sampled countries, across developing countries, and across devel- oped countries.To classify countries as developing or developed, we use theWorld Bank classi- ¢cationofcountries. Forcomparison purposes, we also present t-statistics of tests of di¡erences in the means of the variables across developingand across developed countries.The sources and de¢nitions of the data are reported inTable I. Signi¢cance level corresponds to 1%. a Means across Countries t-Tests of Di¡erence in Means Developed Developing All Developed vs. Developing Countries Countries Countries Countries Property (Freedom) 4.68 3.42 3.96 7.10a Intellectual property (301) 4.47 3.74 4.12 3.97a Patents (GP) 3.33 2.20 2.67 5.44a Property (WEF) 6.11 4.69 5.33 7.66a Intellectual property (WEF) 5.74 3.47 4.51 10.64a Property (ICRG) 9.14 5.42 7.03 11.82a Private credit to GDP 0.49 0.26 0.36 4.37a Market capitalization to GDP 0.24 0.17 0.20 0.64 Law and order 9.23 4.40 6.67 11.74a Accounting standards 0.65 0.66 0.65 0.12 Settler mortality rate 2.49 4.36 4.03 6.25a Human capital 7.92 4.07 5.84 5.72a GDP per capita 9.04 6.84 7.79 10.28a Number of countries 19 25 44 development and property rights on the level of external ¢nancing available and the allocation of investment across di¡erent assets. However, the correlation be- tween the two concepts is not perfect. That is, there exist countries with good property rights and underdeveloped ¢nancial systems. Chile, for example, scores high on the protection of property rights (with a property rights index of ¢ve) but its level of ¢nancial development is only average (re£ected by a level of private credit to GDP of 36%). France, on the other hand, has a relatively well-developed ¢nancial system (re£ected bya level of private credit to GDPof 54%) but the pro- tectionof its property rights is onlyaverage (with a property rights index of four). Calculating the simple correlation between the property rights index and the le- vel of ¢nancial development,0.59, con¢rms that the relationship between the two concepts is high but not perfect.The correlations of the interaction variables are even less perfect, less than 0.20. Our data set includes 45 countries.5 For the growth regressions, as in RZ, we need to drop the benchmark country, the United States, and we are therefore left 5 The countries include Australia, Austria, Bangladesh, Belgium, Brazil, Canada, Chile, Co- lombia, Costa Rica, Denmark, Egypt, Finland, France, Germany, Greece, India, Indonesia, Israel, Italy, Jamaica, Japan, Jordan, Kenya, Korea, Malaysia, Mexico, Morocco, the Nether- lands, New Zealand, Nigeria, Norway, Pakistan, Peru, the Philippines, Portugal, Singapore, South Africa, Spain, Sri Lanka, Sweden, Turkey, the United Kingdom, the United States, Venezuela, and Zimbabwe. Chapter Two 59 Financial Development, Property Rights, and Growth 2413 with 44 countries. As we collected additional data, the number of countries in- cluded in our data set somewhat exceeds that in RZ, who use data on 41countries. Like RZ, we construct benchmark data on an industry basis. We use the benchmark data from RZ for all of our industry variables, but construct our own intangible-to-¢xed assets variable. We assume that the intangible-to-¢xed assets ratio for each industry in the United States forms a good benchmark (like RZ, who use the U.S. external ¢nancial dependence ratio as a benchmark). We refer to the ratio of intangible to ¢xed assets as the intangible intensity. In the same way RZ calculate the external ¢nancial dependence ratios by industry, we calculate the benchmark of intangible intensity using COMPUSTATdata on U.S. ¢rms for the years 1980 to 1989.We measure intangibles by the net value of intan- gible assets, that is, using COMPUSTAT item 33. Generally, intangibles are as- sets that have no physical existence in themselves but represent rights to enjoy some privilege. In COMPUSTAT, this item includes blueprints or building de- signs, patents, copyrights, trademarks, franchises, organizational costs, client lists, computer software patent costs, licenses, and goodwill (except on unconso- lidated subsidiaries). Intangibles in the COMPUSTATdata excludes goodwill on unconsolidated subsidiaries, which are included in investments and advances under the equity method (COMPUSTAT item 31). We measure tangibles by net ¢xed assets, that is, using COMPUSTAT item 8. This represents net property, plant, and equipment, which equals gross property, plant, and equipment (COM- PUSTAT item 7) less accumulated depreciation, depletion, and amortization (COMPUSTAT item 196). Table IVreports the intangible-intensity benchmarks for U.S. ¢rms in di¡erent industrial sectors on a two-digit SIC level.The total number of ¢rms used to cal- culate these benchmarks is 5,241. The average intangible-intensity ratio during the 1980s for U.S. manufacturing ¢rms is 77%.The variation of intangible inten- sityacross industries is large: It ranges from as lowas 2.0% for the petroleum and coal products industry to as high as 454% for the printing and publishing indus- try. The variation concurs with notions of what constitute relatively capital-in- tensive versus more knowledge-intensive industries. The stone, clay, glass, and concrete products industry, for example, relies mainly on ¢xed assets for produc- tion, as would be expected since the technology used in this sector is well-estab- lished and embodied in the ¢xed assets. It has an intangible-intensity ratio of 5%.The chemical and allied products industry and the electrical and electronic industry, in contrast, rely heavily on intangible assets as inputs, such as patents and licenses. They have an intangible-intensity ratio of 96% and 77%, respec- tively.The data show that the various technical and economic reasons that make various types of products require di¡erent input mixes can be benchmarked well at the industry level. III. Empirical Results In this section, the regression results are presented. In the ¢rst set of regres- sions, the dependent variable is the average annual real growth rate of value added in a particular sector in a particular country over the period 1980 to 1989, 60 A Reader in International Corporate Finance 2414 The Journal of Finance TableIV Sectoral Measure of Intangible Intensity The table reports the measure of intangible intensity for each sector based on U.S. ¢rm-level data. Intangible intensity is measured by the ratio of intangible assets to net ¢xed assets.The data are averages for all U.S. ¢rms in the COMPUSTAT (U.S.) database for the period 1980 to 1989. For external ¢nancial dependency benchmarks across sectors, we refer to the original source:Table I in Rajan and Zingales (1998).The table also reports the number of U.S. ¢rms used to construct the benchmark for each industrial sector. As in Rajan and Zingales (1998) we focus on manufacturing ¢rms and use 1980 to 1989 data to construct the benchmarks.The total num- ber of ¢rms is 5,241. SIC Code Industrial Sectors Intangible Intensity Number of Firms 20 Food and kindred products 0.75 304 21 Tobacco manufactures 0.49 21 22 Textile mill products 0.21 131 23 Apparel and other textile products 0.53 139 24 Lumber and wood products 1.20 97 25 Furniture and ¢xtures 0.49 87 26 Paper and allied products 0.20 130 27 Printing and publishing 4.54 202 28 Chemicals and allied products 0.96 556 29 Petroleum and coal products 0.02 86 30 Rubber and miscellaneous plastics 0.46 191 31 Leather and leather products 0.33 41 32 Stone, clay, glass, and concrete products 0.05 96 33 Primary metal industries 0.11 191 34 Fabricated metal products 0.31 277 35 Industrial machinery and equipment 0.25 795 36 Electrical and electronic equipment 0.77 815 37 Transportation equipment 0.24 262 38 Instruments and related products 0.90 660 39 Miscellaneous manufacturing industries 2.29 160 Mean 0.76 Median 0.48 Standard deviation 1.03 with one observation per sector in each country.The speci¢cation for the ¢rst set of regressions is as follows: Growthj ¼ Constant þ C1 Industry dummiesj ;k þ C2 Country controlsk þ c3 Industry share of manufacturing value addedj ;k þ c4 External dependencej Financial developmentk þ c5 Intangible intensityj Property rightsk þ ej ; ;k ð1Þ where each industry is indicated by index j and each country by index k. Upper- case Greek letters indicate vectors of coe/cients, indexed by industry j or Chapter Two 61 Financial Development, Property Rights, and Growth 2415 countryk. Growth is the average annual real growth rate of value added in indus- try j in country k.The industry dummies correct for industry-speci¢c e¡ects.The vector of country control variables di¡ers per speci¢cation and can include the following variables: private credit to GDP, index of property rights, stock market capitalization to GDP, human capital, rule of law, accounting standards, and the logarithm of per capita GDP.The exact vector of country control variables is de- scribed in greater detail in the presentation of the speci¢c empirical results. As a measure of ¢nancial development, we use private credit to GDP. As a measure of external ¢nancial dependence at the sectoral level, we use the data from RZ. As a measure of intangible intensity, we use the ratio of intangible to ¢xed assets for U.S. ¢rms on the sectoral level. For the property rights index, we use the Econom- ic Freedom property rights index. The results are presented in Table V. We ¢rst discuss the basic regression speci¢cations, which are estimated using OLS and include country dummies (columns 1 to 3). Industry dummies (not reported) are used in all regressions. The industry's market share in total manufacturing in a speci¢c country has a negative sign in all regressions, in line with RZ, suggesting that there is some industry-speci¢c convergence. In terms of the main hypotheses, we ¢nd that in- dustrial sectors that rely relatively more on external ¢nance develop dispropor- tionately faster in countries with better-developed ¢nancial markets because the coe/cient for the interactive variable private credit to GDP times external ¢nan- cial dependence is positive and statistically signi¢cant (at the 1% level, column 1). Hence, consistent with the ¢ndings of RZ, we ¢nd that ¢nancial development facilitates economic growth through greater availability of external ¢nancing. As noted by Beck et al. (2000) and others, the quality of the legal system in£u- ences ¢nancial sector development and overall growth. Interacting the external ¢nancial dependencevariable withthe index of the qualityof thelegal framework used by La Porta et al. (1998), instead of the ¢nancial development variable, also leads to a positive coe/cient (not reported). The regression result con¢rms the law and ¢nance view that increased availability of external ¢nancing and better legal systems enhance ¢rm growth. In terms of the asset allocation e¡ect, we ¢nd that industrial sectors using re- latively more intangible assets develop faster in countries with better protection of property rights, because the coe/cient for the interactive variable property rights times intangible intensity is statistically signi¢cant and positive (column 2). Hence, better property rights facilitate economic growth as they favor growth through better asset allocation, that is, in ¢rms that would naturally choose a higher share of investment in intangible assets.6 The asset allocation e¡ect on growth appears to be in addition to the increase in ¢rm growth due to greater external ¢nancing, since in the regressions where both the external ¢nancial de- pendence and the intangible-intensity variables are included (column 3), both in- teractive variables are statistically signi¢cant. Additionally, the coe/cients in 6 Exclusion of sectors with a relatively high estimated usage of intangible assets, such as printing and publishing and/or miscellaneous manufacturing industries, does not qualita- tively alter the results (not reported). 62 24 TableV 16 TheAverage E¡ect of Financial Development and Property Rights on Industrial Growth The dependent variable is the average annual real growth rate of value added in a particular sector in a particular countryover the period1980 to 1989.Table I describes all variables in detail. As a measure for protection of property rights, we use the property rights index from the Index of Economic Freedom from the Heritage Foundation. All regressions include industry dummies and a constant but these are not reported. Regres- sions (1) to (3) and regressions (6) to (8) include country dummies but these are not reported. Regressions (4) and (5) include country-speci¢c variables rather than country dummies. Regression (6) uses legal origin as the instrumental variable (IV) for property rights. Regression (7) uses European settler mortalityas IV for property rights. Robust standarderrors are shown below the coe/cients.The United States is dropped as it is A the benchmark. Signi¢cance levels and correspond to 1% and 5%, respectively. a b Reader (6) (7) (1) (2) (3) (4) (5) IV legal origin IV mortality Th in Fraction of sector in value added 1.041a 0.9721a 1.076a 1.040a 0.4511a 0.9672a 1.463a International e of manufacturing in 1980 (0.2454) (0.2482) (0.2491) (0.2210) (0.1028) (0.2480) (0.3658) Sectoral measure of ¢nancial 0.1401a 0.1354a 0.1376a 0.0509a dependence n private credit to GDP (0.0383) (0.0376) (0.0380) (0.0204) Journ Sectoral measure of intangible 0.0103a 0.0092a 0.0091a 0.0067a 0.0090a 0.0259b intensity n property (freedom) (0.0029) (0.0028) (0.0033) (0.0024) (0.0033) (0.0107) Private credit to GDP 0.0213 0.0488a alofF Corporate (0.0163) (0.0151) Property (freedom) 0.0004 0.0030 inance (0.0050) (0.0058) Stock market capitalization to GDP 0.0253a (0.0068) Finance Human capital 0.0008 (0.0017) Rule of law 0.0019 (0.0022) Accounting standards 0.0428b (0.0180) Log of per capita GDP 0.0205a (0.0043) R2 0.2711 0.2548 0.2757 0.1028 0.2386 0.2547 0.2391 N 1242 1277 1242 1242 830 1277 635 Number of countries 44 44 44 44 33 44 23 Chapter Two 63 Financial Development, Property Rights, and Growth 2417 the regressions including both e¡ects are of similar magnitudes as in the two regressions where each of them was included separately (columns 1 and 2), sug- gesting that the two variables measure complementary e¡ects.7 The e¡ects of external ¢nancial development and property protection on ¢rm growth are not only both statistically signi¢cant but are also equally economic- ally important.We can use the regression coe/cient estimates of TableV to infer how much higher the growth rate of an industryat the 75th percentile of intangi- ble intensity would be compared to an industry at the 25th percentile level, when the industries are located in a country at the 75th percentile of property protec- tion, rather than in a country at the 25th percentile.The industry at the 75th per- centile, instruments and related products, has an intangible-intensity ratio of 0.90. The industry at the 25th percentile, textile mill products, has an intangi- ble-intensity ratio of 0.21.The country at the 75th percentile of property protec- tion has a value of ¢ve for the property rights index and the country at the 25th percentile has a value of three.The estimated coe/cient for the interaction term in regression 2 of TableVequals 0.010 andwe can set the industry's initial share of manufacturing at its overall mean.The regression coe/cient estimates therefore predict the di¡erence in growth rates betweenthe 75th and 25th percentile intan- gible-intensive industry to be 1.4% per year higher in a country with a property rights index of ¢ve compared to one with an index of three. For comparison, the average growth rate is 3.4% per year.Therefore, a di¡erential rate of 1.4% due to an improvement in the property rights index from three to ¢ve represents a large increase. The e¡ect of ¢nancial development on di¡erential real ¢rm growth can be cal- culated in a similar way using the estimated coe/cient for the interaction term of regression 1 in TableVof 0.140.The coe/cient estimate predicts the di¡erence between the growth rate of the 75th and 25th percentile external ¢nancial depen- dence industry to be 1.4% higher in a country at the 75th percentile of ¢nancial development compared to one at the 25th percentile.8 Thus, the e¡ects of prop- erty protection and ¢nancial development on di¡erential ¢rm growth are not only both statistically signi¢cant, but also of similar economic importance. In other words, the asset allocation e¡ect is economically as important as the ¢nance e¡ect. The relative importance of the two e¡ects can also be demonstrated by a com- parison of two countries, Egypt and Finland. Egypt is a country with a relatively low degree of property protection, having avalue of three for the property rights 7The two interacted variables, external ¢nancial dependence and intangible intensity inter- acted with ¢nancial development and property rights indexes, do appear to measure di¡erent concepts as the correlation between these variables is low. The correlation between the exter- nal ¢nancial dependence variable interacted with the ¢nancial development measure and the intangible intensity measure interacted with the property rights index is 0.149. Similar corre- lations are found when the other four property rights indexes are used (not reported). 8RZ used the same approach to compute the e¡ect of ¢nancial development on di¡erential real ¢rm growth. Our estimated e¡ect di¡ers somewhat from the di¡erential growth rate ef- fect estimated in RZ, 1.3%, because our sample is slightly larger and because we use private sector credit instead of total capitalization as our measure of ¢nancial development. 64 A Reader in International Corporate Finance 2418 The Journal of Finance index (at the 25th percentile of property protection), while Finland is a country with a relatively high degree of property protection, having avalue of ¢ve for the property rights index (at the 75th percentile of property protection).The regres- sion coe/cient estimates predict that if Egypt had had the same property rights as Finland, but its actual ¢nancial development, then the growth rate in value added of its industry at the median level of intangible intensity, 0.48, would have been 1.0% per year higher. Egypt is also a country with a relatively low level of ¢nancial development, with a level of private credit to GDP of 21% (at the 25th percentile of ¢nancial development), while Finland is a country with a relatively high level of ¢nancial development, with a level of private credit to GDP of 48% (at the 75th percentile of ¢nancial development). If Egypt had had the same ¢nan- cial development as Finland, but its actual degree of property protection, then the growth rate in value added of its industry at the median level of external ¢- nancial dependence, 0.23, would have been 0.9% per year higher. Again, the two e¡ects are quite large and of comparable magnitude. These numerical interpretations can be compared to the results found by Hall and Jones (1999) and Acemoglu et al. (2001) for the e¡ects of institutions on output and income level. Hall and Jones (1999) explore the e¡ects of di¡er- ences in institutions and government policies, which they call social infrastruc- ture, on output per worker in a cross section of countries. Their ¢ndings imply that the observed di¡erence in social infrastructure between Niger and the United States is more than enough to explain the 35-fold di¡erence in output per worker. Acemoglu et al.'s (2001) ¢ndings imply that improving Nigeria's insti- tutions to the level of Chile could, in the long run, lead to as much as a 7-fold increase in Nigeria's income (in practice Chile is over 11 times as rich as Nigeria). Although these papers study the e¡ects of institutions on the output or income level, rather than the rate of growth, it shows that our results are of comparable orders. Thus far, our speci¢cations have focused on the di¡erential e¡ect on growth of property rights across industries with di¡erent asset mixes (captured by the in- teraction term of property rights and the intangible-intensity measure).To avoid possible biases caused by any omitted country-speci¢c regressors, we have in- cluded country dummies to capture any institutional or other di¡erences a¡ect- ing growth, such as comparative advantage or general levelof development. Since we are less interested in the importance of general country di¡erences, we use this approach rather than a vector of speci¢c country control variables. Still, the use of country dummies could introduce a misspeci¢cation to the extent that any omitted institutional di¡erences important for growth are correlated with our two interaction variables. Examples of such country-speci¢c variables that have been used in the general growth literature, besides ¢nancial depth and property rights, include the level of per capita GDP, human capital, and other institutional variables (Romer (1990), Barro (1991), and Levine and Zervos (1998), among others). Furthermore, we want to analyze the ¢rst-order country e¡ects of property rights to investigate whether property rights a¡ect ¢rm growth mainly through the asset allocation channel or also in other ways. We therefore replace our country dummies with country-speci¢c institutional and Chapter Two 65 Financial Development, Property Rights, and Growth 2419 other variables and thus perform a robustness checkonwhetheranyofour earlier results are a¡ected if we control in other ways for country di¡erences. We start by documenting the fact that the e¡ects of better property rights on growth work mostly through improved asset allocation as opposed through, for example, an improvement in the overall business environment that increases growth opportunities.We show this by including in our basic regression speci¢- cation the property rights index (and private credit to GDP) directly in addition to the interacted variables.The results are reported in column 4 of TableV, where we exclude country dummies.We do not ¢nd a direct, statistically signi¢cant ef- fect of the qualityof acountry's property rights on industrial sector growth. Most important, including the property rights index directly does not change the mag- nitude or the signi¢cance of the coe/cients for the interaction variables in any meaningful way. Both the ¢nancial dependence and the asset mix interaction variables remain statistically signi¢cant and neither changes much in terms of magnitude. This suggests that the major e¡ect of improved property rights on sectoral growth operates through improvements in asset allocation and that the interaction variable does not capture any general e¡ects, for example, of improvements in the business environment leading to greater growth opportu- nities. Forothercountry-speci¢c variables, weuse the ratio of private credit to GDP in 1980, stock market capitalization over GDP in 1980, a measure of the level of hu- man capital in 1980, a measure of the quality of the legal system, an accounting standards indicator, and the logarithm of per capita income in 1980. RZ and Ce- torelli and Gambera (2001) have also used these variables in the same model.We expect a positive e¡ect on growth of private credit to GDP and stock market ca- pitalization to GDP as proxies for the development of the banking system and stock market respectively, and for ¢nancial development more generally.The le- vel of human capital is measured as the average of the number of years of school- ing attained by the population over 25 years of age in 1980 (as in Barro and Lee (1993)) and is expected to have a positive e¡ect on growth in value added. The quality of the legal system is measured by the law and order tradition variable of La Porta et al. (1998) and is also expected to have a positive e¡ect on growth.The accounting standards indicator is an index re£ecting the quality of accounting standards and is taken from RZ.This variable is also expected to have a positive e¡ect on growth since it proxies for the quality of information investors have re- garding ¢rms and that ¢rms have regarding investment prospects. Per capita GDP is included to capture the convergence e¡ects of the economy as a whole to a long-run steady state and is expected to have a negative coe/cient (see, among others, Barro (1991)).The modelcontinues to include industry dummies to control for any sector-speci¢c e¡ects and the property rights indexes. Since the country variables included in the two interaction termsFprivate credit to GDP and an index of property rightsFare now also part of the country controls, we can as- sess both the overall e¡ect of ¢nancial development and property protection on value added growth as well as the ¢nance and asset allocatione¡ects capturedby the two interaction terms. Note that data on accounting standards is missing for some countries, reducing the sample of countries to 33. 66 A Reader in International Corporate Finance 2420 The Journal of Finance The results of this speci¢cation are reported in column 5 of TableV. Except for the human capital variable, the country controls have the expected relationships with growth.The direct e¡ect of the qualityof property rights on growth remains insigni¢cant, however, which suggests that better property rights by themselves do not translate into higher growth rates of sectoralvalue added.The depthof the ¢nancial systemFmeasured by private credit to GDP and the size of the stock market as a ratio to GDPFhas a positive and statistically signi¢cant in£uence on growth in sectoral value added.The degree of human capital in the country, proxied by the average number of years of schooling attained by the population over 25 years of age and the degree to which the rule of law applies, do not have a statistically signi¢cant e¡ect on growth in sectoral value added.The accounting index, however, is statistically signi¢cantly positive.The general level of develop- ment, proxied by the log of income per capita, has a negative sign, con¢rming the convergence e¡ect. The focus of our attention, the interaction between property rights and the al- location of resources, is very robust to these changes in model speci¢cation.The coe/cient on the interaction term between the property rights indexes and the intangible-intensity measure remains positive and statistically signi¢cant in both speci¢cations.The size of the coe/cient is also only somewhat smaller than those in the regressions with country dummies, and the coe/cient remains sta- tistically signi¢cant at the 1% level.The general result about the importance of the asset allocation e¡ect is thus not altered. Also, the interaction term between ¢nancial development and external ¢nancial dependence remains statistically signi¢cant positive. The regression results in columns 4 and 5 thus show that the e¡ect of property rights on growth operates in an important way through asset allocation, and does not have a direct, ¢rst-order e¡ect on growth. Another concern is that the quality of property rights is a¡ected by the invest- ment behavior of ¢rms and the resulting growth patterns. At the macro level, countries that grow faster may demand greater property rights protection, since a larger share of economic output derives from more property-rights-intensive investments. At the more micro level, sectors that are more dependent on prop- erty rights may seek a higher degree of protection of property rights relevant to their industry. Due to these and other concerns about potential endogeneity, we instrument the property rights variable with a number of predetermined institu- tional variables. Following RZ, we use the colonial origin of a country's legal sys- tem (indicating whether the legal origin is English, French, German, or Scandinavian) as reported in La Porta et al. (1998) as one instrument. As also shown by La Porta et al. (1998), legal origin tends to have a long-lasting e¡ect on acountry's institutional structure, whereas thelegalorigin of acountry is largely determined by the country colonizing it. As such, legal origin is a good instru- mental variable and has been used in several other papers. Following Acemoglu et al. (2001), we also use the settler mortality rate of European bishops, soldiers, and sailors stationed in colonies in the 17th,18th, and 19th centuries as an instru- ment. As argued byAcemoglu et al. (2001), the willingness of colonizing powers to settle and develop long-lasting institutions depended greatly on the ability of co- lonizers to survive physically.They show that the settler mortality rate is a good Chapter Two 67 Financial Development, Property Rights, and Growth 2421 instrumental variable for past institutional characteristics that last into today (in their application, the particular institutional characteristic is the risk of ex- propriation of private property). The instrumental variables (IV) results based on the speci¢cation of column 2 are presented in columns 6 and 7, using respectively legal origin or mortality rates as instruments for property rights. Since the European countries had the institutions that they were exporting to their colonies, we can not apply settler mortality rates as an instrumental variable for the European countries, that is, the colonizing countries themselves. This reduces the sample to 23 countries whenusing mortality rates as an instrumentalvariable.The results are neverthe- less very robust to the use of instruments.9 We again ¢nd a statistically signi¢- cant e¡ect of property rights on growth in sectoralvalue added through the asset allocation of resources. Interestingly, the magnitude of the coe/cients for the interaction variable increases when using mortality rates as an instrumental variable (column 7). Because restricting the sample to former colonies results in a large reduction in the number of observations, we will only use legal origin as an instrument for property rights in what follows.10 As an additional investigation into the channels through which ¢nancial de- velopment and property rights a¡ect ¢rm growth, and following RZ, we analyze whether industries in countries with better ¢nancial development and property rights grow faster because new establishments are added to the industry or be- cause existing establishments grow faster.There are two reasons why it is inter- esting to decompose the e¡ects of access to ¢nancing and asset allocation in terms of number and average size of ¢rms. First, as highlighted by RZ, the crea- tion of new establishments is more likely to require external funds, while the ex- pansion of existing establishments may more easily rely on internal funds.Thus, the e¡ect of ¢nancial development could be more pronounced for new ¢rms than for the growth of existing ¢rms. Second, new ¢rms are often set up in reaction to and to take advantage of new technological developments, while established ¢rms tend to grow through expansion of scale, perhaps also because they are slower in reacting to new developments.11 Furthermore, existing ¢rms may be able to preserve the value of their assets inways other than by resorting to formal property rights (e.g., by using their name recognition, distribution or supply net- works, or general economic and political in£uence). Thus, the importance of property rights that protect the returns to (new) technology and help assure a good allocation of an economy's overall resources might be more pronounced for the emergence of new ¢rms than for the growth of existing ¢rms. 9The ¢rst-stage regressions show strong relationships between the instrumented variables and the potentially endogenous variables, that is, between settler mortality and legal origin and property rights and ¢nancial development (not reported). 10The results presented in Table Vare based on all available data (up to 44 countries). As a further robustness test, we also reestimated the regression models using the subset of 41 countries used in RZ, which implied excluding Indonesia, Jamaica, and Nigeria. The results are very similar to those in Table V (not reported). 11In fact, many new ¢rms that take advantage of new technological developments are spun o¡ from existing ¢rms that have developed some elements of these new technologies. 68 A Reader in International Corporate Finance 2422 The Journal of Finance As before, we follow RZ and use data derived from the UN Industrial Statistics Yearbook database for the growth in the number of establishments and the growth in the average size of existing establishments.The growth in the number of establishments is calculated by RZ as the logarithm of the number of end-of- period establishments less the logarithm of the number of beginning-of-period establishments.The average size of establishments in the industry is calculated bydividing the value added in the industryby the number ofestablishments, with the growth in average size again de¢ned as the di¡erence in logarithms. RZ re- port that in their sample of countries roughly two-thirds of the growth in value- added results from an increase in the average size of existing establishments, while the remaining one-third is accounted for by an increase in the number of establishments. Weuse the same speci¢cation as forourbasic regressionbut withthe growth in number ofestablishments or the growth in average size as the dependent variable instead of the growth in total value added by sector.We use again industry dum- mies and do not use country-speci¢c institutional variables, but country dum- mies.The time period studied remains 1980 to 1989.The exact speci¢cation is as follows: Growthj ¼ Constant þ F1 Industry dummiesj ;k þ F2 Country dummiesk þ f3 Industry share of manufacturing value addedj ;k þ f4 External dependencej Financial developmentk þ f5 Intangible intensityj Property rightsk þ ej ; ;k ð2Þ where the dependent variable is either the growth in the average size or the growth in the number of establishments in industry j in country k. Table VI reports the results, with columns 1 and 2 depicting the OLS results and columns 3 and 4 the instrumental variable results. As Table VI indicates, the external ¢nancial dependence interacted with the ¢nancial development variable is statistically signi¢cant in explaining both the growth in average ¢rm size (column 1) and the growth in the number of establishments (column 2).This contrasts with RZ, who do not ¢nd any statistical signi¢cance (see their Table VII), perhaps because they use accounting standards as a measure for ¢nancial development rather than private credit to GDP and do not include the asset allo- cation interaction variable. Interestingly, the asset allocation variable interacted with the property rights variable is not signi¢cant when explaining the growth in the average size of ¢rms but is signi¢cant when explaining the growth in the number of estab- lishments.This ¢nding is consistent across all of our measures of property rights (not reported). It is also not a¡ected by using legal origin as an instrumental variable for property rights (columns 3 and 4). It suggests, in terms of a¡ecting growth through asset allocation, that the protection of property rights is most important through stimulating the growthof new establishments.Well-protected Chapter Two 69 Financial Development, Property Rights, and Growth 2423 TableVI TheAverage E¡ect of Financial Development and Property Rights on Growth in Average Size and Growth in the Number of Establishments The dependent variable is either the average growth in average size or the average growth in the number of establishments of a particular sector in a particular country over the period 1980 to 1989.Table I describes allvariables in detail. All regressions include industry dummies, country dummies, and a constant but these are not reported. Regressions (3) and (4) use legal origin as the instrumental variable (IV) for property rights. Robust standard errors are shown below the coe/cients.The United States is dropped as it is the benchmark. For Costa Rica, France, Indo- nesia, Italy, Jamaica, the Netherlands, South Africa, and Zimbabwe, we do not have data on the growth of the average size and the number of establishments. Signi¢cance levels , , and cor- a b c respond to 1%, 5%, and 10%, respectively. (3) (1) (2) Growth Average (4) Growth Average Growth Size IV Growth Number Size Number Legal Origin IV Legal Origin Fraction of sector in 0.8687a 0.3399b 0.8396a 0.3038c value added of (0.3131) (0.1702) (0.3143) (0.1624) manufacturing in 1980 Sectoral measure of 0.0856a 0.0480b ¢nancial dependence (0.0289) (0.0220) n private credit to GDP Sectoral measure of intangible 0.0001 0.0069a 0.0007 0.0082b intensity n property (freedom) (0.0021) (0.0022) (0.0036) (0.0034) R2 0.4329 0.3656 0.4164 0.3619 N 1071 1104 1100 1133 Number of countries 36 36 36 36 property rights can thus in£uence growth by allowing new ¢rms to come to market in those industries that typically rely less on tangibles in their optimal production mix. For established ¢rms relying more on intangible inputs, growth seems less a¡ected by the strength of property rights in the country. This may be because such ¢rms have other means of protecting their returns from investments. IV. Further RobustnessTests We have already shown that the results are robust to di¡erent control variables, toalternative means ofcontrolling forcountrydi¡erences, totheuse ofinstrumen- tal variables, and to changes in the sample of countries.We next present evidence that the results are also robust to the particular measure of protection of property rights chosen, to di¡erences in growth opportunities related to the level of general development, and to inclusion of data from alternative time periods. First, we use the ¢ve alternative measures of the degree to which countries protect property rights: Special 301, the patent rights index of Ginarte and Park (1997), the property rights index and the intellectual property rights index of the 70 24 Table VII 24 TheAverage E¡ect of Financial Development and Property Rights on Industrial Growth: Alternative Measures of Property Rights The dependent variable in all regressions is the average annual real growth rate of value added in a particular sector in a particular countryover the period1980 to1989.Table I describes allvariables in detail.We use ¢ve alternative measures for protectionof property rights. In regressions (1) and (6), weuse a measure for protectionof intellectual property rights which is calculated using the Special 301placements of the O/ce of the U.S. Trade Representative.We use the median rating during 1990 to 1999. In regressions (2) and (7), we use the patent rights index by Ginarte and Park (1997).We use the rating for the year 1980. A higher rating of the patent rights index indicates more protection of patent rights. In regressions (3) A and (8), we use the property rights index of theWorld Economic Forum.We use the rating for the year 2001. In regressions (4) and (9), we use the Reader intellectual property rights index of theWorld Economic Forum.We use the rating for the year 2001. In regressions (5) and (10), we use the property rights index of Knack and Keefer (1995). Average over 1982 to 1995. All regressions include industry dummies, country dummies, and a constant, but these are not reported. Robust standard errors are shown below the coe/cients.The United States is dropped as it is the benchmark. Sig- Th in ni¢cance levels and correspond to 1% and 5%, respectively. a b International e (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Journ Fraction of sector in value added 0.5225a 0.9592a 1.053a 1.055a 0.9802a 0.5708a 1.064a 1.139a 1.141a 1.082a of manufacturing in 1980 (0.1561) (0.2449) (0.2655) (0.2659) (0.2493) (0.1625) (0.2458) (0.2652) (0.2656) (0.2503) Corporate Sectoral measure of ¢nancial 0.0740a 0.1357a 0.1355a 0.1360a 0.1353a alofF dependence n private credit to GDP (0.0252) (0.0382) (0.0389) (0.0390) (0.0376) Sectoral measure of intangible 0.0062a 0.0052b inance intensity n intellectual property (301) (0.0023) (0.0021) Sectoral measure of intangible 0.0074a 0.0066a Finance intensity n patents (GP) (0.0026) (0.0026) Sectoral measure of intangible 0.0109a 0.0093a intensity n property (WEF) (0.0029) (0.0027) Sectoral measure of intangible 0.0072a 0.0062a intensity n intellectual property (WEF) (0.0019) (0.0018) Sectoral measure of intangible 0.0043a 0.0037a intensity n property (ICRG) (0.0012) (0.0012) R2 0.3269 0.2521 0.2581 0.2575 0.2548 0.3592 0.2734 0.2789 0.2786 0.2755 N 1119 1277 1211 1211 1277 1090 1242 1179 1179 1242 Number of countries 36 44 42 42 44 36 44 41 41 44 Chapter Two 71 Financial Development, Property Rights, and Growth 2425 World Economic Forum, and the property rights index of Knack and Keefer (1995).The regression speci¢cation we use is identical to model (1) in Section III, where we include industry and country dummies and the fraction of sector in va- lue added in manufacturing in 1980.We include the interaction term between in- tangible intensity and the property rights index, varying between the ¢ve property rights indexes. We also estimate speci¢cations that include, besides the interaction term between the property rights index and the intangible-to- ¢xed assets measure, also the interaction term between external ¢nancial depen- dence and private credit to GDP.The estimation technique remains OLS.The de- pendent variable is the same as in TableV, the real growth rate in sectoral value added of a particular country over the period 1980 to 1989. The results are presented inTableVII and are very similar to those of column 2 and 3 of Table V. Both without including the interaction term between external dependence and ¢nancial development (columns 1 to 5) and with including this interaction term (columns 6 to 10), we ¢nd statistically signi¢cant coe/cients on the interaction term between the intangible-intensity measure and all of the ¢ve alternative property rights measures.The results with the alternative measures of the degree of property rights protection are also robust to the use of legal ori- gin and European settler mortality as instruments (not reported).This suggests that the results are not due to the particular property rights index chosen. Second, we want to investigate whether growth opportunities di¡er across in- dustries and countries in such a way that they confound the relationships be- tween our interaction variables and growth in sectoral value added. In particular, it is possible that the external ¢nancial dependence and asset mix variables are proxies for growth opportunities at the sectoral level. Providedthat ¢nancial development is high and property rights are protected, it may not be those industries with a particular external ¢nancial dependence or intangible intensity that grow fast, but rather those with better growth opportunities. If these growth opportunities happen to be correlated with our ¢nancial develop- ment and property rights variables, then a bias in the estimations can arise. In particular, countries with similar levels of ¢nancial development or property rights may experience the same growth patterns across industries because their ¢rms face similar patterns of growth prospects, not because their levels of ¢nan- cial sector development or quality of property rights protection imply a greater supply of resources for ¢rms or a better allocation of resources by ¢rms. Correspondingly, countries with di¡erentlevels of ¢nancial development orprop- erty rights may have di¡erent growth opportunities and consequently grow in di¡erent ways, not because of di¡erences in the supply of external ¢nancing or the protection of property rights. In a recent paper, Fisman and Love (2002) explore this hypothesis using the RZ model, focusing on ¢nancial development.They use the actual U.S. sales growth at the sectoral level as a measure for sectoral growth opportunities at a global level.When they substitute the industry's actual sales growth for the industry's external ¢nancial dependence ratio in the interaction term with ¢nancial devel- opment, they ¢nd a positive coe/cient for this new interactionvariable. Further- more, when including both the old and new interaction variables, that is, the 72 A Reader in International Corporate Finance 2426 The Journal of Finance industries'external ¢nancial dependence times countries' ¢nancial development as well as the industries'actual sales growth times countries' ¢nancial develop- ment, they ¢nd that the interaction variable with external ¢nancial dependence is no longer statistically signi¢cant. This suggests, if indeed actual U.S. sales growth rates are a good proxy for (global) growth opportunities, that it is the similarity (or di¡erence) in growth opportunities for countries at similar (or dif- ferent) levels of ¢nancial development that leads to the positive relationship be- tween growth and the interaction variable external ¢nancial dependence times countries' ¢nancial sector development. A similar possibility may arise with respect to the asset allocation hypothesis and our asset mix variable. If growth opportunities systematically vary across countries with the degree of property rights protection, then a statistically sig- ni¢cant coe/cient for our interaction variable could be inaccurately interpreted as support for the asset allocation hypothesis.To investigate this possibility, we use the same approach as Fisman and Love (2000). Speci¢cally, we interact both the external ¢nancial development and property rights variables with the U.S. sectoral sales growth rates and include these two new interaction variables as well in the regressions. The estimation technique remains OLS, and the depen- dent variable remains the average annual real growth rate of value added in a particular sector in a particular country over the period 1980 to 1989. The new speci¢cation thus becomes Growthj ¼ Constant þ G1 Industry dummiesj ;k þ G2 Country dummiesk þ g3 Industry share of manufacturing value addedj ;k þ g4 External dependencej Financial developmentk þ g5 Growth opportunitiesj Financial developmentk þ g6 Intangible intensityj Property rightsk þ g7 Growth opportunitiesj Property rightsk þ ej : ð3Þ ;k In this extended speci¢cation of the model, we include the interaction between the growth opportunities of industry j and ¢nancial development in country k, and the interaction between the growth opportunities of industry j and property rights in country k. TableVIII shows the results where the speci¢cations vary in how many inter- acted variables they include and which proxy we use for growth opportunities. Columns 2 to 4 inTableVIII show the regression results of adding the interacted U.S. sales growth variable in this way to the model, with column 1 repeating the results of column 3 of TableV. Column 2 con¢rms the result of Fisman and Love, that is, the interaction termbetween ¢nancial development and U.S. sales growth ``dominates''the interaction term between ¢nancial development and external ¢- nancial dependence in terms of sectoral growth, as the coe/cient on the interac- tionterm between ¢nancial development and external ¢nancial dependence is no longer statistically signi¢cant. In column 3, we add the interaction term between Chapter Two 73 Financial Development, Property Rights, and Growth 2427 property rights and U.S. sales growth. Although this new interaction term is also statisticallysigni¢cant, our main resultFa positive relationshipbetween sector- al growth and the interaction term property rights and asset mixFis robust to this change in speci¢cation, although the statistical signi¢cance for our main result decreases somewhat.When we add both new interaction variables, that is, the interaction between U.S. sales growth and ¢nancial development and be- tween U.S. sales growth and property rights, to the model (column 4), our main result still holds, but the RZ and Fisman and Love variables are no longer statis- tically signi¢cant. This suggests that the asset allocation e¡ect remains an im- portant explanation of ¢rm growth. The measure of growth opportunities used in Fisman and Love, that is, the ac- tual sales growth at the sectoral level, is an ex post measure. It is thereforehighly correlated with actual growth in value added, our dependent variable, and as such may not be the best measure to use for growth opportunities and could ex- plain the reduced signi¢cance of the interaction variables in columns 3 and 4. As an alternative, more forward-looking proxy for growth opportunities, we use To- bin's Q ratio, that is, the ratio of the market value of the ¢rm to the book value of its assets. We use COMPUSTAT data to construct the industry-level median of the time-average Tobin's Q of U.S. ¢rms during the period 1980 to 1989.The re- sults of using this alternative measure of growth opportunities in the interaction variables are presented in columns 5 to 7 of TableVIII. In contrast to the actual sales growth measure, we ¢nd that the interactionvariables withTobin's Q do not enter signi¢cantly in any of the regressions, showing that the results are depen- dent on the proxy used for growth opportunities. Our main result is strength- ened, however, as the coe/cients for the interaction variable property rights and asset mix become more statistically signi¢cant. This suggests that growth opportunities, as measured by ¢rms' Tobin's Q, do not vary across countries in such a systematic way with the degree of property rights protection as to a¡ect the relationship between property rights and actual growth that is occurring through improved asset allocation. As a third robustness test, we investigate whether using U.S. sectoral data biases our results in some way. It could be the case, for example, that investment opportunities in poorer countries are di¡erent from those in the United States due to di¡erences in the general level of a country's development rather than dif- ferences in property rights. For a poor country with the same property rights as a rich country, for example, the sectoral measure of intangible intensity may not relate in the same way to relative growth rates because growth opportunities dif- fer due to its general lower level of development. Any relationship between growth and our interaction term of intangible intensity times property rights may then be spurious because it re£ects di¡erences in growth opportunities, and not the asset allocation e¡ect.We test for this possibility by adding an inter- actionvariable between the U.S. sectoral asset mix and countries'per capita GDP to the regression.We use the level of per capita GDP as a measure of the overall level of a country's economic development and of corresponding country-level in- vestment opportunities. The same robustness test was performed by RZ, but then by using an interaction between external dependence and per capita GDP. 74 2428 Table VIII TheAverage E¡ect of Financial Development and Property Rights on Industrial Growth: Di¡erent Growth Opportunities and Income Levels The dependent variable in all regressions is the average annual real growth rate of value added in a particular sector in a particular countryover the period 1980 to 1989.Table I describes all variables in detail. All regressions include industry dummies, country dummies, and a constant, but these are not reported. Robust standard errors are shown below the coe/cients. Regression (9) includes only those observations for which the property rights index takes a low value of three, regression (10) includes only those observations for which the property rights index takes a medianvalue of four, and regression (11) includes only those observations for which the property rights index takes a highvalue of ¢ve.The United A States is dropped as it is the benchmark. Signi¢cance levels , , and correspond to 1%, 5%, and 10%, respectively. a b c Reader (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Property Property Property Th in Index ¼3 Index ¼4 Index ¼5 International e Fraction of sector in value 1.076a 1.071a 1.074a 1.072a 1.068a 1.064a 1.066a 1.077a 1.466a 0.9445a 0.2194c added of manufacturing in 1980 (0.2491) (0.2496) (0.2471) (0.2478) (0.2510) (0.2522) (0.2528) (0.2503) (0.2255) (0.3819) (0.1178) Journ Sectoral measure of ¢nancial 0.1354a 0.0649 0.0896a 0.0617 0.1176a 0.1124a 0.1183a 0.1353a dependence n private (0.0376) (0.0458) (0.0338) (0.0457) (0.0364) (0.0324) (0.0364) (0.0376) Corporate credit to GDP alofF Sectoral measure of sales 1.170c 0.5671 growth n private credit to GDP (0.6806) (0.5426) inance Sectoral measure of Tobin's Qn 0.0318 0.0136 private credit to GDP (0.0430) (0.0363) Finance Sectoral measure of intangible 0.0092a 0.0075a 0.0048c 0.0046c 0.0088a 0.0071a 0.0071a 0.0086b intensity n property (freedom) (0.0028) (0.0025) (0.0026) (0.0026) (0.0028) (0.0028) (0.0028) (0.0038) Sectoral measure of sales 0.3377b 0.2915c growth n property (freedom) (0.1731) (0.1612) Sectoral measure of Tobin's Qn 0.0185 0.0198 property (freedom) (0.0129) (0.0133) Sectoral measure of intangible 0.0005 0.0049 0.0027 0.0056 intensity n per capita GDP 1980 (0.0022) (0.0046) (0.0023) (0.0045) R2 0.2757 0.2793 0.2832 0.2839 0.2761 0.2783 0.2784 0.2757 0.3030 0.3781 0.4546 N 1242 1242 1242 1242 1242 1242 1242 1242 387 381 471 Number of countries 44 44 44 44 44 44 44 44 14 13 15 Chapter Two 75 Financial Development, Property Rights, and Growth 2429 If investment opportunities relate systematically to a country's level of develop- ment and a¡ect the abilityof sectors with di¡erent asset mix to grow, rather than a country's property rights a¡ecting growth through the asset mix chosen, then this new interaction variable should be signi¢cant and our old interaction vari- able should no longer be signi¢cant.The speci¢cation becomes Growthj ¼ Constant þ Y1 Industry dummiesj ;k þ Y2 Country dummiesk þ y3 Industry share of manufacturing value addedj ;k þ y4 External dependencej Financial developmentk þ y5 Intangible intensityj Property rightsk þ y6 Intangible intensityj Per capita GDPk þ ej : ;k ð4Þ In this extended speci¢cation of model (1), we include the interaction between the intangible intensity of industry j and per capita GDP of country k. Controlling for di¡erences in thelevelof development in this waydoes not alter our main result since the new interaction variable is not statistically signi¢cant, while our old interactionvariable still is signi¢cant (column 8 inTableVIII).Thus, variations in property rights across countries that lead to di¡erent growth pat- terns do not seem to be due to simple di¡erences in investment opportunities re- lated to the level of development, but rather to di¡erences in the asset mix chosen in response to variations in property rights. As an alternative robustness test along the same lines, we test whether for countries with the same level of property rights, investment opportunities di¡er in a systematic way with income levels such as to confound the relationship be- tween assets mix and growth. If investment opportunities across sectors do not vary in a systematic way with income level, then for the same level of property rights, we should not ¢nd an e¡ect across countries of the income level variable interacted with the asset mix variable. Columns 9 to 11 in Table VIII show the results of regressions for three subsamples of countries with each having the same degree of protection of property rights (as measured by our main property rights index), but di¡erent levels of per capita GDP. Using this speci¢cation, we do not ¢nd an income level e¡ect since the coe/cients for the interaction term between asset mix and per capita GDP are insigni¢cant in each of the three cases. Finally, we explore the robustness of our result to the time period chosen. Par- ticularly, we explore the sensitivity of results to the inclusion of data from the 1990s. First, we use as the dependent variable the average annual real growth rate of value added in a particular sector in a particular country over the period 1980 to 1999, rather than only the 1980s. Using growth rates over a longer period has some advantages since we are interested in the long-run relationships be- tween property rights, ¢nancial development, and growth. The main drawback of including growth data from the 1990s is that the number of countries drops sharply, from 44 to 19. This is because data on sectoral growth in value added 76 2430 TableIX TheAverage E¡ect of Financial Development and Property Rights on Industrial Growth: Di¡erentTime Periods The dependent variable in all regressions is the average annual real growth rate of value added in a particular sector in a particular country. In A regressions (1) and (3), the average annual real growth rate is calculated over the period1980 to 1989. In regressions (2) and (4), the average annual Reader growth rate is calculated over the period 1980 to 1999. Regressions (1) and (2) use intangible-intensity values based on data from the 1980s, while regressions (3) and (4) use intangible-intensity values basedon data from the1990s.Table Idescribes allvariables in detail. All regressions include industry dummies, country dummies, and a constant, but these are not reported. Robust standard errors are shown below the coe/cients.The Th in United States is dropped from all regressions as it is the benchmark country. Signi¢cance levels , , and correspond to 1%, 5%, and 10%, a b c International respectively. e (1) (2) (3) (4) Journ Growth over 1980^89; Growth over 1980^99; Growth over 1980^89; Growth over 1980^99; Intangibility over 1980s Intangibility over 1980s Intangibility over 1990s Intangibility over 1990s alofF Corporate Fraction of sector in 1.076a 0.2256b 1.047a 0.1973b value added of manufacturing (0.2491) (0.1012) (0.2470) (0.0974) inance in 1980 Sectoral measure of ¢nancial 0.1354a 0.0449c 0.1398a 0.0516b Finance dependence n private (0.0376) (0.0259) (0.0379) (0.0262) credit to GDP Sectoral measure of intangible 0.0092a 0.0074a 0.0078c 0.0056b intensityn property (freedom) (0.0028) (0.0018) (0.0047) (0.0026) R2 0.2757 0.6133 0.2735 0.6061 N 1242 478 1242 478 Number of countries 44 19 44 19 Chapter Two 77 Financial Development, Property Rights, and Growth 2431 are not available for manycountries, since the United Nations database on Indus- trial Statistics includes data on sectoral growth in value added with a lag of sev- eralyears for most countries.The results of using growth rates over the1980s and the 1990s are reported in column 2 of Table IX, where column 1 reports for ease of comparison the results using the same speci¢cation for the 1980s (as already re- ported in TableV, column 3).We ¢nd that our main result is not qualitatively al- tered, because the coe/cients for both the interactive variable external ¢nancial dependence times ¢nancial development and the interactive variable intangible intensity times property rights remain statistically signi¢cant and positive. As a further robustness test of the time period studied, we also reestimated model (1) using the growth data of the 1980s, but with the sectoral intangible-in- tensity variable measured over the 1990s rather than the 1980s.This test investi- gates whether the use of a particular time period for the benchmark, industry levelof intangible intensity, a¡ects our ¢ndings. Our main result does notchange qualitatively either when using this di¡erent benchmark (column 3 in Table IX), although the statistical signi¢cance is reduced somewhat. This robustness should not be a surprise, since the correlation between the sectoral intangible- intensity variables for the two di¡erent time periods is high,0.90. Our results are also robust to using the average growth rates over the period 1980 to 1999 and the intangible-intensity values for the 1990s (column 4 in Table IX). Overall, our re- sults do not seem to be a¡ected by the particular time period chosen. V. Conclusions Countries di¡er from each other in many ways.Two aspects are the degree of their ¢nancial sector development and the quality of their property rights.This paper argues that an environment with poorly developed ¢nancial systems and weak property rights has two e¡ects on ¢rms: First, it reduces the access of ¢rms to external ¢nancing and, second, it leads ¢rms to allocate resources in a subop- timal way.The importance of the lack of ¢nancing e¡ect has already been shown in the law and ¢nance literature. We investigate the importance of property rights for ¢rm growth by studying its impact on ¢rms' allocation of investable resources.We ¢nd evidence suggesting that the e¡ect of insecure property rights on the asset mix of ¢rms, the asset allocation e¡ect, is economicallyas important as the lack of ¢nancing e¡ect, because it impedes the growth of ¢rms to the same quantitative magnitude. Furthermore, the evidence suggests that the asset allo- cation e¡ect is particularly important in hindering the growth of new ¢rms. While we use the ratio of tangibles and intangible assets as a measure of asset mix, the implications of our results probably go beyond this particular asset choice and may imply that an e/cient allocation of ¢rm resources can more gen- erallybe impededby weak property rights. Our results may imply that the degree to which ¢rms allocate resources in an optimal way will depend on the strength of a country's property rights and that ¢rms' asset allocation is an important channel through which property rights a¡ect ¢rm growth. Thus, our results may have the policy implication that, just as it is important to have a good ¢nan- 78 A Reader in International Corporate Finance 2432 The Journal of Finance cial system, requiring in turn a functioning legal system, it is also important to assure the protection of returns to di¡erent types of assets.To the extent that the emergence of the``new economy'' has increased the economic returns to assets on which yields are more di/cult to secure, our results could even underestimate the overall costs of weak property rights. If indeed new economy assets and fu- ture growth opportunities are more related to intangible assets, then any under- allocation of investable resources towards intangible assets may impede the future growth of ¢rms and economies more generally, and even more so going forward. Appendix:TheValues of the InstitutionalVariables by Individual Country Table AI reports the values of the country variables for the countries studied. Property (freedom) is a rating of property rights in each country (on a scale from 1 to 5). The index equals the median rating for the period 1995 to 1999, and the source is the Index of Economic Freedom from the Heritage Foundation.We re- versed the original order of the index. Intellectual property (301) is an index of intellectual property rights (on a scale from 1to 5).The index is calculated using the Special 301 placements of the O/ce of the U.S.Trade Representative.The in- dex equals the median rating for the period 1990 to 1999. Patent rights (GP) is an index of patent rights (on a scale from 0 to 5) in 1980. The source of the patent rights index is Ginarte and Park (1997). Property (WEF) is an index of property rights for the year 2001 (on a scale from 1to 7).The source is theWorld Economic Forum (2002). Intellectual property (WEF) is an index of intellectual property rights for the year 2001 (on a scale from 1to 7).The source is theWorld Economic Forum (2002). Property (ICRG) is a measure of property rights in each country (on a scale from 0 to 10).The index equals the average rating for the period 1982 to 1995.The source is Knack and Keefer (1995). Each property rights index is con- structed such that the more protection property receives, the higher the score of the index. Private credit is private credit dividedby GDP in1980.The source is RZ and the International Financial Statistics of the International Monetary Fund. Market cap is stock market capitalization divided by GDP in 1980.The source is RZ. Accounting is accounting standards in1983 on a scale from 0 to 90, withhigh- er scores indicating more disclosure.The source is RZ. Human capital is the aver- age for 1980 of the years of schooling attained by the population over 25 years of age.The source of the human capital variable is Barro and Lee (1993). Rule of law is an assessment of thelawand order tradition inthe country (on a scale from 0 to 10).The rating is the average of the months of April and October of the monthly index between 1982 and 1995. The source is La Porta et al. (1997). Legal origin identi¢es the legal origin of the Company Law or Commercial Code of each coun- try. There are four origins: (1) English Common Law, (2) French Commercial Code, (3) German Commercial Code, and (4) Scandinavian Commercial Code. The source is La Porta et al. (1999). European settler mortality is the European settler mortality rate, measured in terms of deaths per annum per 1,000 mean strength. The source is Acemoglu et al. (2001). GDP per capita is the logarithm of GDP per capita in 1980. The source is the World Development Indicators of Table AI TheValues of the InstitutionalVariables by Individual Country Intellectual Intellectual Rule European GDP Property Property Patents Property Property Property Private Market Accounting Human of Legal Settler per Country (Freedom) (301) (GP) (WEF) (WEF) (ICRG) Credit Cap Standards Capital Law Origin Mortality Capita F Australia 5.00 4.00 3.23 6.20 6.00 9.30 0.28 0.38 0.70 10.08 10.00 1.00 2.15 9.20 Austria 5.00 5.00 3.81 6.40 6.20 9.45 0.77 0.03 0.48 6.22 10.00 3.00 n.a. 9.16 inanc Bangladesh 2.00 n.a. 1.99 3.70 2.20 2.85 0.07 0.00 n.a. 1.68 n.a. 1.00 4.27 4.79 ia Belgium 5.00 5.00 3.38 5.90 5.50 9.58 0.29 0.09 0.63 8.79 10.00 2.00 n.a. 9.33 lD Brazil 3.00 3.00 1.85 5.00 4.10 6.64 0.23 0.05 0.69 2.98 6.32 2.00 4.26 7.41 Canada 5.00 4.00 2.76 6.20 5.80 9.73 0.45 0.46 0.68 10.16 10.00 1.00 2.78 9.26 Chile 5.00 4.00 2.41 5.60 4.20 6.44 0.36 0.34 0.60 5.99 7.02 2.00 4.23 7.84 evelopm Colombia 3.00 4.00 1.12 4.30 3.00 5.54 0.14 0.05 0.39 4.23 2.08 2.00 4.26 7.05 Costa Rica 3.00 n.a. 1.94 5.20 3.70 6.47 0.26 0.04 n.a. 4.81 n.a. 2.00 4.36 7.68 Chapter Denmark 5.00 5.00 3.62 6.40 6.30 9.80 0.42 0.09 0.62 10.14 10.00 4.00 n.a. 9.41 ent,P Egypt 3.00 3.00 1.99 5.60 4.10 4.96 0.21 0.01 n.a. 2.16 4.17 2.00 4.22 6.33 Finland 5.00 5.00 2.95 6.50 6.40 9.76 0.48 0.06 0.71 9.61 10.00 4.00 n.a. 9.23 rope rt France 4.00 5.00 3.90 6.40 6.60 9.37 0.54 0.10 0.76 5.97 8.98 2.00 n.a. 9.34 Two y Germany 5.00 5.00 3.86 6.50 6.30 9.55 0.78 0.09 0.68 8.46 9.23 3.00 n.a. 9.42 R Greece 4.00 3.00 2.46 5.00 3.90 6.56 0.44 0.08 0.44 6.56 6.18 2.00 n.a. 8.25 India 3.00 3.00 1.62 4.90 3.00 5.80 0.24 0.05 0.71 2.72 4.17 1.00 3.88 5.48 Indonesia 3.00 4.00 0.33 3.80 2.90 4.38 0.20 0.00 n.a. 3.09 3.98 2.00 5.14 6.21 Israel 4.00 4.00 3.57 6.30 4.90 7.22 0.67 0.35 n.a. 9.14 4.82 1.00 n.a. 8.18 ights,and Italy 4.00 4.00 3.71 6.20 5.70 8.07 0.42 0.07 0.69 5.83 8.33 2.00 n.a. 8.77 G Jamaica 4.00 n.a. 2.86 4.90 3.50 5.05 0.15 0.02 n.a. 3.60 n.a. 1.00 4.87 7.11 Japan 5.00 4.00 3.94 6.10 5.50 9.34 0.86 0.30 0.67 8.17 8.98 3.00 n.a. 9.20 row Jordan 4.00 4.50 1.86 5.80 4.60 5.15 0.54 0.50 n.a. 2.93 4.35 2.00 n.a. 7.01 th Kenya 3.00 n.a. 2.57 n.a. n.a. 5.58 0.20 0.00 n.a. 2.44 5.42 1.00 4.98 6.03 Korea, Rep. 5.00 3.00 3.28 4.70 4.00 6.90 0.50 0.08 n.a. 6.85 5.35 3.00 n.a. 7.25 Malaysia 4.00 4.00 2.57 5.20 3.50 7.09 0.48 0.65 0.78 4.49 6.78 1.00 2.87 7.43 Mexico 3.00 4.00 1.40 4.60 3.60 5.76 0.16 0.07 n.a. 3.51 5.35 2.00 4.26 7.88 Morocco 3.50 n.a. 2.38 n.a. n.a. 5.05 0.16 0.02 n.a. n.a. n.a. 1.00 4.36 6.69 24 Netherlands 5.00 5.00 4.24 6.50 6.50 9.87 0.60 0.19 0.73 8.20 10.00 2.00 n.a. 9.32 33 New Zealand 5.00 4.00 3.32 5.90 5.30 9.80 0.19 0.33 0.61 12.14 10.00 1.00 2.15 8.92 79 80 2434 Table AI (continued) Intellectual Intellectual Rule European GDP A Property Property Patents Property Property Property Private Market Accounting Human of Legal Settler per Reader Country (Freedom) (301) (GP) (WEF) (WEF) (ICRG) Credit Cap Standards Capital Law Origin Mortality Capita Nigeria 3.00 n.a. 3.05 3.80 2.50 3.85 0.12 n.a. 0.62 n.a. 2.73 1.00 7.60 6.81 Th in Norway 5.00 5.00 3.29 5.90 5.30 9.69 0.34 0.06 0.71 10.32 10.00 4.00 n.a. 9.51 e International Pakistan 4.00 4.00 1.99 n.a. n.a. 4.21 0.25 0.03 0.69 1.74 3.03 1.00 3.61 5.67 Peru 3.00 4.00 1.02 4.10 3.00 4.19 0.11 0.06 n.a. 5.44 2.50 2.00 4.26 6.74 Journ Philippines 4.00 4.00 2.67 4.30 2.90 3.62 0.28 0.10 0.63 6.00 2.73 2.00 n.a. 6.59 Portugal 4.00 5.00 1.98 5.30 4.90 7.94 0.52 0.01 0.52 3.23 8.68 2.00 n.a. 7.74 Singapore 5.00 4.00 2.57 6.50 5.60 8.69 0.57 1.62 0.73 3.69 8.57 1.00 2.87 8.45 South Africa 3.00 4.00 3.57 5.30 4.50 7.50 0.26 1.20 0.81 4.61 4.42 1.00 2.74 7.97 alofF Corporate Spain 4.00 4.00 3.29 5.90 5.30 7.99 0.76 0.09 0.42 5.15 7.80 2.00 n.a. 8.53 Sri Lanka 3.00 n.a. 2.79 4.20 3.10 4.64 0.21 0.06 n.a. 5.18 1.90 1.00 4.25 5.53 inance Sweden 4.00 4.00 3.47 5.90 5.80 9.80 0.42 0.11 0.81 9.47 10.00 4.00 n.a. 9.57 Turkey 4.00 3.00 1.80 4.20 3.10 5.76 0.14 0.01 n.a. 2.62 5.18 2.00 n.a. 6.99 Finance UK 5.00 5.00 3.57 6.30 6.10 9.40 0.25 0.38 0.80 8.35 8.57 1.00 n.a. 9.17 Venezuela 3.00 4.00 1.35 3.80 3.00 5.82 0.30 0.05 n.a. 4.93 6.37 2.00 4.36 8.29 Zimbabwe 3.00 n.a. 2.90 3.90 2.90 5.09 0.30 0.45 n.a. 2.40 3.68 1.00 n.a. 6.09 Average 3.96 4.12 2.67 5.33 4.51 7.03 0.36 0.20 0.65 5.84 6.67 1.91 4.03 7.79 Chapter Two 81 Financial Development, Property Rights, and Growth 2435 theWorld Bank. More detailonthe de¢nitions and sources of thevariables can be found inTable I. Countries are sorted in ascending alphabetical order.The abbre- viation n.a. stands for not available. REFERENCES Acemoglu, Daron, Simon Johnson, and James A. Robinson, 2001,The colonial origins of comparative development: An empirical investigation, American Economic Review 91, 1369^1401. Acemoglu, Daron, Simon Johnson, and James A. Robinson, 2002, Reversal of fortune: Geography and institutions in the making of the modern world income, QuarterlyJournal of Economics 117, 1231^ 1294. Barro, RobertJ.,1991, Economic growth in a cross sectionof countries, QuarterlyJournal of Economics 106, 407^443. Barro, RobertJ., andJong-Wha Lee,1993, International comparisons of educational attainment,Jour- nal of Monetary Economics 32, 363^394. Beck,Thorsten, Asli Demirgc-Kunt, and Ross Levine, 2003, Law, endowments, and ¢nance, Journal of Financial Economics, forthcoming. Beck,Thorsten, Ross Levine, and Norman Loayza, 2000, Finance and the sources of growth, Journal of Financial Economics 58, 261^300. Besley, Timothy, 1995, Property rights and investment incentives: Theory and evidence from Ghana, Journal of Political Economy 103, 903^937. Booth, Laurence,Varouj Aivazian, Asli Demirgc-Kunt, andVojislav Maksimovic, 2000, Capital struc- tures in developing countries, Journal of Finance 56, 87^130. Bradley, Michael, Gregg A. Jarrell, and E. Han Kim,1984, Onthe existence of anoptimalcapital struc- ture: Theory and evidence, Journal of Finance 39, 857^878. Ceterolli, Nicola, and Michele Gambera, 2001, Banking market structure, ¢nancial dependence and growth: International evidence from industry data, Journal of Finance 56, 617^648. Demirgc-Kunt, Asli, and Vojislav Maksimovic, 1998, Law, ¢nance, and ¢rm growth, Journal of Finance 53, 2107^2137. Demirgc-Kunt, Asli, and Vojislav Maksimovic, 1999, Institutions, ¢nancial markets, and ¢rm debt maturity, Journal of Financial Economics 54, 295^336. Fisman, Raymond, and Inessa Love, 2002, Patterns of industrial development revisited: The role of ¢nance, Mimeo, Columbia University. Fisman, Raymond, and Inessa Love, 2003,Trade credit, ¢nancial intermediary development, and in- dustry growth, Journal of Finance 58, 353^374. Ginarte, Juan Carlos, and Walter Park, 1997, Determinants of patent rights: A cross-national study, Research Policy 26, 283^301. Hall, Robert E., and Charles I. Jones, 1999,Why do some countries produce so much more output per worker than others? QuarterlyJournal of Economics 114, 83^116. Harris, Milton, and Artur Raviv, 1991,The theory of capital structure, Journal of Finance 46, 297^355. Johnson, Simon, Daniel Kaufmann, and Pablo Zoido-Lobaton, 1998, Government in transition: Regu- latory discretion and the uno/cial economy, American Economic Review 88, 387^392. Johnson, Simon, John McMillan, and ChristopherWoodru¡, 2002, Property rights and ¢nance,Amer- ican Economic Review 92,1335^1356. King, Robert G., and Ross Levine, 1993, Finance and growth: Schumpeter might be right, Quarterly Journal of Economics 108,717^737. Knack, Steven, and Philip Keefer, 1995, Institutions and economic performance: Cross-country tests using alternative measures, Economics and Politics 7, 207^227. La Porta, Rafael, Florencio Lopez-de-Silanes, Christian Pop-Eleches, and Andrei Shleifer, 2002,The guarantees of freedom, Mimeo, Harvard University. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and RobertW.Vishny,1997, Legal deter- minants of external ¢nance, Journal of Finance 52, 1131^1150. 82 A Reader in International Corporate Finance 2436 The Journal of Finance La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W.Vishny, 1998, Law and ¢nance, Journal of Political Economy 106, 1113^1155. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and RobertW.Vishny,1999,The quality of government, Journal of Law, Economics and Organization 15, 222^279. Levine, Ross, 1997, Financial development and growth, Journal of Economic Literature 35, 688^726. Levine, Ross, and Sara Zervos,1998, Stock markets, banks, and economic growth, American Economic Review 88, 537^558. Mans¢eld, Edwin,1995, Intellectual property protection, direct investment, and technology transfer, Discussion paper 27, International Finance Corporation,Washington, DC. Myers, Stewart C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5, 146^ 175. North, Douglass C., 1990. Institutions, institutional change, and economic performance (Cambridge Uni- versity Press, Cambridge, MA). Rajan, Raghuram, and Luigi Zingales,1995,What do we know about capital structure: Some evidence from international data, Journal of Finance 50, 661^691. Rajan, Raghuram, and Luigi Zingales, 1998, Financial dependence and growth, American Economic Review 88, 559^586. Romer, Paul M., 1990, Endogenous technological change, Journal of Political Economy 98, S71^S102. Stern, Scott, Michael E. Porter, and Je¡rey L. Furman, 2000,The determinants of national innovative capacity, NBER Working Paper No.7876. Titman, Sheridan, and RobertoWessels,1988,The determinants of capital structure choice, Journal of Finance 43, 1^19. World Economic Forum, 2002, Global Competitiveness Report (Oxford University Press, Oxford, UK). Chapter Three 83 DOES LEGAL ENFORCEMENT AFFECT FINANCIAL TRANSACTIONS? THE CONTRACTUAL CHANNEL IN PRIVATE EQUITY* JOSH LERNER AND ANTOINETTE SCHOAR Analyzing 210 developing country private equity investments, we find that transactions vary with nations' legal enforcement, whether measured directly or through legal origin. Investments in high enforcement and common law nations often use convertible preferred stock with covenants. In low enforcement and civil law nations, private equity groups tend to use common stock and debt, and rely on equity and board control. Transactions in high enforcement countries have higher valuations and returns. While relying on ownership rather than contractual provisions may help to alleviate legal enforcement problems, these results suggest that private solutions are only a partial remedy. I. INTRODUCTION A large literature in economics and finance has documented a systematic relationship between a country's legal system and the development and liquidity of its financial markets. Starting with La Porta et al. [1997, 1998], these works identify legal origin as a crucial determinant of minority shareholder protection against expropriation by corporate insiders, with common law systems providing better protection than civil law ones. Glaeser, Johnson, and Shleifer [2001] and Djankov et al. [2003] suggest that parties in common law countries can more readily enforce commercial contracts. Common law and high enforcement na- tions have broader and more valuable capital markets, more public offerings, dispersed ownership of public firms, and other indicators of financial development (also see Demirgu¨c-Kunt and Levine [2001]). Much less attention, however, has been directed to under- * We thank many private equity groups for making this study possible by providing the transaction information. Teresa Barger, Richard Frank, Felda Har- dymon, Gustavo Herrero, Mario Mahler, Kenneth Morse, Bruce Purdue, Kanako Sekine, and Camille Tang Yeh introduced us to many groups. Zahi Ben-David, Adam Kolasinski, Jiro Kondo, and especially Yok Nam Ng provided excellent research assistance. We also thank our legal research team: Arturo Garcia de Leon, May Fong Yue Lo, Alexander Nadmitov, Rahul Singh, Michiel Vissier, Agata Waclawik, and Feng Wang, as well as Sridhar Gorthi of Trilegal. We thank Erik Bergloff, Nittai Bergman, Peter Henry, Katharina Lewellen, Roberta Ro- mano, Andrei Shleifer, Per Stro¨mberg, Amir Sufi, Yishay Yafeh, and participants at presentations at Harvard University, the London School of Economics, the Stockholm Institute for Financial Research, and the Western Finance Association annual meeting for helpful comments. Harvard Business School's Division of Research provided financial assistance. All errors are our own. © 2005 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology. The Quarterly Journal of Economics, February 2005 223 84 A Reader in International Corporate Finance 224 QUARTERLY JOURNAL OF ECONOMICS standing the specific avenues through which the nature of the legal system affects financial development. The current paper highlights the importance of what we term "the contractual chan- nel": the ability of investors to enter into complex, state-depen- dent contracts. We document that investors in countries with effective legal enforcement rely on specific contracting contingen- cies and securities that shift control rights depending on the performance of the investment and enable investors to separate cash flow and control rights. A large theory literature points to the benefits of these contracting possibilities for entrepreneurs and investors (as we describe in the following section). By way of contrast, investors in countries with difficult legal enforcement seem to be required to secure control rights through majority ownership. These results suggest that a critical impact of the legal system is the way it constrains the ability of private parties to write contracts that are complex or state contingent. Parties cannot easily undo deficiencies of the law through private trans- actions if the legal system does not enforce certain types of contracts. We focus on a specific set of transactions: private equity investments. We concentrate on these transactions since they are better documented than most private financial transactions, and follow a relatively standardized setup. Private equity transac- tions represent a relatively modest share of the absolute value of investments made in most developing countries. But we think that they are representative of the legal and economic consider- ations that private parties face in any contract negotiation. We collect data on the actual contractual relationships between in- vestors and entrepreneurs in 210 transactions from a wide vari- ety of private equity groups and countries. We find that investments in countries with a common law tradition and with better legal enforcement are far less likely to employ common stock or straight debt, and more likely to use convertible preferred stock. Similarly, transactions in these na- tions are generally associated with greater contractual protec- tions for the private equity groups. These contracts look similar to U. S. contracts, which an extensive theoretical literature suggests are a second-best solution to contracting in private equity. In contrast, investors in countries with civil law or socialist legal background and where legal enforcement is difficult rely more heavily on obtaining majority control of the firms they invest in, Chapter Three 85 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 225 use debt more often, and have more board representation. These findings suggest that private equity groups here rely on owner- ship, which may substitute for the lack of contractual protections. We also verify that our results are not driven by the tendency of common law-based funds to invest in common law countries. Finally, we investigate the consequences of these differences: can the parties successfully address the absence of the contrac- tual channel by relying on large ownership stakes? We find that firms' valuations are significantly higher in nations with a com- mon law tradition, and superior legal enforcement and private equity funds investing in common law countries enjoy higher returns. We point out, however, that this evidence is only sug- gestive of any effects of contracting constraints on investment outcomes. These results suggest that systematic differences in legal enforcement impose constraints on the type of contracts that can be written. This inability to separate cash flow rights from control rights has the potential to seriously distort the contracting pro- cess by forcing the parties to rely on large equity stakes. Private equity investors face constraints in diversifying their portfolio, since they have to hold larger stakes of a given firm than they would like for pure control purposes. Entrepreneurs might have reduced incentives since they are forced to give up a substantial amount of cash flow (and control) rights early on. These findings suggest that the lack of contract enforcement may not be easily undone by private contracting arrangements that emphasize ownership. The plan of this paper is as follows. Section II lays out the theoretical motivation for the analysis. Section III describes the construction of the data set. The analysis is in Section IV. The final section concludes the paper. II. THE ECONOMICS OF PRIVATE EQUITY Financial contracts are written to assign cash flow and con- trol rights between contracting parties, e.g., a private equity group and an entrepreneur. An extensive literature on optimal contracting, starting with Holmstro¨m [1979], has analyzed the role of contracts in alleviating principal-agent problems through the contingent allocation of cash flow rights. It relies on the assumption that contracts can be enforced costlessly. 86 A Reader in International Corporate Finance 226 QUARTERLY JOURNAL OF ECONOMICS The literature on incomplete contracting--see Grossman and Hart [1986] and Hart and Moore [1990]--highlights that if courts are unable to enforce or even verify complicated, state-dependent contracts, the allocation of control rights can allow the parties to reach a second-best agreement. Aghion and Bolton [1992] and Hellmann [1998] show that convertible preferred securities allow control rights to be transferred to the party that makes better use of them. In particular, these securities allocate control to the entrepreneur when things are going well, but allow the investors to assert control if the firm is doing poorly. These securities will give stronger incentives to entrepreneurs than majority control based on common stock contracts, since they prevent the holdup of entrepreneurs by investors if the entrepreneurs are running the firm well. In the context of private equity, Kaplan and Stro¨mberg [2003] and Gompers [1998] identify a number of benefits to in- vestors and entrepreneurs from being able to separate cash flow and control rights, typically through the use of convertible pre- ferred securities.1 The ability to maintain control rights without majority cash flow rights allows investors to invest relatively small amounts of capital early on without fearing expropriation, thereby allowing capital diversification. Entrepreneurs benefit since they do not have to give away cash flow rights early on when valuations are still very low. It might well be, however, that private equity groups in certain nations are unable to enforce contracts involving the separation of ownership and control or more complicated contin- gencies, since it may be difficult to educate judges and lawyers about these contract features. In these instances, we envision that firms will employ third-best contracts, which entail the use of controlling blocks of common stock or straight debt. We expect this pattern to be most prevalent in nations where the legal system is less well developed. Moreover, we would predict that control through majority ownership of common stock and control through contract contingencies would be substitutes. Obviously, if courts are so inefficient or corrupt that they cannot enforce any 1. Unlike in public settings, in private equity preferred stock refers to a security that awards liquidation rights to the investor if the company does not achieve a threshold performance level. In the following, we refer to the group of securities as convertible preferred stock to avoid confusion with preferred that only has preferential voting rights. Chapter Three 87 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 227 contract at all, even majority ownership would not protect investors. Bergman and Nicolaievsky [2003] develop a formal model that starts from a similar assumption as put forward here: legal regimes differ in their ability to enforce complicated contingen- cies to prevent investor expropriation. They find supporting evi- dence in Mexico. The focus of the analysis is complementary to the current paper, since the paper aims to contrast the use of contractual contingencies in private versus public firms, where renegotiation between different groups of investors is more difficult. In a contemporaneous paper, Kaplan, Martel, and Stro¨mberg [2003] examine venture capital contracts for a set of high-income European countries. They find that most of the contractual varia- tion between common law and civil law countries in their sample is explained by the fact that private equity groups use contracts that are similar to the ones they employ in their home countries. It is possible that the higher sophistication of the judicial system in these countries allows private equity groups to experiment with contracts that are different from those customarily em- ployed in the local market. One might also conjecture, however, that a perceived sense of similarity between the United States and Continental Europe led investors in some cases to make contracting choices that might ultimately be very difficult to enforce in these countries.2 Our hypothesis was informally corroborated in our conversa- tions with investment professionals at private equity groups. The groups indicated that they place much greater emphasis on hav- ing controlling equity blocks in nations with poor contract en- forcement, largely due to their inability to enforce more complex contracts. One group operating in Latin America, for instance, had initially employed convertible preferred securities in all its transactions. Their enthusiasm for this investment strategy waned, however, when they began litigating with one of their portfolio companies in Peru. The private equity investors found 2. Similarly, Cumming and MacIntosh [2002] examine the types of transac- tions funded and exit routes employed in twelve Asian nations. They argue that the legal regimes affect the types of investments selected and the way in which the private equity groups exit their holdings, but not returns. Qian and Strahan [2004] show that bank loans in countries with better legal protection are less likely to be secured and have more covenants. 88 A Reader in International Corporate Finance 228 QUARTERLY JOURNAL OF ECONOMICS themselves unable to convince the judge that their preferred stock agreement gave them the right to replace a third-genera- tion founder of the company, even if the group's shares were only convertible into 20 percent of the firm's equity. After this experi- ence, the private equity group structured its subsequent invest- ments as common stock deals in which they held the majority of the equity. In many nations, our interviewees asserted, not only were the entrepreneurs unfamiliar with equity investments that used securities other than common stock, but key actors in the legal system--lawyers and judges--were suspicious and indeed hostile to such transactions. As a result, they chose to employ common stock there. These conversations did not yield a consis- tent answer to the question of whether the efforts to address the ineffectiveness of the contractual channel through a reliance on ownership would be successful.3 III. THE DATA We constructed the sample by asking private equity groups that invest in developing nations4 to give us a representative array of their transactions in terms of the type of deal, the location and industry of the firm, and the success of the transac- tion. For each transaction we obtained the investment memoran- dum, the associated stock purchase agreements, and any other documents associated with the structuring of the transaction. We deliberately attempted to recruit as diverse an array of private equity funds as possible. In a study along these lines, selection biases are an almost inevitable consequence. We tried to amelio- rate this concern by obtaining transactions from groups with 3. While there are a few examples, we did not discover many instances where contracting parties in countries with poor legal enforcement relied on private arbitrators instead. See, for example, Johnson, McMillan, and Woodruff [2002] for an analysis of private contract enforcement mechanisms. 4. According to the World Bank, developing nations are those countries that have either low- or middle-level per capita incomes, have underdeveloped capital markets, and/or are not industrialized. It should be noted, however, that the application of these criteria is somewhat subjective. For instance, Kuwait appears on many lists of developing nations despite its high per capita gross domestic product. The reason for its inclusion lies in the income distribution inequality that exists there, which has not allowed it to reach the general living standards of developed countries. For the purposes of this paper, we take an expansive view of what constitutes a developing nation, and simply eliminate any transactions taking place in the 24 nations that were original members of the Organisation for Cooperation and Development or joined within fifteen years of its creation (i.e., through the addition of New Zealand in 1973). Chapter Three 89 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 229 TABLE I CONSTRUCTION OF SAMPLE This table summarizes the key features associated with the construction of the sample of 210 private equity transactions. Private Year of equity group deal Industry of firm Deal type Country of firm Group 1 8 1987 2 Distribution/Retail 14 Buyout 28 Argentina 18 Group 2 6 1988 2 Finance 16 Corp. acquisition 10 Bolivia 2 Group 3 6 1992 3 Food 29 Distress 4 Brazil 18 Group 4 5 1993 4 Health care 9 Expansion 97 Bulgaria 8 Group 5 3 1994 2 Information tech 24 IPO 12 Chile 7 Group 6 3 1995 5 Internet 9 Privatization 10 China 13 Group 7 10 1996 10 Manufacturing 32 Venture capital 49 Estonia 8 Group 8 8 1997 17 Media 8 Ghana 3 Group 9 6 1998 35 Natural resources 11 Hong Kong 13 Group 10 6 1999 31 Real estate 4 India 28 Group 11 11 2000 34 Services 17 Korea 10 Group 12 3 2001 40 Software 10 Indonesia 2 Group 13 2 2002 22 Telecom 14 Latvia 4 Group 14 4 2003 3 Other 13 Malaysia 2 Group 15 10 Mexico 14 Group 16 8 Peru 2 Group 17 6 Poland 13 Group 18 5 Romania 18 Group 19 10 Singapore 6 Group 20 13 South Africa 2 Group 21 14 Taiwan 4 Group 22 8 Tanzania 2 Group 23 5 Thailand 3 Group 24 7 Uruguay 2 Group 25 21 Yugoslavia 6 Group 26 13 Other 5 Group 27 7 Group 28 2 diverse backgrounds. But it is likely that the private equity groups that participated in this study are more Western-oriented and sophisticated than their peers. The presence of this bias should, in fact, reduce the observed variation between legal re- gimes and thus makes the substantial differences that we see even more striking. Table I summarizes the sample. The 210 transactions are from 28 private equity groups, who contributed between 2 and 21 deals for our sample. The transactions occurred between 1987 and 2003, with the bulk of investments between 1996 and 2002. Thirty distinct countries are represented with no single nation or 90 A Reader in International Corporate Finance 230 QUARTERLY JOURNAL OF ECONOMICS region dominating the sample. The industries include a broad array, from food to information technology. We classified the transactions by type using the definitions in European Venture Capital Association [2002]. The investments are dominated by expansion transactions, as well as venture capital and buyout transactions. Panel A of Table II shows that the average GNP per capita for the countries in our sample is $2142 per year. Moreover, 27 percent of the investments are based in countries that have British legal origin, 30 percent have French legal origins, and 42 percent are in former socialist countries. In comparison, 56 per- cent of the investments included in this study are funded by private equity partnerships that are based either in the United States or United Kingdom. While U. K.- and U. S.-based partner- ships in our sample are more likely to invest in countries with British legal origin, we find that they also invest in a large fraction of deals that are not based in common law countries. This heterogeneity is important, since it will allow us to analyze whether a given partnership adjusts the contract terms in re- sponse to the environment of the country where the deal takes place. Panel B of Table II provides an initial overview of the trans- actions. The differences between this sample and U. S. transac- tions are striking. In the United States nearly 80 percent of private equity transactions are dominated by convertible pre- ferred stock (see Kaplan and Stro¨mberg [2003]).5 Common stock is quite rare, found in only a little more than 10 percent of the U. S. deals. In contrast, in our sample 54 percent of the transac- tions employ common stock, while convertible preferred stock is only encountered in 21 percent of the deals.6 Similarly, many of the protections commonly employed by venture capitalists in the United States are rarely found here. Kaplan and Stro¨mberg 5. It should be noted that Kaplan and Stro¨mberg's sample includes only venture capital transactions, which would encompass transactions described as "venture capital" and "expansion" transactions in the developing world. (The category of "expansion" deals is not frequently employed in the United States.) Legal texts (e.g., Bartlett [1995]), however, suggest that we would observe similar patterns if we examined all U. S. private equity transactions. 6. We tried as best as possible to avoid any bias in our coding of contractual terms that are purely based on differences in contractual language. For example, any security structure that has payoff streams equivalent to a convertible pre- ferred would be classified as such, even if the contract did not explicitly use that term. Chapter Three 91 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 231 TABLE II CHARACTERISTICS OF DEVELOPING COUNTRY PRIVATE EQUITY TRANSACTIONS The sample consists of 210 investments in developing countries by private equity groups (PEGs). The first panel describes the features of the transactions; the second panel, the features of the nation and the private equity group involved in the transaction. We do not record the medians and standard deviations of the dummy variables. Panel A: Setting of transactions Mean Median Standard dev Minimum Maximum Per capita gross national product 2142 1743 2561 181 12368 Logarithm of rule of law index 0.22 0.28 0.59 1.25 1.85 English legal family nation 0.27 0 1 French legal family nation 0.30 0 1 Socialist legal family country 0.42 0 1 U. K.- or U. S.-based private equity group 0.56 0 1 Panel B: Nature of transactions Mean Median Standard dev Minimum Maximum Size of financing (1997 $MMs) 4.31 3.29 5.12 0.17 18.53 Implied valuation (1997 $MMs) 5.12 4.18 4.92 0.45 61.38a Straight debt 0.11 0 1 Common stock 0.55 0 1 Straight preferred stock 0.09 0 1 Participating preferred stock 0.05 0 1 Convertible preferred stock 0.21 0 1 Warrants 0.06 0 1 Contingent equity 0.34 0 1 PEG's maximum equity stake 0.47 0.40 0.37 0 1 PEG's minimum equity stake 0.33 0 1 Difference in PEG ownership 0.15 0.01 0.26 0 1 PEG has control when maximum stake 0.37 0 1 PEG has control when minimum stake 0.29 0 1 Antidilution provisions 0.27 0 1 Automatic conversion provisions 0.26 0 1 Maximum board size 6.50 6 2.03 3 12 Minimum board size 5.40 5 1.95 3 11 Maximum PEG board seats 2.66 2 1.89 0 9 Minimum PEG board seats 1.35 1 1.24 0 6 Maximum founder/manager board seats 3.22 3 1.87 0 7 Minimum founder/manager board seats 2.47 2 1.72 0 6 Supermajority sum 18.47 15 12.98 0 57 a. The size of the financing is greater than the valuation in the largest transaction (a leveraged buyout which entailed the purchase of all of the firm's equity) because part of the financing proceeds were used to cover fees to investment bankers, lawyers, and others. 92 A Reader in International Corporate Finance 232 QUARTERLY JOURNAL OF ECONOMICS [2003] find that venture capitalists obtain redemption rights in 84 percent of the transactions, antidilution protection in 95 percent of deals, and founder vesting requirements in 42 percent of trans- actions. The corresponding shares in our sample are much lower: 31 percent, 27 percent, and 5 percent. Finally, the structure of the boards differs little from that seen in the United States. The mean U. S. transaction has a board with 6.2 members, of which two seats were allocated to the founders and managers and two-and-a-half to venture capitalists [Kaplan and Stro¨mberg 2003]. The patterns here are similar, though we see a slightly greater representation of founders and managers on the boards. IV. ANALYSIS We now analyze how contractual choices vary across coun- tries with different legal structure and enforcement. The econo- metric analyses throughout the paper employ a similar structure. We use the existence of different contract provisions as dependent variables: we create a dummy variable equal to one if the deal contains, for example, an antidilution right and zero otherwise. The main explanatory variables we are interested in are the countries' legal origin and, alternatively, the enforcement of con- tracts, measured as the "time-to-contract-dispute-resolution" (see Djankov et al. [2003]). We control for industry, deal type, and year fixed effects.7 We also include per capita gross national product (in current dollars) averaged over the 1990s as a control for the national economic development. We also replicate our results employing logit specifications without industry dummy variables and the results are generally very similar. IV.A. Security Structure In Table III we begin by examining the security structure employed in countries with different legal origins. The economet- 7. We use dummy variables for the observations in three time periods in the reported regressions: the years 1993 to 1997, 1998 to 2000, and 2001 to 2003. These periods correspond, respectively, to the years when many institutions made initial investments into private equity funds focusing on leveraged buyouts in developing nations, the growth of venture capital funding in these nations, and the recent sharp falloff in venture capital and private equity activity there. The results are robust to the use of dummy variables for each year, as well as to the use of controls measuring the annual level of private equity fundraising world- wide and of foreign direct investment into developing nations. LEGAL TABLE III SECURITY STRUCTURE AND LEGAL REGIME The sample consists of 210 investments in developing countries made by private equity groups (PEGs). The dependent variables are dummies denoting whether common stock, straight debt, or convertible preferred stock was employed in the transaction. Independent variables include dummy variables ENFORCEMENT denoting nations with British or socialist legal origin (French legal origin is the omitted category) and the time to resolve commercial disputes in that nation. U. K./U. S.-based PEG is a dummy if the private equity fund is based in the United Kingdom or United States. GNP per capita is the per capita gross national product of the country averaged over the 1990s. All regressions employ ordinary least squares specifications. Standard errors are clustered at the private equity group. Common stock Debt Convertible preferred stock British legal origin 0.19 0.17 0.13 0.11 0.17 0.17 Chapter ***[0.09] **[0.09] ***[0.06] **[0.06] **[0.09] **[0.09] AND Socialist legal origin 0.09 0.07 0.05 0.08 0.05 0.01 [0.09] [0.09] [0.06] [0.06] [0.08] [0.08] FINANCIAL Three Dispute time 0.07 0.10 0.09 [0.05] **[0.05] *[0.05] U. K./U. S.-based PEG 0.03 0.18 0.11 *[0.02] ***[0.05] **[0.06] GNP per capita 0.06 0.05 0.02 0.05 0.06 0.08 0.01 0.03 0.01 TRANSACTIONS [0.07] [0.06] [0.05] [0.06] [0.04] **[0.04] [0.03] [0.05] [0.03] Industry dummies Y Y Y Y Y Y Y Y Y Deal type dummies Y Y Y Y Y Y Y Y Y Year dummies Y Y Y Y Y Y Y Y Y N of observations 210 210 210 210 210 210 210 210 210 Adjusted R2 0.11 0.11 0.09 0.07 0.11 0.06 0.09 0.09 0.07 * Significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level. 233 93 94 A Reader in International Corporate Finance 234 QUARTERLY JOURNAL OF ECONOMICS ric specification follows the description above, with French legal origin as the omitted category. Columns (1), (4), and (7) of Table III show that private equity transactions in common law coun- tries less frequently use common stock or debt in their transac- tion and much more often employ convertible preferred stock compared with those in French or socialist legal origin nations. One concern is that the observed contract structure could be biased due to selection problems. Private equity groups based in common law countries, such as the United States and the United Kingdom, may be disproportionately investing in common law nations, and vice versa for civil law countries. In this case, the structure of the deal might not be driven by the contracting constraints in the country of the transaction, but rather by the familiarity of the private equity group with the contracts in its domestic market. To alleviate this concern, we include a dummy variable equal to one if the private equity group is based in a common law country and zero otherwise. The results in columns (2), (5), and (8) suggest that this potential selection bias does not explain our results. While indeed deals done by private equity groups based in common law countries look more similar to U. S.- style private equity contracts (i.e., they are less likely to rely on common stock or debt and are more likely to use preferred stock), this control does not eliminate the effect on the British legal origin dummy. In fact, the coefficient on the dummy is almost completely unchanged in all specifications. We also repeat the analysis including group fixed effects (not reported). Again, the results on the legal origin of countries are very similar in direc- tion and magnitude. Finally, we use time-to-resolve-contract-dispute as an alter- native proxy for the quality of enforcement of the legal system. We focus on this variable, since it captures more precisely the quality of the enforcement of laws through the court system. We do not include the legal origin indicators in these regressions, since Djankov et al. [2003] show that dispute resolution time is strongly correlated with a country's legal origin. The results in columns (3), (6), and (9) show that countries that take a longer time to resolve contract disputes are less likely to rely on pre- ferred stock and are more likely to use debt. In unreported regressions we repeat this and subsequent analysis excluding any countries that have legal restrictions on private equity transactions. We want to prevent our results from Chapter Three 95 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 235 being "hard wired" by legal rules in different countries (see the Appendix for a summary). For example, in the case of the People's Republic of China, firms can only get permission to use security structures other than common stock in very exceptional cases. We find that the results presented above are qualitatively unchanged when excluding nations restricting security types from the sam- ple. This suggests that our findings reflect the investors' contract- ing choices and not just the constraints imposed by different legal regimes. IV.B. Allocation of Equity and Board Control In Table IV we first examine whether the private equity group controls the company's equity. The dependent variable in columns (1) and (2) is a dummy that takes on the value one if the private equity investors own at least 50 percent of the equity when at their minimum stake. The size of the stake can vary, due to contingent clauses in the main contract that call for supple- mental equity grants to founders and managers in case of good performance and side-agreements regarding vesting. We find that in countries with British legal origins, as well as those with quick dispute resolution, private equity groups are much less likely to have equity control of a firm in the minimum stake scenario. Similarly, in columns (3) and (4) of Table IV, we see that the difference between the maximum and minimum equity stake a private equity group can hold in a given firm is significantly larger in common law countries. In countries with poor enforce- ment, firms avoid contingent equity stakes. The difference in ownership stakes is predominantly driven by the fact that inves- tors in countries with better legal enforcement are willing to invest without a controlling equity stake, since they can achieve minority shareholder protection through other contractual provisions. The last four columns of Table IV investigate the structure of the board as specified in the stock purchase agreements, exam- ining the overall board size as well as the seats assigned to the private equity group. We see that common law nations tend to have larger boards with fewer private equity group representa- tives on the board. Similarly, nations where the time to resolve disputes is shorter have larger boards. (In unreported regressions we show that countries with quick dispute resolution have more 96 236 TABLE IV EQUITY OWNERSHIP, BOARD COMPOSITION, AND LEGAL REGIME IN DEVELOPING COUNTRY PRIVATE EQUITY TRANSACTIONS The sample consists of 210 investments in developing countries by private equity groups (PEGs). The dependent variables in the first four columns are a dummy denoting whether the PEG has control of the firm's equity when it has its minimum contractually specified share of the equity and the difference in the equity ownership stake in the minimum and maximum scenarios. The dependent variables in the last four columns are the logarithms of the number of seats on the board, as well as the seats assigned to the PEG. Independent variables include dummy variables denoting nations with British or socialist legal origin (French legal origin is the omitted category) and the time to resolve commercial disputes in that nation. GNP per capita is the per capita gross national product of the country averaged over the 1990s. All regressions employ ordinary least squares specifications. Standard errors are clustered at the private QUARTERLY A equity group. Reader Does PEG have control when min. ownership Difference between min. Number of PEG in stake? and max. stake Number of board seats board seats International JOURNAL British legal origins 0.20 0.20 0.17 0.06 ***[0.07] ***[0.07] ***[0.08] **[0.03] Socialist legal origins 0.10 0.10 0.05 0.04 Corporate OF *[0.06] *[0.06] [0.08] [0.03] Dispute time 0.11 0.11 0.16 0.09 ECONOMICS ***[0.05] ***[0.05] ***[0.07] [0.07] GNP per capita 0.02 0.09 0.02 0.09 0.01 0.03 0.03 0.04 Finance [0.04] ***[0.04] [0.04] ***[0.04] [0.04] [0.05] [0.04] [0.04] Industry dummies Y Y Y Y Y Y Y Y Deal type dummies Y Y Y Y Y Y Y Y Year dummies Y Y Y Y Y Y Y Y Number of observations 194 194 194 194 197 197 197 197 Adjusted R2 0.08 0.09 0.08 0.09 0.09 0.06 0.07 0.06 * Significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level. Chapter Three 97 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 237 TABLE V CONTROL RIGHTS AND LEGAL REGIME IN DEVELOPING COUNTRY PRIVATE EQUITY TRANSACTIONS The sample consists of 210 investments in developing countries by private equity groups (PEGs). The dependent variables are dummies denoting whether the PEG group has antidilution protection and automatic conversion and the sum of the score of supermajority provisions. (A higher score implies greater use of supermajority provisions.) Independent variables include dummy variables denoting nations with British or socialist legal origin (French legal origin is the omitted category) and the time to resolve commercial disputes in that nation. GNP per capita is the per capita gross national product of the country averaged over the 1990s. All regressions employ ordinary least squares specifications. Standard errors are clustered at the private equity group. Antidilution Automatic rights conversion Supermajority British legal origins 0.20 0.17 1.76 ***[0.09] ***[0.07] ***[0.61] Socialist legal origins 0.08 0.07 1.06 [0.09] [0.08] **[0.56] Dispute time 0.09 0.04 1.01 [0.06] [0.03] **[0.53] GNP per capita 0.05 0.01 0.12 0.10 0.22 0.72 [0.04] [0.04] ***[0.05] **[0.05] [0.35] *[0.40] Industry dummies Y Y Y Y Y Y Deal type dummies Y Y Y Y Y Y Year dummies Y Y Y Y Y Y Number of observations 210 210 194 194 210 210 Adjusted R2 0.09 0.05 0.05 0.02 0.18 0.17 * Significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level. managers on the board.) Table IV suggests that investors use board and equity control to protect their investments in countries with poor legal enforcement. If other methods of enforcing inves- tor rights are effective, equity and board control are less critical. IV.C. Control Rights Table V analyzes control rights that affect the prerogatives of the private equity investors without the need for obtaining a controlling ownership stake. We focus on a number of the most important provisions. The first two columns analyze the existence of antidilution provisions, i.e., the right to have some compensa- 98 A Reader in International Corporate Finance 238 QUARTERLY JOURNAL OF ECONOMICS tion if subsequent financings are done at a lower valuation. This protects investors against losing their equity through dilutive financing rounds. The next two columns focus on the existence of automatic-conversion provisions. Lawyers typically interpret the latter as protecting the lead private equity investor against indi- vidual or smaller private equity investors, who may seek to hold up an IPO or acquisition by refusing to convert their shares. In the last two columns we look at supermajority provisions. These provisions require that a fraction greater than one-half of the investors approves a decision specified in the contract. Typical supermajority provisions include voting on major acquisitions, changes in the business plan that change the nature of the firm, change in top management, etc. These provisions protect minor- ity shareholders from mismanagement or outright fraud by the management of the company.8 A common theme emerges from the analysis in Table V: transactions in common law countries are much more likely to include contractual protections for the private equity investors than those with French or socialist legal origin. This pattern holds whether we examine antidilution, automatic conversion, and supermajority protections. We again replicate these findings using the time-to-resolve-contract-dispute variable as an alterna- tive proxy for the quality of contractual enforcement. We see that dispute resolution time is most strongly related to the use of supermajority provisions. IV.D. Correlation of Different Contract Parts So far, we have analyzed each of the contractual features in isolation. We now want to understand whether the different contract features (security structure, ownership stakes, and other control provisions) are used as complements or substitutes in financial contracting. To undertake this analysis, we regress each of the contract provisions of interest on each other, as well as controls for the logarithm of gross national product and dummy variables for the year, industry, and deal type. We find in Table VI a strong negative correlation between common stock and convertible preferred stock. Moreover, pre- 8. We identify nineteen different types of provisions in these agreements. We score each of these clauses from zero to three, with a higher score representing a more stringent supermajority clause. Instead of using a simple sum of the scores, we also conducted a principal component analysis. Our results are very similar. LEGAL ENFORCEMENT TABLE VI CORRELATION IN THE USE OF CONTRACTING TOOLS OF PRIVATE EQUITY CONTRACTS The sample consists of 210 investments in developing countries by private equity groups (PEGs). We regress the contract provision at the top of the column on the provisions at the beginning of each row. Each cell contains the coefficients from separate regressions of the contract provisions on the right-hand-side variables (standard errors are reported in brackets). We control for log of gross national product and year, industry, deal type dummies. All variables are defined as before. Chapter PEG AND Preferred Automatic equity Debt Common stock stock Antidilution conversion stake FINANCIAL Three Common stock 0.08 [0.04]*** Preferred stock 0.02 [0.04] 0.21 [0.07]*** Antidilution 0.01 [0.04] 0.25 [0.07]*** 0.16 [0.07]*** Automatic conversion 0.09 [0.05]** 0.50 [0.07]*** 0.34 [0.08]*** 0.43 [0.07]*** TRANSACTIONS PEG maximum equity stake 0.22 [0.09]*** 0.02 [0.14] 0.20 [0.16] 0.18 [0.16] 0.07 [0.17] Board size 0.03 [0.08] 0.04 [0.15] 0.38 [0.15]*** 0.16 [0.17] 0.06 [0.14] 0.10 [0.12] 239 99 100 A Reader in International Corporate Finance 240 QUARTERLY JOURNAL OF ECONOMICS ferred stock offerings are more likely to employ other protections such as antidilution and automatic conversion terms, while these provisions are negatively associated with common stock. We also find a strong positive correlation between the maximum owner- ship stakes that the private equity group obtains and the use of debt. The correlations between the minimum ownership stake and the use of debt and between board size and preferred stock are significantly positive. Overall, these results suggest that contracts differ systemat- ically in the way they aim to provide investors with control rights. Preferred security structures and control provisions such as an- tidilution clauses are generally used as complements. Deals with common shares and debtlike securities rely more heavily on con- trolling ownership stakes rather than other control provisions. Taken together, these results suggest that private equity groups rely on either (a) protection of minority shareholders through detailed specification of behavior that is ruled out or (b) control through ownership of a majority of the common stock and board dominance. IV.E. Consequences A natural question, suggested by La Porta et al. [2002], relates to the consequences of these investment choices. We would like to examine this question by looking at the relation- ship between transaction structures and investment outcomes. Given the relative recentness of most of the investments, and the difficulties that investors have recently had in exiting developing country investments, such an analysis would be premature. We focus instead on two proxies: valuations and fund returns. When we look at the valuations of the financings in Table VII, we see that investments in common law countries and those with quick dispute resolution have higher valuations. These results hold even after controlling for the size of the firm, measured by sales in the year of the investment. These find- ings suggest that the differences in legal regime affect not just the structure of transactions, but also have real effects on firms' valuations.9 9. Similarly, we observe that the amount of capital invested is larger in common law countries than civil law countries holding constant firm size. Our Chapter Three 101 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 241 TABLE VII FINANCING VALUATION IN DEVELOPING COUNTRY PRIVATE EQUITY TRANSACTIONS The sample consists of 210 investments in developing countries by private equity groups (PEGs). The dependent variable is the logarithm of the implied "postmoney" valuation of the transaction. Independent variables include dummy variables denoting nations with British or socialist legal origin (French legal origin is the omitted category) and the time to resolve commercial disputes in that nation. GNP per capita is the per capita gross national product of the country averaged over the 1990s. Sales is a control for the size of the firm: the annual sales in the year the investment was made (in 1997 dollars). All regressions employ ordinary least squares specifications. Standard errors are clustered at the private equity group. Implied valuation British legal origins 0.75 *[0.42] Socialist legal origins 1.62 ***[0.43] Dispute time 0.49 *[0.30] GNP per capita 0.27 0.43 [0.25] [0.28] Sales 0.15 0.19 ***[0.06] ***[0.07] Industry dummies Y Y Deal type dummies Y Y Year dummies Y Y Number of observations 193 193 Adjusted R2 0.26 0.18 * Significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level. We also examine the overall returns of funds that are active in developing countries. We use Private Equity Intelli- gence's 2004 Private Equity Performance Monitor, which has data on over 1700 private equity funds (for more details see Lerner, Schoar, and Wong [2004]). We examine all listed funds active primarily in developing countries of a certain type, e.g., excluding funds active in both common and civil law developing countries. Private equity funds that were active in common law developing nations had an average return multiple 19 percent interpretation of these results must be cautious since we only observe realized transactions. Investments that are completed in noncommon law countries, de- spite the many difficulties there, might be particularly promising. Thus, there may not be as many differences in the intensive margin, i.e., the observed amount of financing, as along the extensive margin (the number and types of deals that are done). Since we cannot construct an exhaustive sample of transactions, it is very difficult to draw any conclusions about the extensive margin. 102 A Reader in International Corporate Finance 242 QUARTERLY JOURNAL OF ECONOMICS better than the typical fund established in that subclass and that year, while those in socialist and civil law countries had a multiple 49 percent worse than the benchmark (significantly different at the 1 percent confidence level).10 It must be ac- knowledged that we can analyze only the investors' (private) returns, not the returns to society as a whole. We anticipate, however, that the two measures should be correlated: for ex- ample, there are unlikely to be many social returns from a liquidated company. We hope to explore this question in future work. V. CONCLUSIONS This paper seeks to understand how differences in the en- forcement of commercial laws, measured directly as well as through legal origin, affect financial contracting. We focus on a well-documented and reasonably systematized set of transac- tions, private equity investments. We find that investments in nations with effective legal enforcement are more likely to employ preferred stock and to have more contractual protections for the private equity group, such as supermajority voting rights and antidilution provisions. By way of contrast, contracts in low en- forcement countries tend to rely more heavily on common stock (or even debt) and control the firm via majority ownership and board dominance. Relying on ownership as opposed to contractual protections seems to be only a partial remedy: these investments have lower valuations and returns. The results suggest the importance of a contractual channel between legal enforcement and financial transactions. The legal system appears to profoundly shape the transactions into which private equity groups enter, and efforts to address this problem by relying on ownership rather than contractual protections are only partially successful. Exploring this channel outside of pri- vate equity would be a natural next step. 10. The return multiple is the ratio of the value of distributed invest- ments and undistributed holdings to their cost. These results are also robust to using internal rates of return: the adjusted IRRs are 2.6 percent and 22.6 percent, respectively (significantly different at the 5 percent confidence level). Chapter Three 103 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 243 APPENDIX: KEY LEGAL PROVISIONS AFFECTING PRIVATE EQUITY INVESTORS IN NINE NATIONS MOST FREQUENTLY REPRESENTED IN THE SAMPLE Class of limitation Argentina Brazil Hong Kong Security Type No restrictions, but No restrictions. No restrictions. preferred stock can only have same vote as common stock. Also possible to have common stock with enhanced voting rights (up to 5 votes). Super-Majority No restrictions. No restrictions. No restrictions. Provisions Many corporate events require approval of 75% of shareholders. Management No restrictions. Limitations on types of No restrictions, Equity Ambiguities firms who can issue except that Holdings surround tax stock options. shareholders in treatment of Special disclosure private firms options. requirements for must first offer option-issuing firms. shares to other Disadvantageous tax investors. treatment of options. Reinvestment Equity holders can Equity holders can Equity holders and maintain pro rata maintain pro rata can maintain Antidilution share. Provision share. Restrictions pro rata share. Provisions can be waived with on unreasonably shareholder vote. dilutive financings. Domiciling Entity Could be domiciled Can be domiciled Can be domiciled overseas until overseas, but may be overseas. recently. Now more difficult to substantial enforce corporate difficulties to do so. rights locally. (continued on next page) 104 A Reader in International Corporate Finance 244 QUARTERLY JOURNAL OF ECONOMICS APPENDIX (CONTINUED) India Mexico People's Republic of China Preferred stocks cannot No restrictions, but Most domestic and foreign have any voting some limitations private equity rights, except in on voting rights investments must special of preferred employ common stock- circumstances. shareholders. like structure. Some Limits on extent of large investments may returns preferred use other securities, but shareholders can must receive enjoy. authorities' permission first. No restrictions. Some No restrictions. No restrictions. Some corporate events Some legal corporate events must require approval of protections for have 2/3rds approval by 75% of shareholders. minority investors. For foreign shareholders investments, decisions (e.g., right to must be approved by name at least 2/3rds of directors in one director). many cases. No restrictions on No restrictions. For most investments, not private firms. possible to issue equity to management. May be allowed in certain very large investments, but permission of authorities may be required. Equity holders can Equity holders can Equity holders have maintain pro rata maintain pro preemptive right to share. Provision can rata share. purchase shares, except be waived with Provision can be for certain very large shareholder vote. waived with investments. shareholder vote. Can be domiciled Can be domiciled Cannot be domiciled overseas. overseas. overseas. Chapter Three 105 LEGAL ENFORCEMENT AND FINANCIAL TRANSACTIONS 245 Poland Republic of Korea Romania No restrictions, No restrictions, but No restrictions, but but limitations only common investors cannot on voting (no stock had voting require that more than 2­3 rights until late classes of common stock), 1990s. Now, no shareholders dividend, and restrictions. vote as a block. liquidation preference rights of preferred shareholders. No restrictions. No restrictions. No restrictions. Some corporate events must have 75% approval by investors. No restrictions. No restrictions. No restrictions. Equity holders can Equity holders Equity holders maintain pro have preemptive have preemptive rata share. right to purchase right to purchase Provision can be shares, with shares, except for waived with limited some private 80% shareholder exceptions. firms. vote. Can be domiciled Can be domiciled These restrictions overseas. overseas. May cannot be entail loss of avoided by attractive tax domiciling incentives for company in startups. another country. 106 A Reader in International Corporate Finance 246 QUARTERLY JOURNAL OF ECONOMICS HARVARD UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH MASSACHUSETTS INSTITUTE OF TECHNOLOGY AND NATIONAL BUREAU OF ECONOMIC RESEARCH REFERENCES Aghion, Philippe, and Patrick Bolton, "An Incomplete Contracts Approach to Financial Contracting," Review of Economic Studies, LIX (1992), 473­494. Bartlett, Joseph W., Equity Finance: Venture Capital, Buyouts, Restructurings, and Reorganizations, 2nd edition (New York, NY: John Wiley, 1995). Bergman, Nittai, and Daniel Nicolaievsky, "Investor Protection and the Coasian View," unpublished working paper, Harvard University, 2003. Cumming, Douglas J., and Jeffrey G. MacIntosh, "A Law and Finance Analysis of Venture Capital Exits in Emerging Markets," unpublished working paper, University of Alberta, 2002. Demirgu¨c-Kunt, Asli, and Ross Levine, Financial Structure and Economic Growth: A Cross-Country Comparison of Banks, Markets, and Development (Cambridge, MA: MIT Press, 2001). Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shlei- fer, "Courts," Quarterly Journal of Economics, CXVIII (2003), 453­517. European Venture Capital Association, EVCA Yearbook (Zaventum, Belgium: EVCA and KPMG, 2002). Glaeser, Edward L., Simon Johnson, and Andrei Shleifer, "Coase versus the Coasians," Quarterly Journal of Economics, CXVI (2001), 853­899. Gompers, Paul A., "An Examination of Convertible Securities in Venture Capital Investments," unpublished working paper, Harvard University, 1998. Grossman, Sanford J., and Oliver D. Hart, "The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration," Journal of Political Economy, XCIV (1986), 691­719. Hart, Oliver, and John Moore, "Property Rights and the Nature of the Firm," Journal of Political Economy, XCVIII (1990), 1119­1158. Hellmann, Thomas, "The Allocation of Control Rights in Venture Capital Con- tracts," Rand Journal of Economics, XXIX (1998), 57­76. Holmstro¨m, Bengt, "Moral Hazard and Observability," Bell Journal of Economics, X (1979), 74­91. Johnson, Simon, John McMillan, and Christopher Woodruff, "Courts and Rela- tional Contracts," Journal of Law, Economics and Organization, XVIII (2002), 221­277. Kaplan, Steven N., Frederic Martel, and Per Stro¨mberg, "How Do Legal Differ- ences and Learning Affect Financial Contracts?" Working Paper No. 10097, National Bureau of Economic Research, 2003. Kaplan, Steven N., and Per Stro¨mberg, "Financial Contracting Meets the Real World: An Empirical Analysis of Venture Capital Contracts," Review of Eco- nomic Studies, LXX (2003), 281­316. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, "Legal Determinants of External Finance," Journal of Finance, LII (1997), 1131­1150. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, "Law and Finance," Journal of Political Economy, CVI (1998), 1133­1155. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, "Investor Protection and Corporate Valuation," Journal of Finance, LVII (2002), 1147­1170. Lerner, Josh, Antoinette Schoar, and Wan Wong, "Smart Institutions, Foolish Choices?: The Limited Partner Performance Puzzle," unpublished working paper, Harvard University and MIT, 2004. Qian, Jun, and Philip Strahan, "How Law and Institutions Shape Financial Contracts: The Case of Bank Loans," unpublished working paper, Boston College, 2004. Chapter Four 107 THE JOURNAL OF FINANCE · VOL. LVII, NO. 6 · DECEMBER 2002 Disentangling the Incentive and Entrenchment Effects of Large Shareholdings STIJN CLAESSENS, SIMEON DJANKOV, JOSEPH P. H. FAN, and LARRY H. P. LANG* ABSTRACT This article disentangles the incentive and entrenchment effects of large owner- ship. Using data for 1,301 publicly traded corporations in eight East Asian econ- omies, we find that firm value increases with the cash-flow ownership of the largest shareholder, consistent with a positive incentive effect. But firm value falls when the control rights of the largest shareholder exceed its cash-flow ownership, con- sistent with an entrenchment effect. Given that concentrated corporate ownership is predominant in most countries, these findings have relevance for corporate gov- ernance across the world. THE EFFECTS OF OWNERSHIP STRUCTURES on the value of firms have been re- searched extensively, with the role of large investors receiving special atten- tion. Investors with large ownership stakes have strong incentives to maximize their firms' value and are able to collect information and oversee managers, and so can help overcome one of the principal­agent problems in the modern corporation--that of conflicts of interest between shareholders and man- agers Jensen and Meckling 1976 . Large shareholders also have strong incentives to put pressure on managers or even to oust them through a proxy fight or a takeover. For example, Shleifer and Vishny 1997, p. 754 point * University of Amsterdam and Centre for Economic Policy Research; World Bank and Centre for Economic Policy Research; Hong Kong University of Science and Technology; and Chinese University of Hong Kong, respectively. Joseph P. H. Fan gratefully acknowledges the Hong Kong Government's Earmarked Grant for research support. Larry H. P. Lang gratefully ac- knowledges the financial support of the Hong Kong Government's Earmarked Grant and Direct Grant. The authors are grateful for the helpful comments of Lucian Bebchuk, Erik Berglof, Alexander Dyck, Caroline Freund, Ed Glaeser, Simon Johnson, Tarun Khanna, Florencio Lopez- de-Silanes, Randall Morck, Tatiana Nenova, Raghuram Rajan, Henri Servaes, Daniel Wolfen- zon, and Luigi Zingales, the article's two anonymous referees, seminar participants at the World Bank, International Monetary Fund, Federation of Thai Industries, Georgetown Univer- sity, George Washington University, Hong Kong University of Science and Technology, Korean Development Institute, Korea Institute of Finance, Vanderbilt University, University of Illinois, University of Michigan, University of Amsterdam, 1999 National Bureau for Economic Re- search summer conference on corporate finance, 2000 American Economic Association annual meetings, and especially of Rafael La Porta, Andrei Shleifer, and René Stulz. An earlier version of this article was called "Expropriation of Minority Shareholders: Evidence from East Asia." The opinions expressed here do not necessarily reflect those of the World Bank. 2741 108 A Reader in International Corporate Finance 2742 The Journal of Finance out, "Large shareholders thus address the agency problem in that they have both a general interest in profit maximization, and enough control over the assets of the firm to have their interest respected." Less work has been done on the costs--in terms of lower firm valuation-- associated with the presence of large investors. Again, according to Shleifer and Vishny 1997, p. 758 , "Large investors may represent their own inter- ests, which need not coincide with the interests of other investors in the firm, or with the interests of employees and managers." Empirically, Morck, Shleifer, and Vishny 1988 find an inverse U-shaped relationship between managerial equity ownership and firm valuation for a sample of U.S. firms. One interpretation is that firms' performance improves with higher mana- gerial ownership, but that, after a point, managers become entrenched and pursue private benefits at the expense of outside investors. The costs of large shareholdings and entrenchment are formalized in the model of Stulz 1988 , which predicts a concave relationship between man- agerial ownership and firm value. In the model, as managerial ownership and control increase, the negative effect on firm value associated with the entrenchment of manager-owners starts to exceed the incentive benefits of managerial ownership. In that model, the entrenchment costs of manager ownership relate to managers' ability to block value-enhancing takeovers. McConnell and Servaes 1990 provide empirical support for this relation- ship for U.S. firms. But ownership structures exhibit relatively little concentration in the United States. Elsewhere, most firms are predominantly controlled by a single large shareholder La Porta, Lopez-de-Silanes, and Shleifer 1999 . Thus, study- ing non-U.S. firms can provide evidence about the effects of large share- holders that is difficult to detect in U.S. data. Moreover, the literature indicates that the positive incentive effect relates to the share of cash-flow rights held by large shareholders and that the negative entrenchment effect relates to the share of control rights held by large shareholders. Non-U.S. firms ex- hibit far more divergence between cash-flow rights and control rights than do U.S. firms, because in most countries, the largest shareholder often estab- lishes control over a firm despite little cash-flow rights. Using a sample of corporations outside the United States, we are thus better able to disentan- gle the incentive and entrenchment effects of large ownership that are so difficult to tell apart in U.S. data. To do so, we investigate the valuation of publicly traded East Asian cor- porations relative to their ownership structures. In previous work, we found that more than two-thirds of East Asian firms are controlled by a single shareholder Claessens, Djankov, and Lang 2000 . East Asian firms also show a sharp divergence between cash-flow rights and control rights--that is, the largest shareholder is often able to control a firm's operations with a relatively small direct stake in its cash-flow rights. Control is often en- hanced beyond ownership stakes through pyramid structures and cross- holdings among firms, and sometimes through dual-class shares, with the divergence between cash-flow rights and control rights most pronounced in Chapter Four 109 Incentive and Entrenchment Effects of Large Shareholdings 2743 family-controlled firms.1 Finally, managers of East Asian corporations are usually related to the family of the controlling shareholder. Thus, it is pos- sible to analyze the relative importance of incentive and entrenchment ef- fects in East Asian corporations, because ownership is highly concentrated and the divergence between cash-flow rights and control rights is large, while manager-owner conflicts are generally limited. Our analysis uses data for 1,301 publicly traded corporations from eight East Asian economies: Hong Kong, Indonesia, South Korea, Malaysia, the Philippines, Singapore, Taiwan, and Thailand. Using regression techniques, we find that relative firm value--as measured by the market-to-book ratio of assets--increases with the share of cash-flow rights in the hands of the largest shareholder. This result is consistent with previous studies on the pos- itive incentive effects associated with increased cash-flow rights in the hands of one or a few shareholders. But we find that the entrenchment effect of control rights has a negative effect on firm value. This finding complements that of Morck, Stangeland, and Yeung 2000 . Using data for Canadian pub- lic corporations, they show that concentrated corporate control impedes growth, because entrenched controlling shareholders have a vested interest in pre- serving the value of existing capital. Our work also complements that of La Porta et al. 2002 , who document lower valuations for firms in countries with worse protection of minority shareholders. Such countries tend to have more concentrated ownership structures. Our results also support the predictions of theoretical studies that inves- tigate the effects on firm value of the separation of cash-flow rights and control rights. Grossman and Hart 1988 and Harris and Raviv 1988 show that separating ownership and control can lower shareholders' value and may not be socially optimal. Shleifer and Vishny 1997, p. 759 argue that "as ownership gets beyond a certain point, large owners gain nearly full control of the company and are wealthy enough to prefer to use firms to generate private benefits of control that are not shared by minority share- holders." Bebchuk, Kraakman, and Triantis 2000 argue that separating control rights from cash-flow rights can create agency costs an order of mag- nitude larger than the costs associated with a controlling shareholder who also has a majority of the cash-flow rights in his or her corporation. In this article, we show that, for the largest shareholders, the difference between control rights and cash-flow rights is associated with a value dis- count and that the discount generally increases with the size of the wedge between control rights and cash-flow rights. We do not have strong evidence on which mechanism separating ownership and control is associated with the value discounts. Pyramid schemes, cross-holdings among firms, and the 1Pyramiding is defined as the ultimate ownership of a firm running through a chain of ownership of intermediate corporations. Cross-holdings refer to horizontal and vertical owner- ship links among corporations that can enhance the control of a large, ultimate shareholder. Dual-class shares refer to shares with different voting rights. 110 A Reader in International Corporate Finance 2744 The Journal of Finance issuance of dual-class shares are all associated with lower corporate valua- tion, but none of the associations is individually statistically significant. Finally, we investigate whether a certain type of owner--families, the state, or widely held corporations and widely held financial institutions--drives our results. We find that concentrated ownership in the hands of all types of owners is associated with a higher market-to-book ratio. We also find that the wedge between control and ownership is associated with value discounts for family-controlled firms and somewhat for state-controlled corporations, but not significantly when the principal owner is a widely held corporation or financial institution. The differences in valuation effects by type of owner could arise from the fact that managers at firms owned by widely held cor- porations and financial institutions have fewer ways to divert benefits to themselves compared with managers at firms owned by families and the state. The rest of the paper is structured as follows. Section I describes the se- lection criteria for the data sample and the construction of the industry origin, ownership, control, and corporate valuation variables. Section II in- vestigates the evidence on the incentive and entrenchment effects of large shareholdings and conducts some robustness tests. Section III studies the effects of various mechanisms used for the separation of ownership and con- trol, and the relation between the type of ownership and corporate valua- tion. Section IV concludes. I. Sample Selection and Data This section describes the selection criteria used and the resulting sample of corporations. It also provides details on the construction of the data on ownership and control structures and provides statistics on key variables for the sample. Finally, it describes the valuation measure used for the empir- ical tests that follow. A. Sample Selection Our starting point for the data is Claessens et al. 2000 , who collected 1996 data on ownership for corporations in Hong Kong, Indonesia, Japan, Korea, Malaysia, the Philippines, Singapore, Taiwan, and Thailand. Their main source was Worldscope, supplemented by other sources that provide ownership structures as of December 1996 or the end of fiscal 1996. From a complete sample of 5,284 publicly listed corporations in the nine East Asian economies, ownership data were collected for 2,980 firms. For this analysis, we take a subset of these firms. First, we exclude from the sample all Japanese corporations. We do so for several reasons. World- scope provides data on 1,740 publicly listed Japanese corporations, and Jap- anese corporations also dominate the sample for which we have ownership data 1,240 of 2,980 corporations . Thus, Japanese firms could influence the results too much. An unbalanced outcome is even more likely given the fea- Chapter Four 111 Incentive and Entrenchment Effects of Large Shareholdings 2745 tures of Japanese firms--most have dispersed ownership structures, and ownership and management are separated far more often than in other East Asian economies. The most important shareholders in Japan are widely held financial institutions, again unlike many economies in the region. But these financial institutions and their affiliated firms often work together to in- fluence the governance of the owned corporations, a phenomenon that can- not be captured by formal ownership data. Thus, including Japan in our set of East Asian economies would be less useful for disentangling the incentive and entrenchment effects of concentrated ownership and control. Second, we exclude firms that operate in certain industrial sectors-- specifically, financial corporations and regulated utilities. For financial firms, profitability and valuation data are difficult to calculate and to compare with firms in other sectors. For regulated utilities, profitability and valua- tion can be strongly influenced by government regulations. To determine the primary industry in which each firm operates, we rely on historical segment sales data from Worldscope. If such information is not provided, we rely on information from the Asian Company Handbook 1998 .2 We next determine the sector to which each firm belongs according to the two-digit Standard Industrial Classification SIC system, using the largest share of sales rev- enue among the firm's activity in each sector. We then use Campbell 1996 to classify firms into 11 industries.3 We exclude all financial corporations SIC 6000­6999 and regulated utilities SIC 4900­4999 , making for 304 corporations excluded using those criteria. Third, we need to know whether a firm consolidates its financial state- ments and, if so, the method used, because our valuation measure can be distorted by accounting rules on consolidation.4 Specifically, excessive con- solidation of sales and balance sheet items can result when partly owned subsidiaries are treated like fully owned subsidiaries--the full method of consolidation. This method tends to understate the true market-to-book ratio of the consolidated corporation because the book value includes 100 percent of the assets of the subsidiaries, while the market value includes only the actual stakes owned. The market-to-book ratio of the consolidated corpora- tion is not distorted when the corporation uses cost, proportional, or equity consolidation methods. Under these methods, the parent corporation in- cludes its prorated share of subsidiaries in its balance sheet as well as any dividends received from subsidiaries in its income statement . Accordingly, 2 We still had to exclude 53 firms that do not report their segment sales to Worldscope or the Asian Company Handbook. 3 The industries are petroleum SIC 13, 29 , consumer durables SIC 25, 30, 36, 37, 50, 55, 57 , basic industry SIC 10, 12, 14, 24, 26, 28, 33 , food and tobacco SIC 1, 2, 9, 20, 21, 54 , construction SIC 15, 16, 17, 32, 52 , capital goods SIC 34, 35, 38 , transportation SIC 40, 41, 42, 44, 45, 47 , unregulated utilities SIC 46, 48 , textiles and trade SIC 22, 23, 31, 51, 53, 56, 59 , services SIC 72, 73, 75, 76, 80, 82, 87, 89 , and leisure SIC 27, 58, 70, 78, 79 . 4 La Porta et al. 2000 further discuss the biases resulting from different consolidation methods. 112 A Reader in International Corporate Finance 2746 The Journal of Finance these methods do not distort balance sheet items and so do not understate the market-to-book ratio. Worldscope almost always says whether a firm consolidates its financial statements. When Worldscope does not report that information, we exclude the corporation--making for 82 dropped corporations. More than two-thirds of the remaining corporations have consolidated financial statements.5 World- scope also indicates whether the consolidation covers all significant sub- sidiaries and whether the annual report is on a cost basis unconsolidated . But Worldscope does not indicate at what level the corporation has done the consolidation, and in particular, whether partly owned subsidiaries are treated as fully owned subsidiaries. Lacking that information, we cannot investigate whether the consolidation method used affects the firm valuation. We can only investigate whether the fact that the corporation consolidates or not affects our results. These sample selection criteria leave us with 1,301 corporations in eight East Asian economies--about 37 percent of the sample of 3,544 publicly traded corporations in these economies. B. Ownership and Control Definitions Following La Porta et al. 1999 , we analyze ultimate ownership and con- trol patterns. In most cases, the immediate shareholders of a corporation are corporate entities, nonprofit foundations, or financial institutions. We then identify their owners, the owners of those owners, and so on. We do not consider ownership by individual family members to be separate, and we use total ownership by each family group--defined as a group of people related by blood or marriage--as the unit of analysis. Studying the separation of ownership and control requires data on both cash-flow rights and control rights, which we calculate using the complete chain of ownership. Suppose that a family owns 11 percent of the stock of publicly traded firm A, which in turn has 21 percent of the stock of firm B. We then say that the family controls 11 percent of firm B--the weakest link in the chain of control rights. In contrast, we say that the family owns about 2 percent of the cash-flow rights of firm B, the product of the two ownership stakes along the chain. We make the distinction between cash-flow rights and control rights by using for each firm information on pyramid structures, cross-holdings among firms, and dual-class shares. To determine effective control at any intermediate levels as well as the ultimate level, we need to use a cutoff point above which we assume that the largest shareholder has effective control over the intermediate and final corporations. We use 10 per- cent as the cutoff point in our empirical analysis because that level is com- 5 That number is highest for Hong Kong, Malaysia, and Singapore, where 76, 75, and 75 per- cent of corporations use consolidated accounts, respectively. In contrast, only 34 percent of Korean corporations have consolidated accounts, 51 percent of Indonesian corporations, and 57 percent of Taiwanese corporations. Chapter Four 113 Incentive and Entrenchment Effects of Large Shareholdings 2747 monly used by other studies. But we also provide information using the 20 percent and 40 percent levels, to show the distributions of large owner- ship across economies and types of owners. Information on pyramid structures and cross-holdings among firms is limited because our data cover only listed corporations. Many East Asian corporations affiliated with business groups, and hence with pyramid struc- tures and cross-holdings, are unlisted. At the end of 1996, for example, the three biggest business groups in Korea--Hyundai, Samsung, and LuckyGoldstar--had 46, 55, and 48 affiliated firms, respectively. Of those, only 16, 14, and 11 were publicly listed. Covering only listed corporations may create a bias in terms of ownership structures and firm valuation. Unlisted corporations could have direct and indirect ownership links with listed corporations, resulting in a possible underreporting of our measures for ultimate control and ownership, since we assume that someone other than a related shareholder controls the unlisted corporations. Anecdotal evidence suggests that such underreporting can lead to considerable under- estimates.6 In addition, complex ownership structures and group-affiliated corporations presumably increase opportunities for the entrenchment of large shareholders--even where ownership structures are similar to those of independent corporations. Because we likely underestimate the ultimate ownership and influence of large shareholders for group-affiliated firms, we may underestimate the effect of ownership structures on firm valuation. But group affiliation may also affect firm valuation, because there may be intragroup financial trans- fers that are not market based. The direction of the effect on firm value is unclear. Firm valuations for group-affiliated firms could be lower or higher than for comparable independent firms, depending on the net costs they incur or the net benefits they receive from group affiliation. We control for some firm-specific factors, such as age and size, that may be correlated with the possible net costs or benefits from group affiliation. But these factors likely do not fully control for the influence on firm value of affiliation with specific groups. Thus, we account for the possibility that the valuations of group-affiliated firms are not independent of each other by running regres- sions in which all firms in a business group are considered jointly.7 In terms of dual-class shares, the financial information service Data- stream provides data on all classes of listed shares. For the firms under investigation, 88 cases of dual-class shares are found. Of those, some pre- ferred shares are more like debt instruments because they are redeemable 6 Some Korean firms are illustrative. Samsung Corporation, part of the Samsung chaebol, is partly owned by Samsung Life Insurance, which is not listed. But Samsung Life Insurance is controlled by the same family that has a large direct stake in Samsung Corporation, increasing the family's overall control stake in Samsung Corporation. Similarly, control for Samsung Elec- tromagnetic is underestimated because it is also partly owned by Samsung Life Insurance as well as other Samsung corporations . 7 Still, not being able to cover unlisted firms in a group does not allow us to fully investigate the effect on firm value of variables like the size of business groups. 114 A Reader in International Corporate Finance 2748 The Journal of Finance or callable at the option of the corporation at a preset price, are convertible into common shares, or receive a fixed cumulative dividend unrelated to the profits of the corporation. We consider such preferred shares to be debt-like instruments and do not include them as shares that further separate own- ership and control. Following this methodology, we end up with 43 corpora- tions with dual-class shares--5 in Hong Kong, 37 in Korea, and 1 in the Philippines. Dual-class shares are now legally forbidden in Hong Kong and Singapore, but the corporations in the Hong Kong sample are protected by a grandfather clause. In Indonesia, Malaysia, Taiwan, and Thailand, dual- class shares could exist in principle, but Datastream covers none. C. Sample Characteristics The number of corporations for each economy is shown in Table I. Korea has the largest share of corporations in the sample, 21.6 percent, followed by Hong Kong with 17.3 percent. The Philippine sample is the smallest, ac- counting for 5.9 percent of the corporations. About 20 percent of the corpo- rations in our sample are in the consumer durables industry. Corporations in basic industry, construction, and textiles and trade each account for about 13 percent of the sample. Petroleum companies and unregulated utilities make up the smallest number of corporations in our sample. In terms of ownership structure, we define corporations as being widely held or having large ultimate owners. We apply the commonly used defini- tion of a widely held corporation as one that does not have any owner with 10 percent or more of control rights. Ultimate owners are split into three groups: families, including all related individuals with large stakes; the state or municipality; and the combined group of widely held corporations and widely held financial institutions, such as banks and insurance companies. Ownership types are used in some of the regressions below to investigate whether any of the effects differ by type of owner. We start by reporting aggregate data on the distribution of ultimate con- trol by ownership type Table II . Only four percent of corporations do not have a single controlling shareholder at the 10 percent cutoff level of control rights. Table II also shows ultimate ownership structures at the 20 and 40 per- cent cutoff levels for the share of control rights in the hands of the largest shareholder though these higher cutoff levels are not used in our empirical analysis . These higher cutoff levels show how concentrated ownership struc- tures are. At the 20 percent cutoff level, 18 percent of corporations are widely held. In contrast, 77 percent are widely held at the 40 percent cutoff level-- indicating that in many corporations, the largest shareholder has a control stake of less than 40 percent. At lower control levels, families are the largest shareholders, covering more than two-thirds of corporations at the 10 per- cent cutoff level and three-fifths at the 20 percent level. At the 10 percent cutoff, corporate sectors do not differ much in terms of ownership patterns across the eight economies. The exception is Korea, which has a larger share--13 percent--of widely held corporations. More Incentive Table I The Sample of Publicly Traded East Asian Corporations by Economy and Industry This table shows the distribution of sample corporations across industries and economies. The source of the data is Claessens et al. 2000 , Worldscope, and Asian Company Handbook 1998 . The industrial classification is based on Campbell 1996 . Industries are defined as follows: and petroleum SIC 13, 29 , consumer durables SIC 25, 30, 36, 37, 50, 55, 57 , basic industry SIC 10, 12, 14, 24, 26, 28, 33 , food and tobacco SIC 1, 2, 9, 20, 21, 54 , construction SIC 15, 16, 17, 32, 52 , capital goods SIC 34, 35, 38 , transportation SIC 40, 41, 42, 44, 45, 47 , unregulated Entrenchment utilities SIC 46, 48 , textiles and trade SIC 22, 23, 31, 51, 53, 56, 59 , services SIC 72, 73, 75, 76, 80, 82, 87, 89 , and leisure SIC 27, 58, 70, 78, 79 . The sample excludes financial companies SIC 60-69 and regulated utilities SIC 49 . Total Hong Korea, Percentage Industry Kong Indonesia Rep. of Malaysia Philippines Singapore Taiwan Thailand Number of Total Effects Petroleum 1 1 12 4 6 3 1 1 29 2.2 Chapter Consumer durables 57 17 59 17 7 44 29 29 259 19.9 Basic industry 10 24 55 22 14 16 24 12 177 13.6 of Food and tobacco 13 21 20 17 18 18 15 11 133 10.2 Large Construction 22 4 44 49 11 14 14 16 174 13.4 Four Capital goods 22 12 35 8 3 21 16 6 123 9.5 Transportation 19 4 6 10 1 12 6 5 63 4.8 Shareholdings Unregulated utilities 5 5 3 3 6 4 1 5 32 2.5 Textiles and trade 43 33 35 15 6 9 17 10 168 12.9 Services 7 7 4 15 3 15 4 7 62 4.8 Leisure 26 4 8 11 2 20 2 8 81 6.2 Total 225 132 281 171 77 176 129 110 1,301 100.0 Percentage of total 17.3 10.1 21.6 13.1 5.9 13.5 9.9 8.5 100.0 2749 115 116 A Reader in International Corporate Finance 2750 The Journal of Finance Table II Control of East Asian Corporations by Owner Type and Economy, 1996 (Percentage of Corporations in the Sample) Data for 1,301 publicly traded corporations excluding financial institutions, SIC 60­69, and regulated utilities, SIC 49 , based on Worldscope, supplemented by information from the Asian Company Handbook 1998 . All data are as of December 1996 or the end of fiscal 1996. To determine effective control at any intermediate as well as ultimate level, a cutoff level of 10 percent was used in all empirical analyses. Above that level, the largest shareholder is assumed to have effective control over the intermediate or final corporation. The 20 percent and 40 percent cutoff levels are also used here to show the distribution of large ownership across economies and owner types. The percentages in the last four columns sum to 100, subject to rounding. Percentage of Firms with Ultimate Control Owned by a Percentage Widely Held Number of Firms with Corporation of Firms Dispersed Family- State- or Financial Economy in Sample Control owned owned Institution 10 percent cutoff for effective control of the largest shareholder Hong Kong 225 0 72 3 24 Indonesia 132 1 73 9 17 Korea, Rep. of 281 13 73 2 12 Malaysia 171 1 75 12 12 Philippines 77 4 51 3 43 Singapore 176 1 55 29 15 Taiwan 129 5 59 2 35 Thailand 110 1 72 5 21 Total 1,301 4 68 8 20 20 percent cutoff for effective control of the largest shareholder Hong Kong 225 8 69 1 23 Indonesia 132 6 70 8 16 Korea, Rep. of 281 41 52 0 7 Malaysia 171 11 70 11 9 Philippines 77 19 45 1 34 Singapore 176 9 53 24 14 Taiwan 129 29 47 1 24 Thailand 110 6 68 5 20 Total 1,301 18 60 6 16 40 percent cutoff for effective control of the largest shareholder Hong Kong 225 72 20 0 8 Indonesia 132 50 35 5 10 Korea, Rep. of 281 94 5 0 1 Malaysia 171 80 13 2 5 Philippines 77 83 8 1 8 Singapore 176 71 17 5 8 Taiwan 129 93 5 1 1 Thailand 110 53 35 4 8 Total 1,301 77 16 2 5 Chapter Four 117 Incentive and Entrenchment Effects of Large Shareholdings 2751 pronounced differences emerge at the 20 percent cutoff. In Korea, 41 per- cent of corporations are widely held, while in Indonesia and Thailand only 6 percent of corporations fall into that category, indicating that ownership structures are much more concentrated in Indonesia and Thailand. State control is high in Singapore, at 24 percent, while control by widely held corporations and financial institutions is important in the Philippines, at 34 percent. At the 40 percent cutoff, differences become smaller across economies in terms of type of controlling shareholder except in Indonesia and Thailand, where families still control more than one-third of the sam- ple corporations . D. The Valuation Measure As noted, we use the market-to-book ratio of assets to measure firm val- uation. Researchers have used the market-to-book ratio as well as Tobin's Q to measure variations in market values resulting from different ownership structures. Market value is defined here as the sum of the market value of common stock and the book value of debt and preferred stock. To calculate the value of equity, we use end-1996 shares of common stock and stock prices, both from Worldscope. We do not try to calculate the replacement cost of assets in the denominator, as we would need to do if we were using Tobin's Q, for two reasons. Most important, the data required to calculate replace- ment values are generally not available, and the eight economies have dif- ferent ways of accounting for depreciation of physical assets. In addition, we did not want to impose a fixed depreciation formula, given that the age of assets varies by economy. Instead, we use the book value of assets as reported in firms' balance sheets when calculating the market-to-book ratio. Mean and median market-to-book ratios of the sample corporations are shown in Table III. This table provides insights into the relative value of firms by their main industrial sector and economy of origin. Unregulated utilities have the highest firm valuation, with a mean market-to-book ratio of 1.79 and a median of 1.42. Service and leisure corporations also have high valuations. Firm values are lowest in textiles and trade, with a mean market- to-book ratio of 1.27 and a median of 1.07. The range of median firm valuations across economies is similar in mag- nitude to that across sectors. Malaysian corporations have the highest rel- ative valuations, with a mean of 1.70 and a median of 1.43. They are followed by Singaporean corporations, with a mean of 1.63 and a median of 1.38, and Taiwanese corporations, with a mean of 1.59 and a median of 1.35. Korean and Philippine corporations have the lowest valuations. The valuation data reported here for Hong Kong, Korea, and Singapore are lower than those in La Porta et al. 2002 . Our median values are 1.12, 1.00, and 1.38, respec- tively, compared with their 1.15, 1.06, and 1.52. This difference is likely accounted for by the different year of data coverage--1996 compared with 1995--because East Asian stock markets experienced a decline over this pe- 118 2752 A Reader Table III The Valuations of East Asian Corporations by Economy and Industry, 1996 (Market-to-Book Ratio) in The market-to-book ratio is the ratio of the market value of assets to the book value of assets at the end of 1996. Market value is defined as the International Journal sum of the market value of common stock and the book value of debt and preferred stock. The book value of assets comes from firms' balance sheets. All corporations, including those without an ultimate controlling owner, are included. Industries are defined as follows: petroleum SIC 13, 29 , consumer durables SIC 25, 30, 36, 37, 50, 55, 57 , basic industry SIC 10, 12, 14, 24, 26, 28, 33 , food and tobacco SIC 1, 2, 9, 20, 21, 54 , construction SIC 15, 16, 17, 32, 52 , capital goods SIC 34, 35, 38 , transportation SIC 40, 41, 42, 44, 45, 47 , unregulated utilities SIC 46, of 48 , textiles and trade SIC 22, 23, 31, 51, 53, 56, 59 , services SIC 72, 73, 75, 76, 80, 82, 87, 89 , and leisure SIC 27, 58, 70, 78, 79 . The sample Corporate Finance excludes financial companies SIC 60­69 and regulated utilities SIC 49 . Hong Korea, Industry Kong Indonesia Rep. of Malaysia Philippines Singapore Taiwan Thailand Total Finance Petroleum Mean 0.77 0.37 1.76 1.31 1.19 2.29 1.15 1.20 1.51 Median 0.77 0.37 1.50 1.59 1.01 1.40 1.15 1.20 1.20 Consumer durables Mean 1.31 0.92 1.30 1.94 1.48 1.59 1.67 1.20 1.40 Median 1.08 0.79 0.99 2.00 1.24 1.29 1.64 1.23 1.18 Basic industry Mean 1.63 1.62 1.10 2.00 1.21 1.67 1.69 1.57 1.48 Median 1.47 1.24 0.99 1.78 1.06 1.47 1.34 1.31 1.17 Food and tobacco Mean 1.85 1.65 1.10 1.72 1.13 2.16 1.42 1.40 1.55 Median 1.51 1.45 1.01 1.35 0.92 1.88 1.22 1.31 1.24 Construction Mean 1.12 1.35 1.13 1.52 1.53 1.19 1.53 1.24 1.32 Median 1.11 1.38 0.89 1.25 1.18 1.18 1.40 1.02 1.14 Capital goods Incentive Mean 1.35 1.37 1.27 2.13 0.76 1.61 1.44 1.16 1.41 Median 1.17 1.41 0.91 1.74 0.57 1.58 1.20 1.07 1.17 Transportation Mean 1.10 1.38 1.46 1.41 1.56 1.56 1.79 1.23 1.37 Median 1.12 1.26 0.94 1.30 1.56 1.43 1.53 1.22 1.24 and Unregulated utilities Mean 0.94 1.88 1.93 1.89 1.12 1.76 1.94 3.18 1.79 Entrenchment Median 0.89 1.88 2.08 1.51 1.06 1.51 1.94 1.88 1.42 Textiles and trade Mean 1.38 1.15 1.11 1.47 1.16 1.40 1.50 1.04 1.27 Median 1.08 1.02 1.00 1.43 0.97 1.29 1.18 0.85 1.07 Chapter Services Mean 1.07 1.30 2.43 1.94 1.51 1.66 1.50 1.91 1.68 Median 0.99 1.53 2.34 1.18 1.87 1.58 1.60 1.11 1.36 Effects Leisure Four Mean 1.24 1.65 1.68 1.50 1.32 1.53 2.22 1.13 1.43 Median 1.25 1.58 1.80 1.32 1.32 1.31 2.22 1.21 1.32 of Total Large Mean 1.31 1.36 1.25 1.70 1.25 1.63 1.59 1.38 1.43 Median 1.12 1.13 1.00 1.43 1.06 1.38 1.35 1.22 1.19 Shareholdings 2753 119 120 A Reader in International Corporate Finance 2754 The Journal of Finance riod. Another reason for the difference could be that La Porta et al. 2002 use only the 20 to 30 largest publicly traded corporations in each economy, while our samples are much larger.8 II. Ownership and Control Concentration and Their Effect on Firm Value As noted, we seek evidence about the effects of ownership and control concentration on firm value when there is a controlling shareholder. We want to test two hypotheses. The first is that the more concentrated cash-flow rights in the hands of the largest shareholder are, the stronger is that share- holder's incentive to have the firm run properly, because having the firm running properly would raise his wealth; likewise, his incentive to reduce the value of the firm by extracting private benefits is weaker, because doing so would lower his wealth. Both effects should result in a positive relation- ship between firm values and the largest shareholder's cash-flow rights. In contrast, the second hypothesis holds that the more concentrated con- trol is in the hands of the largest shareholder, the more entrenched the shareholder is and the better able he is to extract value--to the detriment of the firm's value to minority shareholders. This hypothesis suggests a nega- tive relationship between firm values and the largest shareholder's control rights. The agency problem of entrenchment and value extraction will be especially pronounced when there is a big divergence between control rights and cash-flow rights, because the willingness to extract value is less re- strained by the controlling shareholder's cash-flow stake. A. Graphical Evidence To investigate these two hypotheses, we first present figures showing the association between market-to-book ratios and the cash-flow and control stakes of the largest shareholder. We then conduct a series of regressions. 8 In a previous version of this article Claessens et al. 1999a , we used an industry-adjusted valuation measure as our dependent variable. Each firm's valuation was adjusted relative to the economy-wide average for the industries in which the firm operated, taking into account the shares each industry represented in the firm's overall sales. The idea was to take out both economy and industry effects, since the economies in the sample are at different stages of development and since firm valuation can vary widely across industries. The adjustment was burdensome, however, because many publicly listed corporations in East Asia operate in mul- tiple segments. For example, if firms are classified as multisegment if they derive less than 90 percent of their sales from one two-digit SIC code, then more than two-thirds of corporations from Hong Kong, Malaysia, and Singapore have multiple segments. In contrast, less than 20 percent of U.S. corporations operate in multiple segments Claessens et al. 1999b . Adjusting for multisegment firms thus adds an extra layer of complexity in computing industry-adjusted valuation measures. Still, we ran regressions using these industry adjustments and found sim- ilar, even slightly stronger, results as when using the market-to-book ratio; see Claessens et al. 1999a . Chapter Four 121 Incentive and Entrenchment Effects of Large Shareholdings 2755 Figure 1. Company valuation and ownership of the largest shareholder in East Asian corporations, 1996. We start by plotting the association between market-to-book ratios and the cash-flow stake of the largest shareholder Figure 1 . Firm value, as measured by the market-to-book ratio, generally increases with the share of cash-flow rights in the hands of the largest owner. This pattern is consistent with the positive incentive effect of larger cash-flow ownership on firm value. But the relationship is not monotone. Ownership by the largest shareholders of 41 to 50 percent, for example, is associated with lower mean market val- uation than ownership of 36 to 40 percent, and the difference is statistically significant. Ownership of 51 to 55 percent is associated with the highest mean market-to-book ratios, with valuation falling again for ownership con- centration above 55 percent. The association between firm valuation and the separation of control and ownership rights is shown in Figure 2. The figure suggests that the larger the wedge is between control and ownership rights, the lower a firm's val- uation is. Corporations with no separation of control and ownership rights have the highest value. Corporations with a separation of more than 35 percentage points--that is, when the control rights of the largest share- holder exceed his ownership rights by 35 percentage points or more--have the lowest value. Again, the relationship is not monotone. Corporations with moderate levels of separation, such as 11 to 15 percentage points, are valued higher than corporations with separation levels of 1 to 10 percentage points. Once the separation of ownership and control reaches 15 percentage points, however, there is a monotone decrease in firm value. These two figures provide suggestive evidence on our two hypotheses. Fig- ure 1 provides evidence in favor of the incentive effects associated with in- creased cash-flow rights in the hands of the largest shareholder. Figure 2 is 122 A Reader in International Corporate Finance 2756 The Journal of Finance Figure 2. Company valuation and the difference between control and ownership of the largest shareholder in East Asian corporations, 1996. generally consistent with the entrenchment effect. As the control rights of the largest shareholder increase relative to his ownership rights, firm val- uation appears to fall. But in both figures, the association with market-to- book ratios is not monotone, and here we did not control for other factors influencing firm valuation. Thus, multivariate analysis allowing for nonlin- ear relationships is needed to investigate more precisely the incentive and entrenchment hypotheses. B. Regression Results We start by including as control variables several firm-specific variables commonly used in studies of firm valuation. Specifically, we include sales growth in the previous year and capital spending relative to sales in the previous year. We expect both variables to have a positive relationship with firm value, because they proxy for a firm's growth prospects and investment. We also include firm age measured in years since establishment and firm size measured by the log of total assets in the previous year . On the one hand, we expect age and size to be positively related to firm value for the same reasons often mentioned in studies of firms in developed econo- mies: older and larger firms have better disclosure, more liquid trading, more attention from analysts, and more diversified activities leading to lower risk of financial distress. On the other hand, younger and smaller firms may have more growth opportunities. Furthermore, in East Asia, smaller firms may be less diversified, leading to smaller value discounts. Claessens et al. 1999b show that diversification is associated with a value discount for East Asian corporations. Chapter Four 123 Incentive and Entrenchment Effects of Large Shareholdings 2757 We do not expect to introduce significant colinearities in the regressions by using this set of variables, because the correlations between the variables are very low. For example, the correlation between sales growth and capital spending over sales is just 0.0263, and the correlation between firm age and firm size is only 0.1272. We also include industry dummy variables in all the regressions to correct for possible valuation differences among industries. The leisure sector is used as the numeraire. We next want to control for possible within-economy correlations that could bias our analysis. The Breusch and Pagan 1980 Lagrange multiplier test rejects the null hypothesis that errors are independent within country sam- ples, suggesting that a fixed-effects specification cannot be used. To correct for within-economy correlations, we use a random-effects specification that assumes each sample has a common explanatory variable component, which may differ across economies. In other words, we do not treat corporations in a given economy as independent observations. This specification takes ex- plicit account of the correlated errors among our observations within an economy and produces consistent standard errors. Moreover, a random- effects specification is preferable to fixed effects when a subsample of the population is used, as we have done here Greene 1997, p. 623 . Table IV presents regression results that link firm valuation to the own- ership and control of the largest shareholder, with ownership and control as continuous variables. The table presents three specifications, with the first the basic regression, the second the basic regression with a dummy added for whether the firm consolidates its financial statements using either the full or cost method , and the third a specification that investigates possible nonmonotonicity in the relationship. As noted, consolidation tends to under- state the market-to-book ratio with the full consolidation method but not with the cost method. Because we do not know the method of consolidation for each firm, the consolidation dummy will pick up the combined effects of no bias of the market-to-book ratio with the cost method and the understate- ment of the market-to-book ratio with the full method. Thus, we should ex- pect a negative sign for the consolidation dummy. For all three regression specifications, we find that ownership concentra- tion is positive and associated with increased firm valuation at a statisti- cally significant one percent level. The three coefficients for the ownership variable are similar and are economically significant. A one standard devi- ation increase in the ownership stake of the largest shareholder induces a 0.091 increase in the market-to-book ratio, or an increase of more than 6.4 per- cent of the average under regression specification 1 . Increases in control rights over ownership rights are associated with lower firm values for all three specifications. The coefficients on the control minus ownership vari- able are also highly economically significant. A one standard deviation in- crease in the concentration of control over ownership rights in the hands of the largest shareholder lowers relative values by 0.076--more than a 5.3 per- cent drop again under specification 1 . The incentive and entrenchment effects of large shareholdings are thus large and economically significant. 124 A Reader in International Corporate Finance 2758 The Journal of Finance Table IV Regression Results on the Relationship between Firm Value and the Largest Shareholder's Ownership and Control The regressions are performed using a random-effects economy-level specification. Numbers in parentheses are standard errors. The dependent variable is the ratio of the market value of assets to the book value of assets at the end of 1996. Market value is defined as the sum of the market value of common stock and the book value of debt and preferred stock. The book value of assets comes from firms' balance sheets. The main independent variables are the share of cash-flow rights held by the largest shareholder ownership and the share of voting rights held by the largest shareholder control . Control minus ownership is a continuous variable measur- ing the simple difference between the share of control rights and the share of cash-flow rights in the hands of the largest shareholder. Control exceeds ownership is a dummy equal to one if control rights are higher than cash-flow rights; otherwise, it is zero. Control exceeds ownership, high is a dummy equal to one if control rights are higher than cash-flow rights and if this separation is higher than the median separation in corporations where control and ownership differ; otherwise, it is zero. Sales growth, capital spending over sales, firm age, firm size, and industry dummies the leisure sector is the numeraire are included as control variables. The consolidation dummy equals one if the corporation consolidates its financial statements; other- wise, it is zero. Independent variable Specification 1 Specification 2 Specification 3 Ownership 0.0073a 0.0020 0.0073a 0.0020 0.0080a 0.0020 Control minus ownership 0.0103a 0.0033 0.0103a 0.0033 Control exceeds ownership 0.0234 0.0621 Control exceeds ownership, 0.1260a 0.0552 high Sales growth 0.5568a 0.1145 0.5603a 0.1147 0.5574a 0.1148 Capital spending over sales 0.1105 0.1156 0.1100 0.1157 0.1106 0.1162 Firm age years 0.0005 0.0012 0.0005 0.0012 0.0007 0.0012 Firm size log of assets 0.0476a 0.0135 0.0476a 0.0135 0.0463a 0.0135 Consolidation dummy 0.0260 0.0467 Petroleum 0.1126 0.1763 0.1101 0.1764 0.1169 0.1766 Consumer durables 0.0601 0.1042 0.0624 0.1043 0.0560 0.1044 Basic industry 0.0485 0.1098 0.0440 0.1102 0.0557 0.1100 Food and tobacco 0.0625 0.1153 0.0591 0.1155 0.0687 0.1155 Construction 0.1313 0.1100 0.1324 0.1100 0.1242 0.1102 Capital goods 0.0498 0.1172 0.0528 0.1174 0.0438 0.1175 Transportation 0.0501 0.1370 0.0491 0.1371 0.0456 0.1373 Unregulated utilities 0.3752b 0.1708 0.3792b 0.1710 0.3655b 0.1712 Textiles and trade 0.2803c 0.1637 0.2794c 0.1638 0.2806c 0.1641 Services 0.0873 0.1835 0.0861 0.1836 0.0834 0.1839 Constant 0.8532 2.4950 0.8932 2.4967 0.4968 2.4947 R2 0.0716 0.0718 0.0685 Number of observations 1,301 1,301 1,301 aSignificant at the 1 percent level; bsignificant at the 5 percent level; significant at the 10 c percent level. The regression results do not appear to be influenced by whether firms consolidate their financial statements. When the dummy is included for whether a firm consolidates Table IV, specification 2 , the dummy has a Chapter Four 125 Incentive and Entrenchment Effects of Large Shareholdings 2759 negative sign but is not statistically significant. More importantly, the co- efficients for the ownership, control minus ownership, and other variables barely change, if at all. If firms were more likely to have subsidiaries and consolidate their financial statements when ownership is concentrated, our results would be biased against finding a positive effect on firm value of ownership structures. That the coefficients do not change when we include a dummy for whether firms consolidate suggests that consolidation and the methods used to consolidate do not bias our results. Figure 2 suggests that the degree of entrenchment of the largest share- holder to the detriment of firm value and other shareholders might be higher when there is more than a 15 percentage point gap between control rights and cash-flow rights. The importance for this sample of a high level of separation between control rights and cash-flow rights is confirmed in the regression result that includes two dummies specification 3 . The first dummy--control exceeds ownership--equals one when control rights exceed cash-flow rights. The second dummy--control exceeds ownership, high-- equals one when the separation between control rights and cash-flow rights exceeds the median separation for all firms with separation. This median separation is 15.1 percentage points. The first dummy has a negative coefficient but is not statistically signif- icant. The second dummy is statistically significant at the one percent level and has a large economic effect, because it indicates a 12.6 percentage point reduction in the market-to-book ratio. This outcome suggests that, for this sample of firms, a large wedge between control and ownership stakes leads to value losses. This critical wedge of about 15 percentage points contrasts with the find- ings in Morck et al. 1988 , who show that the entrenchment effect for U.S. manager-owners becomes apparent at a low concentration of control, start- ing at just over five percent. This difference may be due to the fact that in Morck et al. and Stulz 1988 , entrenchment arises from managers' ability to prevent takeovers. In the United States, it is possible to prevent takeovers with low ownership concentration. But, in East Asia, takeovers are rare to begin with. Presumably, the valuation discount brought about by entrenched owners in East Asia arises from actions other than blocking value-enhancing takeovers. Such other actions may include private benefits and direct ex- propriation through transfer of financial wealth to affiliated firms, and would require large control stakes. Reducing such behavior by large stakeholders would require strong action by minority shareholders--a difficult task in these economies given their weak corporate governance and poor enforce- ment Johnson et al. 2000 . Among the other explanatory variables, sales growth in the previous year and firm size have significant explanatory power, with sales growth show- ing a positive coefficient and size a negative coefficient. The first finding is common, because higher growth reflects better future growth opportunities and so higher firm valuation. The second suggests that for this sample, be- ing smaller leads to higher relative valuation, suggesting that small firms 126 A Reader in International Corporate Finance 2760 The Journal of Finance have better growth prospects. Given the East Asian context, lower values for large firms may also derive from their more extensive diversification Claes- sens et al. 1999b . The other firm-specific variables are statistically insignificant for all three specifications. This is perhaps not surprising given that their simple correlation coefficients with the market-to-book ratio are low. For example, the correlation coefficient between firm age and the market-to-book ratio is only 0.0413. The industry dummies are jointly statistically significant in explaining firm valuation. Individually, however, the only statistically significant industry dummies are for unregulated utilities, with a coeffi- cient of 0.3752, and textiles and trade, with a coefficient of 0.2803 under specification 1 . C. Tests of Robustness C.1. Accounting for Group Effects Observations within business groups may not be independent due to the common ownership and the sometimes common management of members of such groups, which can lead to intragroup financial transfers that are not necessarily market based. Such transfers could lead to interdependent val- uation measures among firms that are members of the same group. To ad- dress this concern, we treat all observations within each business group as a single observation and rerun the regressions of Table IV. Because defini- tions of business groups vary across East Asia, we identify group member- ship broadly by including all firms in the same group if they are part of a set of firms linked through pyramiding or if they have cross-holdings with other firms. This definition leads to a larger set of affiliated corporations than does the conventional use of ownership links above a certain threshold. As such, this definition should provide a conservative bound on any group effect. We use two alternative regression specifications when collapsing all ob- servations within each business group into a single observation. The first regresses the median market-to-book ratio within a business group on the medians of the explanatory variables of all corporations belonging to that group. Stand-alone firms, that is, firms not belonging to any group, are treated as separate observations in this regression. In the second specifica- tion, we weigh within-group observations with weights equal to the assets contributed by each firm to the group as a share of total group assets, in effect giving more importance to large members of the group. This adjust- ment accounts for the possibility that within-group ownership structures and net financial transfers lead to a size-related bias in the relationship between ownership structures and firm valuation. Claessens et al. 2000 show that smaller firms are more likely to be con- trolled by a single shareholder. If smaller firms also gain more value from group affiliation relative to large firms, as might be expected, then weighing by size would bias our analysis against finding a relationship between own- Chapter Four 127 Incentive and Entrenchment Effects of Large Shareholdings 2761 ership structures and firm valuation. Again, stand-alone firms are treated as separate observations in the weighted regression. The resulting sample for both specifications has 872 observations. Table V shows the regression results using both the basic specification of Table IV and the specification that investigates large differences between ownership rights and control rights. We do not use industry dummies in either specification. Industry dummies would not be meaningful, because we collapse all within-group firm observations to one observation per group and because within each group these firms typically engage in many industries. The main results on ownership and control rights are maintained. The own- ership stake of the largest shareholder in specifications 1 and 3 continues to have a positive and statistically significant relationship with firm value, with coefficients similar to those in Table IV. The coefficients on the control minus ownership variable are again negative and statistically significant and of the same order as in Table IV. In the specifications with the dummy variables, 2 and 4, the coefficients are not statistically significant for the first dummy, control exceeds owner- ship. But they have the same magnitude as the coefficients of the same variable in Table IV. The coefficients are statistically significant for the sec- ond dummy, control exceeds ownership, high, and of somewhat larger mag- nitude than the coefficients of the same variable in Table IV. Comparing the median specifications 1 and 2 and the value-weighted least squares spec- ifications 3 and 4 shows that the coefficients of the ownership variables are similar, suggesting that the distribution of firm size within each busi- ness group does not bias the results. Sales growth is the only statistically significant control variable in these specifications. The magnitude of its coefficient is slightly different from those in Table IV, possibly because of the smaller weight given to firms in business groups. A general comparison of Tables IV and V suggests that entrench- ment effects are equally severe in group-affiliated firms, because the coef- ficients are similar regardless of whether all firms affiliated with a single group are reduced to one observation. Together, the regression results show that the dependence among firms in business groups does not alter our main results for valuation or ownership and control structures. C.2. Results by Economy We also study the relationship between firm valuation and ownership and control in the hands of the largest shareholder at the economy level, using the basic specification of Table IV. We include but do not report the four control variables: sales growth, capital spending over sales, firm age, and firm size. Higher ownership rights in the hands of the largest owner are associated with higher valuations in six economies, and this relationship is statistically significant in all six except the Philippines Table VI . That outcome may be due to the fact that the Philippine sample is the smallest of the eight economies, with just 77 observations. Singapore and Taiwan show 128 2762 Table V Regression Results on the Relationship between Firm Value and the Largest Shareholder's Ownership and Control, by Business Group The regressions are performed using a random-effects specification in which all observations within a business group are collapsed into one observation. Stand-alone corporations are treated as separate observations, that is, each is viewed as its own business group. Specifications 1 and 2 are run on the median value within each business group for both the dependent and independent variables. In specifications 3 and 4, the business group observations are reached by weighing each group affiliate observation by its assets as a share of the group's total assets. Numbers A in parentheses are standard errors. The dependent variable is the ratio of the market value of assets to the book value of assets at the end of Reader 1996. Market value is defined as the sum of the market value of common stock and the book value of debt and preferred stock. The book value of assets comes from firms' balance sheets. The main independent variables are the share of cash-flow rights held by the largest shareholder The ownership and the share of voting rights held by the largest shareholder control . Control minus ownership is a continuous variable measuring in the simple difference between the share of control rights and the share of cash-flow rights in the hands of the largest shareholder. Control International exceeds ownership is a dummy equal to one if control rights are higher than cash-flow rights; otherwise, it is zero. Control exceeds ownership, Journal high is a dummy equal to one if control rights are higher than cash-flow rights and if this separation is higher than the median separation in corporations where control and ownership differ; otherwise, it is zero. Sales growth, capital spending over sales, firm age, firm size, and industry dummies the leisure sector is the numeraire are included as control variables. of Corporate Independent Variable Specification 1 Specification 2 Specification 3 Specification 4 Finance Ownership 0.0077a 0.0024 0.0079a 0.0023 0.0070a 0.0024 0.0072a 0.0025 Control minus ownership 0.0109b 0.0045 0.0095b 0.0046 Control exceeds ownership 0.0211 0.0542 0.0178 0.0517 Finance Control exceeds ownership, high 0.1416a 0.0627 0.1387a 0.0583 Sales growth 0.6494a 0.1452 0.6502a 0.1453 0.6404a 0.1453 0.6411a 0.1462 Capital spending over sales 0.1297 0.1307 0.1292 0.1303 0.1418 0.1315 0.1422 0.1318 Firm age years 0.0006 0.0016 0.0006 0.0016 0.0005 0.0016 0.0005 0.0016 Firm size log of assets 0.0277c 0.0160 0.0275c 0.0159 0.0260c 0.0160 0.0265c 0.0161 Constant 0.4410 3.2895 0.4457 3.2901 0.5200 3.2242 0.5215 3.2245 R2 0.0392 0.0398 0.0396 0.0408 Number of observations 872 872 872 872 aSignificant at the 1 percent level;b significant at the 5 percent level; significant at the 10 percent level. c Chapter Four 129 Incentive and Entrenchment Effects of Large Shareholdings 2763 Table VI Regression Results on the Relationship between Firm Value and the Largest Shareholder's Ownership and Control, by Economy The regressions are performed on each economy sample using an ordinary least squares spec- ification. Numbers in parentheses are standard errors. The dependent variable is the ratio of the market value of assets to the book value of assets at the end of 1996. Market value is defined as the sum of the market value of common stock and the book value of debt and pre- ferred stock. The book value of assets comes from firms' balance sheets. The main independent variables are the share of cash-flow rights held by the largest shareholder ownership and the share of voting rights held by the largest shareholder control . Control minus ownership is a continuous variable measuring the simple difference between the share of control rights and the share of cash-flow rights in the hands of the largest shareholder. Sales growth, capital spending over sales, firm age, and firm size are included as control variables but are not re- ported. Industry dummies are not included, given the smaller sample size at the economy level. Control Minus Number of Economy Constant Ownership Ownership R2 Observations Hong Kong 1.4429a 0.0088a 0.0181b 0.0502 225 0.1877 0.0037 0.0083 Indonesia 0.9852a 0.0252a 0.0133a 0.1583 132 0.2827 0.0072 0.0059 Korea, Rep. of 1.1871a 0.0268a 0.0038 0.0675 281 0.1429 0.0063 0.0107 Malaysia 2.0198a 0.0084b 0.0201c 0.0364 171 0.2743 0.0043 0.0109 Philippines 1.5051a 0.0051 0.0019 0.0056 77 0.2694 0.0091 0.0204 Singapore 2.3004a 0.0111c 0.0090 0.0153 176 0.2237 0.0068 0.0115 Taiwan 2.1297a 0.0070 0.0118 0.0084 129 0.2113 0.0086 0.0152 Thailand 1.2455a 0.0130a 0.0190c 0.0389 110 0.3839 0.0057 0.0105 a Significant at the 1 percent level;b significant at the 5 percent level; significant at the 10 c percent level. a negative relationship between ownership rights and firm valuation, but the relationship is statistically significant only in Singapore.9 Most of the coefficients on ownership rights for the economy-specific sam- ples are larger than those for the overall sample. This is especially the case in economies with weaker corporate governance, such as Indonesia and Ko- rea, suggesting that the incentive effects of concentrated ownership are more important in these settings, consistent with the findings of La Porta et al. 2002 . 9 The result for Singapore disappears when state firms are excluded, and the coefficient on ownership rights then becomes marginally significantly positive at the 10 percent level . This outcome suggests that state-controlled firms are driving the negative coefficient for the sample of Singaporean firms. 130 A Reader in International Corporate Finance 2764 The Journal of Finance The wedge between ownership and control rights is associated with lower valuations in all eight economies, and this relationship is statistically sig- nificant in Hong Kong, Indonesia, Malaysia, and Thailand. Again, the sta- tistically significant coefficients are somewhat larger than those for the whole sample. These four economies also display a positive, statistically significant coefficient for ownership stakes, suggesting that incentive and entrench- ment effects can go together. That the coefficients are larger suggests that while the incentive effects of concentrated ownership can be more important in settings with weak corporate governance, so can the entrenchment ef- fects, leading to unclear net effects of ownership concentration on firm value. C.3. Reverse Causality Another issue that might arise is the possibility of reverse causality in terms of the impact on firm valuation of deviations between ownership and control rights. Suppose that the largest shareholder considers his firm over- valued and wants to invest his money elsewhere. He might then want to lower his ownership rights but maintain his control rights. Firm values would then adjust with a lag to their equilibrium levels. We could then find that as deviations become large, firm valuation becomes lower, but that would tell us little about the possible entrenchment effect of the separation of control and ownership. This possibility would imply changes in ownership and con- trol patterns that are followed with some lag by lower valuations. It seems unlikely, however, that firms can change their ownership struc- tures quickly and frequently in light of temporary overvaluations or under- valuations. La Porta et al. 1999 report that ownership structures for the top 20 to 30 East Asian firms are relatively stable over time. More gener- ally, our regression results are based on cross-sectional relationships. The possibility of reverse causality would thus lead to a bias only if insiders changed their cash-flow rights quickly and frequently in light of temporary overvaluations or undervaluations, while maintaining their control rights, and did so systematically across many corporations. Such behavior seems unlikely. III. Owner Types and Mechanisms for Separating Ownership and Control Previous research has documented that a large shareholding in general and the separation of ownership and control in particular is usually associ- ated with family ownership La Porta et al. 1999 and Claessens et al. 2000 . Thus, we investigate whether a particular type of owner is largely respon- sible for our results. We study separately the effects on firm value of own- ership by families, the state, or widely held corporations and financial institutions. The control stakes of the largest shareholder are used to clas- sify firms into one of these ownership categories. The family is the largest blockholder in 908 firms, or nearly 70 percent of the sample. Few corpora- Chapter Four 131 Incentive and Entrenchment Effects of Large Shareholdings 2765 tions are controlled by the state--111 in total--and most are from Singapore see Table II . Finally, 282 observations have widely held controlling owners, either corporations or financial institutions. We also study the relationship between corporate valuation and divergen- cies in cash-flow rights and control rights for these three types of owners. We use the same specifications as for regressions 1 and 3 in Table IV, with the same firm-specific control variables and industry dummies the latter are not reported . When we consider the effects on corporate value of own- ership and control rights for each type of controlling shareholder, we find that the ownership variable has a similar coefficient for all three types of controlling shareholders Table VII . Only with the state as controlling owner is the coefficient not statistically significant, and then only for the first specification. Still, significance levels are generally lower than in Table IV. The coefficient for the difference between control and ownership stakes is statistically significant at the 5 percent level for family control and at the 10 percent level for state control. Some results are less robust, however. In particular, for specifications using the dummy for high divergence between control and ownership as well as the dummy for any difference between control and ownership specifications 2, 4, and 6 , only the coefficient for the first dummy in the case of state ownership is statistically significant. The other coefficients lose their sta- tistical significance. These weaker results could be due to the smaller set of firms for each regression. Nevertheless, the results suggest that family con- trol, and to some extent state ownership, are driving the main results. This could be because managers at widely held corporations and financial insti- tutions are less able than families and the state to efficiently divert benefits to themselves. So far the results do not yet shed light on which mechanisms separating control rights from ownership rights may be driving the results. As noted, in East Asian corporations, deviations between control and ownership rights come about through different means, including pyramiding, cross-holdings, and dual-class shares. Bebchuk 1999 and Wolfenzon 1999 suggest that pyramiding is associated with value discounts. Cross-holdings could also be associated with value losses because they facilitate nonmarket-based finan- cial transfers among corporations within a group, either horizontally or ver- tically. Besides pyramid structures and cross-holdings, dual-class shares, while not common in East Asia, can separate control from ownership rights and be associated with value loss. For a larger sample of countries, Nenova 2001 highlights the role of dual shares in environments with poor corporate gov- ernance as a mechanism for value transfers. To measure the importance of each of these mechanisms, we construct dummy variables to explain the relative variations in firm valuation Table VIII . Pyramid is a dummy equal to one if the firm is part of a pyr- amid structure including if it is the apex firm at the top of a pyramid , and zero otherwise. Crosshold is a dummy equal to one if the firm is controlled at least partly by a cross-holding, and zero otherwise. Dualclass is a dummy 132 2766 A Reader Table VII The in Regression Results on the Relationship between Firm Value and the Largest Shareholder's International Ownership and Control, by Owner Type Journal The regressions are performed using a random-effects specification. Numbers in parentheses are standard errors. A corporation is family owned if the largest ultimate shareholder is a family group, state owned if the largest shareholder is the state, and company owned if the largest shareholder is a widely held corporation or financial institution. The dependent variable is the ratio of the market value of assets to the book value of assets at the end of 1996. Market value is defined as the sum of the market value of common stock and the book value of debt and of Corporate preferred stock. The book value of assets comes from firms' balance sheets. The main independent variables are the share of cash-flow rights held Finance by the largest shareholder ownership and the share of voting rights held by the largest shareholder control . Control minus ownership is a continuous variable measuring the simple difference between the share of control rights and the share of cash-flow rights in the hands of the largest shareholder. Control exceeds ownership is a dummy equal to one if control rights are higher than cash-flow rights; otherwise, it is zero. Control exceeds ownership, high is a dummy equal to one if control rights are higher than cash-flow rights and if this separation is higher than Finance the median separation in corporations where control and ownership differ; otherwise, it is zero. Sales growth, capital spending over sales, firm age, firm size, and industry dummies are included as control variables but are not reported. Widely Held Corporation Family The State or Financial Institution Independent Variable Specification 1 Specification 2 Specification 3 Specification 4 Specification 5 Specification 6 Ownership 0.0086b 0.0084a 0.0073 0.0121c 0.0086c 0.0075c 0.0026 0.0025 0.0070 0.0062 0.0045 0.0041 Control minus ownership 0.0090b 0.0247c 0.0189 0.0037 0.0130 0.0154 Control exceeds ownership 0.0494 0.1218 0.1086 0.0722 0.1845 0.1816 Control exceeds ownership, high 0.0342 0.4806b 0.3685 0.0828 0.2264 0.3331 Incentive Sales growth 0.6621b 0.6323a 0.1833 0.1847 0.5105b 0.4833c 0.1341 0.1358 0.1241 0.1346 0.2491 0.2557 Capital spending over sales 0.1370 0.0814 0.0043 0.0229 0.0353 0.1959 0.1334 0.1332 0.4329 0.4341 0.2729 0.2726 and Firm age years 0.0011 0.0014 0.0061 0.0043 0.0020 0.0030 0.0014 0.0014 0.0047 0.0050 0.0030 0.0031 Entrenchment Firm size log of assets 0.0358b 0.0373b 0.0512 0.0023 0.0714b 0.0889a 0.0169 0.0165 0.0482 0.0461 0.0284 0.0278 Constant 0.6068 1.1551 11.4143 7.1866 6.0540 8.2864 2.8349 2.8243 9.4312 10.0226 6.0331 6.0979 Chapter R2 0.0523 0.0496 0.0450 0.0855 0.0714 0.0811 Number of observations 908 908 111 111 282 282 Effects aSignificant at the 1 percent level;bsignificant at the 5 percent level; significant at the 10 percent level. c Four of Large Shareholdings 2767 133 134 2768 Table VIII Regression Results on the Relationship between Firm Value and Pyramiding, Cross-Holdings, and Dual-Class Shares The regressions are performed using a random-effects specification. Numbers in parentheses are standard errors. Pyramid is a dummy equal to one if the firm is part of a pyramid structure; otherwise, it is zero. Crosshold is a dummy equal to one if the firm is controlled at least partly A Reader by a cross-holding; otherwise, it is zero. Dualclass is a dummy equal to one if the firm has issued dual-class shares; otherwise, it is zero. The dependent variable is the ratio of the market value of assets to the book value of assets at the end of 1996. Market value is defined as the sum of the market value of common stock and the book value of debt and preferred stock. The book value of assets comes from firms' balance sheets. The in The main independent variables are the share of cash-flow rights held by the largest shareholder ownership and the share of voting rights held International by the largest shareholder control . Sales growth, capital spending over sales, firm age, firm size, and industry dummies are included as control Journal variables but are not reported. Independent Variable Specification 1 Specification 2 Specification 3 Specification 4 of Ownership 0.0119a 0.0020 0.0095a 0.0020 0.0118a 0.0020 0.0091a 0.0020 Corporate Pyramid dummy 0.0571 0.1365 0.0136 0.0524 Finance Crosshold dummy 0.0332 0.0507 0.0077 0.0732 Dualclass dummy 0.0468 0.0703 0.1595 0.1360 Sales growth 0.5754a 0.1138 0.5683a 0.1146 0.5778a 0.1137 0.5622a 0.1149 Finance Capital spending over sales 0.1149 0.1157 0.0897 0.1151 0.1152 0.1157 0.0862 0.1153 Firm age years 0.0012 0.0012 0.0009 0.0012 0.0011 0.0012 0.0008 0.0012 Firm size log of assets 0.0312a 0.0136 0.0400a 0.0134 0.0314b 0.0136 0.0420a 0.0134 Constant 0.7902 2.4572 0.0173 2.4722 0.7560 2.4576 0.1212 2.4780 R2 0.0474 0.0480 0.0467 0.0491 Number of observations 1,301 1,301 1,301 1,301 aSignificant at the 1 percent level; bsignificant at the 5 percent level. Chapter Four 135 Incentive and Entrenchment Effects of Large Shareholdings 2769 equal to one if the firm has issued dual-class shares, and zero otherwise. We run four specifications, using a dummy for each mechanism separately and then combining all three dummies in the final regression. This final regres- sion does not create any collinearity problems, because the three variables are not highly correlated. The simple correlation between Pyramid and Cross- hold is 0.2876, between Pyramid and Dualclass 0.1457, and between Cross- hold and Dualclass 0.0174. All three dummy variables have a negative coefficient, a sign that these mechanisms reduce value, correcting for ownership structures and other fac- tors. But none of the three is statistically significant. The ownership vari- able remains positive and statistically significant, with coefficients similar to those in Tables IV and V. While the entrenchment of the largest share- holders in East Asian corporations may thus be supported by combinations of pyramiding, cross-holdings, and dual-class shares, the evidence suggests that the separation of ownership and control is what leads to value dis- counts, not any mechanism in particular.10 An alternative hypothesis to the two we have explored here could be that value discounts are due to bad management, and the likelihood of bad man- agement is related to the ownership structure. Multiple layers of pyramidal ownership and numerous cross-holdings could mean that the controlling owner- manager at the apex of the pyramid does not have the capacity to monitor the managers of all its affiliated firms. The result could be bad performance and value discounts. But Claessens et al. 2000 show that for more than two-thirds of firms with concentrated ownership, managers come from the controlling families. Controlling owners that are managers are thus not lim- ited to apex firms, but are widespread throughout business groups. As such, managers would have few incentives to mismanage firms for which they are also controlling owner. So, although appealing, this alternative hypothesis does not hold for the average corporation in our sample. Nevertheless, we did split the sample into firms managed by people who belong to the con- trolling shareholder's family and firms with unrelated managers, and we found similar results not reported . IV. Conclusion This article documents the relationships between ownership and control stakes held by the largest shareholder on the one hand, and market valua- tion on the other hand, for a large sample of publicly traded corporations in East Asia. Its main contribution is disentangling the incentive and entrench- ment effects of large ownership that are so difficult to tell apart in U.S. data. We show that firm valuation increases with cash-flow ownership in the hands of the largest shareholder. This result is consistent with a large 10 Including in the regression only firms with families as the largest controlling shareholder, however, we find that, for these firms, pyramid structures are negatively related to firm value at a statistically significant 10 percent level. 136 A Reader in International Corporate Finance 2770 The Journal of Finance literature on the positive incentive effects associated with increased cash- flow rights in the hands of a single or few shareholders. We also find a negative entrenchment effect with large controlling shareholders: Increases in control rights by the largest shareholder are accompanied by declines in firm values. This negative effect is particularly severe for large deviations between control and ownership rights. When investigating individual ownership types, we find that our results appear to be driven by family control. We also provide support for the pre- dictions of theoretical studies that separating control rights and cash-flow rights can create agency costs larger than the costs associated with a con- trolling shareholder who also has a majority of cash-flow rights. Because concentrated corporate ownership is predominant in most countries outside the United States, these findings may have relevance worldwide. The re- sults suggest that the risk of expropriation of minority shareholders by large, controlling shareholders is an important principal­agent problem in most countries. The degree to which certain ownership and control structures are associ- ated with entrenchment discounts likely depends on economy-specific cir- cumstances. These may include the quality of banking systems, the legal and judicial protection of individual shareholders, and the degree of finan- cial disclosure required. This is especially the case for a number of the econ- omies in this study, because they have been identified as having deficient corporate governance and weak institutional development. The exact mag- nitude to which institutional differences across economies affect the valua- tion discount is an important issue for future research. REFERENCES Asian Company Handbook, 1998, winter edition Toyo Keizai Shinposha, Tokyo . Bebchuk, Lucian, 1999, A rent-protection theory of corporate control and ownership, NBER working paper 7203. Bebchuk, Lucian, Reinier Kraakman, and George Triantis, 2000, Stock pyramids, cross- ownership, and dual class equity: The creation and agency costs of separating control from cash flow rights, in Randall K. Morck, ed.: Concentrated Corporate Ownership University of Chicago Press, Chicago, IL . Breusch, Trevor, and Adrian Pagan, 1980, The Lagrange multiplier test and its applications to model specifications in econometrics, The Review of Economic Studies 47, 239­253. Campbell, John, 1996, Understanding risk and return, Journal of Political Economy 104, 298­345. Claessens, Stijn, Simeon Djankov, Joseph Fan, and Larry Lang, 1999a, The expropriation of minority shareholders: Evidence from East Asia, World Bank, Washington, DC. Claessens, Stijn, Simeon Djankov, Joseph Fan, and Larry Lang, 1999b, The patterns and val- uation effects of corporate diversification: A comparison of the United States, Japan, and other East Asian economies, World Bank, Washington, DC. Claessens, Stijn, Simeon Djankov, and Larry Lang, 2000, The separation of ownership and control in East Asian corporations, Journal of Financial Economics 58, 81­112. Greene, William, 1997, Econometric Analysis MacMillan, New York . Grossman, Sanford, and Oliver Hart, 1988, One-share, one-vote, and the market for corporate control, Journal of Financial Economics 20, 175­202. Harris, Milton, and Artur Raviv, 1988, Corporate governance: Voting rights and majority rules, Journal of Financial Economics 20, 203­235. Chapter Four 137 Incentive and Entrenchment Effects of Large Shareholdings 2771 Jensen, Michael, and William Meckling, 1976, Theory of the firm: Managerial behavior, agency costs, and ownership structure, Journal of Financial Economics 3, 305­360. Johnson, Simon, Peter Boone, Alasdair Breach, and Eric Friedman, 2000, Corporate governance in the Asian financial crisis, 1997­1998, Journal of Financial Economics 58, 141­186. La Porta, Rafael, Florencio Lopez-de-Silanes, and Andrei Shleifer, 1999, Corporate ownership around the world, Journal of Finance 54, 471­518. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny, 2002, In- vestor protection and corporate valuation, Journal of Finance 57, 1147­1170. McConnell, John, and Henri Servaes, 1990, Additional evidence on equity ownership and cor- porate value, Journal of Financial Economics 27, 595­612. Morck, Randall, Andrei Shleifer, and Robert Vishny, 1988, Management ownership and market valuation: An empirical analysis, Journal of Financial Economics 20, 293­315. Morck, Randall, David Stangeland, and Bernard Yeung, 2000, Inherited wealth, corporate con- trol and economic growth: The Canadian disease, in Randall K. Morck, ed.: Concentrated Corporate Ownership University of Chicago Press, Chicago, IL . Nenova, Tatiana, 2001, The value of a corporate vote and private benefits: Cross-country analy- sis, Manuscript, Harvard University. Shleifer, Andrei, and Robert W. Vishny, 1997, A survey of corporate governance, Journal of Finance 52, 737­783. Stulz, René, 1988, Managerial control of voting rights: Financing policies and the market for corporate control, Journal of Financial Economics 20, 25­54. Wolfenzon, Daniel, 1999, A theory of pyramidal structures, Manuscript, Harvard University. Chapter Five 139 THE JOURNAL OF FINANCE VOL. LIX, NO. 2 · ·APRIL 2004 Private Benefits of Control: An International Comparison ALEXANDER DYCK and LUIGI ZINGALES ABSTRACT We estimate private benefits of control in 39 countries using 393 controlling blocks sales. On average the value of control is 14 percent, but in some countries can be as low as ­4 percent, in others as high a +65 percent. As predicted by theory, higher private benefits of control are associated with less developed capital markets, more concentrated ownership, and more privately negotiated privatizations. We also ana- lyze what institutions are most important in curbing private benefits. We find evidence for both legal and extra-legal mechanisms. In a multivariate analysis, however, media pressure and tax enforcement seem to be the dominating factors. THE BENEFITS OF CONTROL OVER corporate resources play a central role in mod- ern thinking about finance and corporate governance. From a modeling device (Grossman and Hart (1980)) the idea of private benefits of control has become a centerpiece of the recent literature in corporate finance, both theoretical and empirical. In fact, the main focus of the literature on investor protection and its role in the development of financial markets (La Porta, Lopez-de-Salines, and Shleifer (2000)) is on the amount of private benefits that controlling sharehold- ers extract from companies they run. In spite of the importance of this concept, there are remarkably few estimates of how big these private benefits are, even fewer attempts to document empir- ically what determines their size, and no direct evidence of their impact on financial development. All of the evidence on this latter point is indirect, based on the (reasonable) assumption that better protection of minority sharehold- ers is correlated with higher financial development via its curbing of private benefits of control (La Porta et al. (1997)). The lack of evidence is no accident. By their very nature, private benefits of control are difficult to observe and even more difficult to quantify in a reliable Dyck is from the Harvard Business School and Zingales is from the University of Chicago. Chris Allen, Mehmet Beceren, and Omar Choudhry provided invaluable research assistance in preparing the data. We thank Andrew Karolyi, John Matsusaka, David Moss, Tatiana Nenova, Krishna Palepu, Mark Roe, Julio Rotemberg, Abbie Smith, Debora Spar, Per Stromberg, Rene Stulz, an anonymous referee, Richard Green (the editor), and seminar participants from Georgetown University, Harvard Business School, the NBER corporate finance program, University of Chicago, the University of Pennsylvania (Wharton), and the University of Southern California, and the University of Toronto for helpful comments. We also gratefully acknowledge financial support from the Division of Research, Harvard Business School, the Center for Research on Security Prices, and the George Stigler Center at the University of Chicago. Any errors are our own. 537 140 A Reader in International Corporate Finance 538 The Journal of Finance way. A controlling party can appropriate value for himself only when this value is not verifiable (i.e., provable in court). If it were, it would be relatively easy for noncontrolling shareholders to stop him from appropriating it. Thus, private benefits of control are intrinsically difficult to measure. Two methods have been used in attempting to quantify them. The first one, pioneered by Barclay and Holderness (1989), focuses on privately negotiated transfers of controlling blocks in publicly traded companies. The price per share an acquirer pays for the controlling block reflects the cash flow benefits from his fractional ownership and the private benefits stemming from his controlling position in the firm. By contrast, the market price of a share after the change in control is announced reflects only the cash flow benefits noncontrolling share- holders expect to receive under the new management. Hence, as Barclay and Holderness have argued, the difference between the price per share paid by the acquiring party and the price per share prevailing on the market reflects the differential payoff accruing to the controlling shareholder. In fact, after an adjustment, this difference can be used as a measure of the private benefits of control accruing to the controlling shareholder. The second method relies on the existence of companies with multiple classes of stock with differential voting rights. In this case, one can easily compute the market value of a vote (Lease, McConnell, and Mikkelson (1983, 1984), DeAn- gelo and DeAngelo (1985), Rydqvist (1987)). On a normal trading day market transactions take place between noncontrolling parties who will never have direct access to the private benefits of control. Hence, the market value of a vote reflects the expected price a generic shareholder will receive in case of a control contest. This in turn is related to the magnitude of the private ben- efits of control. Thus, if one is willing to make some assumptions on the prob- ability a control contest will arise, the price of a voting right can be used to estimate the magnitude of the private benefits of control (Zingales (1994, 1995a)). In this paper we use the Barclay and Holderness (1989) method to infer the value of private benefits of control in a large (39) cross section of countries. Based on 393 control transactions between 1990 and 2000 we find that on average corporate control is worth 14 percent of the equity value of a firm, ranging from a ­4 percent in Japan to a +65 percent in Brazil. Interestingly, the premium paid for control is higher when the buyer comes from a country that protects investors less (and thus is more willing or able to extract private benefits). This and other evidence suggest that our estimates capture the effect the institutional environment has on private benefits of control. Given the large number of transactions from countries with different levels of financial development in our data set, we are able to provide a direct test of several theoretical propositions on the effects private benefits of control have on the development of financial markets. Theory predicts that where private benefits of control are larger, entrepreneurs should be more reluctant to go pub- lic (Zingales (1995b)) and more likely to retain control when they do go public (Zingales (1995b) and Bebchuk (1999)). In addition, where private benefits of control are larger a revenue maximizing Government should be more likely Chapter Five 141 Private Benefits of Control 539 to sell a firm through a private sale than through a share offering (Zingales (1995b) and Dyck (2001)). We find strong evidence in support of all these predictions. A one standard deviation increase in the size of the private benefits is associated with a 67 per- cent reduction in the ratio of external market capitalization of equity to GNP, an 11 percent reduction in the percentage of equity held by noncontrolling share- holders, and a 36 percent increase in the number of privatized companies sold in private negotiations rather than through public listings. This evidence gives support to the prominent role private benefits have come to play in corporate finance. While the existence of private benefits is not necessarily bad, their negative effect on the development of security markets raises the question of what affects their average size across countries. Thus far, the literature has emphasized the law as the primary mechanism to curb private benefits by giving investors leverage over controlling shareholders. The right to sue management, for in- stance, limits the discretionary power of management and, with it, the ability to extract private benefits (Zingales (1995a)) and so does any right attributed to minority shareholders (La Porta et al. (1997)). A common law legal origin is similarly argued to constrain management by lowering the standard of proof in legal suits and increasing the scope of management decisions subject to judicial review (Johnson et al. (2000)). Consistent with this literature, we analyze the effect the law has on the size of private benefits. Besides the law, we also consider extra-legal institutions, which have been mentioned in the literature as possible curbs for private benefits: competition, labor pressures, and moral norms. To these well-known mechanisms we add two: public opinion pressure and corporate tax enforcement. Reputation is a powerful source of discipline, and being ashamed in the press might be a pow- erful deterrent (Zingales (2000)), especially where the press is more diffused. Similarly, effective tax enforcement can prevent some transactions (such below market transfer prices) that expropriate minority shareholders. We find that a high level of diffusion of the press, a high rate of tax compliance, and a high de- gree of product market competition are associated with lower private benefits of control. Given the noisiness of the proxies used and the paucity of degrees of freedom, it is impossible to establish reliably which factor is more important. That in a multivariate analysis newspapers' circulation and tax compliance are most important suggests these extra legal mechanisms deserve further study. Our paper complements and expands the existing work in this area that focuses on the voting premia such as Zingales (1998), who assembles estimates of the voting premium across seven countries, and Nenova (2001a), who uses the price of differential voting shares in 18 countries. We complement the existing work by providing an alternative estimate of the private benefits of control, available for a broader cross section of countries. While in a few cases our estimates differ from Nenova's (she finds that both Brazil and Australia have a ratio of value of control to value of equity equal to 0.23, while we find only 0.02 for Australia and 0.65 for Brazil), overall our estimates are remarkably 142 A Reader in International Corporate Finance 540 The Journal of Finance similar. Moreover, we are able to understand the differences between the two sets of estimates in terms of a sample selection bias present in estimates based on differential voting shares. These findings give confidence that the extraction of private benefits is a real phenomenon, which can be consistently estimated. Our paper also expands the existing work. The estimates for 39 countries allow us to test several theoretical propositions on the effects private benefits of control have on the development of financial markets. Our large sample of countries and their institutional variation enable us to test alternative theories of the major factors driving the magnitude of private benefits of control and to identify some new ones. The rest of the paper proceeds as follows. Section I discusses how the mea- sure developed by Barclay and Holderness (1989) relates to the magnitude of the private benefits of control. Section II describes the data used and presents our estimates. Section III uses these estimates to test several theoretical predic- tions regarding the effects private benefits of control have on the development of markets. Section IV analyzes the correlation between the magnitude of the pri- vate benefits of control and the various institutional characteristics. Section V discusses our findings and concludes. I. Theoretical Framework A. What Are Private Benefits of Control? The theoretical literature often identifies private benefits of control as the "psychic" value some shareholders attribute simply to being in control (e.g., Harris and Raviv (1988) and Aghion and Bolton (1992)). Although this is cer- tainly a factor in some cases, it is hard to justify multimillion dollar premia with the pure pleasure of command. Another traditional source of private benefits of control is the perquisites enjoyed by top executives (Jensen and Meckling (1976)). The use of a company's money to pay for perquisites is the most visible but not the most important way in which corporate resources can be used to the sole (or main) advantage of the controlling party. If the law does not effectively prevent it, corporate resources can be appropriated by the large shareholder through outright theft. Fortunately such activities, while documented in a few cases, are generally rare. Nevertheless, there are several reasons why more moderate versions of these strategies might be more pervasive. Educated economists can legitimately dis- agree on what is the "fair" transfer price of a certain asset or product. As a result, small deviations from the "fair" transfer price might be difficult or im- possible to prove in court. If these small deviations are applied to large volume trade, however, they can easily generate sizeable private benefits. Similarly, it is easy to disagree over who is the best provider of an asset or product when the relationship might involve considerations of quality and price. Orconsiderthevalueoftheinformationacorporateexecutiveacquiresthanks to his or her role in the company. Some of this information pertains directly to Chapter Five 143 Private Benefits of Control 541 the company's business while some reflects potential opportunities in other more or less related areas. It is fairly easy for a controlling shareholder to choose to exploit these opportunities through another company he or she owns or is associated with, with no advantage for the remaining shareholders. The net present value of these opportunities represents a private benefit of control. The common feature of all the above examples is that some value, whatever the source, is not shared among all the shareholders in proportion of the shares owned, but it is enjoyed exclusively by the party in control. Hence, the name private benefits of control. Control does not only confer benefits: sometimes it involves costs as well. Maintaining a controlling block, for instance, forces the largest shareholder to be not well diversified. As a result, it might value the controlling block less. At the same time, a fledging company might inflict a loss in reputation to the controlling party and, in some extreme cases, even some legal liabilities. For this reason we do not necessarily expect all our estimates to be always positive. In particular, we expect a higher frequency of negative value of control for financially distressed companies (see also Barclay and Holderness (1989)). Note that the existence of private benefits of control is not necessarily in- efficient. First of all, private benefits might be the most efficient way for the company to capture some of the value created. Imagine, for instance, that a cor- porate executive acquires valuable information about investment opportunities in other lines of businesses, which the company cannot or does not want to pur- sue. The executive could sell this information in the interest of shareholders. But the price she will be able to fetch is probably very low. Thus, it might be efficient that the executive exploits this opportunity on her own. Second, even if the extraction of private benefits generates some inefficiency, their existence might be socially beneficial, because their presence makes value-enhancing takeovers possible (Grossman and Hart (1980)). Given the difficulties in distinguishing whether private benefits are socially costly, consistently in this analysis we shy away from any welfare considera- tion. Even the implications of the effects of private benefits on the development of security markets should be interpreted as a positive statement, not a nor- mative one. In fact, in at least one of the models from where these implications are derived (Zingales (1995b)), the level of private benefits has no efficiency consequences, but only distributional ones.1 B. How to Measure Private Benefits? Unfortunately, it is very difficult to measure the private benefits directly. Psy- chic values are intrinsically difficult to quantify, as is the amount of resources captured by the controlling shareholder to her own benefit. As argued above, a controlling party will find it possible to extract corporate resources to his or her benefit only when it is difficult or impossible to prove that this is the case. In 1Bebchuk and Jolls (1999) discuss additional issues associated with a welfare evaluation of private benefits. 144 A Reader in International Corporate Finance 542 The Journal of Finance other words, if private benefits of control were easily quantifiable, then those benefits would not be private (accruing only to the control group) any longer because outside shareholders would claim them in court. Nevertheless, there are two methods to try to assess empirically the mag- nitude of these private benefits of control. The first one, pioneered by Barclay and Holderness (1989), is simple. Whenever a control block changes hands, they measure the difference between the price per share paid by the acquirer and the price quoted in the market the day after the sale's announcement. As we will show momentarily, this difference (which we shall call the control premium) represents an estimate of private benefits of control enjoyed by the controlling party. The second method of estimating the value of private benefits of control uses the price difference between two classes of stock, with similar or identical div- idend rights, but different voting rights. If control is valuable, then corporate votes, which allocate control, should be valuable as well. How valuable? It de- pends on how decisive some votes are in allocating control and how valuable control is. If one can find a reasonable proxy for the strategic value of votes in winning control--for example in forming a winning coalition block--then one can infer the value of control from the relationship between the market price of the votes and their strategic role. This is the strategy followed by Rydqvist (1987), Zingales (1994, 1995a), and Nenova (2001a). Both methods suffer from a common bias: They capture only the common value component of private benefits. If an incumbent enjoys a psychic benefit from running the family company, this value is unlikely to be shared by any other potential buyer and hence is unlikely to be reflected into the value of a controlling block when this changes hands (and hence in the value of a voting right). If, as it is likely, psychic benefits are more idiosyncratic to the control- ling shareholder, then companies with large nonmonetary private benefits are less likely to change hands (it is more difficult to find somebody that values control more than the incumbent) and when they do, they are likely to exhibit lower control premia.2 Hence, both methods tend to underestimate the value of control, and more so in countries where the major source of private benefits is nonpecuniary. 3 Besides this bias, both methods have pluses and minuses. The estimates ob- tained using the control premia method are relatively model free (albeit, see Section II.C. below). If we are careful in isolating only the transactions that transfer control, we do not have to worry about the proper model of how private benefits will be shared among different parties and what is the probability of a takeover (e.g., Nicodano and Sembenelli (2001)). On the other hand, sales of controlling blocks are relatively rare and might not occur randomly over time. Furthermore, any systematic overpayment or any delay in incorporating 2 The reason why a superior voting share trades at a premium is that its holder expects to receive a differential premium (see Zingales (1995b)). Hence, if a potential buyer is not willing to pay any more for control, the premium disappears. 3 We thank the referee for pointing out this bias. Chapter Five 145 Private Benefits of Control 543 public information can bias the estimates (a problem we will deal with in Section III.E.). Estimates obtained using dual class shares are often based on many firms and therefore are less likely to be driven by outliers. On the other hand, dual class shares are not allowed in every country. Hence, the second method limits the number of countries that can be included in the study. More importantly, the proportion of dual class companies differs widely across countries. Hence, the estimates obtained using the second method represent a differently selected universe of companies in each country. In any case, given the importance of private benefits in our understanding of corporate finance, it makes sense to explore both approaches. Nenova (2001a) has followed the voting rights ap- proach while we use control premia. C. Theoretical Relation between Control Premium and Size of the Private Benefits of Control An implicit assumption in the Barclay and Holderness (1989) approach for estimating private benefits is that the sale price reflects the buyers' willing- ness to pay. However, as Nicodano and Sembenelli (2001) point out, if there is imperfect competition in the market for controlling blocks, the Barclay and Holderness approach can misestimate private benefits. We illustrate this point with a simple bargaining model. Let , on the interval [0, 1], be the bargaining power of the controlling share- holder selling out, Bs the level of private benefits extracted by the seller ,b (buyer), and Ys the level of security benefits generated by the seller (buyer), ,b then the price P paid for a controlling block of shares with cash flow rights, on the interval [0, 1], is P = (Bb + Yb) + (1 - )(Bs + Ys) (1) and the per share price of the controlling block equals P Bb + (1 - )Bs (2) = + Yb + (1 - )Ys. To compute the control premium, Barclay and Holderness (1989) subtract from equation (2) the price prevailing in the market after the announcement that control has changed hands, which should equal to Yb. Thus, they obtain Bb + (1 - )Bs (3) - (1 - )(Yb - Ys). They then multiply this price difference by the size of the controlling block . Hence, their estimate of private benefits of control B^ is B^ = Bb + (1 - )Bs - (1 - )(Yb - Ys). (4) 146 A Reader in International Corporate Finance 544 The Journal of Finance In a perfectly competitive market ( = 1), B^ collapses to Bb and thus the control premium is a legitimate estimate of the private benefits of control the buyer expects to enjoy. When the market is not perfectly competitive, but the security value is the same for the buyer and the seller (Yb = Ys), B^ is still a legitimate es- timate of the private benefits of control, albeit this time it represents a weighted average of the private benefits of the seller and those of the buyer. The problem arises when the security values are different (Yb = Ys). By sub- tracting the price after the announcement from the per share price paid for the controlling block (the step from equation (2) to equation (3) above), Barclay and Holderness implicitly assume that the seller is able to capture the full value of the security benefits produced by the buyer. When this is not true, B^ mises- timates the average value of private benefits, where the extent of this bias is represented by the term (1 - )(Yb - Ys). To understand this bias, consider the other extreme case, where the buyer has all the bargaining power, ( = 0). In this case, B^ collapses to Bs - (Yb - Ys). Intuitively, the sale price of the controlling block does not reflect the differen- tial ability of the new buyer to create security benefits, while the price on the exchange does reflect this ability. Hence, B^ misestimates the value of private benefits by the difference in security value times the amount of security value contained in the controlling block ( ). Since the magnitude of this bias is zero if = 1 and B - (Yb - Ys) when = 0, in general it is (1 - )(Yb - Ys). All the terms in this bias, except for the bargaining power of the seller, are observable. Hence, if we can estimate , we can adjust our estimates. II. Data and Descriptive Statistics An example motivates our sample selection strategy and definition of our dependent variable. In January 1999 Ofer Brothers Investment Limited, an investment vehicle for Sami and Yuli Ofer of Israel, bought 53 percent of the shares and control of Israel Corporation Limited from the Eisenberg family. The price per share for the control block was reported to be 508 shekels per share while the exchange price after announcement of the transfer was 363 shekels per share. The price premium paid per share for the controlling block over the postannouncement price in this case is 40 percent. A better measure of the value of the private benefits of control is the total premium paid divided by the equity value of the firm. In this example, the Ofer brothers paid a 40 percent premium relative to the postannouncement price for 53 percent of the firms' equity, which produces an estimate of private benefits as a percentage of equity of 21 percent. This example turns out to be fairly typical of Israeli deals where we calculate a mean private benefit as a percentage of equity of 27 percent and a median value of 21 percent. As suggested by this example, to construct a measure of private benefits, we need to identify transactions that meet at least three criteria. First, the trans- action must involve a transfer of a block of shares that convey control rights. Second, we need to observe the price per share for the control block. Third, we have to observe the exchange price after the market has incorporated the Chapter Five 147 Private Benefits of Control 545 identity of the new acquirer in its expectation of future cash flow. We also add a fourth criterion, implicit in this choice of an Israeli deal--both the control and the postannouncement market prices should not be restricted by regulation. Many countries do not follow the Israeli (and U.S.) approach of allowing buyers and sellers to determine their own prices but impose some link between the exchange and the control price. As we will explain, we will eliminate all these cases from our sample. A. Identifying Transactions To identify transactions that convey control rights we use the SDC interna- tional mergers and acquisitions database. SDC describes its sources as: "Over 200 English and foreign language news sources, SEC filings and their interna- tional counterparts, trade publications, wires and proprietary surveys of invest- ment banks, law firms, and other advisors." The database provides extensive information on transactions that involve transfers of blocks of shares that may convey control, including details of the parties to the transaction, the value of the transaction, and the date of announcement and conclusion of the transac- tion. SDC provides extensive international coverage with 7,144 transactions in 1990 (including 396 transactions from non-OECD countries) and steadily increasing numbers over the decade, including 21,881 transactions in 1999 (in- cluding 3,300 from non-OECD countries). To identify candidates for control sales, we began with the complete set of control transactions in publicly traded companies during the period 1990 to 2000. We then restricted our attention to completed purchases of blocks larger than or equal to 10 percent of the stock.4 Since we wanted transactions that conveyed control, we further restricted our attention to transactions that result in the acquirers moving from a position where they hold less than 20 percent of the shares to a position where they have assembled more than 20 percent of the shares. We exclude all transactions that were conducted through open market purchases and were identified by SDC as tender offers, spinoffs, re- capitalizations, self-tenders, exchange offers, repurchases, and acquisitions of remaining interest. We further restricted ourselves to transactions where there was a reported transaction value or price per share in the control block. We refined our sample by exploiting additional available qualitative data to screen out transactions that do not involve control transfers (e.g., trans- fer of shares among subsidiaries of common parent, where acquirer is not the largest shareholder) or were problematic for other reasons (e.g., involved re- lated parties, reported price per share based on securities that could not be valued objectively, transfer involved the exercise of options). This step involved reading multiple news stories for every transaction resulting from searches of Lexis-Nexis and Dow-Jones Interactive to confirm the details of the transaction 4We have also explored the robustness of our results if we were to further restrict this criterion and exclude deals where block is less than 15 percent. The results are unchanged although we lose some countries as a result of a lack of observations. 148 A Reader in International Corporate Finance 546 The Journal of Finance collected by SDC and collecting ownership information through use of company annual reports and other sources. This process significantly increased our con- fidence in the observations included in the data set, but inevitably involved greater use of discretion in determining whether an observation was included in our data set. To ensure the availability of exchange prices, we restricted ourselves to trans- actions involving companies available in the Datastream International database. To implement the criterion that the difference between the con- trol price and the exchange price not be driven by legal requirements, we excluded observations driven by legal requirements. We first excluded all in- stances where the controlling block was purchased as part of a public offer, as in this circumstance there are usually laws that require all shareholders be treated equally. We researched rules regarding mandatory tender offers across different countries and only include transactions where there is no forced link- age between prices for the control block and prices on the exchange. For ex- ample, in Britain where the city code on takeovers requires that those who purchase a stake greater than or equal to 30 percent of the shares make an equal offer to all remaining shareholders on the same terms as the block sale, we restrict our attention to block sales less than 30 percent. As an illustration of the importance of this legal threshold, more than one quarter of our obser- vations in Britain are between 29 and 30 percent, with a median block size of 25 percent. Finally, we eliminated all transactions where there are ex ante or ex post indications (in SDC synopsis, news stories, or Datastream) of a tender offer for the remaining stock in the six months following the announcement. This criterion, also used by Barclay and Holderness (1989), is meant to eliminate events where the expectation of a tender offer distorts the value of minority shares. Table I summarizes our variable definitions and sources. The data appendix provides a more complete description of the construction of our sample. Appendix Table AI lists countries and rules regarding control transactions. Appendix Table AII lists the number of equities available for Datastream in each sample year from each of our countries. B. Descriptive Statistics of the Raw Control Premium Table II presents descriptive statistics of the block premia from our sample by country in which the acquired firm is located. After imposing our criteria, we have an unbalanced panel of 393 observations from 39 countries for the time period 1990 to 2000.5 The sample includes more than 40 observations from ac- tive equity markets such as the United Kingdom and the United States. For 5We only include countries in our analysis if there were two or more transactions over our sample period. The final sample is based on all of the data available over the 10-year sample period for every country aside from the U.S. For the U.S., there were many more potential observations and we limited ourselves to an initial sample based on the first 20 transactions for each calendar year over our 10-year sample period that met our sample selection criteria. Chapter Five 149 Private Benefits of Control 547 Table I Description of Variables Variable Description Block premia as a The block premia are computed as the difference between the price per percentage of the share paid for the control block and the price on the Exchange two value of equity days after the announcement of the control transaction, divided by the price on the Exchange after the announcement and multiplied by the proportion of cash flow rights represented in the controlling block. Securities Data Corporation, Datastream International, 20-Fs, Company annual reports, Lexis-Nexis, Dow-Jones interactive, various country sources including ISI Emerging markets and country company yearbooks. The change in The difference between the security value of the buyer (market price at security value t + 2) and of the seller (market price at t - 30) normalized by the market price at t + 2. We subtract from this amount the percentage difference in the level of the market index over the same time period (between date t + 2 and t - 30 normalized by the level of the index at date t + 2). Datastream International. Majority block A dummy variable that takes the value one if the control block includes 50 percent of all shares or 50 percent of all voting shares. Securities Data Corporation, 20-Fs, Company annual reports, Lexis-Nexis, Dow-Jones interactive, various country sources including ISI Emerging markets and country company yearbooks. Another large A dummy variable that takes the value one if there is another shareholder shareholder with a stake in excess of 20 percent after the block sale. Securities Data Corporation, Company annual reports, Lexis-Nexis, Dow-Jones interactive, various country sources including ISI Emerging markets and country company yearbooks. Financial distress A dummy variable that takes the value one if earnings per share in the target are zero or negative in the year of the block trade or the year preceding the block trade. Datastream International. Seller identity Dummy variables to identify seller identity. Includes dummies for individual seller, the company itself (through new share issues), a corporate entity, or unknown. A corporate entity is the most prevalent category and is the excluded category. Securities Data Corporation, Company annual reports, Lexis-Nexis, Dow-Jones interactive, various country sources including ISI Emerging markets and country company yearbooks. Foreign acquirer A dummy variable that takes the value one if the acquirer is from a different country than the target. Where acquirer is unknown, assume acquirer is from same country as target. Securities Data Corporation. Acquirer identity Dummy variables to identify if the acquirer is a public company, subsidiary, the government, or a private company. A public company is the most prevalent group and is the excluded category. Securities Data Corporation. Cross listed Dummy variable that takes the value one if the company's stock is listed in the United States either on an exchange, on Portal under rule 144A, or as an over-the-counter listing. Data provided by Andrew Karolyi based on Citibank Universal Issuance Guide. 150 A Reader in International Corporate Finance 548 The Journal of Finance Table I--Continued Variable Description Industry type Dummy variables that indicate the acquired companies industrial type (two digit SIC). Manufacturing is the most prevalent group and is the excluded category. Securities Data Corporation, Global Access. Agriculture, forestry, & fishing (01­09) Mining (10­14) Construction (15­17) Manufacturing (20­39) Transportation & pub. utilities (40­49) Wholesale trade (50­51) Retail trade (52­59) Finance, insurance, & real estate (60­67) Services (70­89) Tangibility of The median value of the percentage of total assets that are fixed for U.S. assets firms in the same three digit SIC code as the acquired firm. Securities Data Corporation, Standard and Poor's Research Insight (COMPUSTAT) Stock market As a measure of valuation uncertainty we use the average R2 of synchronicity firm-level regressions of bi-weekly stock returns on local and U.S. market indexes in each country in 1995. Returns include dividends and are trimmed at 25 percent. Higher levels indicate that stocks are more likely to move together. Morck et al. (2000). Control premia "Control benefits based on a sample of 661 dual-class firms in 18 based on voting/ countries using data for 1997. Control benefits are extracted from the nonvoting shares total value of the votes in the control block, based on a baseline control contest model in the case of a dual class firm," Nenova (2001a). Nenova (2001a). Log GDP per capita Average log GDP per capita 1970 to 1995. World Bank. Ownership "The average percentage of common shares owned by the three largest concentration shareholders in the 10 largest nonfinancial, privately owned domestic firms in a given country. A firm is considered privately owned if the state is not a known shareholder in it." La Porta et al. (1998). La Porta et al. (1998), derived from: Moodys International, CIFAR, EXTEL, Worldscope, 20-F's, Price-Waterhouse, and various country sources. Initial public "Ratio of the number of initial public offerings of equity in a given offerings/ country to its population (in millions) for the period 1995:7­1996:6." La population Porta et al. (1997). La Porta et al. (1997), derived from: Securities Data Corporation, AsiaMoney, LatinFinance, GT Guide to World Equity Markets, and World Development Report, 1996. Number of listed "Ratio of the number of domestic firms listed in a given country to its firms/ population population (in millions) in 1994." La Porta et al. (1997). La Porta et al. (1997) derived from: Emerging Market Factbook and World Development Report, 1996. External market "The ratio of the stock market capitalization held by minorities to gross capitalization/ national product for 1994. The stock market capitalization held by GNP minorities is computed as the product of the aggregate stock market capitalization and the average percentage of common shares not owned by the top three shareholders in the ten largest nonfinancial, privately owned domestic firms in a given country. A firm is considered privately owned if the State is not a known shareholder in it." La Porta et al. (1997). La Porta et al. (1997), derived from Moodys International, CIFAR, EXTEL, Worldscope, 20-F's, Price-Waterhouse, and various country sources Chapter Five 151 Private Benefits of Control 549 Table I--Continued Variable Description Takeover laws A dummy variable that takes the value one if the transaction takes place in the presence of a legal requirement to make a mandatory offer if the shareholding after acquisition exceeds a threshold, yet the transaction lies below the threshold. Data presented in Appendix Table I. ISSA Handbook, 6th and 7th editions, EIU country commerce guides, exchange web sites, country company handbooks. Accounting "Index created by examining and rating companies' 1990 annual reports standards on their inclusion or omission of 90 items. These items fall into seven categories (general information, income statements, balance sheets, funds flow statement, accounting standards, stock data, and special items). A minimum of three companies in each country were studied. The companies represent a cross section of various industry groups; industrial companies represented 70 percent, and financial companies represented the remaining 30 percent." La Porta et al. (1998). La Porta et al. (1998) derived from: International accounting and auditing trends, Center for International Financial Analysis and Research. Antidirector rights "An index aggregating shareholder rights formed by adding one when (1) the country allows shareholders to mail their proxy vote to the firm, (2) shareholders are not required to deposit their shares prior to the general shareholder's meeting, (3) cumulative voting or proportional representation of minorities in the board of directors is allowed, (4) an oppressed minorities mechanism is in place, (5) the minimum percentage of share capital that entitles a shareholder to call for an extraordinary shareholder's meeting is less than or equal to 10 percent (the sample median), or (6) shareholders have preemptive rights that can be waived only by a shareholders' vote. The index ranges from zero to six." La Porta et al. (1998). La Porta et al. (1998) based on company law or commercial code. Pistor et al. (2000) for Czech Republic and Poland. Rule of law "Assessment of the law and order tradition in the country produced by the country risk rating agency International Country Risk (ICR). Average of the months of April and October of the monthly index between 1982 and 1995. Scale from zero to 10, with lower scores for less tradition for law and order (we changed the scale from its original range going from zero to six)." La Porta et al. (1998). La Porta et al. (1998), derived from: International Country Risk guide. Pistor et al. (2000) for Czech Republic and Poland. Competition laws Response to survey question, "competition laws prevent unfair competition in your country?" Higher scores suggest agreement that competition laws are effective. World Competitiveness Yearbook, 1996. Newspaper Circulation of daily newspapers/population. UNESCO Statistical circulation/ yearbook 1996, as reported in World Competitiveness Report, for population Taiwan based on Editors and Publishers' Association Year Book and AC Nielsen, Hong Kong, as reported in "Asian Top Media--Taiwan" www.business.vu.edu Violent crime This is a proxy for moral norms suggested by Coffee (2001). It is the reported number of murders, violent crimes, or armed robberies per 100,000 population. Interpol and country data for 1993 as reported in World Competitiveness Yearbook, 1995. 152 A Reader in International Corporate Finance 550 The Journal of Finance Table I--Continued Variable Description Catholic This is another proxy for moral norms suggested by Stulz and Williamson (2001). The indicator variable takes the value one if the country's primary religion is Catholic. 2000 CIA World Factbook as reported in Stulz and Williamson (2001). Labor power We use as an index of labor power the extent of statutory employee protections based on the average of indicators on regular contracts (procedural inconveniences, notice and severance pay for no-fault-dismissals, difficulty of dismissal) and short-term contract (fixed-term and temporary) as derived in Pagano and Volpin (2000). An alternate index is the weighted average of indicators on regular contracts, short-term contract and collective dismissals as derived by Pagano and Volpin (2000). The index is from Pagano and Volpin (2000) based on data from OECD 1999. Tax compliance "Assessment of the level of tax compliance. Scale from 0 to 6 where higher scores indicate higher compliance. Data is for 1995." La Porta et al. (1999). The Global Competitiveness Report 1996 as reported in La Porta et al. (1999). Cheating on taxes Response to survey question "cheating on taxes if you have a chance is justified?" Scaled from one to 10 where one is never justified and 10 is always justified. World Values Survey, 1996. Legal origin Identifies the legal origin of the company law or commercial code of each country. Categories include English common law, French commercial code, German commercial code, Scandinavian civil law, and former Soviet bloc country. La Porta et al. (1998), derived from Reynolds and Flores (1989). some countries despite looking at the full population of control transactions available in SDC, we have relatively few observations as a result of the combi- nation of weak coverage by Datastream, few reported prices for control sales, and limited observability of control premia as a result of laws regarding tender offers in case of control sales. The rank ordering of countries by control premia is very similar using mean and median values suggesting that our results are not driven by a few outliers. The first column of Table III presents the average control premium by coun- try, computed as the coefficient of fixed country effects in a regression where the dependent variable is B^ (calculated as in (4)) normalized by Yb. Over- all, the average control premium is 14 percent if each country has an equal weight and 10 percent if each observation receives equal weight. In 10 of our 39 sample countries, we find that the control premia exceeds 25 percent of equity value. These high private benefit countries include Argentina, Austria, Colombia, Czech Republic, Israel, Italy, Mexico, Turkey, and Venezuela (of these Brazil has the highest estimated value of 65 percent. At the other extreme, we have 14 countries where private benefits are 3 percent of the value of equity or less.) These low private benefit countries include Australia, Canada, Finland, France, Hong Kong, Japan, Netherlands, New Zealand, Norway, Singapore, South Africa, Taiwan, United Kingdom, and United States. Chapter Five 153 Private Benefits of Control 551 Table II Block Premium as Percent of Firm Equity This table presents descriptive statistics by country on the block premia in the 393 control block transactions we study. The block premia are computed as the difference between the price per share paid for the control block and the price on the Exchange two days after the announcement of the control transaction, divided by the price on the Exchange after the announcement and multiplied by the proportion of cash flow rights represented in the controlling block. Securities Data Corporation, Datastream International, 20-Fs, Company annual reports, Lexis-Nexis, Dow-Jones interactive, various country sources including ISI Emerging markets and country company yearbooks. Number of Standard Number of Positive Country Mean Median Deviation Minimum Maximum Observations Observations Argentina 0.27 0.12 0.26 0.05 0.66 5 5 Australia 0.02 0.01 0.04 -0.03 0.11 12 8 Austria 0.38 0.38 0.19 0.25 0.52 2 2 Brazil 0.65 0.49 0.83 0.06 2.99 11 11 Canada 0.01 0.01 0.04 -0.02 0.06 4 2 Chile 0.18 0.15 0.19 -0.08 0.51 7 6 Colombia 0.27 0.15 0.34 0.06 0.87 5 5 Czech Republic 0.58 0.35 0.80 0.01 2.17 6 6 Denmark 0.08 0.04 0.11 -0.01 0.26 5 3 Egypt 0.04 0.04 0.05 0.01 0.07 2 2 Finland 0.02 0.01 0.06 -0.07 0.13 14 9 France 0.02 0.01 0.11 -0.10 0.17 4 2 Germany 0.10 0.11 0.14 -0.24 0.32 17 14 Hong Kong 0.00 0.02 0.05 -0.12 0.05 8 6 Indonesia 0.07 0.07 0.03 0.05 0.09 2 2 Israel 0.27 0.21 0.32 -0.01 0.89 9 8 Italy 0.37 0.16 0.57 -0.09 1.64 8 7 Japan -0.04 -0.01 0.09 -0.34 0.09 21 5 Malaysia 0.07 0.05 0.10 -0.08 0.39 40 30 Mexico 0.34 0.47 0.35 -0.04 0.77 5 4 Netherlands 0.02 0.03 0.05 -0.07 0.06 5 4 New Zealand 0.03 0.04 0.09 -0.17 0.18 16 12 Norway 0.01 0.01 0.05 -0.05 0.13 12 8 Peru 0.14 0.17 0.11 0.03 0.23 3 3 Philippines 0.13 0.08 0.32 -0.40 0.82 15 11 Poland 0.13 0.12 0.11 0.02 0.28 4 4 Portugal 0.20 0.20 0.14 0.11 0.30 2 2 Singapore 0.03 0.03 0.03 -0.01 0.06 4 3 South Africa 0.02 0.00 0.03 0.00 0.07 4 2 South Korea 0.16 0.17 0.07 0.04 0.22 6 6 Spain 0.04 0.02 0.06 -0.03 0.13 5 4 Sweden 0.07 0.03 0.09 -0.01 0.22 11 10 Switzerland 0.06 0.07 0.04 0.01 0.15 8 8 Taiwan 0.00 0.00 0.01 -0.01 0.00 3 2 Thailand 0.12 0.07 0.19 -0.08 0.64 12 11 Turkey 0.37 0.11 0.58 0.05 1.41 5 5 United Kingdom 0.01 0.00 0.04 -0.06 0.17 41 21 United States 0.01 0.02 0.09 -0.20 0.25 46 27 Venezuela 0.27 0.28 0.21 0.04 0.47 4 4 Average/Number 0.14 0.11 0.18 -0.04 0.48 393 284 154 A Reader in International Corporate Finance 552 The Journal of Finance These estimates assume the seller has all the bargaining power. If this as- sumption is not valid, these estimates would be downward biased on average, since the bias is proportional to -(Yb - Ys), which on average is negative six percentage points.6 More importantly, the bias can differ across deals and coun- tries, since both the improvement in security value, (Yb - Ys), and the percent- age of voting rights contained in the controlling block, , differ across deals (and thus a fortiori across countries). All the terms of this bias, (1 - )(Yb - Ys), are observable, except for the seller's bargaining power (). Unfortunately, we do not have enough degrees of freedom to estimate reliably a country-specific . Therefore, we initially restrict it to be equal across all transactions, and we estimate (1 - ) as a coefficient of the term (Yb - Ys) inserted in our previous regression (column 1 of Table III), where the dependent variable is B^ and the YB other explanatory variables are the country fixed effects. The estimate of so obtained equals 0.655 and is statistically different from zero at the 10 percent level. Not only does this estimate lie in the [0, 1] interval, as predicted by the model, but it is also very reasonable. It suggests that on average the seller captures two-thirds of the gains from trade. Table III (column 2) presents the estimates of the country fixed effects ob- tained in this way. A few countries see the estimated private benefits of control increase after this adjustment. For example, the estimate for the United States goes from 1.0 to 2.7 percent. The overall ranking, however, remains substan- tially unchanged.7 Of course, the seller's bargaining power is unlikely to be constant across all deals. The question is how potential differences in bargaining power can affect our estimates. If differences in the bargaining power have large effects on our private benefits estimates, then our estimates should be correlated with proxies for the buyer's bargaining power. A proxy for the buyer's bargaining power is the announcement return experienced by the buyer of the controlling block. In our sample, we have 203 observations where the acquirer is a publicly traded company and the stock price is reported in Datastream for 115 of those. As we show later (in Table IV, panel B), we regress the acquirers' cumulative abnormal returns around the transaction on our estimates of private benefits. We find no significant correlation between the two, thus potential biases do not seem to be of the first order. Nevertheless, to address this problem in the next section, we introduce additional control variables, which will proxy for deal-specific differences in the relative bargaining power of the parties involved. Our major concern, however, is not variability across deals, but systematic variability across countries, which might bias our cross-country comparison. In particular, if competition for control is stronger in some countries than others, imposing an equal will artificially inflate the estimates of private benefits in countries with strong competition and reduce them in others. To exclude this 6 With an average controlling block size of 37 percent, the maximum downward bias, on average, in our sample of 2.2 percent if the seller has no bargaining power and there is no bias if sellers have all the bargaining power. 7 While is constrained to be fixed across countries, the term (Yb - Ys) does differ across deals (and a fortiori across countries). Thus, the adjustment introduced in column 2 could alter the relative ranking across countries. Chapter Five 155 Private Benefits of Control 553 possibility, we divide countries in quartiles according to our estimates of private benefits and we re-estimate , imposing it to be equal only within each quar- tile. We find that countries with higher levels of private benefits have lower estimated lambdas than countries with lower levels of private benefits. These results suggest that our assumption of equal across countries tends, if any- thing, to dampen the cross-country differences in the level of private benefits. C. Differences in Deal and Firm Characteristics Cross-country differences in the level of private benefits could be driven by systematic differences in deal characteristics and firm characteristics, which affect the amount of control transferred, the size of the private benefits, and the relative bargaining power of the parties involved. To increase confidence that our estimates of block premia reflect country differences rather than other characteristics, we generate revised estimates based on a regression of our raw data against firm and deal characteristics.8 C.1. Differences in the Extent the Block Carries Control First of all, we assume that all transactions transfer absolute control. This is probably incorrect. The transfer of a 20 percent block does not carry the same amount of control as the transfer of a 51 percent block. Similarly, the transfer of a 30 percent block when there is another shareholder controlling 20 percent carries less control than the transfer of the same block when the rest of the shares are dispersed. Thus, per given size of private benefits control blocks above 50 percent are likely to fetch a higher price. Similarly, the presence of another large shareholder (a stake in excess of 20 percent) should reduce the premium.9 In our sample, 27 percent of the transactions involve sales that exceed 50 per- cent of the votes, and in 16 percent of the cases the acquirer has to deal with another large shareholder with more than a 20 percent stake.10 As shown in Table III, ceteris paribus an absolute majority of votes increases the value of a controlling block by 9.5 percent of the total value of equity, significant at the 5 percent level. Contrary to expectations, the presence of another large shareholder has a positive effect on the premium, but this is not statistically significant. C.2. Differences in the Extent of the Seller's Bargaining Power In estimating the private benefits of control, we assumed that the seller's bargaining power is constant across deals. As we just discussed, variations in 8 Summary statistics for the characteristics of the deals that we use later in our empirical anal- ysis are provided in our earlier working paper, Dyck and Zingales (2002a). 9 In Canada and Australia we used 15 percent since exceeding 20 percent would trigger a manda- tory offer for remaining shares. 10 An alternative approach to identify the likelihood that a stake brings control is to calculate a Shapley value associated with control. Unfortunately, we were not able to collect information on a consistent basis on the ownership status of other shareholders. For example, some countries might report the presence of all shareholders with stakes that exceed 5 percent while other countries might only report holdings that exceed 10 percent or higher. 156 554 Table III Estimated Block Premia by Country The dependent variable is the block premia as a percent of firm equity. Each regression includes country fixed effects. In addition, in column (2) we introduce the buyer's proportion of the difference in security value between the buyer and seller. In column (3) we introduce several deal characteristics: whether it is a majority block, whether there is another large shareholder, whether the firm is in financial distress, whether the block was created by issuing new shares, whether the buyer is foreign, and if the firms' shares are cross listed in the United States. In column (4) we introduce several industry and seller/buyer characteristics: identity of the buyer (individual, government, subsidiary, dispersed), identity of the seller (individual, government, unknown), two-SIC code industry dummies, and the proportion of fixed to total assets. Definitions for each of the variables can be found in Table I. All regressions are estimated by OLS. Robust standard errors are in parentheses. A Reader Dependent Variable: Block Premium Independent Variables (1) (2) (3) (4) T in Buyer's proportion of change in security value -0.345 (0.214) -0.323 (0.211) -0.319 (0.209) he International Stake greater than 50% 0.095 (0.039) 0.095 (0.039) Journal Another large shareholder 0.041 (0.043) 0.018 (0.040) Financial distress in selling firm -0.054 (0.028) -0.043 (0.028) Sold through new share issue 0.041 (0.057) 0.034 (0.059) of Buyer is foreign 0.069 (0.034) 0.065 (0.036) F Corporate Cross-listed in the US -0.062 (0.040) -0.067 (0.039) inance Buyer individual or private -0.042 (0.026) Buyer government 0.008 (0.046) Buyer subsidiary -0.001 (0.049) Finance Buyer dispersed or unknown -0.039 (0.044) Seller individual 0.021 (0.029) Seller government 0.008 (0.100) Seller unknown 0.028 (0.031) Fixed assets as percent of total -0.097 (0.062) Industry--Agriculture, Forestry, Fishing -0.03 (0.050) Industry--Mining -0.071 (0.071) Industry--Construction -0.027 (0.042) Industry--Transportation & utilities 0.066 (0.031) Industry--Wholesale Trade 0.046 (0.047) Industry--Retail Trade -0.057 (0.055) Industry--Finance, Insurance, Real Est. 0.055 (0.045) Industry--Services -0.024 (0.038) Argentina 0.268 (0.111) 0.268 (0.112) 0.158 (0.131) 0.197 (0.123) Australia 0.020 (0.013) 0.029 (0.018) -0.001 (0.034) 0.051 (0.052) Austria 0.383 (0.099) 0.364 (0.082) 0.318 (0.054) 0.309 (0.050) Brazil 0.650 (0.252) 0.653 (0.249) 0.606 (0.229) 0.652 (0.245) Canada 0.013 (0.017) 0.016 (0.009) -0.06 (0.056) -0.055 (0.075) Chile 0.183 (0.069) 0.213 (0.070) 0.149 (0.065) 0.165 (0.067) Colombia 0.273 (0.142) 0.274 (0.129) 0.197 (0.137) 0.242 (0.132) Czech Republic 0.578 (0.312) 0.600 (0.320) 0.462 (0.297) 0.555 (0.325) Denmark 0.077 (0.048) 0.076 (0.045) 0.039 (0.050) 0.036 (0.070) Egypt 0.038 (0.024) 0.035 (0.015) -0.050 (0.061) 0.025 (0.082) Finland 0.025 (0.016) 0.028 (0.018) -0.016 (0.027) -0.010 (0.036) Private France 0.019 (0.052) 0.035 (0.049) 0.040 (0.059) 0.080 (0.077) Germany 0.095 (0.034) 0.090 (0.033) -0.020 (0.052) 0.016 (0.059) Hong Kong 0.003 (0.019) 0.026 (0.021) 0.045 (0.033) 0.040 (0.044) Indonesia 0.072 (0.017) 0.032 (0.025) -0.034 (0.040) 0.043 (0.047) Benefits Israel 0.270 (0.107) 0.284 (0.113) 0.238 (0.108) 0.259 (0.114) Chapter Italy 0.369 (0.199) 0.378 (0.201) 0.323 (0.191) 0.311 (0.192) Japan -0.043 (0.021) -0.041 (0.020) -0.070 (0.044) -0.038 (0.054) Malaysia 0.072 (0.017) 0.072 (0.014) 0.063 (0.018) 0.093 (0.032) of Five Mexico 0.345 (0.146) 0.381 (0.150) 0.296 (0.143) 0.322 (0.144) Control Netherlands 0.016 (0.020) -0.031 (0.047) -0.054 (0.068) -0.015 (0.060) New Zealand 0.027 (0.024) 0.044 (0.027) -0.028 (0.042) 0.026 (0.046) 555 157 158 556 Table III--Continued Dependent Variable: Block Premium Country Fixed Effects (1) (2) (3) (4) A Reader Norway 0.015 (0.014) 0.019 (0.019) 0.007 (0.026) 0.052 (0.041) Peru 0.142 (0.053) 0.121 (0.075) 0.067 (0.080) 0.060 (0.082) Phillipines 0.129 (0.083) 0.169 (0.085) 0.115 (0.081) 0.142 (0.079) T in Poland 0.133 (0.052) 0.134 (0.041) 0.003 (0.081) 0.041 (0.092) he International Portugal 0.203 (0.073) 0.215 (0.075) 0.159 (0.052) 0.197 (0.059) Singapore 0.030 (0.016) 0.027 (0.019) 0.024 (0.035) 0.042 (0.069) Journal South Africa 0.017 (0.015) 0.035 (0.019) -0.045 (0.061) 0.005 (0.072) South Korea 0.157 (0.027) 0.146 (0.036) 0.086 (0.066) 0.088 (0.086) Spain 0.041 (0.027) 0.049 (0.026) 0.021 (0.042) 0.047 (0.058) of Sweden 0.074 (0.027) 0.083 (0.029) 0.033 (0.047) 0.041 (0.057) F Corporate Switzerland 0.063 (0.015) 0.061 (0.016) -0.073 (0.056) -0.067 (0.074) inance Taiwan -0.004 (0.004) -0.011 (0.005) -0.047 (0.039) -0.040 (0.074) Thailand 0.125 (0.054) 0.142 (0.057) 0.073 (0.080) 0.121 (0.084) Turkey 0.371 (0.246) 0.362 (0.226) 0.276 (0.232) 0.346 (0.249) Finance United Kingdom 0.014 (0.007) 0.016 (0.009) 0.000 (0.019) 0.040 (0.033) United States 0.01 (0.013) 0.027 (0.016) 0.002 (0.031) 0.044 (0.038) Venezuela 0.270 (0.094) 0.305 (0.103) 0.256 (0.105) 0.221 (0.112) Number of observations 393 393 393 393 R-squared 0.389 0.399 0.431 0.459 significant at 10% level; significant at 5% level; significant at 1% level. Chapter Five 159 Private Benefits of Control 557 the seller's bargaining power can affect our estimates of the private benefits of control: Per given size of private benefits of control, the lower the seller's bargaining power, the lower our estimates. We try to control for these differences with three proxies. First, if the company is in financial distress, the seller is more likely to be forced to sell. Hence, her bargaining power is smaller. As a proxy for financial distress, we create a dummy variable that takes value one if earnings per share are zero or negative in the year of the block trade or the year preceding the block trade.11 In our sample, 27 percent of the firms are in financial distress in the year of the block trade and 23 percent in the year preceding the block trade. As expected, firms in financial distress exhibit a control premium that is 5.4 per- centage points lower. This effect is statistically significant at the 10 percent level. Similarly, that the acquisition of a controlling block takes the form of an equity infusion probably indicates that a company needs to raise equity, a sign of a weak bargaining position. We insert a dummy if the block was formed by newly issued equity (16 percent). This method is particularly diffused in Japan where in a majority of cases control is transferred by a financially distressed company via a private placement of newly issued equity. This clustering underscores the importance of controlling for industry firms' and deals' characteristics, to avoid attributing to the Japan institutional framework a feature due to the particular economic phase Japan has been going through during our sample period. Contrary to expectations, the fact a block was created through a new equity offering has a positive effect on the premium, but this is not statistically significant. Finally, companies that can be acquired by foreigners are likely to face more competition. We attempt to capture this possibility by introducing a dummy variable equal to one if the acquirer is foreign. As a result of the increased competition, the bargaining power of the seller in these transactions is likely to be bigger. We find that foreign buyers pay a premium of 6.9 percent that is statistically significant at the 5 percent level. C.3. Cross Listing in the United States Coffee (1999), Reese and Weisbach (2001), and Doidge, Karolyi, and Stulz (2001) argue that foreign companies list in the United States to submit them- selves to tougher governance rules and precommit to extract less private ben- efits of control. Since we want to measure the country-specific value of private benefits, we want to control for companies that might have lower than average private benefits due to their borrowing of foreign institutions. To this purpose we insert a dummy variable equal to one for any company that is cross listed in the United States as well as in its home market.12 As expected, cross-listed companies enjoy lower private benefits, although given the paucity of cross 11While other measures of cash flow are preferable, earnings per share is one of the few data items consistently reported in Datastream for the companies in our database. 12We obtained the list of cross listing from Doidge, Karolyi, and Stulz (2001). We thank Andrew Karolyi for kindly providing us with the data. 160 A Reader in International Corporate Finance 558 The Journal of Finance listed companies in our sample (23), the statistical significance of this effect is just below conventional levels (p-value = 12 percent). C.4. Estimates of Private Benefits Controlling for Differences in Deal and Firm Characteristics After inserting all these deals' and firms' characteristics into our basic re- gression, we re-estimate the country fixed effects. The results are reported at the bottom of column 3 in Table III. Since many of the control variables in- cluded capture part of the value of control, the country fixed effects cannot any longer be interpreted as the estimates of the average value of private ben- efits in that country, but only as relative rankings. Including these controls dramatically lowers the ranking for countries characterized by higher than av- erage incidence of foreign acquirers and sales of majority stakes like Germany, Switzerland, Egypt, and Poland. On the one hand, these estimates represent an improvement over our raw data, for they keep constant deal characteristics. On the other hand, they suffer from an econometric problem. To estimate the impact of these deal and firm characteristics, we had to assume that this impact is constant across countries. In some cases this assumption might be untenable. The difference between acquiring a 51 percent stake rather than a 30 percent one might be huge in a country where private benefits of control are large, but it might be small or even irrelevant in a country where the private benefits of control are very tiny.13 The regression, however, imposes the same effect on all the countries, underestimating differences across countries. In the rest of the paper, where we explore the effects and causes of these cross-country differences, we focus on this refined measure that controls for deal (and other) characteristics. But recognizing that this procedure may bias the results because deal characteristics may not be constant across countries, we also test results without controls. D. Differences in Industry and Buyer/Seller Characteristics Cross-country differences could also arise because of other differences in in- dustry and deal characteristics. Private benefits might differ across industry. The media industry, for instance, is often mentioned (Demsetz and Lehn (1985)) as an industry where private benefits are larger. Similarly, individuals might value opportunities to consume prerequisites more highly than corporate block- holders (see e.g., Barclay and Holderness (1989)). We want to make sure our cross-country comparison is not affected by any systematic difference in the industry characteristics of the deals or the nature of the seller and the buyer. 13Since we have enough observations for the U.S. (46), we can assess the realism of our assump- tion by estimating the same specification restricted to U.S. data. While the other coefficients are very similar to the ones reported in Table IV, the coefficient of the majority block dummy is small and insignificant. "Imposing" to the U.S. the same majority dummy effect as other countries, thus, will distort its average level of private benefits upward. Chapter Five 161 Private Benefits of Control 559 For this reason, we re-estimate the country averages, controlling for differences in industry characteristics and identity of the controlling party. To capture industry differences, we introduce an industry dummy based on the two-digit SIC code of the acquired firm. About three quarters of our trans- actions are accounted for by manufacturing (39 percent); finance, insurance, and real estate (24 percent); and services (10 percent). In a crude way these controls capture differences in private benefits linked to product market com- petition. Second, we construct a measure of tangibility of assets (percentage of total assets that are fixed) based on the three-digit SIC code the acquired firm belongs to. The argument for this control is that insiders will have more difficulty diverting resources if assets are tied down and easily observable, as is the case with tangible assets. To avoid potential endogeneity problems, we use U.S. averages (see Rajan and Zingales (1998)).14 Table III column 4 shows that firms with more tangible assets have lower private benefits, and firms in wholesale trade, finance (financial, insurance, and real estate sector), and transportation and utilities have a higher level of private benefits than firms in manufacturing, although these differences are not statistically significant. We also collected information on the identity of the acquirer and the seller. To identify characteristics of the seller, we focus exclusively on the news stories, identifying whether the seller is an individual, the company itself (through new share issues), a corporate entity, or unknown. Here we find the most common seller to be a corporation, followed next by individuals (18 percent), new share issues (16 percent), unidentified (8 percent) and the government (3 percent). We use SDC data to identify whether the acquirer is a public company, subsidiary, the government, or a private company. The typical transaction in our sample involves a public acquirer (41 percent), although private acquirers are also very common (41 percent). We provide a further classification using news stories and the SDC synopsis field. We identify 13 percent of our transactions involving an individual acquirer, using as our criteria whether the stories mention the name of an individual or if the private company involved is identified with a particular individual. We also identify 4 percent of transactions involving a financial intermediary who purchases the shares and then resells the shares to institutional investors. We interpret these acquisitions as the dispersal of the controlling stake. None of these buyer or seller characteristics turns out to be significant. At the bottom of column 4 of Table III, we report the estimates of the country average level of private benefits after we control for the above differences in level of private benefits across industries. The relative ranking, however, does not seem to be affected very much by these industry controls. Finally, the level of private benefits extracted might be endogenous to the size of the controlling block. Large shareholders who retain a larger block of 14We derive U.S. measures in a two-step procedure. First, we computed the average ratio of fixed assets (property plant and equipment) to total assets for all companies that in each three-digit SIC- code for the period 1990 to 1999. Then we took the median value across all companies. We then impute this value for all of the companies in our sample. 162 A Reader in International Corporate Finance 560 The Journal of Finance equity have less of an incentive to dilute minority shareholders, because they internalize more the inefficiency they generate (see Burkart, Gromb and Pa- nunzi (1998)). For this reason, in an unreported regression, we also inserted the size of the controlling block . Since it has no effect on the value of control, we dropped it. E. Alternative Interpretations Thus far, we have interpreted block premia as indicative of private bene- fits. Yet, there are alternative interpretations that we need to consider. The most important alternative interpretation, already considered and rejected by Barclay and Holderness (1989) in their U.S. sample, is that control premia arise from a systematic overpayment, possibly due to a winner's curse problem. As in Barclay and Holderness (1989), we check for this possibility by looking at the announcement effect on the stock price of the acquiring company. If these premia reflect overpayments, acquiring firms should experience negative returns at the announcement of the transaction. In our sample, we have 203 observations where the acquirer is a publicly traded company and the stock price is reported in Datastream for 115 of those. Table IV presents the results of our analysis. Inconsistent with the overpayment hypothesis, the mean value of the announcement effect is slightly positive (0.5 percent) and not statistically different from zero. Another implication of the overpayment hypothesis is that the buyer's an- nouncement return should be negatively related to the size of the control pre- mium. In Table IV, panel B, we regress the acquirers' cumulative abnormal returns around the transaction on the raw control premium. We focus on a 16- day event window (t - 8 to t + 7) to allow for information about the transaction to be leaked in advance or to be communicated slowly to the market although results are not significantly affected by the choice of window. The coefficient is indeed negative, but is neither economically nor statistically significant (coef- ficient of ­0.018, p-value of 0.64). The results above reject the hypothesis that on average the control premium is due to overpayment. It is still possible, thus, that this might be true in some countries. In particular, we are concerned that in less developed countries, where there is more uncertainty about the value of a company, the winner's curse is more severe leading to a higher apparent premium and distorting our international comparisons. While such behavior is inconsistent with a rational bidding process (Milgrom and Weber (1982)), we still want to ensure it is not present in the data.15 As a measure of the degree of company-specific informa- tion available we use the synchronicity measure developed by Morck, Yeung, and Yu (2000). This is a measure of how much stock prices move together. The more they move together, the less company-specific information is revealed. If there is more overpayment in less developed markets, we should observe that the control premium is more negatively correlated with the acquirer's return in 15A rational bidder knows that if he bids his valuation he will overpay, the more so the more uncertainty there is about the fundamental value of the asset. Thus, the more uncertainty there is, the more he will shade his bid. Chapter Five 163 Private Benefits of Control 561 Table IV Does the Control Premium Come from Overpayment? Panel A reports the summary statistics of the cumulative abnormal returns (CAR) of the stock price of the acquiring company around the date the acquisition of the controlling block is announced. We use a window from eight days prior to the announcement to seven days after the announcement. We have 203 transactions involving publicly traded acquirers, of which 115 have stock prices reported in Datastream. Panel B reports the OLS estimates of two regressions, where the dependent variable is the acquirer' CAR from t - 8 and t + 7 and the independent variables are: (1) the raw block premia (Table III column 1); (2) the raw block premia (Table III column 1) interacted with a measure of how much stock prices move together at the country level (see Morck et al. (2000)). Definitions for each of the variables can be found in Table I. Robust standard errors are in parentheses. Panel A: Cumulative Abnormal Returns of the Acquirer from t - 8 to t + 7 Mean 0.005 Median 0.000 Maximum 0.333 Minimum -0.408 Standard deviation 0.110 Number of observations 115 Panel B: Systematic Differences in Cumulative Abnormal Returns Dependent Variable: Cumulative Abnormal Return of Acquirer (from t - 8 to t + 7) Independent Variables (1) (2) Block premia -0.018 (0.040) -0.106 (0.156) Block premia × synchronicity in target nation 0.419 (0.835) Constant 0.007 (0.011) 0.007 (0.012) Number of observations 115 105 R-squared 0.001 0.008 a country with a high level of synchronicity. In fact, the interaction coefficient is positive and not statistically significant. A second alternative interpretation that could potentially explain a larger premia in underdeveloped markets is that the buyer has superior information and there is a delay in incorporating new information. On average, delays in adjusting will spuriously inflate our estimates of private benefits. To test for this possibility we re-estimated the private benefits using the market price 30 days after the announcement rather than two days after. The results (not reported) are virtually identical. If anything, the average premium in devel- oping countries, like Brazil, goes up rather than down. We also examined the cumulative abnormal returns to shareholders in target firms from two days to 30 days after the announcement and tested whether the initial level of private benefits was related to the subsequent cumulative abnormal returns. We found no such effect with an insignificant relationship between control premia and postannouncement returns (coefficient = 0.009, p-value = 0.80). 164 A Reader in International Corporate Finance 562 The Journal of Finance Another alternative interpretation focuses on liquidity differences between developed and less developed markets. Differences in liquidity cannot explain our findings either. While a lack of liquidity reduces the willingness to pay for shares on the exchange and this effect is more pervasive in less developed markets, the lack of liquidity also impacts the price that is paid for large blocks. Large noncontrolling blocks generally sell at a discount to the exchange price (Holthausen, Leftwich, and Mayers (1990)) and the more so the more illiquid is the market for the underlying stock. Thus, if the control value were zero there would be a bigger discount in less liquid markets for large blocks. Therefore liquidity differences suggest that, if anything, more underdeveloped countries should have smaller block premia, not larger ones. We are also concerned about a possible distortion due to selective nondisclo- sure. In fact, one of the criteria we had to impose to obtain our estimates was the observability of the price paid for the controlling block. A worrisome possibility is that in countries with better protection of investors, controlling parties are more fearful to disclose large premia. In such a case, we would estimate lower private benefits in the United States, not because they are indeed lower, but because large premia are less likely to be disclosed. To check for this possibility, we compute the percentage of deals we have to drop because the terms are not disclosed. On average, 33 percent of the deals do not disclose the terms, going from 0 percent in Taiwan and other countries to 70 percent in Austria and 82 percent in the Czech Republic. Contrary to the selective nondisclosure argument, we find that countries with higher premium tend to have a higher percentage of deals that are not disclosed (correlation 0.2, not statistically significant). Similarly, if we use as a proxy of shareholders' protection the antidirector rights index constructed by La Porta et al. (1997), we find (not surprisingly) that in countries that protect shareholders a greater percentage of deals are disclosed. In sum, if selective nondisclosure biases our results it biases them in the direction of attenuating the cross-country differ- ences rather than amplifying them. Finally, if the acquirers of the controlling block, for instance, already owned a large stake in the company beforehand, they might be willing to pay a pre- mium only because they internalize a fraction of the increase in the security value via their toeholds (Grossman and Hart (1980) and Shleifer and Vishny (1986)). Toeholds, however, are unusual in our sample. The average sharehold- ing prior to purchasing the control stake is just 1 percent, in 76 percent of the cases the acquirer has no prior shareholding, and in 86 percent of the cases the prior shareholding is less than 1 percent. Nevertheless, to examine the impact of a toehold we re-estimate the regressions in Table III (not reported) intro- ducing the initial toehold as an additional regressor. The initial toehold has a negative and statistical insignificant impact (p-value of 0.20 to 0.32) on our private benefits' estimates. All of our results are unaffected by the inclusion of this additional regressor. F. Are We Really Estimating Private Benefits? Therefore, we can reject all these alternative interpretations, but what evi- dence do we have that our estimates indeed capture private benefits of control? Chapter Five 165 Private Benefits of Control 563 At the anecdotal level, we have papers documenting the pervasiveness of self- dealing transactions in countries like Italy (Zingales (1994)) and the Czech Republic (Glaeser, Johnson, and Shleifer (2001)). It is reassuring, thus, that our estimated private benefits for these two countries are very high (respec- tively, 37 percent and 58 percent). It is particularly interesting to stress the difference between Poland and the Czech Republic. Both are former social- ist countries, with a similar level of GDP per capita. Nevertheless, our esti- mates are very different (11 percent for Poland and 58 percent for the Czech Republic). At a more systematic level, if our measures reflect the different ability to extract private benefits in different countries, they should be affected in pre- dictable ways by country-specific institutions that restrict the ability to extract private benefits. We will explore these implications in Section IV. One limitation with this approach, however, is that it is difficult to separate specific institu- tions from a broad institutional context. More subtle tests of whether these estimates really reflect the ability to extract private benefits are whether our estimated private benefits depend not only upon the institutional variables of the country of the company whose control has been acquired, but also on in- stitutions of the country of the acquiring company (when this is different) and on institutions of the country where a company's shares are listed (when a company cross lists in the United States). An acquirer coming from a country with less investor protection is better able to siphon out corporate resources from a subsidiary than an acquirer coming from a country with very rigid rules. This should result in a higher willingness to pay and, in a nonperfectly competitive market, at a higher price. Thus, we should observe higher estimated private benefits when the foreign acquirer comes from a country with poor protection of investors. For this reason, in Table V column 1, we re-estimate our basic specification (see Table III) inserting as an additional explanatory variable the interaction between the foreign acquirer's dummy (equal to one if the acquirer comes from a country different from the target) and a measure of the difference in legal protection between the two countries. This measure is the difference between the La Porta et al. (1998) measure of antidirector rights for the country of the acquiring company and the one for the country of the acquired company. As Table V shows, companies coming from more investor friendly countries pay, on average, a control premia that is 2.7 percent less, and this effect is statistically significant. In the bottom of Table V, we present country fixed effects with this control. The finding is interesting per se within the context of the debate on corporate governance convergence. Coffee (1999) predicts that companies from countries with better protection of investors will end up buying companies from countries with weaker protection. Our result suggests that in the presence of controlling blocks this might not be the case. Companies from countries with better investor protection are more limited in their ability to extract private benefits and thus ceteris paribus are able to bid less for the controlling block. This engenders the risk that controlling blocks may end up in the hands of companies from the countries with the worst rules, not the best ones. 166 A Reader in International Corporate Finance 564 The Journal of Finance Table V Does Legal Protection in the Investor's Country of Origin Affect the Acquirer's Willingness to Pay for Control? The dependent variable is the block premia as a percent of firm equity. The explanatory variables include all of the variables introduced in Table II column (4). In column 1, we include the interaction between the foreign acquirer's dummy (equal to one if the acquirer comes from a country different from the target) and a measure of the difference in legal protection between the two countries. This measure is the difference between the La Porta et al. (1998) measure of antidirector rights for the country of the acquiring company and the one for the country of the acquired company. In column 2, we include the interaction between the dummy for cross listing in the U.S. and a measure of the difference in investor protection between the U.S. and the country where the target firm is located. Robust standard errors are in parentheses. Dependent Variable: Block Premium Independent Variables (1) (2) Foreign acquirer dummy 0.063 (0.035) 0.060 (0.036) Cross listed in the US -0.060 (0.039) 0.113 (0.083) Interaction of relative strength of -0.027 (0.011) -0.028 (0.011) antidirector rights (home--target nation) and foreign acquirer Interaction of relative strength of -0.070 (0.034) antidirector rights (home--target nation) and cross listed in the US Variables Controlled for: Buyer's proportion of change y y in security value Ownership variables y y Financial distress y y Buyer identity y y Seller identity y y Industry group y y Tangibility of assets y y Country fixed effects Argentina 0.183 (0.114) 0.183 (0.113) Australia 0.054 (0.051) 0.052 (0.051) Austria 0.309 (0.048) 0.319 (0.051) Brazil 0.655 (0.245) 0.653 (0.245) Canada -0.059 (0.083) -0.052 (0.083) Chile 0.160 (0.065) 0.16 (0.065) Colombia 0.282 (0.131) 0.325 (0.128) Czech Republic 0.563 (0.328) 0.563 (0.330) Denmark 0.028 (0.065) 0.027 (0.065) Egypt 0.077 (0.085) 0.112 (0.093) Finland -0.002 (0.037) 0.002 (0.037) France 0.076 (0.077) 0.084 (0.078) Germany 0.038 (0.058) 0.041 (0.058) Hong Kong 0.039 (0.043) 0.008 (0.048) Indonesia 0.042 (0.046) 0.043 (0.045) Israel 0.254 (0.116) 0.252 (0.116) Italy 0.323 (0.193) 0.349 (0.199) Japan -0.032 (0.052) -0.039 (0.051) Malaysia 0.090 (0.033) 0.089 (0.033) Chapter Five 167 Private Benefits of Control 565 Table V--Continued Dependent Variable: Block Premium Independent Variables (1) (2) Mexico 0.348 (0.129) 0.396 (0.133) Netherlands -0.025 (0.062) -0.015 (0.062) New Zeland 0.027 (0.046) 0.028 (0.045) Norway 0.06 (0.042) 0.061 (0.043) Peru 0.076 (0.075) 0.08 (0.075) Phillipines 0.147 (0.079) 0.148 (0.080) Poland 0.045 (0.092) 0.039 (0.092) Portugal 0.204 (0.060) 0.207 (0.059) Singapore 0.046 (0.064) 0.038 (0.062) South Africa -0.014 (0.075) -0.014 (0.074) South Korea 0.128 (0.080) 0.137 (0.081) Spain 0.058 (0.053) 0.058 (0.052) Sweden 0.044 (0.057) 0.047 (0.056) Switzerland -0.054 (0.073) -0.051 (0.073) Taiwan -0.038 (0.074) -0.038 (0.073) Thailand 0.111 (0.080) 0.107 (0.080) Turkey 0.364 (0.246) 0.363 (0.246) United Kingdom 0.029 (0.034) 0.02 (0.033) United States 0.037 (0.038) 0.035 (0.038) Venezuela 0.234 (0.107) 0.268 (0.119) Number of observations 393 393 R-squared 0.466 0.470 significant at 10% level; significant at 5% level; significant at 1% level. This finding that the owners' identity (as reflected in the home country of the acquirer) is associated with the extent of private benefits also provides one rationale for the approach in many privatizations of not simply selling to the highest bidder and for the consistent finding in central and eastern Europe (Djankov and Murrell (2000)) of superior returns for firms sold to foreigners (most from countries with higher levels of antidirector rights than in the tran- sition countries) after controlling for possible selection issues. Cross-listed companies provide another test of whether these estimates re- flect the ability to extract private benefits. A subtler prediction of the argu- ment that cross-listing in the United States. acts as a precommitment is that the effect of this cross listing should be a function of the difference between the corporate governance rules in the United States and the rules facing the company in its home market.16 To test this hypothesis, we measure the superi- ority in governance as the difference between antidirector rights in the United States and antidirector rights in the target country. In Table V specification 2, 16This is the prediction that Doidge (2002) tests using companies with differential voting stock. He finds that the voting premium of companies cross listed in the United States is significantly lower. This is consistent with our findings and an additional confirmation that different methods lead to the same answer: private benefits exist and are important. 168 A Reader in International Corporate Finance 566 The Journal of Finance we again re-estimate our basic specification (see Table III) and include an inter- action term that is the product of the cross-listing dummy and the measure of the superiority of governance rules. We find a statistically significant negative effect of the superiority of governance rules on the control premia. This means that the reduction in private benefits with cross listing is greater for firms from countries that have weaker investor protections. These results provide direct support for the contention of Coffee (1999), Reese and Weisbach (2001), and Doidge et al. (2001) of a link between cross listing and private benefits. F.1. Comparing Control Premia Measures Another check that our estimates measure the value of control comes from comparing them with estimates of the value of control obtained using different methods. Nenova (2001a) provided the largest set of alternative estimates. By using the prices of shares with different voting rights, she estimates the value of control across 18 countries. Table VI (panel A) reports both her numbers and our numbers. The first two columns report the raw measure of private benefits (both Nenova's and ours) and the second two the adjusted measures, after controlling for extraneous factors, which might bias the estimates. In spite of the different method used, there is a remarkable similarity in findings. Our estimates for countries like Mexico and Germany are identical, and the overall correlation between our measures is 0.59 for the raw mea- sure and 0.62 for the refined measure (statistically different from zero at the 2 percent level).17 There are, however, notable exceptions. Nenova finds that both Australia and Brazil have a ratio of value of control to value of equity equal to 0.23, while we find only 0.02 for Australia and 0.65 for Brazil. What can explain these differences? As we discussed, both sets of measures can have pluses and minuses. One possible sample selection story that could account for these differences goes as follows. Companies are more likely to issue dual class shares when private benefits of control are large (Grossman and Hart (1988) and Zingales (1995b)). Hence, a measure of private benefits of control based on the voting premium of companies that issued dual-class shares tends to overestimate the value of con- trol. Most importantly, this upward bias is not homogeneous across countries, but it is more severe the fewer the percentage of dual class companies in the population of traded companies in a country. And this percentage varies widely across countries. The final column in Table VI reports the percentage of dual-class firms with prices available by Datastream as a percentage of the total population of Data- stream firms in the country in that year. In countries that allow dual-class shares, on average only 14 percent of the firms have two classes of shares traded. There is a wide cross-sectional variation: Brazil has 59 percent of such firms, while Australia and the United Kingdom have only 1 percent. 17If we exclude Brazil, as we should for reasons to be discussed in Section III.G., the correlation increases to 0.69 using the raw data and 0.86 using the refined data. Table VI Comparing Control Premia Measures Panel A reports Nenova's (2001a) estimates of the value of control based on the price difference between classes of shares with differential voting rights and ours, based on control block transactions. The first column reports Nenova's raw voting premium, defined as total vote value (value of a vote times number of votes) as a share of firm's market value. The second column reproduces our raw block premium (Table III column (1)). The third column reports Nenova's fixed effect estimates of the value of control, where she controls for differences in the dividend rights between the two classes of stock, differences in liquidity, and the presence of a conversion option (Nenova, Table VI, Col. 4). The fourth column reports our fixed effect estimates of the value of control (Table V). The fifth column reports the percentage of firms in Datastream sample that have multiple share classes with available price data, where the number of firms with multiple share classes is taken from Nenova and the number of firms with equity prices in Datastream for 1997 is reported in Appendix Table AII. Panel B reports OLS regressions of the difference between Nenova's control premia and Private ours. In column 1 there is the difference between the raw estimates, in column 2 the difference between the fixed effect estimates. The explanatory variable is the percentage of firms that have dual-class shares and price data available in each country (column 5 of Panel A). Robust standard errors are in parentheses. Benefits Panel A: Data Comparisons Chapter Raw Data Estimated Country Fixed Effects Premia Using Premia Using Percentage of Equities of Voting/Nonvoting Voting/Nonvoting with Dual-Class Control Five Shares (Nenova, Block Premia Shares (Nenova, Block Premia Shares and Available Country Table V (2000)) (Table III, Col 1) Table V, Col. 4 (2000)) (Table V, Col 2) Price Data Australia 0.232 0.020 0.185 0.052 0.01 Brazil 0.232 0.650 0.180 0.653 0.59 Canada 0.028 0.013 0.035 -0.052 0.04 Switzerland 0.054 0.063 0.054 -0.051 0.19 Chile 0.231 0.183 0.231 0.16 0.07 Germany 0.095 0.095 0.148 0.041 0.14 Denmark 0.008 0.077 0.009 0.027 0.20 Finland -0.050 0.025 0.058 0.002 0.24 France 0.281 0.019 0.282 0.084 0.02 United Kingdom 0.096 0.014 0.090 0.02 0.02 567 169 170 568 Table VI--Continued Panel A: Data Comparisons Raw Data Estimated Country Fixed Effects Premia Using Premia Using Percentage of Equities Voting/Nonvoting Voting/Nonvoting with Dual-Class A Shares (Nenova, Block Premia Shares (Nenova, Block Premia Shares and Available Reader Country Table V (2000)) (Table III, Col 1) Table V, Col. 4 (2000)) (Table V, Col 2) Price Data T Hong Kong -0.029 0.003 -0.029 0.008 0.01 in Italy 0.294 0.369 0.345 0.349 0.31 he International South Korea 0.289 0.157 0.338 0.137 0.11 Journal Mexico 0.364 0.345 0.460 0.396 0.06 Norway 0.058 0.015 0.058 0.061 0.11 Sweden 0.010 0.074 0.010 0.047 0.19 of United States 0.020 0.010 0.016 0.035 0.08 F Corporate South Africa 0.067 0.017 0.063 -0.014 0.07 inance Panel B: Can Differences between Benefits-Estimates Be Explained by Potential Selection Bias in Voting Rights Approach? Dependent Variable Finance Nenova Measure - Our Measure Refined Nenova Measure - Our Refined Measure Percentage of dual-class firms in country -0.873 (0.109) -0.816 (0.218) Constant 0.127 (0.029) 0.144 (0.032) Number of observations 18 18 Adjusted R-squared 0.76 0.63 significant at 1% level. Chapter Five 171 Private Benefits of Control 569 We test the possible effects of the sample selection described above by regress- ing the difference between Nenova's estimates and our estimates against the percentage of companies with dual class shares. If there exists a bias, we expect Nenova's estimates to exceed ours in countries with few dual-class stocks like Australia and the United Kingdom (i.e., a negative coefficient in the regression). This is indeed what we find. In countries where dual class shares are more rare Nenova's number significantly exceeds ours. The effect is economically very important. A one standard deviation increase in the percentage of dual class shares leads Nenova's estimates to exceed ours by 22 percentage points.18 This variable alone explains 76 percent of the difference in raw estimates and 63 percent of the difference in refined estimates. Overall, these results give confidence that the Barclay and Holderness method to estimate private benefits indeed measures private benefits (and not overpayment) and it does so introducing smaller biases than the alternative method. That the two sets of estimates differ in the way predicted by theory is also a strong indication these estimates are indeed measuring the value of private benefits of control. F.2. An Analysis of Outliers Another way to verify that we are indeed measuring private benefits of con- trol is an in-depth analysis of the outliers. In Brazil, we estimate private ben- efits to be 65 percent of the value of equity. Could private benefits really be this large, or is this finding the result of some problem in the way we in- fer private benefits? Nenova (2001b), as part of a study of the impact of le- gal reform on private benefits in Brazil, independently collected information on control sales in Brazil between 1995 and 2000, identifying eight transac- tions that meet our initial sample selection criteria, including six transactions not in our database.19In the sample of eight transactions (Nenova (2001b), Table III) she reports an average value of private benefits of 42 percent, not too dissimilar from our estimate. In addition, we asked a Brazilian investment bank to give us all the privatization data where the Government sold a con- trolling block of a firm already listed.20Their search produced 23 privatization transactions with the requisite data, including 21 transactions that were not 18Using differential voting shares to estimate the value of control can induce also another bias. When ownership is highly concentrated, the price of voting shares tends to underestimate the value of votes, because control is securely held in the hands of the largest shareholder. There is some weak evidence this might be the case if we use Nenova's (2001a) raw estimates. Nenova, however, is aware of this problem and in her regressions she controls for ownership concentration. Consistently, her refined measure seems completely unaffected by this bias. 19Her approach, albeit very similar, is not strictly comparable with our own, as she uses the price on the date of sale and compares the sale price with the price of voting shares on the exchange. 20This sample only includes transaction where sale price is cash. That is, we excluded privatiza- tions where sale price could include so-called "privatization currencies" that included government debt that was trading at a discount. 172 A Reader in International Corporate Finance 570 The Journal of Finance included in our original data set.21The average control premia in this sample is 129 percent. In sum, independent estimates lead to a very similar conclusion: Private benefits of control in Brazil are extremely high. F.3. Within-country Variation in Private Benefits Another check to verify whether our method captures private benefits is to see whether our estimates change when external conditions, which affect the ability to extract private benefits, change. While the fact that we have relatively few transactions from many countries limits our ability to systematically explore time series variation, at least for three events, we have this possibility. The first event we explore is the passage in Italy of a corporate governance reform in 1998, also known as the Draghi reform. Among other things, this reform made it easier for minority shareholders to sue management appointed by the controlling shareholder. Such reform should limit the ability to extract private benefits. When we segment our data into those observations before and after July 1998, we find that before the reform the average value of private benefits is 47 percent, while after the reform it is only 6 percent.22 The second event we explore focuses on Brazil in the 1990s where, as Nenova (2001b) reports, there were two important changes in the legal environment. The first change occurred on May 5, 1997 when Law 9457 was adopted. This law, designed to enhance government revenues from selling State-owned con- trolling blocks, eliminated several protections of minority investors: the right to be bought out at book value in case of major transactions, such as mergers and spinoffs, the requirement for acquirers to make a mandatory offer to other hold- ers of voting shares at the same price as the control block, etc. The elimination of these protections makes control more valuable. The second change was the passage of Instruction 299 by the Brazilian securities and exchange commission (CVM), reinstating these rights and adding new disclosure requirements. These legal changes suggest that private benefits will differ depending on which legal regime is in effect, with private benefits expected to be greatest in the period when Law 9457 was in effect, and lower both before and after. This is in fact what we find in our sample of transactions: the premia are highest during the period of law 9457 at 119 percent, with lower levels in the pre- 9457 period at 53 percent, and in the postinstruction 299 period of 37 percent. Similar findings are found using our methods in the Nenova sample (27 percent for the pre-9457 period to 61 percent in the 9457 period to 37 percent in the postinstruction 299 period). A similar trend is revealed in our privatization 21They identified 12 transactions where the stake sold was 19.26 percent, which we excluded because this level was below our selection criteria, but in Brazil accounted for 50.1 percent of the voting shares in the company. In addition, they were able to identify stock market prices for a number of firms that we were not able to collect using Datastream or were not identified by SDC. 22The p-value for the equality of the two means is only 21 percent, but this is not surprising given we have only six observations before and two afterward. Chapter Five 173 Private Benefits of Control 571 sample where we just have data for the first two periods (with values of 109 percent in the pre-9457 period increasing to 131 percent in the 9457 period). The third event we explore focuses on changes in the economic environment rather than changes in the legal regime to protect investors. It has been sug- gested that stealing will increase when the expected return on investment de- clines and that the Asian crisis presents such an event (Johnson et al. (2000)). We test for this, examining whether the levels of private benefits are different for emerging markets in Asia during the Asian crisis, where following Johnson et al. we define the crisis to be 1997 and 1998.23 Based on a regression of private benefits with country fixed effects we find that the Asian crisis period is indeed associated with higher private benefits (coefficient of 0.068), although this is not significant at conventional significance levels (p-value = 0.162). In sum, in all three instances, our estimates move as theory predicts private benefits should move. Having established some degree of confidence in our estimates, we now move to use them in international comparisons. III. Effects of Private Benefits on Financial Development A. Theoretical Predictions Wehaveshownthatthemagnitudeofprivatebenefitsofcontrolvariesgreatly across countries. We have not shown, however, that larger private benefits are necessarily more inefficient. Can we derive any implication on the effects of larger private benefits of control on the development of financial markets that is independent of their characterization as efficient of inefficient? The answer is yes. In countries where a controlling party can appropriate a larger share of the value of a company, entrepreneurs will be more reluc- tant to take their companies public. If they sell a minority position, outside investors will be willing to pay less for it than what it is currently worth to the entrepreneur, because they factor in the possibility a new acquirer will dilute the value of the company in the future. As a result, entrepreneurs are reluctant to sell (Zingales (1995b)). At the same time, when control value is high they do not want to sell a majority of votes in the market because they will not receive an adequate compensation for it. Atomistic shareholders will pay for the voting rights they expect to receive in a future tender offer. If, as it is likely to be the case, the market for corporate control is not perfectly competitive, atomistic shareholders will receive less in a tender offer than what a controlling share- holder would have obtained in a private negotiation (Zingales (1995b)). Hence, three implications follow: (1) Since fewer companies will list in countries with high private benefits of control, the importance of the equity market relative to GDP should be smaller; 23Specifically, countries included in this test include Hong Kong, Indonesia, Korea, Malaysia, Philippines, Singapore, Taiwan, and Thailand. 174 A Reader in International Corporate Finance 572 The Journal of Finance (2) Since incumbents are more likely to retain control after they take their company public in countries with high private benefits of control, the percentage of companies widely held should be smaller; (3) Since it is more profitable to sell control in a private negotiation in coun- tries with high private benefits of control, a revenue maximizing govern- ment should prefer to sell control in private transactions rather than in public offerings. All these predictions are independent of the direct welfare implications of pri- vate benefits of control. In fact, they are derived from Zingales (1995b), where private benefits of control have no efficiency consequences, but only distribu- tional ones. B. Test In Table VII we test these three predictions using our private benefits mea- sure as an independent variable. We focus on our estimated country fixed effects from Table V. Since our explanatory variable is estimated, OLS estimates are biased and inconsistent. Thus, we also report instrumental variable (IV) es- timates, where we use the family of origin of a country's legal system as an instrument for the extent of private benefits. As we show below in Table XI, legal origin is highly correlated with our private benefit measure. All of the reported results are robust to using the raw measure of private benefits from Table II in place of the estimated country fixed effects from Table V. We begin by focusing on the relation between the size of private benefits and ownership concentration (specification 1). As a measure of ownership concen- tration that is available for almost all of the countries in our data set we use the percentage of equity controlled by the three largest shareholders in the 10 largest nonfinancial firms where the state is not a shareholder (La Porta et al. (1998)). To control for other possible factors, we insert in all the regressions the log GDP per capita. As predicted, countries with higher private benefits have more concentrated ownership. A one standard deviation increase in the size of private benefits translates into 11 percent more of the equity held by the largest three share- holders in the instrumental variables specification. This simple specification seems to also have a very high explanatory power (r-squared = 0.45). In specification 2, we test the effect of private benefits on the way firms are privatized. Our dependent variable is the percentage of privatizations that took place as a private asset sale, rather than as a share offering from Megginson et al. (2000). Asset sales almost always involve the sale of a majority (or 100 per- cent) of the shares to a controlling shareholder or group. Share offerings dis- perse ownership to a greater extent. To control for other factors, we include not only the per capita GDP, but also the importance of the equity market, on the basis that governments are more likely to sell shares in public offerings if the market is more developed.24 24The results are robust to excluding this variable. Table VII Testing the Theoretical Predictions on the Effects of Private Benefits on Financial Market Development In specification 1 of Panel A the dependent variable is the average concentration of ownership as measured by the combined stakes of the three largest shareholders in the 10 largest nonfinancial, nonforeign corporations where the state is not a shareholder (see La Porta et al. (1997)). In specification 2 the dependent variable is the percentage of privatization transactions that took the form of an asset sale rather than a share offering (Megginson et al. (2000)). In Panel B the dependent variables are: (1) the number of initial public equity offerings in 1995 to 1996; (2) the number of listed domestic firms; (3) the ratio of the stock market capitalization held by minority investors to GNP (all from La Porta et al. (1997)). The explanatory variables are the average log GDP per capita 1970 to 1995 (World Bank) and our fixed effect estimates of the country average level of value of control from Table Private V. More complete variable descriptions and sources are provided in Table I. The instruments are the families of origin of a country's legal system (English, French, German, Scandinavian, and Soviet). Robust standard errors are in parentheses. Benefits Panel A: Dependent Variables: Ownership Structure Dependent Variables (1) (2) of Chapter Ownership Concentration Percentage of Privatizations as Asset Sales Control (3 largest) (not share offerings) Instrumental Instrumental Five Independent Variables OLS Variables OLS Variables Country control premia 0.365 (0.124) 0.591 (0.261) 0.999 (0.240) 2.005 (0.797) Log per capita income -0.047 (0.015) -0.033 (0.021) -0.024 (0.057) 0.022 (0.061) Constant 0.807 (0.127) 0.659 (0.207) 0.554 (0.505) 0.037 (0.583) Number of obs. 36 36 36 36 R-squared 0.445 0.276 573 175 176 574 Table VII--Continued A Reader Panel B. Dependent Variable: Capital Market Structure Based on Aggregate Data Dependent Variables T (1) (2) (3) he in International Initial Public Offerings Number of Listed Domestic Equity Market Journal in 1996/Population Firms/Population Capitalization/GNP Instrumental Instrumental Instrumental Independent Variables OLS Variables OLS Variables OLS Variables of Corporate Country control premia -2.753 (1.263) -12.66 (5.609) -24.03 (26.74) -199.3 (94.21) -1.265 (0.413) -3.747 (1.307) F inance Log per capita income 0.451 (0.195) -0.082 (0.419) 8.643 (3.079) -0.327 (5.711) -0.041 (0.065) -0.168 (0.103) Constant -2.315 (1.543) 3.472 (4.064) -45.60 (24.69) 51.57 (57.79) 0.943 (0.614) 2.319 (0.988) Number of obs. 34 34 37 37 37 37 Finance R-squared 0.203 0.168 0.213 significant at 10% level; significant at 5%; significant at 1% level. Chapter Five 177 Private Benefits of Control 575 We find that in countries with large private benefits, governments are more likely to divest companies through private sales. A one standard deviation in- crease in the size of private benefits translates into 36 percent more firms being privatized through private negotiations in the instrumental variables specifica- tion. These results are consistent with evidence from privatizations in specific countries. In Brazil, for example, government interest in receiving the control premia at the time of privatization led them to weaken existing protections for minority investors so that minority holders of voting shares no longer had the right to an equal offer at the same price as the control block. In Mexico, Lopez- de-Silanes (1997) reports that the price per share for sales that did not involve control were just one quarter of the prices for sales of control blocks, helping to explain the fact that 87 percent of all sales in his sample of Mexican firms involved sales of control. In Table VII, panel B, we test the link between private benefits and capi- tal market development, beginning with the various aggregate indicators of financial development introduced by La Porta et al. (1997): number of IPOs/ population, the number of listed firms/population, and the external market capitalization relative to GDP. Private benefits also explain a significant frac- tion of the cross-sectional variation in these measures. Our measure of private benefits is significant in all regressions with the exception of the OLS specifica- tion with the number of listed firms, where the single data point of Israel, with an unusually high level number of firms, reduces our level of significance. All the regressions include log per capita GDP as a regressor, to control for other possible factors.25 A one standard deviation increase in private benefits trans- lates into a 67 percent decline in the percent of external equity capitalization/ GNP. IV. What Curbs Private Benefits of Control? A. Theoretical Predictions Since the extent of private benefits of control seems to matter for security market development, the question of what curbs them becomes of central im- portance for any attempt to foster security market development. The evidence of systematic differences in legal rules and the correlation be- tween these rules and features of financial development La Porta et al. (1997, 1998, 1999) has focused the attention on the importance of the legal system. To capture the effect of the legal framework, we use three empirical proxies: (1) the formal rights of minority shareholders, (2) the degree of accounting disclosure (which allows minority shareholders to identify abuses), and (3) the quality of legal enforcement. 25Similar results obtain if we follow La Porta et al. (1997) and include GDP growth to capture future growth prospects and log GDP to capture any economies of scale in financial development. 178 A Reader in International Corporate Finance 576 The Journal of Finance A.1. Legal Institutions (i) The legal environment. The ability of a controlling shareholder to appro- priate some of the value generated is limited by the possibility of being sued. Thus, a greater ability to sue should translate into smaller private benefits of control (Zingales (1995a)). The same reasoning applies to any legal right at- tributed to noncontrolling shareholders (La Porta et al. (1997)). Accordingly, we examine the explanatory power of legal rights that give minority investors leverageoverinsidersinfirmsfocusingontheso-calledantidirectorrightsindex developed by La Porta et al. (1997) and used by Pistor, Raiser, and Gelfer (2000) for the transition countries. We focus our attention on the level of shareholder rights in the country of the target firm. As seen above, we also examine the impact of shareholder rights in the acquirer's country based on the hypothesis that these might also constrain private benefits (Dyck (2000)). (ii) Disclosure standards. Disclosure standards regulate the information available to noncontrolling shareholders. The more accurate this information is, the more difficult it is for a controlling shareholder to appropriate value without incurring legal penalties or, at least, reputational costs. Thus, mea- sures of quality of disclosure should be negatively correlated with the size of private benefits of control. (iii) Enforcement. The strength of legal protections depends upon the expec- tations of speedy and predictable enforcement. Thus, we include as one of our contractual variables a measure of the strength of a country's law and order tradition as measured by the country risk rating agency, International Country Risk. This rule of law index is scaled from zero to 10. A.2. Extra-legal Institutions The possibility of extracting private benefits is intrinsically related to man- agerial discretion, a discretion that courts cannot easily restrict. As a result, extra-legal institutions may play an important role in constraining private ben- efits (Dyck (2000)), both in settings with legal protections as well as in settings where legal protections are nonexistent or not enforced. The potential constraints imposed by extra-legal institutions have not been prominent in current debates, at least in part because of a lack of empirical examination. We focus our attention on five institutional factors that, at least in theory, have the potential to raise expectations of penalties for activities that produce private benefits for controlling shareholders. Some of these factors that can raise the costs to the controlling shareholder for diverting activities (such as the penalties produced by product market competition and by public opinion pressure) are constraints external to the firm. Other factors (such as the sanctions that can be introduced by moral norms, labor, and the government as tax collector) are more "internal" to the firm. (iv) Product market competition. The degree of product market competition affects the opportunity to appropriate private benefits in two dimensions. First, the more competitive markets are, the more verifiable prices become. When prices are more "objective," it is more difficult for a controlling shareholder to Chapter Five 179 Private Benefits of Control 577 tunnel out resources through manipulated transfer prices without incurring legal and/or reputational costs. Second, in a competitive market the distortions produced by the extraction of private benefits are more likely to jeopardize the survival of the firm. Hence, competition represents a natural constraint to the extraction of private benefits. The extent of product market competition is based both on industry and on country characteristics. In our regressions we include controls for industry char- acteristics, which we constrain to be constant across countries. The extent of product market competition is also influenced by country level characteristics, particularly government policies regarding entry and competition. We use as our proxy for the extent of product market competition at the national level the response to the survey question, "competition laws prevent unfair competition in your country?" as reported by the World Competitiveness Yearbook for 1996. This variable, which is available for all of our countries, captures cross-country differences in the extent to which national policy makers allow for barriers to competition over and above those constraints associated with industry. (v) Public opinion pressure. Controlling shareholders might limit their ef- forts to divert firm resources not out of fear of legal sanction but rather out of concern for their reputation. As Dyck and Zingales (2002b) argue, reputation to reduce diversion, the information about improper behavior must be publicized. For example, shareholders' activist Robert Monks succeeded in initiating some major changes at Sears, not by means of the norms of the corporate code (his proxy fight failed miserably), but through the pressure of public opinion. He paid for a full-page announcement in the Wall Street Journal where he exposed the identities of Sears' directors, labeling them the "non-performing assets" of Sears (Monks and Minnow (1995)). The embarrassment for the directors was so great that they implemented all the changes proposed by Monks. Similarly, re- cent efforts to stem diversionary practices by the powerful Korean Chaebol have also come not from court cases but through the public identification and dis- semination of behavior through the media by shareholder activists. Public hu- miliation is not only a tool of activists, but is also viewed as an important tool of regulators. In Hong Kong, for example, the main sanction available to securities regulators was not financial penalties but the threat and use of publishing those who violate listing requirements through the press.26 Critically, for reputation to work, though, it is necessary to have a "public opinion: that is, a combination of an independent press that publicizes the facts and of a large set of educated investors, who read the newspapers and sanction improper behavior" (Zingales (2000)). We try to capture this idea with an indi- cator of newspapers' diffusion, measured as the circulation of daily newspapers normalized by population. (vi) Internal policing through moral norms. Regardless of the reputational cost and/or the legal punishment the appropriation of private benefits trigger, 26While public opinion pressure is likely to act as a restrain in the extraction of private benefits, it does not necessarily push managers in the direction of shareholders' value maximization. In fact, in Dyck and Zingales (2002b) we show that media pressure also induces companies to be more environmentally conscious even if this does not necessarily benefit shareholders. 180 A Reader in International Corporate Finance 578 The Journal of Finance a controlling shareholder might choose not to appropriate value for moral con- siderations. But what constitutes a measure of the strength of such an internal policeman? Coffee (2001) proposes the violent crime rate as a proxy for these moral norms, noting that this at least captures an important difference between Scandinavian and other countries. Stulz and Williamson (2001) focus on culture as an indicator of norms. They use religion as their proxy for cultural norms and hypothesize that certain religious traditions will be more antagonistic to investor rights, such as the historical antagonism Catholics and Muslims had toward the payment of interest. To test for an impact of moral norms we use both proposed measures: (1) the number of violent crimes reported by the World Competitiveness Yearbook based on Interpol data for 1993 and (2) Stulz and Williamson's classification of countries by their primary religious orientation. (vii) Labor as monitor. Additional constraints on controlling shareholders might come from the presence of economic entities with a direct interest in firm decisions that could penalize efforts to extract private benefits directly without having to turn to the courts. From this perspective, it is clear that labor has the potential to monitor controlling shareholders and the ability to penalize diversions without resorting to legal sanctions. Labor is privy to inside infor- mation on customers and suppliers and can hold up the controlling shareholder by threatening to withhold services and in some cases, through their position on the board of directors. Stiglitz (1985), for example, suggests that unions have both the potential for low cost monitoring and have a strong incentive to monitor. "Labor is also motivated to take actions that protect the long-term sur- vival of the firm, and particularly where employees are also owners through the investment of their pension funds in company stock, there interests are not narrowly focused on wages," (Stiglitz (1985)). At the same time, it is the- oretically ambiguous how labor might act for it does not necessarily have the incentive to constrain private benefits, possibly aligning itself with the con- trolling shareholder against outside investors and labor's information access might not include critical information that is the source of private benefits. We test for the effect of labor on private benefits using as a cross-country measure of the extent of potential labor power the degree of employee protection. This measure is available for all OECD countries. (viii) Government as monitor through tax enforcement. There is one de facto minority shareholder that is common to all companies: the Government. As for minority shareholders, the Government has an interest in ascertaining the value produced by a company and getting a share of it. Transfer pricing, for instance, is disciplined by the tax code. In the United States, intracorporate transfers should take place at the price the two units would have charged in a competitive market. Hence, how tax authorities enforce their rules on trans- fer pricing affects the incentives to transfer profits to related companies. The stricter the enforcement, the less controlling shareholders will use transfer prices to siphon out value at the expense of minority shareholders.27 27Tax authorities should be particularly concerned about diversions of revenues from taxed to nontaxed entities, be those entities domestic or foreign. Chapter Five 181 Private Benefits of Control 579 Unlike noncontrolling shareholders, however, the tax authority does not face any free-rider problem in monitoring and enforcing its right. On the contrary, by aggressively prosecuting a company the Government sets an example that induces all others to behave. Thus, it has an incentive to prosecute cases even when the cost of prosecution is higher than the money recoverable. Further- more, the Government has the benefit of disciplinary powers that are simply not available to dispersed shareholders. Therefore, better tax enforcement can have an important role in reducing the private benefits of control. Note that this effect is true only for the quality of the enforcement not for the level of the tax rates. In fact, a higher tax rate increases a company's benefit from hiding income. In so doing, it subsidizes the siphoning out activity of the largest shareholder. For any dollar siphoned out by the majority shareholder, minority shareholders lose only (1­t) dollars, where t is the corporate tax rate. Hence, the higher the t, the lower the incentives of minority shareholders to stop this activity. For this reason we want a measure of tax compliance, not of tax revenues. To this purpose, we use an index developed by the World Competitiveness Report, which assesses the level of tax compliance. The index goes from zero to six where higher scores indicate higher compliance. That an effective corporate taxation system might have this positive exter- nality has not been emphasized in the corporate finance literature, or, to our knowledge, in the public finance literature.28Any evidence in this direction would be an important element in the debate on the costs and benefits of corpo- rate income taxation, particularly in countries with high private benefits. B. Test The large panel data set of 393 transactions from 39 countries provides a unique sample to try and identify the main institutional curbs of private bene- fits of control discussed above. In what follows, we describe the empirical proxies used and their effect on the private benefits of control. The definition for all these proxies is reported in Table I. Table VIII reports their actual values. In Table IX we test the impact of each institution in isolation and in Table X we try to test them one against the other. For these regressions, we include all of the control variables used in Table V as well as an indicator variable that identifies countries that have any form of tender offer requirement. We start with the impact of "legal" factors, that is, factors that directly or indirectly rely on the court enforcement of certain rights. Information disclo- sure is the prerequisite for any legal action. Thus, we start (column 1) with the quality of the accounting standards, as measured by the CIFAR index. Firms in countries with better accounting standards have lower private ben- efits of control. This effect is both statistically and economically significant. 28 For example, Gresik's (2001) recent review of the literature on rationales for and effects of corporate income taxation in the context of transnationals does not mention any spillovers between government actions and agency costs. 182 580 Table VIII Institutional Variables This table presents summary statistics of the institutional variables used in Tables IX to XI. Variable definitions and sources can be found in Table I. Panel A A Reader Legal Institutions Extra Legal Institutions Rule of Law at Serious T in Accounting Country Newspaper Crime/ Labor Tax Acceptability he International Legal Standards Antidirector Level Competition Circulation/ 100,000 Protection Compliance of Cheating on Primary Journal Country Origin (0­90) Rights (0­6) (1­10) Laws Pop Population Measure (1­6) Taxes (1­10) Religion Argentina French 45 4 5.35 4.85 1.2 8.2 2.41 1.97 Catholic Australia English 75 4 10 5.52 3.0 57.5 0.9 4.58 2.16 Protestant Austria German 54 2 10 5.29 2.9 57.3 2.2 3.6 1.97 Catholic of Brazil French 54 3 6.32 4.9 0.4 2.14 3.11 Catholic F Corporate Canada English 74 5 10 5.37 1.6 122.3 0.6 3.77 2.34 Catholic inance Chile French 52 5 7.02 5.4 1.0 53.7 4.2 1.98 Catholic Colombia French 50 3 2.08 4.71 0.5 129.1 2.11 1.92 Catholic Czech Soviet 2 8.3 4.89 2.5 177.2 2.54 Atheist Finance Republic Denmark Scand. 62 2 10 5.16 3.1 46.1 3.7 2.48 Protestant Egypt French 24 2 4.17 4.6 0.4 3.57 Muslim Finland Scand. 77 3 10 5.26 4.6 47.1 2.0 3.53 2.63 Protestant France French 69 3 8.98 5.83 2.2 126.8 3.0 3.86 3.28 Catholic Germany German 62 1 9.23 5.91 3.1 74.1 2.5 3.41 2.94 Protestant Hong Kong English 69 5 8.22 5.85 8.0 190.8 4.56 Local beliefs Indonesia French 2 3.98 4.42 0.2 4.6 2.53 Muslim Israel English 64 3 4.82 5.11 2.9 68.9 3.69 Judaism Italy French 62 1 8.33 5.14 1.0 61.7 3.3 1.77 2.28 Catholic Japan German 65 4 8.98 5.64 5.8 2.7 2.4 4.41 1.49 Buddhist Malaysia English 76 4 6.78 4.84 1.6 34.5 4.34 Muslim Mexico French 60 1 5.35 4.93 1.0 100.8 2.46 3.35 Catholic Netherlands French 64 2 10 5.53 3.1 122.8 2.1 3.4 3.08 Catholic New Zealand English 70 4 10 5.4 2.2 52.3 1.0 5 Protestant Norway Scand. 74 4 10 4.96 5.9 26.9 2.6 3.96 3.10 Protestant Peru French 38 3 2.5 5.05 0.8 2.66 2.15 Catholic Phillipines French 65 3 2.73 4.61 0.8 90.9 1.83 3.00 Catholic Poland Soviet 3 8.7 5.06 1.1 99.6 2.19 2.61 Catholic Portugal French 36 3 8.68 4.81 0.8 12.4 3.7 2.18 3.82 Catholic Singapore English 78 4 8.57 5.21 3.2 45.2 5.05 Buddhist South Africa English 70 5 4.42 4.89 0.34 225.2 2.4 2.44 Protestant South Korea German 62 2 5.35 4.9 3.9 8.5 3.29 1.64 Protestant Spain French 64 4 7.8 5.07 1.0 169.6 3.1 1.91 2.57 Catholic Sweden Scand. 83 3 10 5.08 4.5 80.1 2.2 3.39 2.30 Protestant Switzerland German 68 2 10 5.22 3.3 38.3 1.0 4.49 2.50 Catholic Taiwan German 65 3 8.52 5.56 2.7 34 3.25 1.98 Buddhist Private Thailand English 64 2 6.25 4.77 0.6 70.4 3.41 Buddhist Turkey French 51 2 5.18 5.14 1.1 69.2 2.07 1.24 Muslim United English 78 5 8.57 5.74 3.3 96.4 0.5 4.67 2.65 Protestant Kingdom Benefits Chapter United States English 71 5 10 5.96 2.12 272.5 0.2 4.47 1.95 Protestant Venezuela French 40 1 6.37 4.24 2.06 86.5 1.56 1.98 Catholic Panel B: Correlation Matrix Five Accounting Rule of Law at Serious Labor Tax Acceptability of of Standards Antidirector Country Level Competition Newspaper Crime/100,000 Protection Compliance Cheating on Control (0­90) Rights (0­6) (1­10) Laws Circulation/Pop Population Measure (1­6) Taxes (1­10) Accounting standards 1.00 Antidirector rights 0.32 1.00 Rule of law 0.53 0.06 1.00 Competition laws 0.49 0.26 0.59 1.00 Newspaper circulation 0.54 -0.01 0.62 0.35 1.00 Serious crime 0.19 0.33 -0.09 0.26 -0.13 1.00 Labor protection -0.57 -0.55 -0.54 -0.46 -0.10 -0.35 1.00 Tax compliance 0.58 0.40 0.65 0.74 0.65 -0.08 -0.78 1.00 Cheat 0.08 -0.11 0.17 -0.03 -0.11 0.07 0.46 -0.10 1.00 581 183 184 A Reader in International Corporate Finance 582 The Journal of Finance A one standard deviation increase in accounting standards reduces the value of control by 9.0 percentage points. Together with the other control variables, accounting standards explain 21 percent of the variation in private benefits of control (the firm-specific control variables alone explain just 15 percent). Our second variable (column 2) is the extent of legal protections for minority investors, measured using La Porta et al. (1998) index of antidirector rights. Countries with more antidirector rights have lower private benefits of con- trol. A one standard deviation increase in antidirector rights reduces the value of control by 4.4 percentage points. Together with the firm-specific variables, antidirector rights explain 17 percent of the variation in private benefits of control. Finally, we use the quality of law enforcement, which we measure using the IBR index of the quality of the law enforcement in a country. Countries with better law enforcement have lower private benefits of control. A one standard deviation increase in our law enforcement measure reduces the value of control by 7.0 percentage points. Together with the firm-specific variables, rule of law explains 20 percent of the variation in private benefits of control. In sum, we find that legal institutions are strongly associated with lower levels of private benefits. When we combine the two legal variables that are available for our full sample in one regression (Table X, column 1), both are statistically significant and the R-squared is 21 percent. We also test the explanatory power provided by extra-legal institutions, which are suggested by a functional rather than an institutional perspective. Here we focus on crude country-wide measures of product market competition, scope of reputational penalties, moral norms, employee protections, and diligence of tax authorities. Table IX, columns 4 to 9, explores the explanatory power of these factors one at a time. In column 4 we test the effect of competition. After having controlled for industry type, we find that countries with more competitive product mar- kets, at least as measured by this survey of the World Competitiveness Report, have lower private benefits of control. A one standard deviation increase in our measure of competition reduces the value of control by 6.0 percentage points. Together with the firm specific variables, competition explains 20 percent of the variation in private benefits of control. In column 5 of Table IX, we explore the idea that public opinion pressure might curb the amount of private benefits extracted. We measure the im- portance of this pressure with the diffusion of newspapers (number of copies sold per 100,000 inhabitants). Diffusion captures both the importance of public opinion and the credibility of newspapers (less credible newspapers sell less).29 Countries where newspapers are more diffused have lower private benefits 29In Dyck and Zingales (2002b), we study the determinants of newspapers' diffusion. We find that the type of dominant religion and the degree of ethnolinguistic fractionalization explain 41 percent of the variation in press diffusion. When we use these as instruments for press diffusion, the results are unchanged. Table IX Institutional Determinants of Private Benefits of Control--Univariate Analysis The dependent variable is the block premia as a percent of firm equity. The explanatory variables include all of variables introduced in Table V except the country fixed effects, but including a dummy to indicate the presence of a mandatory tender offer law. In place of the country fixed effects, we introduce one at a time several institutional variables: (1) accounting standards index; (2) antidirector rights index; (3) rule of law index; (4) tax compliance index; (5) diffusion of the press as measured by the newspaper circulation/population; (6) an index of the extent of competition laws; (7) incidence of violent crimes; (8) extent of legal protections for labor; (9) a dummy variable if primary religion is Catholicism. More complete descriptions of variables are provided in Table I. Standard errors, which are reported in parentheses, are robust and clustered by country. Dependent Variable: Block Premium Private Legal Institutions Extra Legal Institutions Independent Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Benefits Accounting standards -0.007 Chapter (0.002) Antidirector rights -0.036 (0.015) of Five Rule of law -0.029 Control (0.010) Competition laws -0.147 (0.046) Newspaper circulation/pop -0.036 (0.014) Violent crime incidence 0.000 (0.000) Labor protection 0.038 (0.023) Catholic is primary religion 0.118 (0.066) Tax compliance -0.085 (0.025) 583 185 186 584 Table IX--Continued Dependent Variable: Block Premium Legal Institutions Extra Legal Institutions Independent Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) A Reader Variables Controlled for: Buyer bargaining power y y y y y y y y y Ownership variables y y y y y y y y y T in Financial distress y y y y y y y y y he International Foreign acquirer y y y y y y y y y Crosslisted in the U.S. y y y y y y y y y Journal Buyer identity y y y y y y y y y Seller identity y y y y y y y y y Industry group y y y y y y y y y of Tangibility of assets y y y y y y y y y F Corporate Interaction of relative strength y y y y y y y y y inance of antidirector rights (home--target nation) and foreign acquiror dummy Interaction of relative strength y y y y y y y y y Finance of antidirector rights (US--target nation) and crosslisted in the US dummy Presence of takeover law y y y y y y y y y Constant y y y y y y y y y Number of observations 381 393 393 393 393 377 233 393 393 Countries included 36 39 39 39 39 36 18 39 39 R-squared 0.213 0.174 0.203 0.203 0.200 0.175 0.208 0.184 0.230 significant at 5% level; significant at 1% level. Chapter Five 187 Private Benefits of Control 585 Table X Institutional Determinants of Private Benefits of Control--Multivariate Analysis The dependent variable is the block premia as a percent of firm equity. The explanatory variables include all of the variables introduced in Table V except the country fixed effects, but including a dummy to indicate the presence of a mandatory tender offer law. As institutional variables in specification (1), we use antidirector rights index and rule of law index. In specification (2), a dummy variable if primary religion is Catholicism, a tax compliance index, the diffusion of the press as measured by the newspaper circulation/population and the index of the extent of competition laws. The independent variables in specification (3) are antidirector rights index, rule of law index, tax compliance index, diffusion of the press as measured by the newspaper circulation/population. More complete descriptions of variables are provided in Table I. Standard errors, which are reported in parentheses, are robust and clustered by country. Dependent Variable: Block Premium Independent Variables (1) (2) (3) Antidirector rights -0.026 (0.012) -0.003 (0.019) Rule of law -0.026 (0.010) -0.006 (0.011) Catholic 0.019 (0.056) Tax compliance -0.064 (0.021) -0.061 (0.033) Newspaper circulation/ popu- -0.020 (0.009) -0.018 (0.010) lation Competition laws -0.042 (0.036) Variables Controlled for: Buyer bargaining power y y y Ownership variables y y y Financial distress y y y Buyer characteristics y y y Seller characteristics y y y Foreign acquirer y y y Crosslisted in the U.S. y y y Industry type y y y Tangibility of assets y y y Interaction of relative y y y strength of antidirector rights (home--target nation) and foreign acquirer dummy Interaction of relative y y y strength of antidirector rights (US--target nation) and cross listed in the US dummy Presence of takeover law y y y Constant y y y Number of observations 393 393 393 Countries included 39 39 39 R-squared 0.213 0.245 0.243 significant at 10% level; significant at 5%; significant at 1% level. 188 A Reader in International Corporate Finance 586 The Journal of Finance of control. A one standard deviation increase in newspapers' diffusion reduces the value of control by 6.4 percentage points. Together with the firm-specific variables, newspapers' diffusion explains 20 percent of the variation in private benefits of control. Columns 4 and 5 suggest that institutions external to the firm are associated with private benefits. In columns 6 and 8, we test the idea that countries with higher moral norms have lower private benefits. Consistent with Coffee's prediction, countries with worse norms as proxied by a higher violent crime rate have higher private ben- efits of control, but the effect is economically and statistically insignificant. To investigate moral norms, we introduce indicator variables for the four main religions (Buddhist, Catholic, Muslim, and Protestant), which differ in their impact on moral attitudes (Guiso, Sapienza, and Zingales (2003)). As a country religion we use the dominant one (see Stulz and Williamson (2001)). We find that Catholic countries have significantly higher private benefits, and Protes- tant ones significantly lower (estimate not reported). The effect of the Muslim and Buddhist religion is not significant. In columns 7 and 9, we test whether the strength of other entities that have a direct economic interest in firm decision making is associated with lower levels of private benefits. In column 7 we examine the impact of labor as a moni- tor of private benefits. As an index of potential labor strength, we use both an unweighted and a weighted (not reported) index of employee protections based on average indictors on regular contracts and short-term contracts from OECD data compiled in Pagano and Volpin (2000). The restriction to OECD countries unfortunately limits our number of countries and observations, but is perhaps a purer test of the contention that labor can work as monitors, since this literature has focused on organized labor in developed economies. Incon- sistent with the hypothesis that labor is an effective monitor, and consistent with Pagano and Volpin's counter contention that entrepreneurs and workers will align themselves against the interests of minority investors, we find that increased labor power is associated with higher private benefits, although this result is not statistically significant (p-value of 0.204 for employee protections, 0.13 for weighted employee protections). In column 9 we investigate the possibility that a government interested in enforcing tax rules can reduce private benefits. This column shows that those countries with a higher degree of tax compliance, as measured by the World Competitiveness Report, have lower private benefits of control. A one standard deviation increase in our measure of tax compliance reduces the value of control by 8.6 percentage points, a significant amount. Together with the firm specific variables, tax compliance explains 23 percent of the variation in private benefits of control. Tax compliance is an equilibrium outcome, affected both by tax enforcement and by the attitude of citizens toward cheating on their taxes. To try to identify the impact of tax enforcement in an unreported regression, we include a mea- sure of willingness to cheat on taxes as measured in the World Value Survey. Chapter Five 189 Private Benefits of Control 587 In this survey people are asked to rate from one to 10 the statement "cheating on taxes if you have a chance is...," where one is never justifiable and 10 is always justifiable. We find this variable to be insignificant, and the coefficient on tax compliance to remain significant, suggesting the effect of tax compliance comes from tax enforcement and not from differences in moral values across countries. We also examine the robustness of this result to the inclusion of the marginal tax rate and our results are unchanged. In Table X (column 2), we combine the four extra-legal institutions that in- dividually had a statistically significant effect. All four variables retain the predicted sign, but the magnitudes of their coefficients drop and only tax com- pliance and newspaper diffusion remain statistically significant at the 5 per- cent level. Together these four variables are able to explain 24 percent of the variation in private benefits. The evidence, thus far, is consistent with both the legal and the extra-legal institutions playing a role in constraining private benefits. In fact, a crude R- squared test suggests they have roughly the same explanatory power. Can we distinguish which one is more important? There are two obstacles to doing so. First, many of these institutional vari- ables are highly correlated, as panel B of Table VIII shows. Shareholder's pro- tection, though, is not correlated with newspapers' circulation and has a corre- lation of only 0.4 with tax compliance. Second, and most important, all these proxies are measured with error. Hence, their statistical significance in a mul- tivariate analysis might be more related to the level of noise in these measures than to their actual importance. Nevertheless, we think it is interesting to try and put all these variables in one regression. This is what we do in column 3 of Table X. When all the institutional variables we found to be significant in the previous regressions are simultaneously included, only newspapers' diffusion and tax compliance remain significant. The paucity of observations and the high degree of multi- collinearity caution us against drawing any strong conclusion from this compar- ison. We can say, however, that the results are inconsistent with an exclusive focus on legal variables as institutional curbs to private benefits. C. The Effect of Legal Families Since LLSV's (1998) seminal paper, the origin of a country's legal system has played an important role in all the institutional explanations of cross-country differences. LLSV claim that legal traditions differ in their respect for prop- erty rights and, hence, in their ability to protect minority shareholders. We should have already accounted for this effect by inserting the LLSV index of antidirector rights. Nevertheless, it is possible that the origin of a country's legal system is a better indicator of the degree of protection of outside investors than the antidirector index. For this reason, we repeat some of the previous 190 A Reader in International Corporate Finance 588 The Journal of Finance estimates substituting the country of origin of the legal system for the antidi- rector rights variable. As Table XI, panel A shows, the average level of private benefits differs substantially across different legal families. Private benefits are highest in former communist countries (36 percent), then countries with a French code (21 percent), and countries with a German, English, and Scandinavian code seem to have the lowest level of private benefits (respectively, 11, 5.5, and 4.8 percent). Panel B, column 1, shows that the levels of private benefits are sig- nificantly lower in countries with German, English, and Scandinavian legal origins than in French legal origin countries. Thus, the distinction is not in terms of civil law versus common law, but it is more complex. In Table XI, panel B, we report how these results are changed after we control for the most significant extra-legal institutions (diffusion of readership and tax enforcement). Any distinction between English-based legal systems and the others disappears. If anything, common law countries have higher (not lower) private benefits of control once these extra-legal institutions are taken into consideration, but this effect is not statistically significant. Only Scandinavian countries have lower private benefits of control even after controlling for extra- legal institutions. Overall, these results confirm the previous ones: Extra-legal institutions are important and they should be controlled for in any cross-country analysis. Table XI Private Benefits of Control and Legal Origin Panel A presents descriptive statistics of block premia by legal origin, first presenting aver- ages at the country level and second presenting averages based on the full set of 393 trans- actions. Panel B provides OLS regressions of block premia on legal origin and our other ex- planatory variables. The independent variables examined are those included in Table IX with (1) legal origin; (2) tax compliance and newspaper circulation; (3) English origin to capture the difference between common and civil law origin, tax compliance and newspaper circulation; (4) all legal origin dummies, tax compliance, and newspaper circulation. More complete descrip- tions of variables are provided in Table I. Robust standard errors clustered by country are in parentheses. Panel A: Block Premium by Legal Origin Groups of Legal Origin All Transactions Standard Number of Standard Number of Law Origin Mean Deviation Countries Mean Deviation Observations Scandinavian origin 0.048 0.033 4 0.041 0.075 42 English origin 0.055 0.080 11 0.045 0.123 196 German origin 0.109 0.152 6 0.051 0.138 57 French origin 0.212 0.171 16 0.251 0.439 88 Soviet origin 0.356 0.314 2 0.400 0.639 10 Chapter Five 191 Private Benefits of Control 589 Table XI--Continued Panel B: Investigating Explanatory Power of Legal Origin Dependent Variable: Block Premium Independent Variables (1) (2) (3) (4) English origin -0.155 0.043 -0.024 (0.067) (0.044) (0.062) Soviet origin 0.128 0.141 (0.201) (0.207) German origin -0.228 (-0.121) (0.097) (0.084) Scandinavian origin -0.189 -0.098 (0.058) (0.053) Tax compliance -0.070 -0.087 -0.066 (0.021) (0.027) (0.022) Newspaper circulation -0.021 -0.015 -0.003 (0.010) (0.011) (0.008) Variables controlled for: Buyer bargaining power y y y y Ownership variables y y y y Financial distress y y y y Buyer identity y y y y Seller identity y y y y Industry group y y y y Tangibility of assets y y y y Foreign acquirer y y y y Crosslisted in the U.S. Interaction of relative strength of y y y y antidirector rights (home--target nation) and foreign acquirer dummy Interaction of relative strength of y y y y antidirector rights (US--target nation) and crosslisted in the US dummy Constant y y y y Number of observations 393 393 393 393 Number of countries (clusters) 39 39 39 39 R-squared 0.243 0.242 0.244 0.260 significant at 10% level; significant at 5%; significant at 1% level. V. Conclusions In this paper we apply the Barclay and Holderness (1989) approach to mea- sure the magnitude of private benefits of control across countries. That we obtain estimates very consistent with previous studies, using different ap- proaches, indicates that the extraction of private benefits is a very real phe- nomenon that can be consistently measured. We then use these estimates to test several theoretical predictions from the corporate finance literature on the negative effects that large private benefits have on financial development. In countries where private benefits of control are large, ownership is more concentrated, privatizations are less likely to take 192 A Reader in International Corporate Finance 590 The Journal of Finance place as public offerings, and capital markets are less developed by several measures. These results vindicate the emphasis that, since Shleifer and Vishny (1997), corporate finance research has put on the importance of protecting out- side investors against expropriation by insiders. They also suggest the impor- tance of gaining a better understanding of what are the institutions that help curb private benefits. We find that many institutional variables, taken in isolation, seem to be associated with a lower level of private benefits of control: better accounting standards, better legal protection of minority shareholders, better law enforce- ment, more intense product market competition, a high level of diffusion of the press, and a high rate of tax compliance. The possible role of tax enforcement in reducing private benefits, and thus indirectly enhancing financial development, is probably the most important new fact that emerges from our analysis. Improving the corporate taxation sys- tem is well within the range of feasible reforms. If this is indeed a primary mechanism by which private benefits of control can be curbed and financial markets fostered, the benefits of financial development might be within reach for many more countries. Before jumping to any conclusion, though, more re- search is needed. In particular, it would be useful to show that within a country changes in the level of tax enforcement lead to changes in the size of private benefits. Our results suggest also other avenues for future research. We find that pub- lic opinion pressure helps to curb private benefits of control. A strong pressure from the media on corporate managers, however, will not always increase share- holders value. In fact, in Dyck and Zingales (2002b) we find that strong media also induce corporate managers to bow to environmental pressures, which are not necessarily in the shareholders' interest. The broader question, then, which awaits future research, is how media pressure interacts with social norms in shaping corporate policy. We also do not discuss, in this context, what are the incentives of the media to expose bad corporate practices and how these incen- tives may vary over the business cycle. We address this in a separate paper (Dyck and Zingales (2003)). Finally, in this paper we do not try to distinguish between the three potential sources of private benefits: psychic value, perquisites, and dilution. That private benefits are smaller in a country with better protection of investors, better tax enforcement, and more media pressure suggests that not all private benefits are psychic. Further work, however, is needed to establish the importance of dilution and its welfare implications. Appendix A.1. Steps to Identify Transactions We used the following approach to implement the first criterion that a trans- action be a control transaction between unrelated parties: (1) The transac- tion had to be identified in the SDC database and through the transaction the Chapter Five 193 Private Benefits of Control 591 acquirer had to move from a shareholding position of less than 20 percent to shareholding of more than 20 percent shareholding.30(2) The block involved in the transaction had to be 10 percent or greater. (3) The block had to be the largest block in the company. (4) News stories surrounding the transaction had to confirm a transfer of control from the seller to the acquirer, with news stories identified by using the company name and transaction date in Nexis-Lexis and Dow-Jones Interactive search engines, often with the use of both English and foreign language media. Illustrative of the steps we took to identify control transactions is our ex- clusion of related party transactions. With related parties it is questionable whether control is transferred and the price of the deal is unlikely to reflect the value of control. Systematically, we excluded transactions where SDC re- ported that the acquirer involved management, as management already has control rights prior to sale. Using qualitative data we identified further re- lated party transactions excluding transfers of shares between subsidiaries and parents of the same company and other deals that don't transfer con- trol. For example, we excluded the sale of 36 percent of the shares of Shin Corp in Thailand in September 2000. News stories reported that "Telecoms Tycoon turned politician Thaksin Shinawatra and his wife have sold their 35.4 percent stake in their flagship Shin Corporation at a deep discount, in what appears to be an attempt to comply with the laws on ministers' own- ership of companies. The stake was sold to their son and relatives at just 10 baht a share, less than 6 percent of the stocks closing price yesterday of 177 baht.... Analysts said the move was purely political and would have no impact on shareholders or on the company."31 To implement the second criterion, that a control price be available and reflect the value of control, we restricted our attention to SDC transactions that met three additional criteria: (1) There had to be data in SDC to identify a control price. In many cases SDC reports a price per share in a separate data field where they value cash offers at face value and offers of shares at the exchange price on the day prior to the announcement of the transaction. In other instances, the price per share is not reported in the data field but can be derived by combining information in available data fields and information from other data sources on the number of shares outstanding. For example, SDC would report the total price paid and the percentage of shares sold and we would construct an estimate of the per share price involved in the offer by collecting information on the number of shares outstanding at the time of the transaction. For many transactions, SDC reported that no 30 For Australia and Canada we used a 15 percent cutoff due to the presence of takeover rules for stakes exceeding 20 percent. 31 "Thaksin, wife sell entire stake in flagship," Harish Mehta, Business Times Singapore, September 7, 2000. 194 A Reader in International Corporate Finance 592 The Journal of Finance terms were disclosed or that the reported price was only one component of the compensation. We are unable to use such transactions. (2) The form of sale had to involve purchases where assets used to establish a per share sale price include securities that could be priced objectively (we exclude transactions that involve warrants, convertible bonds, notes, liabilities, debt-equity swaps, etc.), and where the terms of sale were not determined by exercising an option or included an option to buy additional shares in addition to the shares purchased. (3) The synopsis field and news stories had to confirm the price per share and to ensure that the reported price was not misleading. We excluded observations where news stories identified other considerations, and ad- justed the price per share from the SDC reported price if two news stories reported a price that deviated from the SDC price. To implement the third criterion that an exchange price be available we begin by restricting our attention to those transactions where the company whose shares are being acquired is covered by Datastream international, the data provider with the most extensive coverage of international firms.32 We also are interested in identifying the exchange price after the market is aware of the purchase of shares by the new controlling shareholder. A traditional approach in the finance literature of focusing on the share price on the day of announcement is not warranted with our database. In many cases, the transfer of control leads to a suspension of trading of the company shares either because there is a need for time for the information about the control transfer to be communicated broadly or there are limits to movement of the exchange price per day. While the suspension is of limited duration in established markets like the United States and the United Kingdom, the suspension can last for a day or more in other settings. Consequently, we use as a standard approach the control price two days after announcement. Where news stories indicated a longer delay, we used the first date after restrictions on trading or pricing of securities. This produced modifications in 17 cases where we use a later date for all of our calculations. A.2. The Special Case of Dispersions of Control Blocks In 17 transactions we identify through reading news stories that the control- ling block is not sold intact but rather sold to a financial intermediary that then sells the block to a variety of institutional investors. We elected to include these deals in our data set. In the Barclay and Holderness (1989) data set such trans- actions were excluded by construction of their sample, but as they argued, such 32We attempted to access additional information sources for price information for local stocks not covered during our time period by Datastream through direct contacts with country stock exchanges and through appealing to news reports that often reported share price information for large local companies. These efforts produced 26 additional observations. Chapter Five 195 Private Benefits of Control 593 transactions should be included if a private benefit measure is to reflect the general benefits and costs of control. Such transactions are only likely if there is a limited benefit to control of enterprises and costs to control. Our data set includes nine transactions from the United Kingdom, three from Germany, and one from Finland, Japan, New Zealand, Norway, and Taiwan. Our results are robust to the exclusion of these transactions, with small increases in our raw measures of private benefits for the United Kingdom (from 1.6 to 2.4 percent), Germany (from 9.5 to 11.8 percent) and New Zealand (2.6 to 3.6 percent). A.3. The Special Case of Companies with Dual Class Shares We identify all transactions that involve firms with multiple classes of shares. When this is the case we measure the control premium for the shares with vot- ing power relative to the shares that lack voting power, where Datastream provides price information for both classes. For example, we have 11 observa- tions from Brazil that involve firms with dual class shares and Datastream has price series for both classes for 10 of these 11 observations. In Brazil, the principle difference between the two classes is the voting right with largely equal rights to cash flow. Our data set includes 38 dual class firms altogether, including companies from Canada, Denmark, Finland, Germany, Italy, Mexico, Norway, Sweden, and the United States. A.4. Biases from Not Reporting Terms of Sale We made some steps to investigate this bias. When the SDC field reported other considerations we made efforts using stories from local media to see if subsequent to the announcement the other considerations became known. For almost all cases we were unsuccessful. However, for Malaysia, a country with an active business press, we were able to identify additional information. For the years 1995 and 1996, we identified all stories regardless of whether SDC included a transaction price or not. Using this technique we identified nine transactions not identified in our original sample and we were able to identify prices reported in the local press for eight of these transactions. Comparing the estimated private benefits from these transactions and from our reported transactions is revealing. The average control premia is similar between the initial sample used and this new SDC sample with unreported prices with a control premia as a percentage of equity of 6.9 percent for our core sample and 4.5 percent for our sample of "unreported prices." 196 594 Table AI Laws Regarding Control Transactions Law Requiring Voluntary Code Shareholding Year of Mandatory Requiring that Triggers Passage of Purchase of Purchase of Mandatory Dominant Additional Additional Purchase of Legal Country Shares Shares Shares Statute Legal and Regulatory Bases on Takeovers A Argentina N -- -- Resolution 227, National Securities Commission Reader Australia Y 20 1989 Corporations Law Austria Y 30 1999 Council of Vienna Stock Exchange, State Commissioner Brazil (1) Y 50 1976 Law 6404, law 9457, CVM rule #299 T in Canada Y 20 1975 Canada Business Corporations Act, Provincial legislation he International Chile (2) N -- 1994 Law 18.045 Journal Colombia N -- 1979 Act No. 32 Czech Republic Y 50 1991 Czech Commercial Code Denmark Y 50 n/a Danish Securities Trading Act, Stock Exchange Ethics Rules Egypt ? of Corporate Finland Y 67 1989 Securities Market Act F inance France Y 33 1992 COB regulations, Stock Exchange Council Germany (3) N Y 50 1995 Voluntary takeover code (Ubernahmekodex) Hong Kong N Y 35 1975 Hong Kong code on Takeovers and Mergers Indonesia Y 20 1995 Decree of Capital Market Supervisory Agency No. 22/PM/1 Finance Israel N -- Italy Y 30 1998 Law no. 149 Japan N -- -- Securities and Exchange Law Ch. II.2 Kenya N -- 1985 Company Act, Capital Markets Authority Act Malaysia Y 33 1993 Malaysian Code on takeovers and mergers, Companies Act Mexico N -- -- Corporation Law, Credit Law, other regulatory acts Netherlands N -- 1970 Merger Code of the Social Economic Council New Zealand N -- 1986 Companies Act 1986 Norway Y 45 1985 Securities Trading Act Peru N -- -- Stock Market Law Phillipines (4) Y -- 1998 Revised tender-offer rules, Securities and exchange commission Poland Y 33 1991 Act on Public Trading in Securities and Trust Funds Portugal Y 50 1986 Securities Act Singapore N Y 25 1985 Singapore Code on Takeovers and Mergers South Africa Y 30 1991 Securities Regulation Code on Takeovers and Mergers South Korea Y 25 ? Securities and Exchange Law Spain Y 25 1991 Law No. 24, Royal Decree 1197 Sweden N -- 1991 Financial Instruments Trading Act Switzerland Y Y 33 1998 Federal Act on Stock Exchanges and Securities Trading Taiwan N -- 1988 Securities and exchange Law, company law 1983 Thailand Y 25 1992 Securities and Exchange Act Turkey Y 25 1986 Capital Market Law United Kingdom N Y 30 1968 City code on Takeovers and Mergers United States N -- 1934 Securities and Exchange Act Private Venezuela N -- -- Capital Markets Law Sources: ISSA All data from ISSA Handbook, 6th and 7th edition. (1) Prior to 1997, Brazil law 6404 required equal offer to minority investors with voting shares (but not nonvoting preferred shares). This protection Benefits Chapter eliminated in May 1997 (Law 9457) with reform to enhance privatization proceeds. In 1999, CVM rule #299 reintroduces protections for minorities, now extending to voting and nonvoting class an equal price offer. (2) In December 2000 (after our observations) Chile has a new law, ley de OPSAS, governing control transactions. of Five (3) Germany has a voluntary takeover code (Ubernahmekodex) in place since 1995. This code "was deemed a failure in early 2000, when both stock market supervisors and the takeover commission appointed by Mr. Schroder demanded a mandatory law." EIE Country Commerce, section 2.2. 2000. Control (4) The Securities and Exchange Commission "issued tender-offer rules in October 1998 outlining the requirements for acquiring majority control in existing companies through open-market purchases or private negotiations. The new rules implement Section 33 of the Revised Securities Code and require bidders for majority control of listed companies to make the same offer of purchase to minority share holders. (EIU March 1999). The SEC generally failed to enforce tender-offer rules in major deals involving mergers and acquisitions from 1998 to 2000 because of loop holes in the old regulations (EIU March 2001). Securities Regulation Code (RA 8799 effective August 2000, implementing rules January 2001) requires those assembling >15% to make offer. Note: Canada has both federal and provincial legislation, where Ontario is most important. Rules require mandatory offer if >20% of voting shares, whereby at least a pro-rata offer for % bought although usually either for 2/3 or 90% of voting rights. 595 197 198 596 A Reader T Table AII he in International Number of Firms with Equities Priced in Datastream, by Year Journal Country Code 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 1990­2000 of AR 12 13 14 23 61 66 70 73 78 80 84 93 655 F Corporate AU 452 474 487 515 594 697 741 1095 1176 1178 1287 1506 9750 inance BD 618 700 735 742 777 800 835 893 926 1015 1188 1348 9959 BR 75 143 234 314 371 371 456 481 500 540 557 4042 CB 104 109 133 122 113 108 98 85 72 944 CL 115 127 134 151 164 170 177 196 209 220 219 221 1988 Finance CN 1502 1971 1976 2024 2195 2353 2473 2653 2981 3222 3352 3759 28959 CZ 38 57 91 114 129 128 131 126 814 DK 205 210 249 252 261 267 277 296 295 303 291 304 3005 ES 110 125 133 138 149 152 154 165 184 211 234 264 1909 EY 10 12 65 72 82 101 103 445 FN 58 67 70 71 76 123 128 153 178 198 228 258 1550 FR 508 628 640 663 693 793 859 1084 1143 1287 1229 1416 10435 HK 251 266 322 371 429 494 518 567 676 723 756 1072 6194 ID 107 120 133 150 192 213 225 259 261 292 313 2265 IS 196 202 201 267 475 542 558 567 569 593 667 710 5351 IT 283 309 319 323 325 348 366 388 406 419 443 535 4181 JP 2011 2321 2520 2592 2677 2954 3136 3347 3552 3582 3829 4304 34814 KN 1 1 43 42 45 45 47 51 52 53 50 47 476 KO 604 660 678 685 694 739 796 1017 1135 1140 1299 1569 10412 MX 46 52 75 99 132 151 145 152 172 163 170 160 1471 MY 346 404 447 493 544 607 663 757 847 872 738 776 7148 NL 237 260 271 273 277 288 303 326 361 408 437 482 3686 NW 80 97 100 114 128 160 182 217 273 287 269 274 2101 NZ 68 74 80 91 111 126 130 145 154 153 157 176 1397 PE 22 48 76 98 100 105 102 99 104 96 850 PH 64 96 103 114 137 160 187 212 231 229 222 225 1916 PO 6 11 12 22 27 51 103 167 200 221 820 PT 101 110 116 135 140 149 144 148 155 152 143 148 1540 SA 161 458 454 469 483 526 535 606 641 724 770 736 6402 Private SD 177 197 202 213 232 296 318 362 442 484 532 633 3911 SG 142 172 176 195 222 255 273 293 337 348 409 534 3214 SW 259 295 295 290 309 325 343 375 390 401 426 459 3908 TA 161 178 199 240 271 305 340 451 515 620 738 856 4713 Benefits Chapter TH 244 291 350 410 441 521 537 576 602 579 546 531 5384 TK 70 100 125 135 152 178 209 236 270 298 302 387 2392 UK 1812 1872 1749 1713 1782 1841 1932 2084 2222 2272 2301 2625 22393 Five US 274 393 415 419 427 438 462 662 929 1235 2671 4743 12794 of VE 10 10 11 14 19 21 22 23 25 29 53 237 Control Grand Total 11,168 13,315 13,979 14,803 16,116 17,771 18,795 21,298 23,378 24,809 27,469 32,692 224,425 597 199 200 A Reader in International Corporate Finance 598 The Journal of Finance REFERENCES Aghion, Philippe, and Patrick Bolton, 1992, An incomplete contract approach to financial contract- ing, Review of Economic Studies 59, 473­494. Barclay, Michael, and Clifford Holderness, 1989, Private benefits of control of public corporations, Journal of Financial Economics 25, 371­395. Bebchuk, Lucien, 1999, A rent-protection theory of corporate ownership and control, Working paper 7203, NBER. Bebchuk, Lucien, and Christine Jolls, 1999, Managerial value diversion and shareholder wealth, Journal of Law, Economics and Organization 2, 487­502. Burkart, Mike, Denis Gromb, and Fausto Panunzi, 1998, Why higher takeover premia protect minority shareholders, Journal of Political Economy 106, 172­204. Coffee, John, 1999, The future as history: The prospects for global convergence in corporate gover- nance and its implications, Northwestern University Law Review 93, 641­708. Coffee, John, 2001, Do norms matter? A cross-country examination of private benefits of control, Mimeo, Columbia University Law School. DeAngelo, Harry, and Linda De Angelo, 1985, Managerial ownership of voting rights, Journal of Financial Economics 14, 36­39. Demsetz, Harold, and Kenneth Lehn, 1985, The structure of corporate ownership: Causes and consequences, Journal of Political Economy 93, 1155­1175. Djankov, Simeon, and Peter Murrell, 2000, The determinants of enterprise restructuring in tran- sition: An assessment of the evidence, Working paper, University of Maryland. Doidge, Craig, 2002, U.S. cross listings and the private benefits of control: Evidence from dual class firms, Working paper, University of Toronto. Doidge, Craig, Andrew Karolyi, and Renee Stulz, 2001, Why are foreign firms listed in the U.S. worth more? Working paper, Ohio State University. Dyck, I.J. Alexander, 2000, Ownership structure, legal protections and corporate governance, in Boris Pleskovic, and Nicholas Stern, eds.: Annual World Bank Conference on Development Economics. Dyck, I.J. Alexander, 2001, Privatization and corporate governance: principles, evidence and future challenges, World Bank Research Observer 1, 59­84. Dyck, I.J. Alexander, and Luigi Zingales, 2002a, Private benefits of control: An international com- parison, Working paper 8711, NBER. Dyck, I.J. Alexander, and Luigi Zingales, 2002b, The corporate governance role of the media, in Roumeen Islam, ed.: The Right to Tell: The Role of Mass Media in Economic Development, Chap. 7 (The World Bank, Washington). Dyck, I.J. Alexander, and Luigi Zingales, 2003, The bubble and the media, in Peter Cornelius, and Bruce Kogut, eds.: Corporate Governance and Capital Flows in a Global Economy, Chap. 7 (Oxford University Press, New York). Glaeser, Edward, Simon Johnson, and Andrei Shleifer, 2001, Coase vs. the Coasians, Quarterly Journal of Economics 116, 853­900. Gresik, Thomas, 2001, The taxing task of taxing transnationals, Journal of Economic Literature 39, 800­838. Grossman, Sanford, and Oliver Hart, 1980, Takeover bids, the free rider problem, and the theory of the corporation, Bell Journal of Economics 11, 42­69. Grossman, Sanford, and Oliver Hart, 1988, One share one vote and the market for corporate control, Journal of Financial Economics 20, 175­202. Guiso, Luigi, Paola Sapienza, and Luigi Zingales, 2003, People's opium? Religion and economic attitudes, Journal of Monetary Economics 50, 225­282. Harris, Milton, and Artur Raviv, 1988, Corporate governance: Voting rights and majority rules, Journal of Financial Economics 20, 203­235. Holthausen, Robert, Richard Leftwich, and David Mayers, 1990, Large-block transactions, the speed of response, and temporary and permanent stock price effects, Journal of Financial Economics 26, 71­95. Chapter Five 201 Private Benefits of Control 599 Jensen, Michael, and William Meckling, 1976, Theory of the firm: Managerial behavior, agency costs and ownership structure, Journal of Financial Economics 3, 305­360. Johnson, Simon, Peter Boone, Alasdair Breach, Eric Friedman, 2000, Corporate governance in the Asian financial crisis, Journal of Financial Economics 58, 141­186. Johnson,Simon,RafaelLaPorta,FlorencioLopez-de-Salines,andAndreiShleifer,2000,Tunneling, American Economic Review 90, 22­27. La Porta, Rafael, Florencio Lopez-de-Salines, and Andrei Shleifer, 1999, Corporate ownership around the world, Journal of Finance 54, 471­517. La Porta, Rafael, Florencio Lopez-de-Salines, and Andrei Shleifer, 2000, Investor protection and corporate governance, Journal of Financial Economics 59, 3­27. La Porta, Rafael, Florencio Lopez-de-Salines, Andrei Shleifer, and Robert W. Vishny, 1997, Legal determinants of external finance, Journal of Finance 1, 1131­1150. La Porta, Rafael, Florencio Lopez-de-Salines, Andrei Shleifer, and Robert W. Vishny, 1998, Law and finance, Journal of Political Economy 6, 1113­1155. Lease, Ronald C., John J. McConnell, and Wayne H. Mikkelson, 1983, The market value of control in publicly traded corporations, Journal of Financial Economics 11, 439­471. Lease, Ronald C., John J. McConnell, and Wayne H. Mikkelson, 1984, The market value of differ- ential voting rights in closely held corporations, Journal of Business 4, 443­467. Lopez-de-Silanes, Florencio, 1997, Determinants of privatization prices, The Quarterly Journal of Economics 4, 965­1025. Megginson, William L., Robert C. Nash, Jeffrey M. Netter, and Annette B. Poulsen, 2000, The choice between private and public markets: Evidence from privatizations, Working paper, University of Georgia. Milgrom, Paul, and Robert Weber, 1982, A theory of auctions and competitive bidding, Econometrica 50, 1089­1122. Monks, Robert, and Nell Minnow, 1995, Watching the Watchers: Corporate Governance for the 21st Century (Blackwell Publishing, Malden, MA). Morck, Randall, Bernard Yeung, and Wayne Yu, 2000, The information content of stock markets: Why do emerging markets have synchronous stock price movements? Journal of Financial Economics 58, 215­260. Nenova, Tatiana, 2001a, The value of corporate votes and control benefits: A cross-country analysis, Mimeograph, Harvard University. Nenova, Tatiana, 2001b, Control values and changes in corporate law in Brazil, Working paper, The World Bank. Nicodano, Giovanna, and Alessandro Sembenelli, 2001, Private benefits, block transaction premi- ums, and ownership structure, Working paper, Universita' di Torino. Pagano, Marco, and Paolo Volpin, 2000, The political economy of corporate governance, Working paper, London Business School. Pistor, Katherina, Martin Raiser, and Satanislaw Gelfer, 2000, Law and finance in transition economies, Economics of Transition 82, 325­368. Rajan, Rajan, and Luigi Zingales 1998, Financial dependence and growth, American Economic Review 88, 559­586. Reese, William, and Michael Weisbach, 2001, Protection of minority shareholder interests, cross- listings in the United States, and subsequent equity offerings, Working paper, NBER. Reynolds, Thomas H., and Arturo A. Flores, 1989, Foreign Law: Current Sources of Codes and Basic Legislation in Jurisdictions of the World (F.B. Rothman, Littleton, CO). Rydqvist, Kristian, 1987, Empirical investigation of the voting premium, Working paper No. 35, Northwestern University. Shleifer, Andrei, and Robert Vishny, 1986, Large shareholders and corporate control, Journal of Political Economy 94, 461­488. Shleifer, Andrei, and Robert Vishny, 1997, A survey of corporate governance, Journal of Finance 52, 737­783. Stiglitz, Joseph, 1985, Credit markets and the control of capital, Journal of Money, Credit and Banking 17, 133­152. 202 A Reader in International Corporate Finance 600 The Journal of Finance Stulz, Renee, and Rohan Williamson, 2001, Culture, openness and finance, Working paper 8222, NBER. Zingales, Luigi, 1994, The value of the voting right: A study of the Milan stock exchange experience, Review of Financial Studies 7, 125­148. Zingales, Luigi, 1995a, What determines the value of corporate votes? Quarterly Journal of Eco- nomics 110, 1047­1073. Zingales, Luigi, 1995b, Insider ownership and the decision to go public, Review of Economic Studies 62, 425­448. Zingales, Luigi, 1998, Why it's worth being in control, in George Bickerstaffe, ed.: The Complete Finance Companion (FT Pitman Publishing, London). Zingales, Luigi, 2000, In search of new foundations, Journal of Finance 55, 1623­1653. Chapter Six 203 FERRETING OUT TUNNELING: AN APPLICATION TO INDIAN BUSINESS GROUPS* MARIANNE BERTRAND PARAS MEHTA SENDHIL MULLAINATHAN Owners of business groups are often accused of expropriating minority share- holders by tunneling resources from rms where they have low cash ow rights to rms where they have high cash ow rights. In this paper we propose a general methodology to measure the extent of tunneling activities. The methodology rests on isolatingand then testing the distinctive implicationsof the tunneling hypothe- sis for the propagation of earnings shocks across rms within a group. When we apply our methodology to data on Indian business groups, we nd a signi cant amount of tunneling, much of it occurring via nonoperating components of pro t. I. INTRODUCTION Weak corporate law and lax enforcement mechanisms raise fears of expropriation for minority shareholders around the world. These fears seem especially warranted in the presence of business groups, a common organizational form in many devel- oped and developing countries. In a business group, a single shareholder (or a family) completely controls several indepen- dently traded rms and yet has signi cant cash ow rights in only a few of them.1 This discrepancy in cash ow rights between the different rms he controls creates strong incentives to expro- priate. The controlling shareholder will want to transfer, or tun- nel, pro ts across rms, moving them from rms where he has * We thank Abhijit Banerjee, Simon Johnson, Tarun Khanna, Jayendra Nayak, Ajay Shah, Susan Thomas, two anonymous referees, the editor (Edward Glaeser), and seminar participants at the MIT Development and Public Finance Lunches, the Harvard/MIT Development Seminar, the NBER-NCAER Conference on Reforms, the Harvard Business School Conference on Emerging Markets, the University of Michigan, the London Business School, the London School of Eco- nomics, the University of Chicago Graduate School of Business, and Princeton University for their useful comments. The second author is also grateful for nancial support from a National Science Foundation Graduate Fellowship. 1. In many cases, control is maintained through indirect ownership. For example, the ultimate owner may own rm A, which in turn owns rm B, which in turn owns rm C. Such ownership structures, which are quite common accord- ing to La Porta, Lopez-d-Silanes,Shleifer, and Vishny [1999], are called pyramids. It is the chain of ownership in pyramids that generates the sharp divergence between control and cash ow rights. Dual class shares are another way to generate such a divergence. In India, the country we study below, dual class shares have not been allowed so far, although recent legislation has attempted to change this. © 2002 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology. The Quarterly Journal of Economics, February 2002 121 204 A Reader in International Corporate Finance 122 QUARTERLY JOURNAL OF ECONOMICS low cash ow rights to rms where he has high cash ow rights.2 Cash can be transferred in many ways: the rms can give each other high (or low) interest rate loans, manipulate transfer prices, or sell assets to each other at above or below market prices, to list just a few. If prevalent, tunneling may have serious conse- quences. By reducing the returns to being an outside shareholder, it can hinder equity market growth and overall nancial devel- opment. Illicit pro t transfers may also reduce the transparency of the entire economy, clouding the accounting numbers and complicating any inference about rms' health. In fact, several observers argued that tunneling made it hard to assess solvency during the emerging market crises of 1997­1998, and possibly exacerbated the crisis.3 Anecdotes of tunneling are easy to nd. In India, for example, one group rm, Kalyani Steels, had more than two-thirds of its net worth invested in other companies in its group. Yet these investments yielded less than a 1 percent rate of return, fueling speculation that they were merely a way to tunnel pro ts out of Kalyani Steels. However, hard evidence of tunneling beyond an- ecdotes of this kind remains scarce, perhaps because of the illicit nature of this activity. The strongest statistical evidence so far is cross-sectional: group rms where the controlling shareholder has higher cash ow rights have higher q-ratios and greater pro tability.4 While informative, this cross-sectional relationship is not a test of tunneling since it could also result from differences in preexisting ef ciency or any number of other unobservable factors. This paper introduces a general procedure to quantify tun- neling. It is based on tracing the propagation of earnings shocks 2. Johnson, La Porta, Lopez-de-Silanes, and Shleifer [2000] argue that the expropriation threat is especially big in business groups. Bebchuk, Kraakman, and Triantis [2000], Wolfenzon [1999], and Shleifer and Wolfenzon [2000] provide theoretical models of various forms of tunneling. In the United States something akin to business groups existed historically, although cartelization was the major issue surrounding them. In modern times, expropriation of shareholders in large U. S. rms is thought to occur through poor decision making [Berle and Means 1934; Jensen and Meckling 1976] or high executive compensation [Bertrand and Mullainathan 2000, 2001]. 3. Johnson, Boone, Breach, and Friedman [2000] show that countries with better legal protection against tunneling were less affected by the crisis. 4. Examples of papers that have documented such correlations include Bian- chi, Bianco, and Enriques [1999], Claessens, Djankov, Fan, and Lang [1999], and Claessens, Djankov, and Lang [2000]. A broader literature has studied groups more generally [Khanna and Palepu 2000; Hoshi, Kashyap, and Scharfstein 1991]. Other papers have documented differences in the price of voting and nonvoting shares [Zingales 1995; Nenova 1999]. Chapter Six 205 FERRETING OUT TUNNELING 123 through a business group. Consider a group with two rms: rm H, where the controlling shareholder has high cash ow rights, and rm L, where he has low cash ow rights. Suppose that rm L experiences a shock that would (in the absence of tunneling) cause its pro ts to rise by 100 dollars. Because some of this increase will be tunneled out of rm L, the actual pro ts of rm L will rise by less than 100 dollars, with the shortfall measuring the amount of diversion. Since the shortfall is being tunneled to H, we would also expect H to respond to L's shock even though H is not directly affected by it. Moreover, we would not expect this pattern if instead H were to receive the shock: there is no incen- tive to tunnel from a high- to a low-cash- ow-right rm.5 We develop a general set of tests based on these observations and use variation in mean industry performance as a source of pro t shocks.6 As an illustration, we apply this test to a panel of Indian rms. We nd evidence for the full set of predictions implied by tunneling. Other results suggest that these ndings are not due to mismeasurement of a rm's industry, simple coinsurance within groups or internal capital markets. Moreover, the magni- tudes of the effects we nd are large: more than 25 percent of the marginal rupee of pro ts in low-cash- ow-right rms appears to be dissipated.7 Our procedure further allows us to examine the mechanics of tunneling. Indian groups appear to tunnel by manipulating non- operating components of pro ts (such as miscellaneous and non- recurring items). In fact, there is no evidence of tunneling on operating pro ts alone. Rather, nonoperating losses and gains seem to be used to offset real pro t shocks or transfer cash from other rms. Finally, we examine whether market prices incorpo- rate tunneling. We nd that high market-to-book rms are more 5. This asymmetry is important. Money ows only from low- to high-cash- ow-right rms, not vice versa. As we will see, this is a crucial distinction between tunneling and other theories of why shocks might propagate through a group, most notably risk sharing. 6. Other papers have used shocks in a related way. Blanchard, Lopez-de- Silanes, and Shleifer [1994] examine how U. S. rms respond to windfalls (win- ning a law suit) to assess agency models. Lamont [1997] uses the oil shock to assess the effects of cash ow on investment. Bertrand and Mullainathan [2001] use several shock measures to assess the effects of luck on CEO pay. 7. It is worth noting that business groups may add social value in other ways that offset the social costs they may impose through tunneling. They might help reduce transaction costs, solve external market failures, or provide reputational capital for their members. We will not, therefore, be attempting to test whether groups are on net bad but merely whether, and if so how much, they tunnel. 206 A Reader in International Corporate Finance 124 QUARTERLY JOURNAL OF ECONOMICS sensitive to both their own shock and shocks to the other rms in their group. Firms whose group has a high market-to-book are also more sensitive to their own shock, but are not signi cantly more sensitive to the group's shock. This suggests that the stock market at least partly penalizes tunneling activities. II. A TEST FOR TUNNELING We begin by describing the exact implications of tunneling for the propagation of shocks.8 Let us return to the ctional example of two group rms, high-cash- ow-right rm H and low-cash- ow-right rm L. Consider again a 100-dollar pro ts shock affecting rm L. Because the controlling shareholder would bene t more if these 100 dollars were in H, he will look for a way to divert them out of L. This gives the rst prediction: group rms should on average underrespond to shocks to their own pro ts. Of course, since tunneling may be costly (either because of resource dissipation or because of a risk of being caught), the controlling shareholder may transfer only some of the 100 dollars out of rm L. How much he transfers will be a function of his cash ow rights in L. The less his cash ow rights in L, the less he values the extra dollar left in L and the more of the pro ts he will want to tunnel out of L. This gives the second prediction: the underresponse to shocks to own pro ts should be larger in low- cash- ow-right rms. The cash tunneled from rm L eventually ends up in rm H. So H will appear to respond to L's shock even though H is not directly affected by L's shock. This gives the third prediction: group rms will on average be sensitive to shocks affecting other rms in the group.9 We know from above that when cash ow rights in rm L are low, more money will be tunneled out of L. But this also implies that more money will be tunneled into H when cash ow rights in L are low. This gives the fourth prediction: group rms will be more sensitive to shocks affecting low-cash- ow-right rms in their group than to shocks affecting high-cash- ow-right rms. 8. Bertrand, Mehta, and Mullainathan [2000] present a model that formal- izes these implications. 9. This prediction distinguishes tunneling from a pure mismanagement in- terpretation of the pro ts shortfall. The rst two predictions could simply re ect a dissipation of resources through inef cient operation rather than a diversion to other group rms. Chapter Six 207 FERRETING OUT TUNNELING 125 Finally, suppose that a 100-dollar shock were now to affect rm H instead of rm L. Since the controlling shareholder has more cash ow rights in H than in L, he will have no incentives to tunnel from H to L. This means that H will respond one for one to its own shock, which is just another way to understand the second prediction above. It also means that L will not be sensitive to H's shock. A more general version of this observation gives the fth prediction: low-cash- ow-right rms will be less sensitive to shocks affecting other rms in their group. To transform these general predictions into testable implica- tions, we need to isolate speci c shocks using available data. Industry shocks provide an ideal candidate since they affect in- dividual rms but are to a large extent beyond the control of individual rms. Some notation will be helpful in de ning these mean industry movements. Let perfktI be a level measure of reported performance for rm k in industry I at time t (in our case pro ts before depreciation, interest, and taxes). AktI be a measure of the rm k's assets (in our case, total book value of assets), and rktI 5 perfktI/AktI be a measure of return on assets for that rm. To isolate the industry shock, we compute the asset-weighted average return for all rms in industry I: r^It 5 Sk AktIrktI/ Sk AktI.10 Given this industry return, we can predict what rm k's performance ought to be in the absence of tunneling by calcu- lating predktI 5 AktI p r^It. Our empirical test will then consist of regressing a rm's actual reported performance on its predicted performance and on the predicted performance of other rms in its group.11 More speci cally, we can test the ve implications above: (1) group rms should be less sensitive to shocks to their industry than nongroup (stand-alone) rms; (2) low-cash- ow-right group rms will show smaller sensitivities to shocks to their industry than high-cash- ow-right ones; (3) group rms should be sensitive to industry shocks affecting other rms in their group; (4) group rms should be especially sensitive to shocks affecting the low cash- ow-right rms in their group; (5) low-cash- ow-right group 10. A mechanical correlation arises if we include a rm itself in estimatingits industry return and then use that industry return to predict the rm's own return. To prevent this, we exclude, for every rm, the rm itself in computing its industry return. In this sense, r^It should actually be indexed by k, but we drop this subscript for simplicity. 11. Given that this is a predicted level of performance, our terminology of shocks may seem inappropriate. But since we include rm xed effects, we will in fact be identifying the effect of industry shocks. 208 A Reader in International Corporate Finance 126 QUARTERLY JOURNAL OF ECONOMICS rms should show smaller sensitivities in predictions 3 and 4. These ve predictions form a simple test of tunneling, one that requires only rm-level data on earnings, industry, group mem- bership, and ownership structure.12 III. AN APPLICATION TO NDIAN I BUSINESS GROUPS We now apply this test to Indian data. As in many other countries, group rms in India are often linked together through the ownership of equity shares. In most cases, the controlling shareholder is a family; among the best-known business families in India are Tata, Bajaj, Birla, Oberoi, and Mahindra.13 Nominally, corporate governance laws in India are quite good, consistent with its English colonial past and its common law heritage [Sarkar and Sarkar 1999]. In reality, however, cor- ruption makes these laws dif cult to enforce and shareholder expropriation a major concern in India. In recent years the role of corporate governance in nancial development has received sig- ni cant attention from the Indian business press and central government. Business groups have come under particular scru- tiny for advancing their private interests at the expense of out- side shareholders.14 Tunneling is also allegedly a problem.15 In- deed, greater oversight of related party transactions was one of 12. A notable feature of these tests is their symmetry. One might have thought that there should be no tunneling for negative groups. This is in fact not clear. For example, suppose that an industry earns a 10 percent natural rate of return and a negative shock reduces it to 5 percent. Since this reduces the amount that can be tunneled out, we will see just as much sensitivity to this shock (for example, among high-cash- ow-right group rms) as to a positive one. Rather than asymmetry in changes, one might expect that below some nominal rate of return, tunneling would cease. A priori, it is unclear where this threshold lies. We tried some thresholds (e.g., zero nominal rate of return) and found standard errors that were too large to reject either linearity or signi cant nonlinearity. Johnson and Friedman [2000] provide further discussion of asymmetry. 13. Piramal [1996] and Dutta [1997] provide accounts of groups in India. 14. One Financial Times Asia article charges that the "boards of Indian companies, especially the family-owned ones, are prime examples of crony capi- talism. They are invariably lled with family members and friends. . . . In such an environment, the promoter can operate to further his own interests even as he takes the other shareholders for a ride." 15. A 1998 Financial Times Asia article reports that "[c]hanneling funds to subsidiaries and group companies in the form of low or nil interest loans or low-yield investments is not new. Such a lockup of costly funds often results in poor nancial performance. JCT, Kalyani Steels, Bombay Burmah Trading Com- pany; and DCM Shriram Industries are examples. JCT's average return over the last four years on outstanding loans and advances of Rs. 270 crores is is just 4 percent. Similarly Kalyani Steels' 1996­97 investments in group companies was worth Rs. 196.80 crores--more than two-thirds its net worth--while the company earned just 1.45 crores as dividends." Chapter Six 209 FERRETING OUT TUNNELING 127 the speci c recommendations made by a government committee organized to study corporate governance.16 Thus, with its weak corporate governance and allegations of impropriety, India pro- vides an ideal location to test for tunneling. III.A. Data Source We use Prowess, a publicly available database maintained by the Centre for Monitoring Indian Economy (CMIE). Prowess in- cludes annual report information for companies in India between 1989 and 1999. It provides much of the information needed for this analysis: nancial statements, industry information, group af liation for each rm, and some corporate ownership data. We exclude state-owned and foreign-owned rms from our sample since these may not be comparable to the private sector domestic rms that interest us. Our sample contains about 18,500 rm- year observations, although sample sizes vary because of missing variables for some rms.17 We rely on CMIE classi cation of rms into group and non- group rms, and of group rms into speci c group af liation. CMIE classi cation is based on a "continuous monitoring of com- pany announcements and a qualitative understanding of the groupwise behavior of individual companies" (Prowess Users' Manual, v.2, p.4). Note also that CMIE assigns each company to a unique ownership group, based on the group most closely asso- ciated with that company. Conversations with local experts cor- roborate these classi cations; which group a rm belongs to is widely known. 16. The Kumar Mangalam Committee recommended measures to strengthen the board of directors' role in "reduc[ing] potential con ict between the speci c interests of managementand the wider interests of the company and shareholders including misuse of corporate assets and abuse in related party transactions." These measures included guidelines for strengthening the independence of boards and for the establishment of an audit committee by the board of directors to review, among other things, "[a]ny related party transactions, i.e. transactions of the company of material nature with promoters or the management, their sub- sidiaries or relatives, etc. that may have potential con ict with the interests of the company at large." 17. Prowess does not use consolidated accounting data, which implies that our ndings are not caused by accounting mechanics. In fact, during the sample period under study, Indian accounting standards did not require disclosing con- solidated accounts for group rms. Very few rms used consolidated nancial statements in practice [Price, Waterhouse & Co. 1999]. 210 A Reader in International Corporate Finance 128 QUARTERLY JOURNAL OF ECONOMICS III.B. Measurement of Controlling Shareholder's Cash Flow Rights A key variable in our analysis is the cash ow rights of the controlling shareholder in a particular rm. There are two com- ponents to cash ow rights. First are direct rights, which are derived from shares that the controlling shareholder (or his fam- ily) has in the company. Second are indirect rights, which are derived from shares held by another company in which the con- trolling shareholder has some shares. Prowess provides two reasonable proxies for direct cash ow rights. Both are derived from data on equity holding patterns, which is available for about 60 percent of rms (all of them publicly traded). For these rms, CMIE reports the shares of equity held by foreigners, directors, various nancial institutions, banks, various governmental bodies, the top fty shareholders, corporate bodies, and others.18 As in many countries, Indian families typically control the rms they have nancial stakes in by appointing family members or family friends to the board of directors and to top managerial positions. Since the company shares held by these board members bene t the controlling shareholder in some sense, the information on director ownership provides a rst proxy for direct cash ow rights.19 The equity held by "other shareholders," where others are de ned as shareholders that are neither directors, nor banks, nor foreigners, not nancial institutions, nor government bodies, nor corporate bodies, nor the top fty shareholders, provides a second proxy. By measuring the shares held by small, minority share- 18. The exact ownership categories reported by CMIE are Foreigners, Insur- ance Companies, Life Insurance Corporation, General Insurance Corporation, Mutual Funds, Unit Trust of India, Financial Institutions (Industrial Financial Corporation of India, Industrial Development Bank of India, Industrial Credit and Investment Bank of India, Industrial Credit and Investment Corporation, Commercial Banks), Government Companies (Central Government Companies, State Government Companies), State Finance Corporation, Other Government Organizations, Corporate Bodies, Directors, Top Fifty Shareholders, and Others. 19. For example, the Financial Times Asia reports that "the boards of Indian companies . . . are invariably lled with family members and friends, whether or not they are quali ed for the position" [Financial Times Asia Intelligence Wire, October 10, 1999]. The article goes on to say: "In such an environment, the promoter can operate to further his own interests even as he takes the other shareholders for a ride." Of course, if some of the directors are not family members or friends, this proxy will overstate the direct cash ow rights. Chapter Six 211 FERRETING OUT TUNNELING 129 holders, it captures the amount of cash ow rights the family does not own.20 Although both variables are good proxies for direct cash ow rights, they do little to capture indirect cash ow rights. Because Prowess only provides information by ownership category, it is impossible to back out of such indirect cash ow rights.21 Conse- quently, our ranking of rms (in terms of cash ow rights) within a group is noisy. For example, suppose that the ultimate owner owns 10 percent of rms A and B and rm B owns 40 percent in rm A. The ultimate owner seemingly has a 10 percent direct cash stake in both rms but actually has a 14 percent stake in rm A. If we modify the example so that the direct ownership stake in rm A is actually 9 percent, then adding indirect cash ow rights reverses the ranking.22 Three points should be noted about this important measure- ment issue. First, indirect cash ow rights by their very nature should be smaller than direct rights because they are diminished as they pass through the chain of ownership. In the above exam- ple, despite the large indirect ownership of A by B (40 percent), the nal difference is only 4 percent since A has only a 10 percent direct stake in B. Moreover, when our ranking of rms was wrong in the second example above, this was because both B and A were very close in terms of direct cash ow rights (10 percent versus 9 percent).23 Second, to the extent that any signi cant error is introduced into our rankings of rms, there will be an attenuation bias. This will bias our estimates toward zero, raise standard errors, and make it more dif cult to nd evidence of tunneling. Finally, although these imperfect measures may make the CMIE 20. The two measures, the equity stake of directors and the equity stake held by minority shareholders, correlate negatively. The correlation is imperfect, how- ever, (about 2.35 for group rms), suggesting that these are not redundant proxies. Besides measuring the absolute level of director and other equity hold- ings, we also measure their relative levels within each group. Finally, because we use within-group differences in director and other ownership levels to identify the direction and magnitude of money ows across rms in a business group, we exclude from the sample all groups where there is no difference between the maximum and the minimum level of direct ownership or between the maximum and minimum level of other ownership. 21. Indian disclosure laws do not mandate release of this information. We have attempted to gather this information in many other ways, from investment bankers to the groups themselves; our attempts have been fruitless. 22. We are grateful to an anonymous referee for providing variants of these examples. 23. This is not to say that one cannot construct examples where indicted ownership matters, but rather that because of the multiplication by the direct ownership in rms, indirect ownership will have on average a smaller effect on cash ow rights. 212 A Reader in International Corporate Finance 130 QUARTERLY JOURNAL OF ECONOMICS data a less than perfect place to apply our test, it is highly representative of the typical data available to implement our test in most countries. Detailed data on ownership between rms are usually hard to get, whereas many countries have readily avail- able categorical ownership data of the kind provided by CMIE. III.C. Measurement of Performance The CMIE data were collected with a focus on accounting numbers. Consequently, we cannot use it to compute reliable annual stock return measures for many rms between 1989 and 1999. More speci cally, we lack dividend data for many observa- tions, which is especially troubling since dividend payments would be the most direct way for a controlling shareholder to affect nal returns.24 Moreover, comparisons with both aggregate data and data on speci c rms from the Bombay Stock Exchange show that the stock prices reported on CMIE are themselves noisy. In several cases, the returns we computed lagged or led true returns.25 These problems constrain us to use the more reliable "pro ts before depreciation, interest and tax" as our speci c performance measure, perfktI. Our asset measure, As- setsktI, is total assets. Each rm's industry comes from CMIE's classi cation of rms into industries. Our sample contains 134 different "four-digit" industries.26 III.D. Summary Statistics Table I reports summary statistics for the full sample and for group and nongroup rms separately. In this table, and through- out the remainder of the paper, nongroup rms are referred to as "stand-alones." Group rms and stand-alones, respectively, ac- count for about 7,500 and 11,000 of the observations in our full sample. All nominal variables in the sample are de ated using 24. By examining the rms with some, not necessarily reliable dividend data, we see that dividends are a sizable fraction of returns. 25. Despite the noisiness, we did estimate the regressions below using mar- ket value as a dependent variable, and the results are quite similar. But, because of ths noisiness of the data, we do not have great faith in these results. They are available as Table B in the unpublished appendix, available from the authors upon request. The average level of market capitalization appears much more reliable, however, and we use it in subsection IV.B. to relate q ratios to the extent of tunneling. 26. They can be found in Table A of the appendix available from the authors upon request. The breakdown is at roughly the level of the four-digit SIC code in the United States. Chapter Six 213 FERRETING OUT TUNNELING 131 TABLE I SUMMARY STATISTICS Sample: All Groups Stand-alones Total assets 131.80 252.76 49.69 (525.91) (741.6) (272.66) Total sales 94.39 188.16 30.73 (305.66) (459.77) (57.84) Pro t before depreciation, interest, and 16.84 32.90 5.94 taxes (63.84) (90.99) (30.48) Ratio of PBDIT to total assets .126 .142 .115 (.128) (.115) (.134) Ratio of operating pro t to total assets .284 .328 .254 (.285) (.312) (.261) Ratio of nonoperating pro t to total assets 2.157 2.186 21.38 (.259) (.288) (.235) q ratio .537 .645 .447 (.818) (.916) (.714) Year of incorporation 1974.55 1967.51 1979.33 (20.03) (22.89) (16.18) Director equity 16.70 7.45 22.99 (18.33) (13.05) (18.72) Other ownership 29.90 27.57 31.48 (17.39) (16.06) (18.07) Director equity spread -- 15.19 -- (14.88) Other ownership spread -- 33.31 -- (21.66) Sample size 18600 7521 11079 a. Data Source: Prowess, Centre for Monitoring Indian Economy (CMIE), for the years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. b. Standard deviations are in parentheses. c. "Operating pro t" refers to manufacturing sales revenue minus total raw material expenses, energy expenses, and wages and salaries. "q ratio" is the ratio of market valuation to total assets. "Director equity spread" is the difference between the minimum and maximum level of director equity in a group; "Other ownershipspread" is the differencebetweenthe minimum and maximum level of other ownershipin a group. Ownership and ownership spread variables are measured in percentages and so range from 0 to 100. the Consumer Price Index series from the International Financial Statistics of the International Monetary Fund (1995 5 100). The average group rm in the sample belongs to a group with fteen rms. Many groups in our data, however, consist of two or three rms.27 Group rms are, on average, twelve years older than nongroup rms: the typical group rm was created in 1967, 27. Some ownership groups have several smaller companies that are set up for taxation or retail business purposes. It is much more dif cult for CMIE to get access to the annual reports of these smaller companies. CMIE also tracks sub- 214 A Reader in International Corporate Finance 132 QUARTERLY JOURNAL OF ECONOMICS TABLE II SENSITIVITY TO OWN SHOCK: GROUP VERSUS STAND-ALONE DEPENDENT VARIABLE: PROFIT BEFORE DIT (1) (2) (3) (4) Own shock 1.05 .10 24.58 25.10 (.02) (.05) (.48) (.47) Own shock* 2.30 2.30 2.26 2.27 group (.02) (.02) (.02) (.02) Ln assets .16 2.98 2.33 2.47 (.32) (.34) (.33) (.34) Own shock* ln -- .10 -- 1.0 assets (.00) (.01) Own shock* -- -- .003 .003 year of incorp. (.000) (.000) Sample size 18600 18600 18588 18588 Adjusted R2 .93 .93 .93 .93 a. Data Source: Prowess, Centre for Monitoring Indian Economy, for years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. Sample includes both stand- alone and group rms. b. All regressions also include year xed effect and rm xed effects. c. Standard errors are in parentheses. the typical stand-alone rm in 1979. More importantly, group rms tend to be much larger than stand-alones. The average group rm has total assets of Rs. 253 crores, while the average stand-alone has total assets of Rs. 52 crores. Stand-alones also have lower levels of sales and pro ts. We will control for these size and age differences in our analysis. The average level of director ownership among group rms is 7.5 percent. The average level of ownership by other shareholders is 27.5 percent. The gap in director ownership between the top and bottom of a group (i.e., the gap between the rm with the highest level of director ownership and the rm with the lowest level of director ownership) is 15 percent on average. The average gap in other ownership is 33 percent. III.E. Sensitivity to Own Shock In Table II we test the rst prediction of tunneling: group rms should be less sensitive to shocks to their own industry than stand-alones. We estimate sidiary companies with small turnover but does not include them in the database we use in this paper. Chapter Six 215 FERRETING OUT TUNNELING 133 (1) perfkt 5 a 1 b~ predkt! 1 c~ groupk p predkt! 1 d~controlskt! 1 Firmk1 Timet, where groupk is a dummy variable for whether rm k is in a group or not, controlskt are other variables that might affect rm performance (speci cally age and log assets), Firmk are rm xed effects, and Timet are time dummies.28 The coef cient b mea- sures the general sensitivity of rms to industry performance; the interaction term groupk p predkt captures the differential sensi- tivity of group rms. If group rms are less sensitive, as tunnel- ing would predict, then c should be negative. Note that because the regression is expressed in performance levels, the magnitude of the effects can easily be interpreted. Column (1) displays our basic result. A one-rupee shock leads to about a one-rupee (1.05) increase in earnings for a stand-alone rm. For a group rm, it leads to .3 rupee smaller increase, or only a .75 rupee increase.29 This suggests that 30 percent of all the money placed into a group rm is somehow dissipated. In Table I we saw that stand-alone rms are smaller and older on average than group rms. This could confound our esti- mate of the effect of group af liation if size or age affects a rm's responsiveness to shocks. In column (2) we include an interaction between the logarithm of total assets and the industry shock. In column (3) we do the same for age. In column (4) we include both interactions simultaneously. The direct effects are always in- cluded. From these, it is clear that both size and age do affect the responsiveness to shocks. But it is also clear that the difference between group and stand-alone rms remains signi cant even in the presence of additional controls.30 In short, the data support the rst prediction. 28. The inclusion of rm xed effects deals with several issues. First, even though we are using level of predicted performance, we are identifying off of changes in predicted performance, hence our use of the term "shocks" throughout the paper. Second, the xed effects account for any inherent, xed differences between rms. Third, because rms do not change groups in our sample, the rm xed effects also account for any xed differences between groups. 29. We have also estimated this and all regressions below excluding small groups, which we de ne as groups with less than ve rms in the CMIE data. The results were not affected when we restrict ourselves to that subsample. 30. We have also attempted more exible speci cations by allowing for more nonlinear terms for size and age in the interaction. These produced identical results. 216 A Reader in International Corporate Finance 134 QUARTERLY JOURNAL OF ECONOMICS The second prediction provides a more stringent test: within- group rms, high-cash- ow-right rms should show greater sen- sitivity to own shocks. We estimate for the set of group rms (2) perfkt 5 a 1 b~ predkt! 1 c~cashk p predkt! 1 d~controlskt! 1 Firmk1 Timet, where cashk is the cash ow rights of the controlling party in rm k, measured either with director or other ownership. The inter- action term, cashk p predkt, measures differential sensitivity by level of cash ow rights. Under the tunneling hypothesis, we would expect c . 0.31 Panel A of Table III uses director equity as the proxy for cash ow rights. Column (1) shows that group rms where director equity is higher are more sensitive to their own industry shock. Each one-percentage point increase in director equity increases the sensitivity to a one-rupee industry shock by .03 rupee. Recall that among group rms, the average difference in director own- ership between the rm with the greatest and the rm with the lowest director ownership was about 15. Thus, for each rupee of industry shock, the typical rm with the highest director owner- ship is .45 rupee more sensitive than the typical rm with the lowest director ownership. This suggests that group rms with high controlling party's cash ow rights may be as sensitive to the marginal rupee as stand-alone rms. The magnitude of this effect is striking and suggests that ownership plays a large role in the extent of the sensitivity. To assess whether the ndings in column (1) capture some aspects of director ownership that are unrelated to group mem- bership, we reestimate equation (2) in column (3) on the sub- sample of stand-alone rms. We nd that director ownership also increases the responsiveness to shocks for stand-alone rms. The effect, however, is quantitatively much smaller, only a sixth of the size of the effect for group rms (.004 versus .025 for group rms). In columns (2) and (4) we allow for the effect of own industry shock to differ by rm size and rm age. These additional controls do not alter the estimated coef cient on "Own shock z director equity" for the sample of group rms (column (2)). They do, however, lead to an increase in the coef cient on "Own shock z 31. When we use "Other ownership" in the interaction, we expect a negative term since this measure is negatively related to cash ow rights. Chapter Six 217 FERRETING OUT TUNNELING 135 TABLE III SENSITIVITY TO OWN SHOCK BY DIRECTOR AND OTHER OWNERSHIP DEPENDENT VARIABLE: PROFIT BEFORE DIT Panel A: Director equity Sample: Stand- Stand- Groups Groups alones alones (1) (2) (3) (4) Own shock .713 25.075 1.058 24.316 (.009) (.742) (.006) (.518) Own shock p director equity .025 .030 .004 .019 (.003) (.003) (.001) (.001) Ln assets .052 4.261 2.590 1.568 (.733) (.807) (.176) (.178) Own shock p ln assets -- .118 -- .201 (.008) (.006) Own shock p year of incorp. -- .002 -- .002 (.000) (.000) Sample size 7521 7510 11079 11078 Adjusted R2 .92 .93 .95 .96 Panel B: Other ownership Sample: Stand- Stand- Groups Groups alones alones (1) (2) (3) (4) Own shock .919 25.764 1.033 23.983 (.023) (.743) (.052) (.603) Own shock p other ownership 2.007 2.007 .001 .002 (.001) (.001) (.000) (.000) Ln assets 1.616 5.189 2.292 2.049 (.724) (.806) (.166) (.180) Own shock p ln assets -- .103 -- .154 (.008) (.006) Own shock p year of incorp. -- .003 -- .002 (.003) (.000) Sample size 7521 7510 11079 11078 Adjusted R2 .92 .93 .95 .96 a. Data Source: Prowess, Centre for Monitoring Indian Economy, for years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. b. All regressions also include year xed effect and rm xed effects. c. Standard errors are in parentheses. 218 A Reader in International Corporate Finance 136 QUARTERLY JOURNAL OF ECONOMICS director equity" in the sample of stand-alone rms (.019 instead of .004). Because standard errors are relatively small, we can still reject that the effect of director ownership on industry shock sensitivity is the same between group rms and stand-alone rms. More director equity increases the responsiveness of a rm to its own industry shock, and this effect is signi cantly larger among group rms. In Panel B of Table III we use our other proxy for direct cash ow rights, the ownership stake of other small sharehold- ers. As predicted, we nd that the sensitivity of a group rm to its own industry shock decreases with its level of other own- ership. A one-percentage point increase in other ownership decreases the responsiveness of a group rm to a one-rupee shock by about .01 rupee (column (1)). Given that the average spread between highest and lowest other ownership among group rms is about 33, the implied magnitude of the effect is the same as in Panel A. Among stand-alone rms (column (3)) the effect of other ownership is of the opposite sign and eco- nomically small. Finally, note that the coef cient on "Own shock z other ownership" is roughly unaffected by the inclusion of controls for rm age and rm size interacted with own industry shock (columns (2) and (4) for group and stand-alone rms, respectively). In summary, these results in Table III are consistent with the idea that fewer resources are tunneled out of the group rms where the promoting family has higher equity stakes and where there are fewer minority shareholders to expropriate. In fact, group rms where the controlling party has a large stake show the same sensitivity to their own industry shocks as stand-alone rms. III.F. Sensitivity to Group Shocks We now examine whether a rm responds to shocks affecting other rms in its group (prediction 3). We estimate (3) perfkt 5 a 1 b~ predkt! 1 c~opredkt! 1 d~controlskt! 1 Firmk1 Timet, where opredkt 5 Sj Þk predjt, the sum being over all other rms in the same business group (excluding the rm itself). A positive Chapter Six 219 FERRETING OUT TUNNELING 137 TABLE IV SENSITIVITY OF GROUP FIRMS TO GROUP AND SUBGROUP SHOCKS DEPENDENT VARIABLE: PROFIT BEFORE DIT (1) (2) (3) (4) (5) Own shock .730 .732 .732 .732 .732 (.009) (.009) (.009) (.009) (.009) Group shock .011 -- -- -- -- (.001) Shock below median -- .016 -- -- -- (director equity) (.002) Shock above median -- 2.002 -- -- -- (director equity) (.005) Shock below 66th pctile -- -- .015 -- -- (director equity) (.002) Shock above 66th pctile -- -- 2.001 -- -- (director equity) (.001) Shock above median -- -- -- .014 -- (other ownership) (.002) Shock below median -- -- -- .007 -- (other ownership) (.004) Shock above 33rd pctile -- -- -- -- .017 (other ownership) (.002) Shock below 33rd pctile -- -- -- -- 2.002 (other ownership) (.004) Sample size 7521 7521 7521 7521 7521 Adjusted R2 .93 .92 .92 .92 .92 a. Data Source: Prowess, Centre for Monitoring Indian Economy, for years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. b. Sample is group rms only. c. "Shock below median (director equity)" is a variable that sums the industry shocks to all the rms in the same group (excluding the rm itself) that have below median level of director ownership in their group. All the other subgroup shocks are de ned accordingly. d. Also included in each regression are the logarithm of total assets, year xed effects, and rm xed effects. e. Standard errors are in parentheses. coef cient on opredkt suggests that rms within a group are in fact sensitive to each other's shocks.32 In column (1) of Table IV we nd a moderate response of group rms to each other's shocks. The coef cient on "Group shock" of .011 suggests that for each rupee earned by the group, an average rm in the group receives .011 rupee. Since we know that group rms underreact by about 1 2 .73 5 .27 rupee to a 32. Note that we control for the rm's own shock, predkt. This control means that we do not confuse an overlap of industry between rms in the same group with a ow of cash within that group. 220 A Reader in International Corporate Finance 138 QUARTERLY JOURNAL OF ECONOMICS one-rupee shock and since there are about fteen rms in each group, this coef cient implies that about 61 percent of the money that is tunneled out reappears elsewhere in the group.33 The next prediction of tunneling (prediction 4) is that the source of the shock matters: rms should respond more to groups affecting low-cash- ow-right rms than to groups affecting high- cash- ow-right rms. We study this prediction in columns (2) to (5). We de ne Hopredkt as the sum of shocks affecting all high cash- ow-right rms in k's group and Lopredkt as the equivalent sum for low-cash- ow-right rms. We then estimate (4) perfkt 5 a 1 b~ predkt! 1 cL~Lopredkt! 1 cH~Hopredkt! 1 d~controlskt! 1 Firmk1 Timet. If group rms are in fact more sensitive to groups to the rms with low cash ow rights, we should nd that cL . cH. In column (2) we classify a group's rms as low- or high-cash- ow-right using the median director equity in that group as a threshold. We nd that rms show greater sensitivity to shocks affecting the low-cash- ow-right rms in their group. A one-rupee shock to rms below group median in terms of director ownership increases the average group rm's earnings by .02 rupee. By contrast, the average group rm's earnings do not respond to industry shocks to rms in the high-cash- ow-right group. Col- umn (3) instead contrasts shocks to rms below and above the sixty-sixth percentile of director equity in their group. This iso- lates a smaller group of rms in the high-cash- ow-right group and allows resources to be equally skimmed from a larger number of rms. The results are very similar. In column (4) we classify a group's rms as low- or high-cash- ow-right using the median other shareholders' equity in that group as a threshold. In this case, we nd that the average group rm is equally sensitive to shocks to the two subgroups. In col- umn (5) we isolate a larger set of rms with low cash ow rights by using the thirty-third percentile of other shareholders' equity as the breaking point. The results suggest that few to no re- sources are transferred from the subgroup of rms with low levels of other equity. In contrast, the coef cient on the shock to rms 33. The remaining 39 percent may be a dissipation factor, suggesting real costs of redistribution. Alternatively, it may re ect redistribution to rms that are not in our sample. Most notably, tunneling may occur through nonpublic rms such as holding companies, which are not represented in our data set. Chapter Six 221 FERRETING OUT TUNNELING 139 TABLE V SENSITIVITY TO GROUP SHOCK BY LEVEL OF DIRECTOR OWNERSHIP IN GROUP DEPENDENT VARIABLE: PROFIT BEFORE DIT (1) (2) (3) (4) (5) (6) (7) (8) Below topmost Level in group: Lower 2 3 Top 1 3 rm Topmost rm Own shock .62 .89 .63 .63 .63 1.01 1.01 1.01 (.01) (.02) (.01) (.01) (.01) (.02) (.02) (.02) Group shock .013 .010 .012 -- -- .020 -- -- (.002) (.002) (.001) (.008) Shock below 66th pctile -- -- -- .015 -- -- .032 -- (director equity) (.002) (.012) Shock above 66th pctile -- -- -- .003 -- -- .007 -- (director equity) (.006) (.018) Shock below 33rd pctile -- -- -- -- 2.000 -- -- 2.013 (other ownership) (.004) (.025) Shock above 33rd pctile -- -- -- -- .017 -- -- .034 (other ownership) (.002) (.011) Sample size 4905 2616 5780 5780 5780 1741 1741 1741 Adjusted R2 .90 .95 .90 .97 .97 .97 .97 .97 a. Data Source: Prowess, Centre for Monitoring Indian Economy, for years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. b. Firms are separated into different"Level in group"based on their within-group level of director equity. For example, "Topmost Firm" are the set of rms that have the highest level of director ownership in their group. c. Also included in each regression are the logarithm of total assets, year xed effects, and rm xed effects. d. Standard errors are in parentheses. with high levels of other equity is large (about .02) and statisti- cally signi cant. These results complement the ndings in Table III: not only are more resources "disappearing" from low-cash- ow right rms, these resources are also the ones more likely to "show up" elsewhere in the group. III.G. Does Money Go to the Top? In Table V we test the nal prediction of tunneling: resources should disproportionately ow toward high-cash- ow-right rms. We rank rms based on their within-group level of director equity and construct four different subsamples: rms with below the sixty-sixth percentile of director equity in their group, rms with above the sixty-sixth percentile of director equity in their group, rms with strictly less than the highest level of director equity in their group, and rms with the highest level of director equity in their group. We compare sensitivity to group shocks and sub- 222 A Reader in International Corporate Finance 140 QUARTERLY JOURNAL OF ECONOMICS group shocks for rms in the four different samples by reestimat- ing equations (3) and (4) separately for these samples. In addition to the variables reported in the table, each regression includes the logarithm of total assets, year xed effects, and rm xed effects. The dependent variable in all regressions is still pro t before depreciation, interest, and taxes. When we contrast rms above and below the sixty-sixth percentile in director equity (columns (1) and (2)), we nd no statistically signi cant differences in their sensitivity to the over- all group shock. In fact, the point estimate on "Group shock" is higher for rms with low levels of director ownership (.013 versus .010).34 In columns (3) to (6), we contrast the sensitivity to the group shock for the rms with the highest level of director own- ership in their group compared with that for all other rms in the group. With this split of the data, the theoretically expected patterns emerge. Firms at the very top gain about .02 rupee for every one-rupee shock to their group (column (6)). All the other rms gain only .012 rupee for the same one-rupee shock (column (3)). Because standard errors are rather large in column (6), however, these two estimates are not statistically different. Interestingly, when we break down the overall group shock into two subshocks, the results become even more suggestive. We nd that top rms gain between .032 and .034 rupee for every one-rupee shock to group rms either below the sixty-sixth per- centile in terms of director equity or above the thirty-third per- centile in terms of other ownership (columns (7) and (8)). All the other rms gain between .015 and .017 rupee on average for the same subshocks (columns (4) and (5)). To summarize, these re- sults give some evidence that the rms with the highest level of director equity in their group seem to bene t most from shocks to the rest of the group. Moreover, these rms bene t the most from shocks to rms with low director equity or higher other share- holders' ownership. III.H. Alternative Explanations Although these ndings match the predictions of the tunnel- ing hypothesis, other possible explanations need to be consid- ered.35 First, suppose that group rms are more diversi ed than 34. Similar results follow if we use median cutoffs. 35. A purely mechanical explanation could be that cross-ownership between rms generate dividend payments that look like tunneling. This effect, however, Chapter Six 223 FERRETING OUT TUNNELING 141 stand-alones and low-cash- ow-right ones are more diversi ed than high-cash- ow-right ones. Then the reduced sensitivity to the industry shock could re ect mismeasurement of these rms' industries. We investigate these questions directly by using de- tailed product data to construct diversi cation measures. For these measures, we nd no difference between group and non- group rms. Nor do we nd any difference between high- and low-cash- ow-right group rms in the extent of their diversi ca- tion. This suggests that differences in industry mismeasurement do not drive our ndings.36 Another possibility is that coinsurance between group rms generates both reduced sensitivity to own shock and redistribu- tion between rms. Such coinsurance may be common in coun- tries such as India, where capital markets are still nascent [Khanna and Palepu 2000]. Insurance may also take a nancing form in which a rich group rm invests in other rms' products, essentially forming a groupwide internal capital market. A sim- ple coinsurance scheme, however, could not generate all of our results. Speci cally, why do high-cash- ow-right rms systemati- cally receive less insurance or nancing? More generally, why does cash ow in only one direction, from low- to high-cash- ow- right rms? For an insurance story to accommodate our ndings, high- cash- ow-right rms within a group would have to be better providers of insurance or nancing. We test this hypothesis in several ways and nd no evidence for it. First, we nd no differ- ence in cash richness (a proxy for ease of insurance provision) between high- and low-cash- ow-right group rms. Second, we nd that adding an interaction of industry cash richness with the various shock measures does not affect the results. Finally, to examine the possibility that these results re ect internal capital markets, we control for the extent of borrowing between rms in a group. This also does not affect the results. As a whole, we nd little support for these alternative explanations. would be too small to explain our results. Moreover, our results do not change when we exclude "earnings from dividends" from our measure of earnings. 36. All the results in this section are described in detail in Bertrand, Mehta, and Mullainathan [2000] as well as in Tables C and D of the unpublished appendix. 224 A Reader in International Corporate Finance 142 QUARTERLY JOURNAL OF ECONOMICS TABLE VI SHOCK SENSITIVITY: AN ACCOUNTING DECOMPOSITION Panel A: Sensitivity to own shock Sample: Groups Stand-alones Dep. variable: Operating pro ts 1.22 1.17 (.018) (.009) Nonoperating pro ts 2.478 2.103 (.014) (.006) Panel B: Sensitivity to own shock by director ownership Sample: Groups Stand-alones Dep. variable: Operating pro ts .0123 .0082 (.0056) (.0013) Nonoperating pro ts .0131 20.0038 (.0043) (.0008) Panel C: Sensitivity to group shock by level of director ownership in group Sample: Topmost rm Below topmost rm Dep. variable: Operating pro ts .0066 .0114 (.0128) (.0026) Nonoperating pro ts .0134 .0006 (.0078) (.0020) a. Data Source: Prowess, Centre for Monitoring Indian Economy, for years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. b. Each coef cient contains the result of a separate regression in which the dependentvariable is either operating pro ts or nonoperating pro ts, as indicated. In Panel A the reported coef cient is the coef cient on "Own shock." In Panel B the reported coef cient is the coef cient on "Own shock z director equity." In Panel C, the reported coef cient is the coef cient on "Group Shock." Also indicated in each regression are the logarithm of total assets, year xed effects, rm xed effects, and "Own shock" (Panels B and C). c. In Panel C the subsamples are for group rms only. Topmost rm and below topmost rms are de ned using director's equity. For example, "Topmost rm" are the set of rms that have the highest level of director ownership in their group. d. Standard errors are in parentheses. IV. OTHER RESULTS IV.A. An Accounting Decomposition If business groups in India are indeed tunneling resources, as the evidence so far strongly suggests, how are they doing it? We address this question in Table VI where we replicate the previous analysis but replace our standard pro ts measure with other Chapter Six 225 FERRETING OUT TUNNELING 143 balance sheet items. More formally, we decompose pro ts into two components. Profits 5 Operating Profits 1 Nonoperating Profits. Operating pro ts are de ned as sales minus total raw material expenses minus energy expenses minus wages and sal- aries.37 Nonoperating pro ts are the "residual." They include such diverse items as write-offs for bad debts, interest income, amortization, extraordinary items, and unspeci ed items. Panel A of Table VI compares the sensitivity of group and stand-alone rms to their own shock for these two measures (as in Table II). Each entry in this panel is the coef cient on "Own shock" from a separate regression. We see in the rst row that group rms' operating pro ts are, if anything, more sensitive to their own industry shock.38 It is on nonoperating pro ts that group rms are far less sensitive to their own shock. More spe- ci cally, nonoperating pro ts seem to fall when there is a positive shock to a rm's industry. Although nonoperating pro ts decline moderately in stand-alone rms, the fall is much larger for group rms. In Panel B we examine the differential sensitivity to own industry shock by the controlling party's cash ow rights (as in Table III). Each entry in this panel belongs to a separate regres- sion. For simplicity, we only report in this table the coef cient on "Own shock p director equity." Each regression also includes the logarithm of total assets, rm xed effects, year xed effects, and the direct effect of "Own shock." As a benchmark, we report in the second column the equivalent regressions for stand-alone rms. The rst row shows that there is little evidence of tunneling in operating pro ts. While group rms' sensitivity rises with direc- tor equity, stand-alone rms show a nearly equivalent rise. The difference is only about .004. In the second row, however, we see a much greater effect on nonoperating pro ts. The difference between group and stand-alone rms is around .017, or four times the difference on operating pro ts. In Panel C we examine how each of the two pro t measures respond to the group shock (as in Table V). Each entry represents the coef cient on "Group shock" from a separate regression which includes year and rm xed effects, the logarithm of total assets, 37. Total raw material expenses include raw material expenses, stores and spares, packaging expenses, and purchase of nished goods for resale. 38. In all regressions in Table VI, the shock measure relates as before to total industry pro ts (operating and nonoperating). So, the shock measures have not changed, only the dependent variables have. 226 A Reader in International Corporate Finance 144 QUARTERLY JOURNAL OF ECONOMICS and own shock. These results complement those of Panels A and B since they tell us about the mechanisms for tunneling money into a rm. We nd a pattern very similar to that in Panels A and B. Much of the differential sensitivity of high- and low-cash- ow- right rms to the group shock occurs on nonoperating pro ts. Hence, according to the ndings in Table VI, the tunneling of money both into and out of rms in India occurs through nonop- erating pro ts.39 This implies that transfer pricing (which would affect operating pro ts) is not an important source of tunneling in India. Moreover, it suggests that nonoperating pro ts may be a force that moves in the opposite direction of operating pro ts and serves to dampen nal earnings. In unreported regressions, we examine this by simply regressing a rm's nonoperating pro ts on its operating pro ts, while controlling for size, year dummies, and rm xed effects. As expected, we nd a strong negative coef cient. When we interact operating pro ts in this regression with a variety of variables, we nd results quite similar to our tunneling ndings. Group rms show a much more negative relationship between operating and nonoperating pro ts. Also, among group rms, the ones with low cash ow rights show the most negative relationship. This evidence reinforces the view that manipulation of nonoperating pro ts is a primary means of re- moving cash from and placing cash into group rms in India. IV.B. Market Valuation Given our ndings so far, it is natural to ask whether stock prices re ect the extent of this tunneling. Does the market pe- nalize rms or groups which show more evidence of tunneling? To address this issue, we compute for each rms an average "q" ratio. We do this by rst regressing standard rm level market- to-book ratios on log(total assets), year xed effects, industry xed effects, and rm xed effects. The value of the rm xed effect in this regression is the variable we call "Firm Q." Our q measure is, therefore,the market premium for the rm relative to other rms in its industry, size class, and year. We also compute an average q ratio for each group. To do this, we estimate a similar regression at the rm level but include group xed effects instead of rm xed effects. The group xed effects from these 39. We have attempted further decomposition of nonoperating pro ts and found no consistent pattern. No one subcomponent of nonoperating pro ts is systematically more important. This may be because different rms tunnel in different ways. Chapter Six 227 FERRETING OUT TUNNELING 145 TABLE VII SENSITIVITY TO OWN AND GROUP SHOCK BY FIRM AND GROUP Q RATIOS DEPENDENT VARIABLE: PROFIT BEFORE DIT (1) (2) (3) (4) Own shock 2.046 .388 .600 .049 (.056) (.027) (.017) (.060) Own shock p rm Q .178 -- -- .143 (.013) (.016) Own shock p relative Q -- .143 -- -- (.011) Own shock p group Q -- -- .414 .171 (.037) (.044) Group shock 2.008 .010 .011 2.008 (.003) (.002) (.003) (.004) Group shock p rm Q .012 -- -- .012 (.001) (.001) Group shock p relative Q -- .008 -- -- (.001) Group shock p group Q -- -- .006 2.001 (.007) (.006) Adjusted R2 .94 .94 .93 .94 a. a. Data Source: Prowess, Centre for Monitoring Indian Economy, for years 1989­1999. All monetary variables are expressed in 1995 Rs. crore, where crore represents 10 million. b. Sample is group rms only. c. "Firm Q" is a variable that represents the estimated rm xed effects in a regression of rm-level q ratios (market valuation over total assets) on log(total assets), year xed effects, industry xed effects, and rm xed effects. "Group Q" is a variable that represents the estimated group xed effects in a regression of rm-level q ratios on log(total assets), year xed effects, industry xed effects, and group xed effects. "Relative Q" is the difference between "Firm Q" and the mean of "Firm Q" within groups. d. Also included in each regression are the logarithm of total assets, year xed effects, and rm xed effects. e. Standard errors are in parentheses. regressions de ne the variable we call "Group Q." Finally, we form a "Relative Q" measure for each rm, which equals its own q minus its group q, and captures a rm's performance relative to the rest of the group. In Table VII we examine how these new variables in uence the sensitivity of a rm to its own shock and to the group shock. In column (1) we show that rms with higher q are more sensitive to both their own shock and to the group shock. Under the tunneling interpretation, this suggests that rms that have more money transferred to them and less money taken away from them have higher q ratios. In column (3) we see the same pattern for relative q. In column (3) we see that the groups with the highest q ratios are those with rms that show higher sensitivity to their own shock, and thus have less money taken away from them. The 228 A Reader in International Corporate Finance 146 QUARTERLY JOURNAL OF ECONOMICS coef cient on group shock interacted with "Group Q" is positive but insigni cant. In column (4) we include interactions of the shock measures with both "Firm Q" and "Group Q." The results are qualitatively similar. The ndings in this section suggest that the stock market (at least partly) recognizes tunneling and incorporates it into pricing. Firms that have more resources tunneled to them are valued more by the market. Firms that have less money tunneled away from them are also valued more. Finally, groups that tunnel less money are valued more. These results complement previous em- pirical ndings that market valuations positively correlate with the controlling shareholders' cash ow rights.40 V. CONCLUSION We have developed a fairly general empirical methodology for quantifying tunneling in business groups. We examined whether shocks propagate between rms in a business group in accord with the controlling shareholder's ownership in each rm. We applied the methodology in Indian data and found signi cant amounts of tunneling, mostly via nonoperating components of pro ts. We also found that market prices partly incorporate tunneling. These results raise some questions. If groups expropriate minority shareholders so much, how do they persist? Why do minority shareholders buy into them in the rst place? We feel that there are three broad possibilities. First, groups may grow through acquisitions. If this is the case, and markets are ef cient, then the act of takeover would generate a one-time drop in share price amounting to the extent of tunneling. Second, shareholders may not recognize the extent of tunneling that takes place in groups. For example, the lack of detailed ownership information may make it dif cult for shareholders to gure out with great reliability which group rms are high- and which are low-cash- ow-right rms. Finally, groups may provide other bene ts, which offset the costs imposed by tunneling. To cite one example, they may provide important political contacts, which are quite 40. For example, Bianchi, Bianco, and Enriques [1999], Claessens, Djankov, Fan, and Lang [1999], and Claessens, Djankov, and Lang [2000]). In the Indian data we nd that rms with a higher level of other equity within a group have a lower q ratio. We do not, however, nd a signi cant relationship between level of director ownership and q ratio within groups. Chapter Six 229 FERRETING OUT TUNNELING 147 valuable in a heavily regulated economy. Given the extent of tunneling found here, assessing the relevance of each of these possibilities appears to be an important direction for future research. UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS, NATIONAL BUREAU OF ECONOMIC RESEARCH, AND CENTRE FOR ECONOMIC AND POLICY RESEARCH MASSACHUSETTS INSTITUTE OF TECHNOLOGY MASSACHUSETTS INSTITUTE OF TECHNOLOGY AND NATIONAL BUREAU OF ECONOMIC RESEARCH REFERENCES Bebchuk, Lucian, R. Kraakman, and G. Triantis, "Stock Pyramids, Cross-Owner- ship, and Dual Class Equity: The Mechanisms and Agency Costs of Separat- ing Control from Cash Flow Rights," in Concentrated Corporate Ownership, A National Bureau of Economic Research Conference Report. R. Morck, editor (Chicago, IL: University of Chicago, 2000). Berle, A., and G. Means, The Modern Corporation and Private Property (New York, NY: Macmillan, 1934). Bertrand, Marianne, and Sendhil Mullainathan, "Agents with and without Prin- cipals," American Economic Review Papers and Proceedings, XC (2000), 203­208. Bertrand, Marianne, and Sendhil Mullainathan, "Are CEOs Rewarded for Luck? The Ones without Principals Are," Quarterly Journal of Economics, CXVI (2001), 901­932. Bertrand, Marianne, Paras Mehta, and Sendhil Mullainathan, "Ferreting out Tunneling: An Application to Indian Business Groups," National Bureau of Economic Research Working Paper No. 7952, 2000. Bianchi, M., M. Bianco, and L. Enriques, "Pyramidal Groups and the Separation between Ownership and Control in Italy," mimeo, Bank of Italy, 1999. Blanchard, Olivier, Florencio Lopez-de-Silanes, and Andrei Shleifer, "What Do Firms Do with Cash Windfalls," Journal of Financial Economics, XXXVI (1994), 337­360. Claessens, Stijn, Simeon Djankov, and Harry Lang, "The Separation of Owner- ship and Control in East Asian Countries," mimeo, World Bank, 2000. Claessens,Stijn, Simeon Djankov, Joseph Fan, and Harry Lang, "Expropriation of Minority Shareholders: Evidence from East Asia," Policy Research Paper No. 2088, World Bank, 1999. Dutta, Sudipt, Family Business in India (New Delhi: Response Books, Sage Publications, 1997). Hoshi, Takeo, Anil Kashyap, and David Scharfstein, "Corporate Structure, Li- quidity, and Investment: Evidence from Japanese Industrial Groups," Quar- terly Journal of Economics, CVI (1991), 33­60. Jensen, Michael, and William Meckling, "Theory of the Firm: Managerial Behav- ior, Agency Costs, and Ownership Structure," Journal of Financial Econom- ics, III (1976), 305­360. Johnson, Simon, P. Boone, A. Breach, and E. Friedman, "Corporate Governance in the Asian Financial Crisis," Journal of Financial Economics, LVIII (2000), 141­186. Johnson, Simon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, "Tunneling," American Economic Review Papers and Proceedings, XC (2000), 22­27. Johnson, Simon, and Eric Friedman, "Tunneling and Propping," MIT Working Paper, 2000. Khanna, Tarun, and Krishna Palepu, "Is Group Membership Pro table in Emerg- ing Markets? An Analysis of Diversi ed Indian Business Groups," Journal of Finance, LV (2000), 867­891. 230 A Reader in International Corporate Finance 148 QUARTERLY JOURNAL OF ECONOMICS Lamont, Owen, "Capital Flows and Investment: Evidence from the Internal Capi- tal Markets," Journal of Finance, LII (1997), 83­109. La Porta, Rafael, Florencio Lopez-de-Silanes,Andrei Shleifer, and Robert Vishny, "Corporate Ownership around the World," Journal of Finance, LIV (1999), 471­517. Nenova, Tatiana, "The Value of a Corporate Vote and Private Bene ts: Cross- Country Analysis," mimeo, Department of Economics, Harvard University, 1999. Piramal, G. Business Maharajas (Bombay: Viking Press, 1996). Price, Waterhouse & Co., Doing Business in India (New York: Price Waterhouse, 1999). Sarkar, Jayati, and Subrata Sarkar, "The Governance of Indian Corporates," India Development Report, 1999­2000, Kirit S. Parikh, editor (New Delhi: Oxford University Press, 1999). Shleifer, Andrei, and Daniel Wolfenzon, "Investor Protection and Equity Mar- kets," National Bureau of Economic Research Working Paper No. 7974, 2000. Wolfenzon, Daniel, "A Theory of Pyramidal Ownership," mimeo, Department of Economics, Harvard University, 1999. Zingales, Luigi, "What Determines the Value of Corporate Votes," Quarterly Journal of Economics, CX (1995), 1047­1073. Chapter Seven 231 Journal of Financial Economics 74 (2004) 277­304 Cross-country determinants of mergers and acquisitions$ Stefano Rossi, Paolo F. Volpin* London Business School, Regent's Park, London NW1 4SA, UK Received 7 August 2002; accepted 6 October 2003 Available online 13 May 2004 Abstract We study the determinants of mergers and acquisitions around the world by focusing on differences in laws and regulation across countries. We find that the volume of M&A activity is significantly larger in countries with better accounting standards and stronger shareholder protection. The probability of an all-cash bid decreases with the level of shareholder protection in the acquirer country. In cross-border deals, targets are typically from countries with poorer investor protection than their acquirers' countries, suggesting that cross-border transactions play a governance role by improving the degree of investor protection within target firms. r 2004 Elsevier B.V. All rights reserved. JEL classification: G28; G32; G34 Keywords: Mergers and acquisitions; Corporate governance; Investor protection $ We thank Richard Brealey, Ian Cooper, Antoine Faure-Grimaud, Julian Franks, Denis Gromb, Ernst Maug, Thomas Noe, Antoinette Schoar, Henri Servaes, Oren Sussman, David Webb, an anonymous referee, and participants at the 2004 AFA meetings in San Diego, at the 2003 EFA meetings in Glasgow and at seminars at Humboldt University, London Business School, London School of Economics, Norwegian School of Economics and Business, Norwegian School of Management, and Tilburg University. Paolo F. Volpin acknowledges support from the JP Morgan Chase Research Fellowship at London Business School. *Corresponding author. Tel.: +44-20-72625050; fax: +44-20-77243317. E-mail address: pvolpin@london.edu (P.F. Volpin). 0304-405X/$ -see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jfineco.2003.10.001 232 A Reader in International Corporate Finance 278 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 1. Introduction In a perfect world, corporate assets would be channelled toward their best possible use. Mergers and acquisitions (M&A) help this process by reallocating control over companies. However, frictions such as transaction costs, information asymmetries, and agency conflicts can prevent efficient transfers of control. Recent studies on corporate governance employ measures of the quality of the legal and regulatory environment within a country as proxies for some of these frictions, and show that differences in laws, regulation, and enforcement correlate with the development of capital markets, the ownership structure of firms, and the cost of capital (see, e.g., La Porta et al., 1997, 1998; Bhattacharya and Daouk, 2002). In this paper we analyze a sample of mergers and acquisitions announced in the 1990s and completed by the end of 2002. Our sample comprises firms in 49 major countries and shows that differences in laws and enforcement explain the intensity and the pattern of mergers and acquisitions around the world. The volume of M&A activity is significantly larger in countries with better accounting standards and stronger shareholder protection. This result holds for several measures of M&A activity, and also when we control for other characteristics of the regulatory environment such as antitrust legislation and takeover laws. Our findings indicate that a more active market for mergers and acquisitions is the outcome of a corporate governance regime with stronger investor protection. We also show that hostile deals are relatively more likely in countries with better shareholder protection. One explanation is that good protection for minority shareholders makes control more contestable by reducing the private benefits of control. Next, we provide evidence on cross-border mergers and acquisitions. We show that the probability that a given deal is cross-border rather than domestic decreases with the investor protection of the target's country. Even after we control for bilateral trade, relative GNP per capita, and cultural and geographical differences, we find that targets are typically from countries with poorer investor protection compared to their acquirers. This result suggests that cross-border M&A activity is an important channel for effective worldwide convergence in corporate governance standards, as argued by Coffee (1999). Selling to a foreign firm is a form of contractual convergence similar to the decision to list in countries with better corporate governance and better-developed capital markets. Pagano et al. (2002) and Reese and Weisbach (2002) show that firms from countries with weak legal protection for minority shareholders list abroad more frequently than do firms from other countries. We show that firms in countries with weaker investor protection are often sold to buyers from countries with stronger investor protection. We also analyze the determinants of the takeover premium and the method of payment in individual transactions. We show that the premium is higher in countries with higher shareholder protection, although this result is driven by deals with US and British targets. We find that the probability of an all-cash bid decreases with the degree of shareholder protection in the acquirer country, indicating that acquisitions paid with stock require an environment with high shareholder protection. Chapter Seven 233 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 279 Our paper belongs to the growing literature exploring cross-country variation in governance structures around the world. Recent studies show that better legal protection of minority shareholders is associated with more developed stock markets (La Porta et al., 1997), higher valuation (La Porta et al., 2002), greater dividend payouts (La Porta et al., 2000b), lower concentration of ownership and control (La Porta et al., 1999), lower private benefits of control (Dyck and Zingales, 2004; Nenova, 2003), lower earnings management (Leuz et al., 2003), lower cash balances (Dittmar et al., 2003), and higher correlation between investment opportunities and actual investments (Wurgler, 2000). Our paper shows that better investor protection is correlated with a more active market for mergers and acquisitions. We structure the paper as follows. Section 2 describes the data. Section 3 contains the analyses of the determinants of M&A activity. Section 4 discusses the main results. Section 5 concludes. 2. Data Our sample contains all mergers and acquisitions announced between January 1, 1990 and December 31, 1999, completed as of December 31, 2002, and reported by SDC Platinum, a database from Thomson Financial. Because we wish to study transactions clearly motivated by changes in control, we focus on mergers (business combinations in which the number of companies decreases after the transaction) and acquisitions of majority interests (when the acquirer owns less than 50% of the target company's stock before the deal, and more than 50% after the deal). A second reason for this sample selection is that the coverage of transfers of minority stakes (below 50%) is likely to be severely affected by cross-country differences in disclosure requirements. By selecting only transfers of stakes above 50%, we minimize these disclosure biases. However, in interpreting the results, we note that the availability and quality of the data might be better in some countries (such as the US and UK) because of broader SDC coverage. A related concern is that the coverage of small countries improves over time. To address this concern, we replicate our analysis on the subsample of deals announced in the second half of the 1990s and find similar results. The availability of empirical measures of investor protection limits our set to 49 countries. The sample from SDC includes 45,686 deals, 22% of which have a traded company as the target. Excluded deals represent about 6% of the original dataset in number and 1% in value. The appendix describes the variables we use in this paper and indicates their sources. These variables can be classified into three broad categories corresponding to three different levels of analysis. The first set of variables is at the country level. It includes measures of M&A activity from the target's perspective, as well as broad macroeconomic conditions and proxies for the legal and regulatory environment. We use these variables in our cross-country analysis of the determinants of international mergers and acquisitions. Our second category of variables measures the flow of M&A activity and cultural differences and similarities between any ordered pairs of 234 A Reader in International Corporate Finance 280 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 acquirer and target countries (there are 49 48 or 2,352 ordered pairs). The third set of variables is at the individual deal level and includes data on the premium paid, the value of the deal, and the means of payment. We use these data, together with the country-level variables defined above, in our analysis of the determinants of the premium and the means of payment. 2.1. M&A activity Tables 1 and 2 show the data on M&A activity sorted by target country. We define volume as the percentage of traded firms that are targets of successful mergers or acquisitions. We interpret this variable as a measure of the ability of an economy to reallocate control over corporate assets. We also use other measures of volume, such as the total number of completed deals divided by population, the value of all completed deals divided by GDP, and the value of completed deals among traded companies divided by stock market capitalization. The qualitative results do not change. As is apparent from Table 1, the market for corporate control plays a different role in different countries. For example, volume is very low in Japan (only 6.4% of Japanese traded companies are targets of a completed deal during the 1990s) and very high in the US (65.6% of US traded companies are targets in a completed deal). The table also shows some similarities across countries. For example, volume in France, Italy, and the United Kingdom is similar, although their governance regimes are quite different. Of all mergers and acquisitions, we focus on hostile deals, since they are likely to play an important governance role. We examine the number of attempted hostile takeovers as a percentage of the total number of traded companies. The intuition is that the disciplinary role of hostile takeovers is related to the threat they represent to incumbent managers. In other words, it is likely that attempted (but failed) hostile takeovers play just as important a role in disciplining management as hostile takeovers that are eventually completed. In all countries, the frequency of hostile takeovers is very small. According to SDC, they are absent in 21 out of 49 countries, and when present they never exceed the 6.44% observed in the United States. Therefore, according to SDC Platinum, hostile takeovers are rare. However, this conclusion could be unwarranted, because our source might fail to record all unsuccessful takeovers. Moreover, in some countries the corporate governance role of hostile takeovers could be performed by hostile stakes, as Jenkinson and Ljungqvist (2001) show for Germany. We define the cross-border ratio as the percentage of completed deals in which the acquirer is from a different country than the target. In the case of mergers, we follow our data source to distinguish acquirers from targets. For example, in the merger between Daimler and Chrysler, Thomson codifies Daimler as the acquirer and Chrysler as the target. The number of cross-border mergers and acquisitions is 11,638, corresponding to 25% of the total. Table 1 shows that different countries play different roles in the cross-border M&A market. For instance, 51% of the acquirers in Mexican deals are foreign, compared to only 9.1% in the United States. Chapter Seven 235 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 281 Table 1 Data on international mergers and acquisitions sorted by target country Volume is the percentage of traded companies targeted in a completed deal. Hostile takeover is the number of attempted hostile takeovers as a percentage of domestic traded firms. Cross-border ratio is the number of cross-border deals as a percentage of all completed deals. Country Volume (%) Hostile takeover (%) Cross-border ratio (%) Argentina 26.80 0.65 53.73 Australia 34.09 4.60 27.16 Austria 38.14 1.03 51.55 Belgium 33.33 0.56 45.14 Brazil 23.08 0.00 52.03 Canada 30.05 2.73 22.66 Chile 10.57 0.42 64.79 Colombia 19.42 0.00 66.67 Denmark 24.03 0.81 38.26 Ecuador 10.53 0.00 68.97 Egypt 1.46 0.00 47.62 Finland 45.45 0.91 22.67 France 56.40 1.68 33.81 Germany 35.51 0.30 26.05 Greece 12.66 0.00 23.13 Hong Kong 33.91 0.41 38.52 India 2.01 0.02 56.02 Indonesia 10.60 0.48 61.03 Ireland 28.90 4.62 52.73 Israel 9.43 0.23 46.94 Italy 56.40 3.04 36.13 Japan 6.43 0.00 13.25 Jordan 0.00 0.00 55.56 Kenya 1.80 0.00 28.57 Malaysia 15.23 0.19 11.27 Mexico 27.51 0.00 51.02 Netherlands 26.49 1.32 43.43 New Zealand 49.82 0.70 46.15 Nigeria 0.61 0.00 58.33 Norway 61.24 5.86 36.76 Pakistan 0.48 0.00 55.56 Peru 12.21 0.00 56.88 Philippines 21.41 0.00 37.97 Portugal 31.37 1.96 40.00 Singapore 34.06 0.40 31.41 South Africa 23.89 0.45 24.65 South Korea 4.81 0.00 53.85 Spain 15.72 0.17 37.55 Sri Lanka 4.83 0.00 42.86 Sweden 62.06 3.74 35.48 Switzerland 38.48 1.43 43.59 Taiwan 0.89 0.00 49.37 Thailand 17.14 0.00 43.24 Turkey 6.12 0.00 45.45 United Kingdom 53.65 4.39 23.46 United States 65.63 6.44 9.07 Uruguay 7.55 0.00 85.00 Venezuela 14.91 0.00 56.60 Zimbabwe 6.35 0.00 46.15 World average 23.54 1.01 42.82 236 A Reader in International Corporate Finance 282 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Table 2 Summary statistics on the sample of individual deals sorted by target country Premium is the bid price as a percentage of the closing price of the target four weeks before the announcement. All-cash bid is a dummy variable that equals one if the acquisition is entirely paid in cash, and zero otherwise. Country Premium All-cash bid N obs. Mean Std. dev. Mean Std. dev. Australia 129.5 37.4 0.60 0.49 212 Austria 129.8 25.2 0.83 0.41 6 Belgium 137.2 56.1 0.86 0.38 7 Brazil 110.5 0.0 0.00 0.00 1 Canada 132.9 40.1 0.36 0.48 157 Chile 149.9 24.5 1.00 0.00 3 Denmark 142.2 41.2 0.83 0.41 6 Finland 149.7 53.2 1.00 0.00 7 France 133.4 53.6 0.88 0.32 112 Germany 116.7 35.3 0.77 0.44 13 Greece 165.5 112.8 0.67 0.58 3 Hong Kong 129.8 56.1 0.93 0.25 46 India 178.6 113.2 0.67 0.50 9 Indonesia 222.5 150.1 1.00 0.00 2 Ireland 121.1 22.7 0.78 0.44 9 Israel 220.2 153.2 0.50 0.71 2 Italy 127.7 26.8 0.88 0.33 26 Japan 99.0 41.7 0.36 0.48 73 Malaysia 151.7 76.8 0.91 0.29 23 Mexico 124.5 17.0 1.00 0.00 2 Netherlands 144.7 37.9 0.50 0.52 16 New Zealand 129.2 17.6 0.94 0.25 16 Norway 136.0 37.6 0.76 0.43 37 Philippines 157.7 81.0 0.56 0.53 9 Portugal 149.9 57.1 1.00 0.00 4 Singapore 152.9 79.3 0.85 0.37 39 South Africa 129.5 63.2 0.68 0.48 28 South Korea 145.1 102.7 0.50 0.58 4 Spain 119.8 30.0 0.70 0.48 10 Sweden 141.7 40.6 0.71 0.46 45 Switzerland 111.0 33.3 0.89 0.33 9 Thailand 126.0 79.3 0.92 0.28 13 Turkey 127.5 0.0 1.00 0.00 1 United Kingdom 145.8 41.9 0.64 0.48 614 United States 144.3 42.4 0.37 0.48 2443 Total 141.6 44.7 0.48 0.50 4007 To study the cross-country variations in the premiums and means of payment, we use transaction-level data. The premium is the bid price as a percentage of the closing price four weeks before the announcement. We characterize the means of payment of an individual deal with a dummy variable that equals one if the acquisition is entirely Chapter Seven 237 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 283 paid in cash, and zero otherwise. We compute these variables using data available from SDC Platinum. After excluding deals with incomplete information, we have 4,007 observations from 35 countries. As shown in Table 2, the data are highly concentrated: the target is a US firm in 60% of the sample and a UK firm in 15% of the sample. The bid price ranges from 99.6% of the pre-announcement price (in Japan) to 227.1% (in Indonesia). In Italy, 88% of the acquisitions of Italian targets are paid entirely in cash. In the US, only 37% of the deals are paid wholly in cash. 2.2. Investor protection By reshuffling control over companies, mergers and acquisitions help allocate corporate assets to their best possible use. Investor protection can affect the volume of mergers and acquisitions because it affects the magnitude of frictions and inefficiencies in the target country. As proxies for investor protection, we use several indexes developed by La Porta et al. (1998): an index of the quality of the accounting standards, an index of shareholder protection that combines an index of the quality of law enforcement (rule of law) and an index of the rights that shareholders have with respect to management (antidirector rights), and a dummy variable for common-law countries. These indexes are highly correlated (their pair-wise correlations range between 40% and 60%) because they all reflect to some degree the underlying quality of investor protection in a country. However, they measure different institutional characteristics. Accounting standards measure the quality of the disclosure of accounting information. The accounting standards quality index is created by the Center for International Financial Analysis and Research and rates the 1990 annual reports of at least three firms in every country on their inclusion or omission of 90 items. Thus, each country obtains a score out of 90, with a higher number indicating more disclosure. This variable affects M&A activity because good disclosure is a necessary condition for identifying potential targets. Accounting standards also reflect corporate governance, because they reduce the scope for expropriation by making corporate accounts more transparent. Our second measure is an index of shareholder protection that ranges between zero and six. It captures the effective rights of minority shareholders with respect to managers and directors and is defined as an antidirector rights index multiplied by a rule of law index and divided by ten. When minority shareholders have fewer rights, they are more likely to be expropriated. As a consequence, the stock market is less developed, and raising external equity, particularly to finance a takeover, is more expensive. At the same time, with low shareholder protection, the private benefits of control are high and the market for corporate control is relatively less effective, because incumbents will try to entrench themselves via ownership concentration and takeover deterrence measures (Bebchuk, 1999). The common law measure is a dummy variable that equals one if the origin of the company law is the English common law, and zero otherwise. La Porta et al. (1998) argue that legal origin is a broad indicator of investor protection and show that 238 A Reader in International Corporate Finance 284 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 countries with common law as the legal origin better protect minority shareholders than do countries with civil law as the legal origin. Although common law should not directly affect mergers and acquisitions, we include this variable because it is correlated with other proxies of investor protection and is truly exogenous. Hence, it is a good instrument for investor protection. We note that the number of observations in our empirical analysis varies with the measure of investor protection used, because accounting standards are not available for Ecuador, Indonesia, Ireland, Jordan, Kenya, Pakistan, Sri Lanka, and Zimbabwe. 3. Determinants of M&A activity We examine five dimensions of mergers and acquisitions: the volume, the incidence of hostile takeovers, the pattern of cross-border deals, the premium, and the method of payment. 3.1. Volume We start with the relation between the volume of M&A activity and investor protection at the target-country level. Our specification is Volume ¼ a þ bX þ g investor protectionþe; ð1Þ where the dependent variable, volume, is the percentage of traded firms that are targets of successful mergers or acquisitions. The variables for common law, accounting standards, and shareholder protection are proxies for investor protection. Control factors (X) in all specifications are GDP growth, which proxies for the change in economic conditions, and the logarithm of the 1995 per capita GNP, which proxies for the country's wealth. Table 3 reports the coefficients of six Tobit models derived from specification (1). We estimate Tobit models because the dependent variable (volume) is bounded between zero and 100 by construction. Column 1 shows that the frequency of mergers among traded companies is 7.5% higher in common-law countries than in civil-law countries. The results in Column 2 show that accounting standards are positively and significant correlated with volume. A 12-point increase in the accounting standards measure (from the quality of accounting standards in Italy to that in Canada) correlates with a 5% increase in the volume of mergers and acquisitions. Column 3 finds a similar result for shareholder protection. A one-point increase in shareholder protection (for instance, the adoption of voting by mail in a country like Belgium) is associated with 4% more volume. Thus, we find that there are more mergers and acquisitions in countries with better investor protection. We note that a one-point increase in the index of antidirector rights (such as the adoption of voting by mail) translates into a one-point increase in shareholder protection only in a country like Belgium, which also scores ten in the index of rule of law. In a country like Italy, which scores 8.33 in the index of rule of law, the same Chapter Seven 239 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 285 Table 3 Determinants of the volume across countries The table presents the results of six Tobit models estimated by maximum likelihood for the sample of 49 target countries. The dependent variable is volume, the percentage of traded companies targeted in a completed deal. The independent variables are: common law, a dummy variable that equals one if the origin of the company law is the English common law, and zero otherwise; accounting standards, an index of the quality of accounting disclosure; shareholder protection, a measure of the effective rights of minority shareholders; ownership concentration, the average equity stake owned by the three largest shareholders in the ten largest nonfinancial domestic firms in 1994; mandatory bid rule, a dummy variable that equals one if acquirers are forced to make a tender offer to all shareholders when passing a given ownership threshold, and zero otherwise; market return, the average annual stock market return in the 1990s; and market dominance, a survey-based measure of product market concentration. The logarithm of GNP per capita and GDP growth are included in all regressions as control variables. Standard errors are shown in parentheses. (1) (2) (3) (4) (5) (6) Log (GNP per capita) 9.00ÃÃÃ 5.61ÃÃÃ 6.40ÃÃÃ 4.49ÃÃ 4.75ÃÃ 8.81ÃÃÃ (1.24) (1.94)` (1.48) (2.04) (2.02) (2.05) GDP growth 2.42 2.57Ã 2.42ÃÃ 3.05ÃÃ 3.11ÃÃ 2.33 (1.12) (1.12) (1.07) (1.32) (1.36) (1.48) Common law 7.52Ã 9.06Ã (3.97) (5.06) Accounting standards 0.47ÃÃ 0.35Ã 0.43ÃÃ (0.18) (0.20) (0.20) Shareholder protection 4.27ÃÃÃ 2.96 4.65ÃÃ (1.69) (2.01) (2.32) Ownership concentration 0.38Ã (0.20) Mandatory bid rule 0.58 (4.10) Market return 0.21 (0.15) Market dominance 3.40 (3.57) Constant 48.1ÃÃÃ 43.1ÃÃÃ 31.8ÃÃÃ 30.8Ã 58.4ÃÃÃ 38.3ÃÃ (12.0) (16.5) (12.5) (18.1) (22.1) (17.7) Pseudo R2 0.10 0.08 0.10 0.09 0.09 0.09 N observations 49 41 49 41 39 41 ÃÃÃ ÃÃ Ã , , indicate significance at 1% percent, 5%, and 10% levels, respectively. change in minority shareholders' rights implies only a 0.833-point increase in shareholder protection. In Column 4, we estimate a joint regression with accounting standards and shareholder protection and find that only the former is statistically significant. This result suggests that disclosure rules are more relevant for takeovers than are shareholder rights. In Column 5, we add ownership concentration, which is potentially an important explanatory variable. Ownership concentration in a country is the average equity stake owned by the three largest shareholders in the ten largest nonfinancial domestic firms in 1994, from La Porta et al. (1998). We find 240 A Reader in International Corporate Finance 286 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 that, as in the individual regressions, the coefficients on accounting standards and shareholder protection are positive and significant. The coefficient on ownership concentration is also positive and significant. This finding indicates that, when we control for investor protection, countries with more concentrated ownership have more mergers and acquisitions. This result is consistent with Shleifer and Vishny (1986), who argue that transfers of control are easier in companies with more concentrated ownership structure because they overcome the free-rider problem in takeovers. The results in Column 5 help explain why shareholder protection is not significant in Column 4. On the one hand, shareholder protection reduces the costs of raising external equity, thereby increasing the volume of mergers. On the other hand, it decreases ownership concentration, which makes friendly transfers of control less likely. By controlling for ownership concentration, we are able to disentangle the two effects. In Column 6, we evaluate the robustness of the results on investor protection by adding further control variables to capture cross-country differences in the regulatory environment. We show the results only with the common law variable as our proxy for investor protection, although we obtain similar results for accounting standards and shareholder protection. A mandatory bid rule, which we capture with a dummy variable that equals one if acquirers are forced to make a tender offer to all shareholders when passing a given ownership threshold and zero otherwise, might reduce the volume of mergers and acquisitions because it imposes further costs on the potential bidder. The market return, calculated as the average annual stock market return during the 1990s, might affect M&A activity because of valuation waves (Shleifer and Vishny, 2003). However, there are two opposing effects when the stock market is booming. Targets could become too expensive, reducing the volume of deals, but acquirers enjoy low takeover costs because they can pay with more highly valued stock, leading to a high takeover volume. Market dominance, a measure of product market concentration in 1995 from the 1992 Global Competitiveness Report (published by the World Economic Forum), could reduce the volume because of lower availability of targets. The results in Column 6 show that common law is still significant and its coefficient is virtually unchanged from Column 1. None of the control variables are statistically significant. Note that the number of observations decreases from 49 to 41 because market return is not available for Taiwan and Uruguay and market dominance is not available for Ecuador, Kenya, Nigeria, Pakistan, Sri Lanka, Uruguay, and Zimbabwe. 3.2. Hostile takeovers Many financial economists argue that hostile takeovers play an important governance role (for instance, see Manne, 1965; Jensen, 1993; and Franks and Mayer, 1996). To analyze cross-country differences in the frequency of hostile takeovers, we estimate Hostile takeover ¼ a þ bX þ g investor protectionþe; ð2Þ Chapter Seven 241 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 287 Table 4 Incidence of hostile takeovers The table presents the results of six Tobit models estimated by maximum likelihood on the sample of 49 target countries. The dependent variable is hostile takeover, or attempted hostile takeovers as a percentage of traded firms. The independent variables are: common law, a dummy variable that equals one if the origin of the company law is the English common law, and zero otherwise; accounting standards, an index of the quality of accounting disclosure; shareholder protection, a measure of the effective rights of minority shareholders; ownership concentration, the average equity stake owned by the three largest shareholders in the ten largest nonfinancial domestic firms in 1994; cross-border regulation, a dummy variable that equals one if foreign buyers need government approval, and zero otherwise; market return, the average annual stock market return in the 1990s; and mandatory bid rule, a dummy variable that equals one if acquirers are forced to make a tender offer to all shareholders when passing a given ownership threshold, and zero otherwise. The logarithm of GNP per capita and GDP growth are included in all regressions as control variables. Standard errors are shown in parentheses. (1) (2) (3) (4) (5) (6) Log (GNP per capita) 1.30ÃÃÃ 0.93ÃÃ 0.75ÃÃÃ 0.61Ã 0.64ÃÃ 1.08ÃÃ (0.26) (0.35) (0.27) (0.32) (0.32) (0.26) GDP growth 0.08 0.04 0.06 0.10 0.05ÃÃ 0.09 (0.19) (0.21) (0.17) (0.18) (0.19) (0.19) Common law 1.53ÃÃ 1.57ÃÃ (0.68) (0.70) Accounting standards 0.07ÃÃ 0.02 0.02 (0.03) (0.03) (0.03) Shareholder protection 0.88ÃÃÃ 0.84ÃÃ 0.73ÃÃ (0.25) (0.26) (0.31) Ownership concentration 0.01 (0.03) Cross-border regulation 1.80Ã (0.93) Market return 0.02 (0.02) Mandatory bid rule 0.04 (0.59) Constant 12.0ÃÃÃ 12.2ÃÃÃ 8.34ÃÃÃ 7.93ÃÃ 7.06Ã 9.75ÃÃÃ (2.63) (3.32) (2.53) (3.09) (3.61) (2.62) Pseudo R2 0.20 0.17 0.24 0.23 0.22 0.23 N observations 49 41 49 41 39 47 ÃÃÃ ÃÃ Ã , , indicate significance at 1%, 5%, and 10% levels, respectively. where the hostile takeover variable is the number of attempted hostile takeovers in the 1990s as a percentage of the number of domestic traded companies. Common law, accounting standards, shareholder protection, and ownership concentration are proxies for investor protection, as described in Section 2.2. We include GDP growth and the logarithm of GNP per capita as control factors in all specifications. The results are presented in Table 4. The first three columns show that common law, accounting standards, and shareholder protection are positively and signifi- cantly correlated with hostile takeovers. To interpret these results, note that hostile takeovers require that control be contestable, a feature that is less common in countries with poorer investor protection. 242 A Reader in International Corporate Finance 288 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Column 4 shows that shareholder protection dominates accounting standards. A one-point increase in shareholder protection (e.g., the introduction of voting by mail in Belgium) is associated with 0.8 percentage points more hostile takeovers. Shareholder protection makes control more contestable by reducing the private benefits of control. In Column 5, we add ownership concentration as a control variable. This variable is not significant. It marginally reduces the coefficient on shareholder protection without affecting its statistical significance. This result compares with Table 3, in which ownership concentration is positive and significant. According to Shleifer and Vishny (1986), ownership concentration facilitates only friendly transfers of control, not hostile takeovers. Hence, the insignificant coefficient in Column 5 of Table 4 is not surprising. To evaluate the robustness of the main result that hostile takeovers are more common in countries with better investor protection, in Column 6 we add some control variables to the specification in Column 1 to capture cross-country differences in the regulatory environment. As in Table 4, we control for mandatory bid rules and market returns. We also incorporate cross-border regulation with a dummy variable that equals one if a foreign buyer needs government approval before acquiring control of a domestic firm, and zero otherwise. Because of cultural differences, deals initiated by foreign bidders are more likely to be hostile. Hence, we expect cross-border regulation to reduce the frequency of hostile takeovers. The results in Column 6 show that common law is significant and that its coefficient is virtually unchanged from Column 1. The frequency of attempted hostile takeovers among traded companies is 1.6% higher in common-law than in civil-law countries. Cross-border regulation is also significant and negative, as predicted. The requirement of government approval for foreign acquisitions reduces the frequency of attempted hostile takeovers by 1.8%. Market returns and mandatory bid rules are not statistically significant. 3.3. Cross-border mergers and acquisitions La Porta et al. (2000a, p. 23) write that ``When a British firm fully acquires a Swedish firm, the possibilities for legal expropriation of investor diminish. Because the controlling shareholders of the Swedish company are compensated in such a friendly deal for the lost private benefits of control, they are more likely to go along. By replacing the wasteful expropriation with publicly shared profits and dividends, such acquisitions enhance efficiency.'' This statement implies two testable hypotheses that we address in this section: first, the probability that a deal is cross-border rather than domestic is higher in countries with lower investor protection; and second, the acquirers in cross-border deals will come from countries that have higher investor protection than the targets' countries. 3.3.1. Target-country analysis As before, we adapt specification (1) by changing the dependent variable Cross-border ratio ¼ a þ bX þ g investor protectionþe; ð3Þ Chapter Seven 243 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 289 where the cross-border ratio is the number of cross-border deals as a percentage of all completed deals by target country. Common law, accounting standards, and shareholder protection are our proxies for investor protection. We expect the cross- border ratio to decrease with investor protection. As before, we control for the logarithm of GNP per capita, as a measure of a country's wealth, and GDP growth as a proxy for the change in macroeconomic conditions. Table 5 reports the coefficients of six Tobit models derived from specification (3). The results confirm our prediction: the probability that a completed deal is cross-border rather than domestic is higher in countries with lower investor protection. The coefficients on common law, accounting standards, and shareholder protection are all negative and significant at the 1% level. In economic terms, the probability that a completed deal is cross-border is 14.5% higher in civil-law than in common-law countries. Raising the accounting standards measure by 12 points (from Italy's to Canada's accounting standards) decreases cross-border deals by 5%. An increase in shareholder protection by one point (for instance, the adoption of voting by mail in Belgium) decreases the cross-border ratio by 4%. Ownership concentration, which we add in Column 5 as a control variable, is not statistically significant. To evaluate the robustness of the results, in Column 6 we augment the specification in Column 1 with some control variables. We add cross-border regulation because we expect fewer cross-border deals when there are more regulatory requirements. We control for market returns because we expect fewer cross-border deals when the stock market is booming and the target firms' stocks are (potentially) overvalued. At the same time, this variable will not be significant if the acquirer's stock market is also thriving. We include openness, a measure of the cultural attitude towards cross-border deals (from the 1996 Global Competitiveness Report) because such deals are more likely if the country is friendlier to foreigners.1 Our results show that common law is still significant and that its coefficient is unaffected. Openness is negative and significant, as predicted. The coefficients on market return and cross-border regulation are not significant. 3.3.2. Ordered-pair analysis The results in Table 5 indicate that cross-border mergers and acquisitions play a governance role by targeting firms in countries with lower investor protection. To explore this hypothesis, we arrange our dataset to produce a worldwide matrix of (49 48) matched pairs. In these pairs, we define each entry, cross-border dealss ; as ;b the number of deals in which the acquirer comes from country b (for buyer) and the target is in country s (for seller), as a percentage of the total number of deals in country s. 1Another potential determinant of international mergers and acquisitions is tax competition across countries. For instance, taxes can affect M&A activity if it is easier for domestic firms to take advantage of investment tax credits and accelerated depreciation in the target country than for foreign firms. Moreover, the tax treatment of foreign income differs across countries. However, we do not control for taxes in our study because the complexity of the issue requires a paper on its own. 244 A Reader in International Corporate Finance 290 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Table 5 Cross-border versus domestic deals The table presents the results of six Tobit models estimated by maximum likelihood on the sample of 49 target countries. The dependent variable is cross-border ratio, or cross-border deals as a percentage of all completed deals. The independent variables are: common law, a dummy variable that equals one if the origin of the company law is the English common law, and zero otherwise; accounting standards, an index of the quality of accounting disclosure; shareholder protection, a measure of the effective rights of minority shareholders; ownership concentration, the average equity stake owned by the three largest shareholders in the ten largest nonfinancial domestic firms in 1994; cross-border regulation, a dummy variable that equals one if foreign buyers need government approval, and zero otherwise; market return, the average annual stock market return in the 1990s; and openness, a survey-based measure of the cultural attitude towards cross-border deals. The logarithm of GNP per capita and GDP growth are included in all regressions as control variables. Standard errors are shown in parentheses. (1) (2) (3) (4) (5) (6) Log (GNP per capita) 5.32ÃÃÃ 1.99 1.47 0.64 1.21 4.77ÃÃÃ (1.20) (1.74) (1.50) (1.79) (1.72) (1.51) GDP growth 1.75 0.90 1.44 1.48 1.38 3.48ÃÃÃ (1.08) (1.17) (1.08) (1.15) (1.16) (1.19) Common law 14.5ÃÃÃ 16.1ÃÃÃ (3.83) (4.02) Accounting standards 0.67ÃÃÃ 0.53ÃÃÃ 0.41ÃÃ (0.16) (0.17) (0.17) Shareholder protection 6.03ÃÃÃ 3.55ÃÃ 4.14ÃÃ (1.71) (1.76) (1.98) Ownership concentration 0.11 (0.17) Cross-border regulation 5.05 (4.36) Market return 0.15 (0.13) Openness 7.77ÃÃÃ (2.84) Constant 87.7ÃÃÃ 96.5ÃÃÃ 62.7ÃÃÃ 81.7ÃÃÃ 85.0ÃÃÃ 38.1Ã (11.7) (14.8) (12.7) (15.9) (18.8) (20.0) Pseudo R2 0.06 0.07 0.05 0.09 0.08 0.09 N observations 49 41 49 41 39 41 ÃÃÃ ÃÃ Ã , , indicate significance at 1%, 5%, and 10% levels, respectively. With the newly arranged dataset, we can study the pattern of cross-border mergers and acquisitions by simultaneously controlling for the characteristics of target and acquirer countries. The specification is Cross-border dealss ¼ bXs þ gD ðinvestor protectionÞs þ db þ zs þ es ; ;b ;b ;b ;b ð4Þ where the dependent variable is the number of cross-border deals in which the acquirer comes from country b and the target from country s ðbasÞ as a percentage of the total number of deals (cross-border and domestic) in country s. Our hypothesis is that the volume of cross-border M&A activity between country b (the Chapter Seven 245 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 291 acquirer) and country s (the target) correlates positively with the difference in investor protection between the two countries. The proxies for investor protection are accounting standards and shareholder protection. We note that our specification also includes fixed effects for target and acquirer countries. These fixed effects control for all cultural and institutional characteristics of the two countries, including the level of investor protection in the individual countries. We control for differences in the logarithm of GNP per capita of the acquirer and target countries as a measure of the relative economic development of the two countries. We also include two dummy variables equal to one if the acquirer and target share the same cultural background, that is, if they have the same official language and if they belong to the same geographical area. Table 6 reports our results. In Columns 1 and 2, we include only one measure of investor protection per regression. We find that the volume of M&A activity between two countries is positively correlated with their difference in investor protection. This result means that acquirers typically come from countries with better accounting standards and stronger shareholder protection than the targets' countries. In Column 3, we estimate the marginal impact of each variable by estimating a joint regression with the two measures. We find that only the difference in shareholder protection is statistically significant. On average, shareholder protection increases in the target company via the cross-border deal. This finding is consistent with the view that such acquisitions enhance efficiency because the increase in shareholder protection curbs the expropriation of minority shareholders and, therefore, reduces the cost of raising external equity. We also find that richer countries are more likely to be acquirers than targets, and that most cross-border deals happen between countries sharing the same language and geographical area. In Column 4, we add the difference in market return between acquirer and target countries as a control variable. We would expect more deals when the acquirer's stock market is booming relatively to the target's stock market, but we find no such evidence. A potentially important missing variable in the analysis is the volume of trade between two countries. In fact, companies that export to a given country might engage in M&A activity in that country for reasons that have nothing to do with governance. To control for this alternative explanation, in Column 5 we add bilateral trade to our regression. We define bilateral trades as imports from country b to ;b country s as a percentage of total imports of country s. Bilateral trade is not available for six countries: Belgium, Brazil, Israel, Nigeria, Switzerland and Zimbabwe. The number of observations in Column 5 changes accordingly. The results for shareholder protection are unchanged. The acquirer typically has stronger share- holder protection than the target. As we expected, bilateral trade is positive and significant, confirming that trade is an important motive for cross-border mergers and acquisitions. Same language and the difference in the logarithm of GNP per capita are no longer significant once bilateral trade is added to the baseline specification. 246 A Reader in International Corporate Finance 292 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Table 6 The governance motive in cross-border M&A The table presents the results of five OLS regressions for the sample of matched country pairs. The dependent variable is cross-border dealss ; or the number of cross-border deals where the target is from ;b country s and the acquirer is from country b ðsabÞ as a percentage of the total number of deals in country s. The independent variables are the difference between acquirer and target countries' investor protection as measured alternatively by accounting standards, an index of the quality of accounting disclosure, and by shareholder protection, a measure of the effective rights of minority shareholders. We include as control variables the difference between the acquirer's and the target's logarithm of GNP per capita; same language, a dummy variable that equals one if the target and acquirer come from countries with the same official language, and zero otherwise; and same geographical area, a dummy variable that equals one if the target and acquirer come from the same geographical area. In Column 4, we add the difference between country b and country s in market return, the average annual stock market return in the 1990s. In Column 5, we add bilateral trades ; the value of imports by country s from country b as a percentage of total ;b imports by country s. The regressions contain fixed effects both for target and acquirer country (not shown). The standard errors shown in parentheses are adjusted for heteroskedasticity using Huber (1967) and White (1980) corrections. (1) (2) (3) (4) (5) DðAccounting standardsÞb 0.02ÃÃÃ 0.01 s (0.01) (0.00) DðShareholder protectionÞb 1.93ÃÃÃ 1.89ÃÃÃ 1.89ÃÃÃ 1.21ÃÃÃ s (0.19) (0.21) (0.20) (0.23) DðLogðGNP per capitaÞÞb 0.10Ã 0.97ÃÃÃ 0.40ÃÃÃ 0.95ÃÃÃ 0.06 s (0.05) (0.10) (0.05) (0.10) (0.04) Same language 0.86ÃÃ 0.97ÃÃÃ 0.86ÃÃ 1.02ÃÃ 0.08 (0.36) (0.30) (0.36) (0.31) (0.22) Same geographical area 1.30ÃÃÃ 1.12ÃÃÃ 1.30ÃÃÃ 1.13ÃÃÃ 0.36ÃÃÃ (0.14) (0.11) (0.14) (0.12) (0.15) DðMarket returnÞb 0.00 s (0.00) Bilateral trades;b 0.67ÃÃÃ (0.10) Adjusted R2 0.53 0.50 0.53 0.51 0.67 N observations 1640 2352 1640 2162 1677 ÃÃÃ ÃÃ Ã , , indicate significance at 1%, 5% and 10% levels, respectively. 3.4. Premium We use the sample of individual transactions to analyze the cross-country determinants of the takeover premium. We estimate the specification Log ðpremiumÞ ¼ a þ bX þ g shareholder protectionþe; ð5Þ where premium is the bid price as a percentage of the target's closing price four weeks before the announcement of the deal, shareholder protection is measured at the target country level, and X is a set of control factors. Control variables at the deal level are target size, the logarithm of the target's market capitalization four weeks before the announcement, a dummy variable (cross-border) that equals one if the Chapter Seven 247 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 293 deal is cross-border and zero otherwise; a dummy variable (hostile bid) that equals one if the deal is hostile and zero otherwise; a dummy variable (tender offer) that equals one if the deal involves a tender offer and zero otherwise; and a dummy variable (contested bid) that equals one if the number of bidders is larger than one and zero otherwise. Table 7 shows the results of six regressions based on specification (5). In all regressions, the standard errors shown in parentheses are adjusted for hetero- skedasticity, using the Huber (1967) and White (1980) corrections, and for clustering at the country level following Huber (1967). We correct for clustering because observations within a country are likely to be correlated with each other. We also include year and industry (at one-digit SIC-code level) dummies, but we do not report their coefficients. In Column 1, we find that shareholder protection is positively correlated with the takeover premium. An increase in the level of shareholder protection by one point (e.g., the introduction of voting by mail in Belgium) is associated with a 0.04 increase in the logarithm of the premium, which translates into an average increase of 6% in the premium. Target size is negative and significant, that is, larger deals are associated with lower premiums. In Column 2, we add the deal-level dummy variables for cross-border, hostile bid, tender offer, and contested bid. The result on shareholder protection does not change and the new controls are all positive, as expected. All but hostile bids are statistically significant. We interpret the finding on tender offers as evidence of the free-rider hypothesis: that is, the bidder in a tender offer needs to pay a higher premium to induce shareholders to tender their shares. This theory would also predict that the premium paid should be higher the more diffuse the target's ownership structure. However, we cannot test this hypothesis directly because we do not have data on ownership structure for individual target companies. Contested bids are associated with a 0.1 increase in the logarithm of the premium, which translates into an average premium increase of 15%, consistent with the view that competition for targets is associated with higher premiums. Cross-border deals are associated with a 0.03 increase in the logarithm of the premium, which translates into an average premium increase of 3%. Our finding that takeover premiums are higher in countries with higher shareholder protection can be interpreted by noting that the takeover premium measures the gain available to all target shareholders. There are two reasons why the premium might be higher in countries with stronger shareholder protection. First, shareholder protection reduces the cost of capital and therefore increases (potential) competition among bidders and the premium paid by the winning bidder. Second, diffuse ownership is more common in countries with higher shareholder protection. In turn, diffuse ownership exacerbates the free-rider problem in takeovers by forcing bidders to pay a higher takeover premium than otherwise (Grossman and Hart, 1980). A concern with this interpretation is the possibility that the premium measures the private benefits of control. To explore this issue, in Column 3 we add the difference between the acquirer and target countries' shareholder protection as a further 248 A Reader in International Corporate Finance 294 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Table 7 Determinants of the takeover premium The table presents the results of six OLS regressions for the sample of individual deals. The dependent variable is the natural logarithm of premium, or the bid price as a percentage of the closing price of the target four weeks before the announcement. Independent variables at the country level are shareholder protection, a measure of the effective rights of minority shareholders, and mandatory bid rule, a dummy variable that equals one if in 1995 there was a legal requirement to make a tender offer when shareholdings after the acquisition exceed a given ownership threshold, and zero otherwise. The control variable at the cross-country level is the difference between the acquirer and target countries' shareholder protection. Control variables at the deal level are: target size, the logarithm of the target's market capitalization four weeks before the announcement; cross-border, a dummy variable that equals one if the deal is cross- border, and zero otherwise; hostile bid, a dummy variable that equals one if the deal is hostile, and zero otherwise; tender offer, a dummy variable that equals one if the deal involves a tender offer, and zero otherwise; contested bid, a dummy variable that equals one if the number of bidders is larger than one, and zero otherwise; and bidder M/B, the equity market-to-book ratio of the bidder four weeks before the announcement. In all regressions, we also include year and industry (at one-digit SIC-code level) dummies (not shown). In Column 6 we add two dummy variables that identify deals where the target firm is from the US (US targets) and from the UK (UK targets), respectively. The standard errors (in parentheses) are adjusted for heteroskedasticity using Huber (1967) and White (1980) corrections and for clustering at country level using the Huber (1967) correction. (1) (2) (3) (4) (5) (6) Shareholder protection 0.04ÃÃÃ 0.05ÃÃÃ 0.05ÃÃÃ 0.07ÃÃÃ 0.04ÃÃÃ 0.01 (0.01) (0.01) (0.01) (0.02) (0.01) (0.02) Target size 0.01ÃÃÃ 0.01ÃÃÃ 0.01ÃÃÃ 0.02ÃÃ 0.02ÃÃÃ 0.02ÃÃÃ (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) Cross-border 0.03Ã 0.03Ã 0.02 0.03ÃÃ 0.04ÃÃ (0.02) (0.02) (0.03) (0.01) (0.02) Hostile bid 0.04 0.04 0.03 0.04 0.06ÃÃÃ (0.03) (0.03) (0.06) (0.03) (0.02) Tender offer 0.05ÃÃÃ 0.05ÃÃÃ 0.04 0.07ÃÃÃ 0.08ÃÃÃ (0.01) (0.01) (0.02) (0.01) (0.01) Contested bid 0.10ÃÃ 0.10ÃÃ 0.05 0.10ÃÃ 0.11ÃÃÃ (0.04) (0.04) (0.05) (0.04) (0.04) DðShareholder protectionÞb 0.00 s (0.01) Bidder M/B 0.01 (0.00) Mandatory bid rule 0.06ÃÃ 0.01 (0.02) (0.04) US targets 0.16ÃÃ (0.07) UK targets 0.09ÃÃÃ (0.03) R2 0.03 0.04 0.05 0.08 0.05 0.06 N observations 4007 4007 4007 1005 4007 4007 N countries 35 35 35 27 35 35 ÃÃÃ ÃÃ Ã , , indicate significance at 1%, 5%, and 10% levels, respectively. Chapter Seven 249 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 295 control variable. If the premium measures the private benefits of control, we expect to find a negative and significant coefficient on this control variable, as in Dyck and Zingales (2004). The reason is that an acquirer coming from a country with lower shareholder protection is better able to extract private benefits of control than an acquirer coming from a country with stricter rules. In Column 3, we find that the difference between acquirer and target countries' shareholder protection is not statistically significant. This result indicates that premium is not a proxy for the private benefits of control but for the total premium available to all shareholders. This finding also indicates that acquirers from countries with better shareholder protection do not need to pay more than acquirers from countries with weaker shareholder protection in cross-border deals. According to Rau and Vermaelen (1998), glamour firms (as measured by high market-to-book ratios) will tend to overestimate their ability to create synergies in the target and should therefore be willing to pay more than managers of value firms (as measured by low market-to-book ratios). Therefore, in Column 4, we add the equity market-to-book ratio (M/B) of the bidder four weeks before the announce- ment. We obtain this information from Datastream. As a result of the matching procedure, the number of observations in Column 4 drops to 1,005. Contrary to the prediction, our results show that the bidder M/B is not correlated with the premium. Comment and Schwert (1995) show that takeover laws are an important determinant of the takeover premium. Therefore, in Column 5 we control for differences in takeover laws across countries. The mandatory bid rule variable equals one if in 1995 there was a legal requirement to make a tender offer when shareholdings after the acquisition exceed a given ownership threshold, and zero otherwise. For instance, the mandatory bid variable rule equals one in the United Kingdom, where the threshold is 30%, and zero in the United States, where only a few states have a similar provision. We find a negative and significant coefficient for the mandatory bid rule, perhaps because a mandatory bid rule increases the cost of takeovers and therefore reduces competition among bidders. However, a mandatory bid rule might also increase the premium, because only high-premium takeovers that compensate the bidders for the high takeover costs succeed. To distinguish between the two effects, in an unreported regression we add the interactive term of mandatory bid rule multiplied by target size. The coefficient on this interactive term should measure the impact on the premium that is due to reduced competition, because larger deals are more likely to be deterred. The coefficient on the mandatory bid rule should reflect the fact that low-premium takeovers do not go through. We find that the coefficient on the mandatory bid rule is negative and significant, and that the coefficient on the interactive term is not significant. This result suggests that the mandatory bid rule variable captures an institutional difference across countries. Because 75% of the deals have a US or UK target, in Column 6 we check the robustness of our findings by using two dummy variables that identify deals with US and UK targets, respectively. The results show that higher premiums are a feature of US and UK targets. The logarithm of the premium is 0.16 higher in the US and 0.09 higher in the UK than in the other countries. Note that the mandatory bid rule is no 250 A Reader in International Corporate Finance 296 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 longer significant. This finding suggests that the mandatory bid rule is significant in Column 5 only because it captures the difference between US and UK targets. 3.5. Means of payment Legal protection of investors may also affect the means of payment used in mergers and acquisitions. In a country with low investor protection, target shareholders are likely to prefer cash over the bidder's equity as the takeover currency, due to the risk of expropriation for being minority shareholders. We therefore expect less equity financing and more cash financing in countries with lower shareholder protection. We estimate the following regression for the method of payment: Prob ðall-cash bidÞ ¼ a þ bX þ g shareholder protectionþe: ð6Þ In this regression, which is similar to Eq. (3), our control variables are the same as those in Table 6: target size, cross-border, hostile bid, tender offer, contested bid, bidder M/B, and mandatory bid rule. We expect that larger deals are less likely to be paid entirely with cash. Cross-border deals might more often be paid in cash because shareholders dislike receiving foreign stocks as compensation. To entice shareholders to tender, hostile bids, tender offers, and contested bids are likely to be in cash. Table 8 reports the results of six regressions based on specification (6). In all regressions, the standard errors shown in parentheses are adjusted for hetero- skedasticity using Huber (1967) and White (1980) corrections, and for clustering at the country level following Huber (1967). We also include year and industry dummies (at the one-digit SIC-code level), but we do not report their coefficients. Across all specifications, we find that shareholder protection is negatively correlated with all-cash bids. We note that a one-point increase in the level of shareholder protection is associated with a reduction of between 13% and 18% in the probability of using only cash as the means of payment. Our interpretation of this result is that stocks are a less popular means of payment in countries with lower shareholder protection because stocks entail a higher risk of expropriation. Among the control variables, target size is negative and significant, and cross- border, hostile bid, and tender offer are positive and significant, as we expected. Contested bids are not associated with more cash as a method of payment. The probability of using only cash as the method of payment is 17% higher in cross- border deals. To deepen the analysis of the means of payment in cross-border deals, in Column 3 we add the difference between acquirer and target countries' shareholder protection as a further control variable. We expect that the use of stocks as a method of payment will be positively correlated with the degree of investor protection in the acquirer country, when acquirer and target countries are different. We find evidence in favor of this prediction because the coefficient on the difference between acquirer and target countries' shareholder protection is negative and significant. Chapter Seven 251 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 297 Table 8 Means of payment The table reports estimates of six Probit models for the sample of individual deals. The dependent variable is all-cash bid, or a dummy variable that equals one if the acquisition is entirely paid in cash, and zero otherwise. Independent variables at the country level are shareholder protection, a measure of the effective rights of minority shareholders, and mandatory bid rule, a dummy variable that equals one if in 1995 there was a legal requirement to make a tender offer when shareholdings after the acquisition exceed a given ownership threshold, and zero otherwise. The control variable at the cross-country level is the difference between the acquirer and target countries' shareholder protection. Control variables at the deal level are: target size, the logarithm of the target's market capitalization four weeks before the announcement; cross- border, a dummy variable that equals one if the deal is cross-border, and zero otherwise; hostile bid, a dummy variable that equals one if the deal is hostile, and zero otherwise; tender offer, a dummy variable that equals one if the deal involves a tender offer, and zero otherwise; contested bid, a dummy variable that equals one if the number of bidders is larger than one, and zero otherwise; and bidder M/B, the equity market-to-book ratio of the bidder four weeks before the announcement. In all regressions, we also include year and industry (at one-digit SIC-code level) dummies (not shown). In Column 6 we add two dummy variables that identify deals where the target firm is from the US (US targets) and from the UK (UK targets), respectively. Displayed coefficients are the change in probability for an infinitesimal change in the independent variables. The standard errors (in parentheses) are adjusted for heteroskedasticity using Huber (1967) and White (1980) corrections and for clustering at country level using the Huber (1967) correction. (1) (2) (3) (4) (5) (6) Shareholder protection 0.18ÃÃÃ 0.13ÃÃÃ 0.14ÃÃÃ 0.08ÃÃ 0.15ÃÃÃ 0.16ÃÃÃ (0.03) (0.03) (0.03) (0.03) (0.02) (0.04) Target size 0.06ÃÃÃ 0.07ÃÃÃ 0.07ÃÃÃ 0.02 0.08ÃÃÃ 0.08ÃÃÃ (0.01) (0.02) (0.02) (0.02) (0.02) (0.02) Cross-border 0.17ÃÃÃ 0.14ÃÃ 0.21ÃÃÃ 0.14ÃÃÃ 0.14ÃÃÃ (0.04) (0.05) (0.05) (0.04) (0.05) Hostile bid 0.10ÃÃÃ 0.09ÃÃ 0.08 0.10ÃÃ 0.09ÃÃ (0.04) (0.04) (0.08) (0.04) (0.04) Tender offer 0.33ÃÃÃ 0.32ÃÃÃ 0.36ÃÃÃ 0.34ÃÃÃ 0.37ÃÃÃ (0.08) (0.08) (0.11) (0.09) (0.08) Contested bid 0.04 0.04 0.12Ã 0.05 0.04 (0.04) (0.04) (0.07) (0.04) (0.04) DðShareholder protectionÞb 0.06ÃÃÃ 0.01 0.06ÃÃÃ 0.05ÃÃÃ s (0.01) (0.03) (0.01) (0.02) Bidder M/B 0.00 (0.00) Mandatory bid rule 0.06 (0.08) US targets 0.04 (0.10) UK targets 0.10 (0.06) Pseudo R2 0.11 0.18 0.19 0.20 0.19 0.19 N observations 4007 4007 4007 1005 4007 4007 N countries 35 35 35 27 35 35 ÃÃÃ ÃÃ Ã , , indicate significance at 1%, 5%, and 10% levels, respectively. 252 A Reader in International Corporate Finance 298 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Bidder M/B might be correlated with the use of stocks as means of payment because the bidder could try to take advantage of market booms, as argued by Shleifer and Vishny (2003). In Column 4, we add the bidder M/B, but we find that its coefficient is not significantly different from zero. The mandatory bid rule might require the bidder to make a cash offer or an offer with a cash alternative, as in the UK. If so, mandatory bid rules should be positively correlated with all-cash bids. However, UK bidders often avoid the mandatory tender offer by bidding for 29.9% of the shares, which is just below the 30% threshold for the mandatory tender offer, and then acquiring the remaining shares via a share offer. In this case, mandatory bid rules should not be correlated with all- cash bids. In Column 5, we control for mandatory bid rules, and find that the coefficient is not statistically significant. In Column 6, we show that our results are not driven by deals involving US and UK firms. The coefficient on shareholder protection is even larger in absolute terms than in Column 1, and equally significant in statistical terms when we include two dummy variables for deals in which the target is a UK or US firm, respectively. As a further robustness check (not reported), we estimate the specification in Column 2 with weighted least squares, in which the weights are the inverse of the number of observations by country. With this procedure, all countries have the same impact on the final results. The coefficient on shareholder protection is identical to that in Column 2. One concern is that the control variables used in regressions (5) and (6) (tender offer, hostile bid, and cross-border) are themselves endogenous. As a result, our estimates could be inconsistent. To address this issue, we estimate a recursive system with five equations, one for each endogenous variable: premium, all-cash bid, tender offer, hostile bid, and cross-border. Exogenous variables are target size, bidder M/B, shareholder protection, and mandatory bid rule. We do not present the results of these regressions here, because the coefficients on shareholder protection are similar to those in Tables 7 and 8. 4. Discussion The results presented in Section 3 have implications for the impact of investor protection on M&A activity and the role of cross-border takeovers as a catalyst for convergence in corporate governance regimes. We discuss both implications below. 4.1. M&A activity and investor protection Overall, the results in Section 3 characterize M&A activity as correlating with investor-friendly legal environments. We interpret these findings along the lines of La Porta et al. (2000b) and argue that a more active market for mergers and acquisitions is the outcome of a corporate governance regime with stronger investor protection. Chapter Seven 253 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 299 With low shareholder protection, there are large private benefits of control (Nenova, 2003; Dyck and Zingales, 2004), and therefore the market for corporate control does not operate freely. Conversely, with high investor protection, there are low private benefits of control, and there is an active market for corporate control. Moreover, better accounting standards increase disclosure, which helps acquirers identify potential targets. Hence, there are more potential targets in countries with better shareholder protection and accounting standards. This view yields two testable predictions: across target countries, both the volume of takeovers and the takeover premium should increase with better shareholder protection and account- ing standards. The results on volume, reported in Table 3, are strongly consistent with this view. The results on the premium, reported in Table 7, are weakly consistent with this view. Table 7 shows that higher shareholder protection in the target company is associated with higher premiums, although US and UK firms drive the results. Our results reject the alternative view that the market for corporate control is a substitute for legal protection of shareholders. According to Manne (1965) and Jensen (1993), if the market for corporate control works efficiently, firms with poor corporate governance become the targets of takeovers from more efficient firms. Extending their argument across countries, the volume of M&A activity and the premium paid should be greater in countries with lower investor protection. These predictions are inconsistent with our findings. 4.2. Convergence in corporate governance The results in Table 6 relate to the ongoing debate among legal scholars on the possibility of effective worldwide convergence in corporate governance standards. Coffee (1999) argues that differences in corporate governance will persist but with some degree of functional convergence. Hansmann and Kraakman (2001) believe that formal convergence will happen soon. Bebchuk and Roe (1999) question the idea of rapid convergence because political and economic forces will slow down any change. Gilson (2001) argues that convergence will happen through all three channels (formal, contractual, and functional). Our findings are consistent with the prediction by Coffee (1999) that companies from countries with better protection of investors will end up buying companies from countries with weaker protection. The case for target shareholders to sell out to bidders with higher governance standards is clear. Targets stand to gain from the lower cost of capital associated with higher investor protection. However, it is not obvious why acquirers seek to take over a poorly governed company. The results in Table 7, Column 3, show that acquirers from countries with better investor protection do not pay higher takeover premiums than acquirers from countries with weaker investor protection. Hence, they share part of the surplus created by improving the corporate governance of the target. One concern is that they might import the poorer governance of their targets (poor accounting and disclosure practices, board structures, and so on). However, anecdotal evidence of cross-border deals with high press coverage suggests that 254 A Reader in International Corporate Finance 300 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 this is not the case. The targets almost always adopt the governance standards of the acquirers, whether good or bad. In Daimler's acquisition of Chrysler, for instance, the resulting company has adopted a two-tier board structure, as required by German law. Thus, if convergence occurs, it is towards the acquirers' governance standards. A related issue is that a deal could be motivated by the agency and hubris problems of the acquirer rather than by the desire to improve the governance regime in the target company. If so, the deal might not create value. Assessing this issue requires a study of the performance of the target and acquirer after the acquisition, which we cannot do with our large sample. Instead, we indirectly test this issue. If countries with poorer investor protection (in particular, lower governance standards, as measured by lower shareholder protection) have more severe agency problems, the hypothesis predicts more acquisitions by companies in countries with lower shareholder protection. This is not what we observe. If we sort our data by acquirer country, we find rather the opposite (not reported): more acquisitions by companies in countries with higher shareholder protection. Our analysis also sheds light on the question as to whether cross-border deals might lead to greater international stock market integration and to a reduction of the home bias in equity investment in target countries. If the foreign bidder pays with stock, target shareholders face the problem of disposing of a new investment domiciled abroad. As a result, they might choose to keep the foreign stocks. In aggregate, these individual decisions would imply a reduction of the home bias in equity investment in target countries. We show in Table 8, Column 3, that target shareholders accept the acquirer's shares more often if the investor protection in the acquirer's country is greater than in the target's country. Hence, the reduction of the home bias puzzle goes together with a convergence in corporate governance regime. In this sense, our findings are consistent with Dahlquist et al. (2003). 5. Conclusion Using a large sample of deals in 49 major countries, announced in the 1990s and completed by the end of 2002, we find that better investor protection is associated with more mergers and acquisitions, more attempted hostile takeovers, and fewer cross-border deals. We also find that better investor protection is associated with the greater use of stock as a method of payment, and with higher takeover premiums. These results indicate that domestic investor protection is an important determinant of the competitiveness and effectiveness of the market for mergers and acquisitions within a country. In cross-border deals, we find that acquirers on average have higher investor protection than targets, that is, firms opt out of a weak governance regime via cross-border deals. This result indicates that the international market for corporate control helps generate convergence in corporate governance regimes across countries. Chapter Seven 255 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 301 Appendix A. Description of the variables included in our study and their sources A.1. Country-level variables Volume Percentage of domestic traded companies targeted in completed deals in the 1990s. Sources: SDC Platinum, provided by Thomson Financial Securities Data, and the World Development Indicators. Hostile takeover Attempted hostile takeovers as a percentage of domes- tic traded companies. Sources: SDC Platinum and the World Development Indicators. Cross-border ratio Number of cross-border deals as target as a percentage of all completed deals. Source: SDC Platinum. GDP growth Average annual real growth rate of the gross domestic product in the 1990s. Source: World Development Report. GNP per capita Gross national product in 1995 (in US$) divided by the population. Source: World Development Report. Common law Equals one if the origin of the company law is the English common law and zero otherwise. Source: La Porta et al. (1998). Accounting standards Index created by the Center for International Financial Analysis and Research to rate the quality of 1990 annual reports on their disclosure of accounting information. Source: La Porta et al. (1998). Rule of law Assessment of the law and order tradition in the country produced by the risk-rating agency Interna- tional Country Risk (ICR). Average of the months of April and October of the monthly index between 1982 and 1995. It ranges between zero and ten. Source: La Porta et al. (1998). Antidirector rights The index is formed by adding one when (i) the country allows shareholders to mail their proxy vote to the firm, (ii) shareholders are not required to deposit their shares prior to the general shareholders' meeting, (iii) cumu- lative voting or proportional representation of mino- rities in the board of directors is allowed, (iv) an oppressed minorities mechanism is in place, (v) the minimum percentage of share capital that entitles a shareholder to call for an extraordinary shareholders' meeting is less than or equal to 10% (the sample median), or (vi) shareholders have preemptive rights that can be waived only by a shareholders' vote. Source: La Porta et al. (1998). 256 A Reader in International Corporate Finance 302 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Shareholder protection Measure of the effective rights of minority shareholders computed as the product of rule of law and antidirector rights divided by ten. It ranges between zero and six. Ownership concentration Average equity stake owned by the three largest shareholders in the ten largest nonfinancial domestic firms in 1994. Source: La Porta et al. (1998). Cross-border regulation Equals one if in 1995 a foreign buyer needed govern- ment approval before acquiring control of a domestic firm and zero otherwise. Source: Economist Intelligence Unit, Country Surveys. Market return Average annual stock market return in 1990s adjusted for inflation with the Consumer Price Index. Source: WorldScope. Market dominance Response to survey question: ``Market dominance by a few enterprises is rare in key industries (1=strongly disagree, 6=strongly agree).'' Source: The Global Competitiveness Report, 1996. Mandatory bid rule Equals one if in 1995 there was a legal requirement to make a tender offer when shareholding after the acquisition exceeds a given ownership threshold and zero otherwise. Source: Economist Intelligence Unit, Country Surveys. Openness Response to survey question: ``Foreign investors are free to acquire control of a domestic company (1=strongly disagree, 6=strongly agree).'' Source: The Global Competitiveness Report, 1996. A.2. Cross-border variables Cross-border dealss ;b Number of deals in which the target is from country s and the acquirer is from country b, shown as a percentage of the total number of deals with target in country s. Source: SDC Platinum. Same language Equals one when target and acquirer's countries share the same main language and zero otherwise. Source: World Atlas 1995. Same geographical area Equals one when target and acquirer's countries are from the same continent and zero otherwise. We classify all countries into four areas (Africa, America, Asia, and Europe). Source: World Atlas 1995. Bilateral trades;b Value of imports by country s from country b as a percentage of total import by country s. Source: World Bank Trade and Production Database. Chapter Seven 257 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 303 A.3. Deal-level variables Premium Bid price as a percentage of the closing price of the target four weeks before the announcement. Source: SDC Platinum. All-cash bid Equals one if the acquisition is entirely paid in cash and zero otherwise. Source: SDC Platinum. Target size Logarithm of the market capitalization of the target four weeks before the announcement of the deal in US$ million. Source: SDC Platinum. Tender offer Equals one if the acquisition is done through a tender offer and zero otherwise. Source: SDC Platinum. Cross-border Equals one if the target country differs from the acquirer country and zero otherwise. Source: SDC Platinum. Hostile bid Equals one if the bid is classified as unsolicited and zero otherwise. Source: SDC Platinum. Contested bid Equals one if the number of bidders is larger than one and zero otherwise. Source: SDC Platinum. Bidder M/B Equity market-to-book ratio of the bidder computed four weeks before the announcement. Source: Datastream. References Bebchuk, L., 1999. A rent-protection theory of corporate ownership and control. NBER Working Paper 7203, National Bureau of Economic Research, Cambridge, MA. Bebchuk, L., Roe, M., 1999. A theory of path dependence in corporate governance and ownership. Stanford Law Review 52, 127­170. Bhattacharya, U., Daouk, H., 2002. The world price of insider trading. Journal of Finance 57, 75­108. Coffee, J., 1999. The future as history: the prospects for global convergence in corporate governance and its implications. Northwestern University Law Review 93, 641­708. Comment, R., Schwert, W., 1995. Poison or placebo? Evidence on the deterrence and wealth effects of modern antitakeover measures. Journal of Financial Economics 39, 3­43. Dahlquist, M., Pinkowitz, L., Stulz, R., Williamson, R., 2003. Corporate governance and the home bias. Journal of Financial and Quantitative Analysis 38, 87­110. Dittmar, A., Mahrt-Smith, J., Servaes, H., 2003. International corporate governance and corporate cash holdings. Journal of Financial and Quantitative Analysis 38, 111­133. Dyck, A., Zingales, L., 2004. Private benefits of control: an international comparison. Journal of Finance 59, 537­600. Franks, J., Mayer, C., 1996. Hostile takeovers and the correction of managerial failure. Journal of Financial Economics 40, 163­181. Gilson, R., 2001. Globalizing corporate governance: Convergence of form or function. American Journal of Comparative Law 49, 329­363. Grossman, S., Hart, O., 1980. Takeover bids, the free-rider problem and the theory of the corporation. Bell Journal of Economics 11, 42­64. Hansmann, H., Kraakman, R., 2001. The end of history for corporate law. Georgetown Law Journal 89, 439­468. Huber, P., 1967. The behavior of maximum likelihood estimates under non-standard conditions. In: Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Vol. 1. University of California Press, Berkeley, CA, pp. 221­233. 258 A Reader in International Corporate Finance 304 S. Rossi, P.F. Volpin / Journal of Financial Economics 74 (2004) 277­304 Jenkinson, T., Ljungqvist, A., 2001. The role of hostile stakes in German corporate governance. Journal of Corporate Finance 7, 397­446. Jensen, M., 1993. The modern industrial revolution, exit, and the failure of internal control systems. Journal of Finance 48, 831­880. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1997. Legal determinants of external finance. Journal of Finance 52, 1131­1150. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and finance. Journal of Political Economy 101, 678­709. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 1999. Corporate ownership around the world. Journal of Finance 54, 471­517. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000a. Investor protection and corporate governance. Journal of Financial Economics 58, 3­27. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000b. Agency problems and dividend policies around the world. Journal of Finance 55, 1­33. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2002. Investor protection and corporate valuation. Journal of Finance 57, 1147­1170. Leuz, C., Nanda, D., Wysocki, P., 2003. Earnings management and investor protection. Journal of Financial Economics 69, 505­527. Manne, H., 1965. Mergers and the market for corporate control. Journal of Political Economy 75, 110­126. Nenova, T., 2003. The value of corporate voting rights and control: a cross-country analysis. Journal of Financial Economics 68, 325­351. Pagano, M., Roell, A., Zechner, J., 2002. The geography of equity listing: Why do companies list abroad? Journal of Finance 57, 2651­2694. Rau, R., Vermaelen, T., 1998. Glamour, value and the post-acquisition performance of acquiring firms. Journal of Financial Economics 49, 223­253. Reese, W., Weisbach, M., 2002. Protection of minority shareholder interests, cross-listing in the United States, and subsequent equity offerings. Journal of Financial Economics 66, 65­104. Shleifer, A., Vishny, R., 1986. Large shareholders and corporate control. Journal of Political Economy 94, 461­488. Shleifer, A., Vishny, R., 2003. Stock market driven acquisitions. Journal of Financial Economics 70, 295­311. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817­838. Wurgler, J., 2000. Financial markets and the allocation of capital. Journal of Financial Economics 58, 187­214. Chapter Eight 259 Journal of Financial Economics 72 (2004) 357­384 The effects of government ownership on bank lending$ Paola Sapienzaa,b,* aKellogg School of Management, Northwestern University, 2001 Sheridan Rd., Evanston, IL 60208, USA b CEPR, 90-98 Goswell Road, London EC1V 7RR, UK Received 4 February 2002; accepted 25 October 2002 Abstract This paper uses information on individual loan contracts to study the effects of government ownership on bank lending behavior. State-owned banks charge lower interest rates than do privately owned banks to similar or identical firms, even if firms are able to borrow more from privately owned banks. State-owned banks mostly favor large firms and firms located in depressed areas. The lending behavior of state-owned banks is affected by the electoral results of the party affiliated with the bank: the stronger the political party in the area where the firm is borrowing, the lower the interest rates charged. r 2003 Elsevier B.V. All rights reserved. Keywords: Banking; Government; Ownership JEL classification: G10; H11; L32 $ This paper is a revised version of ``What do state-owned firms maximize? Evidence from Italian banks. I am indebted to Andrei Shleifer for guidance and encouragement. I also thank Alberto Alesina, Paul Armstrong-Taylor, Richard Caves, Riccardo DeBonis, Xavier Freixas, Anil Kashyap, Randy Kroszner, Janet Mitchell, Patricia Ledesma, Anna Paulson, Luigi Zingales, an anonymous referee, and seminar participants at 1999 European Symposium in Financial Markets, NBER Universities Research Conference on ``Macroeconomic Effects of Corporate Finance'', American Finance Association, the Federal Reserve of New York, and the CEPR conference on ``Will Universal Banking Dominate or Disappear? Consolidation, Restructuring and (Re)Regulation in the Banking Industry'' for helpful comments and suggestions. Laura Pisani provided excellent research assistance. All remaining errors are my responsibility. *Corresponding author. Kellogg School of Management, Northwestern University, 2001 Sheridan Road, Evanston, 22 60208, USA. E-mail address: paola-sapienza@northwestern.edu (P. Sapienza). 0304-405X/$ -see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.jfineco.2002.10.002 260 A Reader in International Corporate Finance 358 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 1. Introduction La Porta et al. (2002) document that government ownership of banks is pervasive worldwide. In 1995 state ownership in the banking industry around the world averaged about 41.6% percent (38.5% if we exclude former socialist countries). Mayer (1990) shows that bank financing is the main source of outside financing in all countries. Yet despite the prevalence of government-owned banks in many countries, the prominent role of bank financing, and the importance of efficient financial markets for growth, there is very little evidence on how government ownership affects bank lending. In this paper I use a unique dataset on state-owned banks in Italy, where lending by state-owned banks represents more than half of total lending. Using data on interest rates charged on individual loans, I study the efficiency of the allocation of credit by state-owned banks. Furthermore, I combine data on lending with the political affiliation of the bank and recent election results to study the impact of political power on bank lending behavior. The debate concerning the role of ownership in banking is framed along the three alternative theories of state ownership: social, political, and agency. The social view (Atkinson and Stiglitz, 1980), which is based on the economic theory of institutions, suggests that state-owned enterprises (SOEs) are created to address market failures whenever the social benefits of SOEs exceed the costs. According to this view, government-owned banks contribute to economic development and improve general welfare (Stiglitz, 1993). In contrast, recent theories on the politics of government ownership (Shleifer and Vishny, 1994) suggest that SOEs are a mechanism for pursuing the individual goals of politicians, such as maximizing employment or financing favored enterprises. The political view is that SOEs are inefficient because of the politicians' deliberate policy of transferring resources to their supporters (Shleifer, 1998). The agency view shares with the social view the idea that SOEs are created to maximize social welfare but can generate corruption and misallocation (Banerjee, 1997; Hart et al., 1997). Agency costs within government bureaucracy can result in weak managerial incentives in SOEs. According to this view, the ultimate efficiency of SOEs depends on the trade-off between internal and allocative efficiency (Tirole, 1994). These theories cannot be disentangled by looking at bank profitability: it is not clear whether government-owned banks are less profitable because they maximize broader social objectives, because they have lower incentives, or because they inefficiently cater to politicians' wishes. My empirical strategy addresses these problems. Instead of looking at overall bank performance, where the mix of activities performed by banks might change under government ownership, I focus on the lending relationships of the banks. My data include information on the balance sheets and income statements of over 37,000 Italian firms. The data are collected by Centrale dei Bilanci (CdB), an institution created to provide its members (mainly banks) with economic and financial information for screening Italian companies. For a large subset of the 37,000 companies, CdB members receive a numerical score, which CdB calculates through traditional linear discriminant analysis, to identify the Chapter Eight 261 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 359 risk profile of the companies (Altman, 1968). I merge this information with data on the credit relationships of the firms surveyed in the CdB database. The information in this database is available to all the banks prior to lending and has proven to be very accurate in predicting the success or failure of a company (see Altman et al., 1994). Since both privately owned and state-owned banks have access to the same information, I can use this system to check the differences in the credit policies of the various banks. I look at the individual loan contracts of the two types of banks and compare the interest rate charged to two sets of companies with identical scores that borrow from either state- or privately owned banks, or both. My main result is that, all else equal, state-owned banks charge lower interest rates than do privately owned banks. On average, the difference is about 44 basis points. I claim that this difference can best be explained by the political view of state- ownership. First, my results show that even companies that are able to access private funds benefit from cheaper loans from state-owned banks. Second, companies located in the south of Italy benefit more by borrowing from state-owned banks than do companies located in the north, consistent with the view that political patronage is more widespread in the south (Ginsborg, 1990). This result holds even after controlling for the presence of credit constraints. Finally, contrary to the social view, state-owned banks are more inclined to favor large enterprises. Overall, my results support the political view of government ownership. However, I note that some of these results could also be consistent with some versions of the social or agency views. For example, one could argue that firms located in the south receive cheaper funds because a socially maximizing government wants to channel funds to depressed areas of the country. My findings that larger firms get cheaper funds could also be consistent with the agency view. To further distinguish among the different theories, I analyze the relation between interest rates, the political affiliation of the bank, and electoral results. I find that the lending behavior of state-owned banks is affected by the electoral results of the party affiliated with the bank: the stronger the political party in the area where the bank is lending, the lower are the interest rates charged. This result is not driven by omitted bank and firm characteristics, since I show that it is robust to including both bank and firm fixed effects. Overall, my results support the political view of SOEs and suggest that state- owned banks serve as a mechanism to supply political patronage. These results relate to important policy debates. My findings show that government ownership of banks has distorting effects on the financial allocation of resources. This is consistent with findings that widespread state ownership of banks is correlated with poor financial development (Barth et al., 2000). In turn, a highly politicized allocation of financial resources may have deleterious effects on productivity and growth, as recent research by La Porta et al. (2002) shows. The paper is organized as follows. The next section outlines the theories of SOEs and their predictions. Section 3 provides a description of the institutional environment. Section 4 describes the data, the sample, and the methodology. Section 5 presents the empirical evidence. Section 6 concludes. 262 A Reader in International Corporate Finance 360 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 2. Theoretical issues The three main views of state-owned enterprises--social, agency, and political-- have different implications for both the existence and the role of state-owned banks. The social view sees SOEs as institutions created by social welfare maximizing governments to cure market failures. According to this view, private and state- owned enterprises differ because the first maximize profits and the latter maximize broader social objectives. In this literature, the reason for creating public financial institutions is the existence of market failures in financial and credit markets (Stiglitz and Weiss, 1981; Greenwald and Stiglitz, 1986). Thus, state-owned banks or programs of direct credit have often been justified on the grounds that private banks fail to take social returns into account. For example, private banks might not allocate funds to projects with high social returns or to firms located in specific industries (Stiglitz, 1993). Under the social theory, the objective of state-owned banks should be to channel resources to socially profitable projects or to firms that do not have access to other funds. The agency view shares with the social theory the idea that governments seek to maximize social welfare. Under the agency hypothesis, governments design public financial institutions to cure market failures. However, since SOEs maximize multiple nonmeasurable objectives, managers of SOEs have low-powered incentives (Tirole, 1994). Of course, low powered incentives are not always bad; Laffont and Tirole (1993) show that, under some circumstances, a concern for quality calls for low-powered incentives. But given the incentive problems associated with the control of SOEs, the agency view concludes that decisions on government in-house provision of public goods should depend on the tradeoff between internal and allocative efficiency. Under this hypothesis, state-owned banks channel resources to socially profitable activities, but public managers exert less effort (or divert more resources) than would their private counterparts. The agency view predicts that in general, state-owned banks serve social objectives and allocate resources where private markets fail. However, public managers of state-owned banks exert low effort or divert resources for personal benefits, such as career concerns, with an eye toward future job prospects in the private sector. According to both the social and agency views, the government role in the economy emerges and evolves to perform the economic functions that markets either cannot handle or cannot perform well. Fundamentally different, the political view is based on the assumption that politicians are self-interested individuals who pursue their own personal, political, and economic objectives rather than maximizing social welfare. The main objective of politicians is to maintain voting support. Hence, SOEs provide jobs for political supporters, and direct resources to friends and supporters (Shleifer and Vishny, 1998). According to this view, politicians create and maintain state-owned banks not to channel funds to economically efficient uses, but rather to maximize their own personal objectives. Though the agency and the political views make very different assumptions about government objectives, the difference in the empirical implications is not so clearly defined. The merit of the agency view is to show that misgovernance can exist even Chapter Eight 263 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 361 when the government has the best of intentions (see Banerjee, 1997). Under both views, we would observe some misallocation of resources, but for different reasons. The agency view claims that the misallocation takes place because managers shirk or divert resources for their private use, but under the political view, the misallocation of resources is a political objective, rather than the result of a lack of incentives. State-owned banks will divert resources to areas where there is more political patronage, will finance friends and supporters of politicians, and will maximize political support, e.g., by maximizing employment at the bank level or at other firms. 3. State ownership of banks in Italy Though there are more privately owned banks than state-owned banks in Italy (864 compared to 117), 58% of total assets in Italy were held by state-owned in 1995, among the highest percentages in the industrialized world.1 But Italy is not unique in this dimension in most Continental Europe: in Germany, the proportion is 50%, and in France, it is 36%. Latin American countries also show a very high percentage of state ownership. Data suggest that in general, Italian state-owned banks have a different lending focus from privately owned banks. De Bonis (1998) shows that state-owned banks make more than 11% of their loans to state or local authorities (compared to 1.6% loaned by private banks). The percentage loaned to companies is similar between state- and privately owned banks (55.1% and 57.5%, respectively), but no analysis has investigated the differences within each class of borrowers for the two groups of banks. De Bonis (1998) also finds that state-owned banks are less profitable in making loans than are privately owned banks. In 1995, bad loans represented 57.2% of bank capital for state-owned banks, almost double of that of private institutions (30.2%). The degree of political influence on state-owned banks is evident in the procedure used to appoint the chairpersons and top executives of state-owned banks. Until 1993, the appointments of the directors and management of the banks was made by a specific Parliamentary commission, specifically the Comitato Interministeriale per il Credito e il Risparmio (CICR), a permanent Parliamentary commission in which the political groups are represented according to their relative strength in Parliament. In 1992, for example, this commission met three times: on October 30th it appointed 72 of chairpersons, vice-chairpersons, and CEO of state-owned banks; on December 1st it appointed other 26; and on December 30th an additional 33. Over time there could have been differences in the level of political interference. Some authors (e.g., Barca and Trento, 1994; Ginsborg, 1990) claim that close personal ties between party leaders and the managers of SOEs were introduced after the mid-1950s. De Bonis (1998) claims that the management of state-owned banks became independent from the central government after 1993, when the public entities that previously owned the banks were transformed into foundations. Nonetheless, some observers believe that the practice of political appointments of top executives in state-owned banks has survived (e.g., Visentini, 2000). 1I consider foreign banks operating in Italy as being privately owned. 264 A Reader in International Corporate Finance 362 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 4. Data and methodology The two main databases come from the Company Accounts Dataset (CAD) and the Credit Register (CR) compiled by Centrale dei Bilanci (CdB). The CAD reports balance sheets and income statements for more than 50,000 Italian companies. The CR collects information about any individual loan contracts over 80 million lire (about 41,300 Euro) granted by banks to any customer. This information is readily available to the CdB membership, which is mainly composed of banks. Starting in 1991, CdB also developed what it called the Diagnostic System, which was designed to provide banks with a tool for quickly identifying the soundness of the companies included in the database. This system applies traditional linear discriminant analysis based on two samples of businesses of healthy and unsound companies. A numerical score is obtained from two discriminant functions. This score summarizes the ``risk profile'' of the business. This system has proven to be very successful: it correctly classified in the year immediately prior to distress 87.6% of healthy companies and 92.6% of unsound companies (see Altman et al., 1994). Appendix A provides more details on the numerical score. For each firm, the CR reports the amount of credit granted by each bank, together with the amount used (outstanding balance). In addition, 90 banks (accounting for over 80% of total bank lending) agreed to file detailed information about the interest rates charged on each loan. These data, collected for monitoring purposes, are highly confidential. A subset of CR data includes all the companies that were surveyed for at least one year in the CAD. Data on loan contracts are quarterly, but data on balance sheets and income statements are annual. Aggregate information on bank balance sheets and income statements comes from the Bank of Italy's prudential supervision statistical data, where it is reported on a quarterly basis. I constructed the data on bank ownership by using the Bank of Italy legal classification prior to 1990.2 Local election results for three national elections, 1989, 1992, and 1994, are from the archives of the Interior Ministry. The local unit is the province (similar to U.S. counties). The archives provide the total valid votes and the votes collected by all the parties running in the elections in each of Italy's 95 provinces. I collect the data on the political affiliation of the top management of the bank from newspapers. For 36 state-owned banks, I am able to identify the political affiliation of the chairperson. Appendix B provides additional details about these data and the electoral data. 4.1. Private and state-owned banks The data on loan contracts come from the subset of the CR and CAD datasets previously described. The sample period begins in 1991 and ends in 1995, because 2Prior to 1990, all state-owned banks had a different legal status from private banks (either corporations and cooperatives). After 1990, all the state-owned banks' charters were modified by law and banks were transformed into corporations. Chapter Eight 265 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 363 1991 is the first year in which CdB distributed the information on the score to its members. I restrict my attention to privately owned and state-owned banks that file information about interest payments and are members of the CdB. This criterion introduces a potential sample selection, since the banks that file interest rate information are generally larger than the average Italian bank. However, it turns out that the sample of state-owned banks represents more than 90% of state-owned loans. I exclude small state-owned banks from the sample, but they are not the typical state-owned bank. Privately owned banks selected with these criteria are larger than the average privately owned bank, but because I use them as benchmarks for comparisons with the behavior of state-owned banks, it is important that they are similar in size to the state-owned banks. These selection criteria restrict the total number of banks to 85: 40 have always been privately owned, 43 are state-owned banks, and two were privatized during the period of observation. Table 1 shows descriptive statistics of the banks in the sample. The median state- owned bank has a ratio of nonperforming loans to total loans of 6.91%, as opposed to 5.25% for privately owned banks. The mean for the two sub samples is statistically different at 1% level of significance. State-owned banks also have a lower return on assets (0.34% for the median state-owned bank, as opposed to 0.51% for the median privately owned bank) and higher operating costs relative to assets (3.05% for the median state-owned bank, as opposed to 2.87% for the median privately owned bank). These differences reflect differences between state-owned and privately owned banks and represent a potential problem in comparing their credit policies of these two subsamples of banks. 4.2. Companies borrowing from privately owned and state-owned banks Ideally, I would like to compare the entire loan portfolios of state-owned and privately owned banks and, using the balance sheet and income statement information for the companies, compare the credit decisions of these two types of banks. Unfortunately, the information on firm characteristics is available for only a subset of the companies that receive credit from the banks. To deal with this lack of information, I take a different approach. I compare two matching samples of companies that borrow from state- and privately owned banks, respectively. The advantage of this approach is that I can compare the interest rate charged in the same period to the same company, or to very similar companies, by state- and privately owned banks (many firms have credit ties with both types of banks). Since state- and privately owned banks have access to the same information for evaluating these companies, any difference in the price of the loan is likely to reflect differences in the objectives of the banks, rather than differences in the evaluation skills of the bank's loan officers. To select the sample of companies in this study, I use the following criteria: from the sample of companies included in CdB, I select a subsample of companies that have a loan with at least one bank in the sample for at least one year and for which there is a numerical score. Because loan characteristics (collateral, etc.) 266 A Reader in International Corporate Finance 364 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 Table 1 Summary statistics: the bank sample Panel A shows summary statistics for the sample of privately owned banks (bank­years). Panel B shows summary statistics for state-owned banks. Return on assets is earnings over total assets. Operating costs include wages and other operating costs. Variable Mean Median Std dev. Min Max Obs. Panel A: Privately owned banks Total assets (bill. of lire) 11,856 6,318 16,674 186 110,531 192 Total loans (bill. of lire) 4,917 2,570 6,604 80 44,820 192 Percentage of loans over total assets 42.90 43.00 5.14 30.00 56.00 192 Percentage of nonperforming loans to total loans 6.14 5.25 4.02 1.42 23.63 192 Return on assets (%) 0.46 0.51 0.62 6.47 1.28 192 Operating costs over total assets (%) 3.04 2.87 0.78 1.56 5.96 192 Panel B: State-owned banks Total assets (bill. of lire) 27,314 6,070 40,192 547 188,944 199 Total loans (bill. of lire) 12,292 2,586 18,750 218 79,011 199 Percentage of loans over total assets 41.75 42.00 7.59 24.00 70.00 199 Percentage of nonperforming loans to total loans 8.41 6.91 6.23 1.63 39.55 199 Return on assets (%) 0.28 0.34 0.73 7.19 1.32 199 Operating costs over total assets (%) 3.05 3.05 0.65 1.73 5.65 199 can affect loan rates (Petersen and Rajan, 1994), I focus on homogeneous loan contracts. Specifically, I analyze credit line contracts, the most common loan contract in Italy, in which banks set the amount of the loan and an interest rate. The loans analyzed here exclude long- term, collateralized, and subsidized loans.3 From these data I select the subset of companies that have been borrowing from state-owned banks. For each observation in this sample (company­bank­year) I identify a matching company that borrows from a privately owned bank in the same year. Whenever the company borrows in the same year from both state- and privately owned banks, I match the company with itself. If a company borrows only from state-owned banks, I choose a similar firm that borrows from privately owned banks in the same year. In these cases, I identify the matching company as a firm operating in the same industry and in the same geographical area (north, center, and south), with an identical risk profile based on Altman's z-score (Altman, 1968, 1993), and similar size (measured by sales). I also require that if the company that receives loans from state-owned banks is itself state owned (privately owned), then the matching company must be state owned (privately owned) as well. These selection 3There are other characteristics of the relationship between banks and borrowers that I cannot control for. Although the loan contracts included in the sample have homogeneous characteristics, borrowers might have contemporaneous contracts with the bank (deposits, collateralized loans) that might affect the cost of the loan. Also, the quality of service or the probability that the loan is revoked can vary across banks. Unfortunately, I cannot rule out any of these possibilities; therefore my results should be interpreted with these caveats in mind. Chapter Eight 267 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 365 criteria reduce the sample to a total of 6,968 companies, corresponding to 110,786 company­bank­year observations; 55,393 observations refer to borrowers of state- owned banks and 55,393 to borrowers of privately owned banks. By construction, the companies that borrow from both state- and privately owned banks have identical scores, operate in identical industries, and are located in the same geographical area. Table 9 in Appendix A shows the scores for the two subsamples of companies. The summary statistics on the two subsamples of firms (Table 2) also show that the selected companies borrowing from state- and privately owned banks are very similar. None of the differences in the means of the relevant variables are statistically significant. In both subsamples, the median firm has 58 employees, 21 billion lire in sales, 18 billion lire in assets and a coverage ratio (interest expenses divided by EBITDA) of 1.47. The median firm has a leverage ratio of 71%. I define leverage as the book value of short- plus long-term debt divided by sum of the book value of short- plus long-term debt and the book value of equity. Return on sales is slightly below 8%. The majority of the companies are privately held. Seventy are state- owned companies. Table 2 Summary statistics: the company sample Summary statistics for the two subsamples of company­bank­years. Panel A shows the summary statistics for the subsample of companies that borrow from privately owned banks (company­bank­year). Panel B shows the summary statistics for the companies that borrow from state-owned banks. Total Assets is beginning-of-year total assets in lire. Sales is beginning-of-year sales in lire. Employees is the number of employees at the beginning of the year. Return on sales is earning before interest, taxes, and depreciation ðEBITDAÞ over sales. Age is the number of years since incorporation. Leverage is book value of short- plus long-term debt divided by the sum of book value of short- plus long-term debt and book value of equity. Coverage is interest expense divided by EBITDA (I truncate values above 100 at 100 and values below zero at zero). Variable Mean Median Std dev. Obs. Panel A: Companies borrowing from privately owned banks Total assets (bill.) 101 18 546 55,393 Sales (bill.) 108 21 654 55,393 Employees 231 58 928 54,782 Return on sales 8.48 7.98 7.25 54,799 Age 25 18 36 55,168 Leverage 68.11 70.69 17.59 54,638 Coverage 1.85 1.47 2.56 55,351 Panel B: Companies borrowing from state-owned banks Total assets (bill.) 101 18 547 55,393 Sales (bill.) 108 21 654 55,393 Employees 231 58 928 54,789 Return on sales 8.51 7.95 7.37 54,817 Age 25 18 36 55,173 Leverage 68.12 70.71 17.60 54,646 Coverage 1.86 1.47 2.63 55,349 268 A Reader in International Corporate Finance 366 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 5. Empirical analysis 5.1. Differences in interest rates To learn whether state- and privately owned banks behave differently, I examine the interest rate charged to similar companies by these two types of banks. I first compare the average interest rates charged by both types of banks and then present a regression analysis that controls for bank and firm characteristics. Table 3 reports the average interest rate, minus the prime rate, charged by both types of banks. I define the interest rate as the ratio of the quarterly payments made by the firm to its bank (interest plus fixed fees) to the firm's quarterly loan balance. Of course, this measure of interest rate overestimates the interest rate of a firm with small average balances. For this reason, I eliminate the rates linked to credit lines with less than 50 million lire in average daily balances. The same criterion has been used by Pagano et al. (1998) and Sapienza (2002). Table 3 Interest rates charged by state-owned and privately owned banks by loan risk category I define the interest rate paid by the firm to the bank as the ratio of the quarterly payment (interest plus fees) to its quarterly average balance minus the prime rate. The loan risk category is based on the numerical score of the company (see details in Appendix A). Difference is the average difference between the second column (interest rates charged by state-owned banks) and the third column (interest rates charged by privately owned banks). I test the statistical significance of the difference using the t-statistic with reference to a mean of zero. ÃÃÃ; ÃÃ indicate statistically significant at the 1% and 5% level, respectively. Risk category State-owned Privately Difference Obs. banks owned banks Panel A: Whole sample Highly secure 2.53 2.75 0:22ÃÃÃ 1,420 Secure 2.75 2.97 0:22ÃÃÃ 15,262 Vulnerable 2.84 3.25 0:41ÃÃÃ 409 Highly vulnerable 3.05 3.28 0:24ÃÃÃ 11,743 Uncertainty between vulnerability and risk 3.18 3.43 0:25ÃÃÃ 13,471 Risk of bankruptcy 3.36 3.58 0:22ÃÃÃ 10,472 High risk of bankruptcy 3.69 3.80 0:11ÃÃ 2,616 All borrowers 3.07 3.31 0:23ÃÃÃ 55,393 Panel B: Firms borrowing from both state-owned and privately owned banks Highly secure 2.52 2.74 0:22ÃÃÃ 1,360 Secure 2.72 2.94 0:22ÃÃÃ 13,373 Vulnerable 2.85 3.26 0:41ÃÃÃ 394 Highly vulnerable 3.02 3.27 0:24ÃÃÃ 10,248 Uncertainty between vulnerability and risk 3.15 3.40 0:25ÃÃÃ 11,899 Risk of bankruptcy 3.33 3.54 0:21ÃÃÃ 9,184 High risk of bankruptcy 3.66 3.79 0:13ÃÃ 2,438 All borrowers 3.05 3.27 0:23ÃÃÃ 48,896 Chapter Eight 269 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 367 Column 3 of Table 3 presents the differences in rates for the two subsets of banks. For the overall sample, these comparisons show that for similar companies, the average interest rate charged by state-owned banks is 23 basis points lower than that charged by private banks. The differences are statistically significant in a t-test. Table 3 also presents the differences in interest rates for various risk profiles of the companies, demonstrating that these differences are not driven by few outliers. Also, to make sure that the differences are not driven by any incorrect matching, Panel B of Table 3 presents the same statistics for the subsample of firms that borrow from both state- and privately owned banks during the same year (the matching firm is itself). My main finding is that for any risk category, the interest rate charged by state- owned banks is lower than that charged by privately owned banks. The difference in interest rates is statistically significant at the 1% level. However, the comparisons presented above are not conditioned on other characteristics of the banks, such as differences in the size and riskiness of the portfolio. Table 1 shows that state- and privately owned banks are different in size, profitability, and riskiness. To address this issue, I use a regression model to estimate the difference in interest rates charged by the two types of banks. I regress rikt; the relative interest rate charged at time t by bank k to company i (defined as the interest rate minus the prime rate) on a dummy variable, STATEk ; ;t that equals one if at time t bank k is a state-owned bank. The coefficient measures the impact of state ownership on interest rates. A negative (positive) value means that state-owned banks charge a lower (higher) interest rate than do privately owned banks. I also include several regressors to control for firm, market, and bank characteristics. Finally, I include a vector of time fixed effects and a vector of firm fixed effects. By using a firm fixed effect, I compare the interest rate charged by various banks to the same company. Panel A of Table 4 reports the regression results. Heteroskedasticity-robust standard errors are shown in parentheses. I also adjust the standard errors for within-year clustering. Column 1 reports the estimates of the interest rate regressed on the STATEk dummy and time and firm fixed effects. These results are directly ;t comparable to the simple differences in the last row of Table 3: state-owned banks charge interest rates 23 basis points lower than do privately owned banks. As mentioned before, the coefficients measuring state ownership might capture specific characteristics of state-owned banks and local market structure. To overcome this problem, Columns 2­5 of Table 4 include several other controls. In column 2, I introduce a proxy for the size of the bank, measured by the logarithm of the bank's total assets. Aside for the role size plays in determining market concentration measures, a bank's size should affect prices according to the theoretical literature. For example, in a standard Cournot model with capacity constraints (increasing returns to scale), the bank with lower capacity would supply loans equal to capacity at a lower price than the other bank with higher capacity (see Tirole, 1989). Size could also reflect some implicit characteristics of the loan. Loans from large banks might carry an implicit guarantee of not being revoked, if large banks are perceived to be less likely to fail. 270 A Reader in International Corporate Finance 368 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 Table 4 Interest rates charged by state-owned and privately owned banks The dependent variable is the interest rate charged to firm i by bank k at time t minus the prime rate at time t: STATEk is a dummy variable equal to one if at time t bank k is a state-owned bank. I measure the ;t size of the bank by logarithm of total assets. The percentage of nonperforming loans is the ratio of non- performing loans to total loans. I measure market concentration at the province level by the Herfindahl- Hirschman Index (HHI) on total banking lending. The size of the firm is the logarithm of sales. All regressions include year and firm dummies. Heteroskedasticity-robust standard errors are in brackets. The standard errors are corrected for within-year clustering.ÃÃÃ; ÃÃ indicate statistically significant at the 1% and 5% level, respectively. The table also reports the p-value of an F-test for the hypothesis that the joint effect of all the variables equals zero. Panel A reports the results for the whole sample. Panels B and C report the results for the subsample of firms that borrow from both state-owned and privately owned banks, respectively. (1) (2) (3) (4) (5) Panel A Statek;t 0:2378ÃÃÃ 0:4589ÃÃÃ 0:5019ÃÃÃ 0:4417ÃÃÃ 0:4424ÃÃÃ (0.0274) (0.0166) (0.0180) (0.0218) (0.0218) Size of the bank 0:1936ÃÃÃ 0:1730ÃÃÃ 0:1723ÃÃÃ 0:1728ÃÃÃ (0.0078) (0.0037) (0.0041) (0.0040) Percentage of nonperforming loans 0:0337ÃÃÃ 0:0338ÃÃÃ 0:0336ÃÃÃ (0.0014) (0.0014) (0.0015) Concentration of loans (HHI) 2:6681ÃÃÃ 3:0753ÃÃÃ 2:8267ÃÃÃ (0.4417) (0.3514) (0.4197) Concentration of loans if Statek ¼ 1 ;t 0:8677ÃÃÃ 0:8561ÃÃÃ (0.3223) (0.3183) Size of the firm 0:2453ÃÃÃ (0.0051) Score of the firm 0:0365ÃÃÃ (0.0081) Firm fixed effect Yes Yes Yes Yes Yes Time fixed effect Yes Yes Yes Yes Yes Observations 110,786 110,786 110,752 110,752 110,752 Adjusted R-squared 0.407 0.420 0.425 0.425 0.428 p-Value of F-test for total effect equal to zero 0.0000 0.0000 0.0000 0.0000 0.0000 Panel B Statek;t 0:2293ÃÃÃ 0:4510ÃÃÃ 0:4980ÃÃÃ 0:4374ÃÃÃ 0:4376ÃÃÃ (0.0261) (0.0185) (0.0168) (0.0221) (0.0220) Size of the bank 0:1895ÃÃÃ 0:1694ÃÃÃ 0:1689ÃÃÃ 0:1691ÃÃÃ (0.0087) (0.0042) (0.0046) (0.0045) Percentage of nonperforming loans 0:0341ÃÃÃ 0:0343ÃÃÃ 0:0340ÃÃÃ (0.0014) (0.0014) (0.0014) Concentration of loans (HHI) 2:8282ÃÃÃ 3:2309ÃÃÃ 2:9066ÃÃÃ (0.4835) (0.4297) (0.4958) Concentration of loans if Statek ¼ 1 ;t 0:8756ÃÃÃ 0:8685ÃÃÃ (0.3341) (0.3322) Size of the firm 0:2648ÃÃÃ (0.0065) Score of the firm 0:0335ÃÃÃ (0.0077) Firm fixed effect Yes Yes Yes Yes Yes Chapter Eight 271 P. Sapienza / Journal of Financial Economics 72 (2004) 357­384 369 Table 4. (Continued) (1) (2) (3) (4) (5) Time fixed effect Yes Yes Yes Yes Yes Observations 97,792 97,792 97,760 97,760 97,760 Adjusted R-squared 0.407 0.420 0.425 0.423 0.427 p-Value of F-test for total effect equal to zero 0.0000 0.0000 0.0000 0.0000 0.0000 Panel C (1) (2) (3) Statek;t 0:4402ÃÃÃ 0:4382ÃÃÃ 0:4373ÃÃÃ (0.0220) (0.0225) (0.0220) Size of the bank 0:1690ÃÃÃ 0:1691ÃÃÃ 0:1691ÃÃÃ (0.0045) (0.0045) (0.0045) Percentage of nonperforming loans 0:0340ÃÃÃ 0:0340ÃÃÃ 0:0340ÃÃÃ (0.0014) (0.0014) (0.0014) Concentration of loans (HHI) 2:9151ÃÃÃ 2:9104ÃÃÃ 2:9071ÃÃÃ (0.4914) (0.4976) (0.4946) Concentration of loans if the bank is state-owned 0:8838ÃÃÃ 0:8751ÃÃÃ 0:8672ÃÃÃ (0.3284) (0.3258) (0.3323) Size of the firm 0:2646ÃÃÃ 0:2647ÃÃÃ 0:2649ÃÃÃ (0.0065) (0.0066) (0.0064) Score of the firm 0:0335ÃÃÃ 0:0335ÃÃÃ 0:0334ÃÃÃ (0.0076) (0.0076) (0.0077) Statek ¼ 1 if the firm has more than 8% of credit line usage ;t 0.0150 (0.0127) Statek ¼ 1 if the firm has more than 15% of credit line usage ;t 0.0214 (0.0303) Statek ¼ 1 if the firm has more than 37% of credit line usage ;t 0:0364 (0.0456) Firm fixed effect Yes Yes Yes Time fixed effect Yes Yes Yes Observations