Banking, Capital Markets and Corporate Governance

Banking, Capital Markets and Corporate Governance Edited by Hiroshi Osano and Toshiaki Tachibanaki

Banking, Capital Markets and Corporate Governance

Also by Hiroshi Osano ECONOMICS OF CORPORATE GOVERNANCE (in Japanese)

Also by Toshiaki Tachibanaki CAPITAL AND LABOUR IN JAPAN THE ECONOMIC EFFECTS OF TRADE UNIONS IN JAPAN FROM AUSTERITY TO AFFLUENCE WHO RUNS JAPANESE BUSINESS? PUBLIC POLICIES AND THE JAPANESE ECONOMY WAGE DETERMINATION AND DISTRIBUTION IN JAPAN WAGE DIFFERENTIALS: An International Comparison INTERNAL LABOUR MARKETS, INCENTIVES AND EMPLOYMENT LABOUR MARKET AND ECONOMIC PERFORMANCE SAVINGS AND REQUESTS

Banking, Capital Markets

and Corporate Governance

Edited by

Hiroshi Osano Professor of Economics Institute of Economic Research Kyoto University Japan and

Toshiaki Tachibanaki Professor of Economics Institute of Economic Research Kyoto University Japan

Editorial matter and selection and Chapter 1 © Hiroshi Osano and Toshiaki Tachibanaki 2001 Chapter 4 © Hiroshi Osano 2001 Chapter 5 © Toshiaki Tachibanaki and Hideo Okamura 2001 Chapters 2, 3, 6–12 © Palgrave Publishers 2001 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP. Any person who does any unauthorised act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The authors have asserted their rights to be identified as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2001 by PALGRAVE Houndmills, Basingstoke, Hampshire RG21 6XS and 175 Fifth Avenue, New York, N. Y. 10010 Companies and representatives throughout the world PALGRAVE is the new global academic imprint of St. Martin’s Press LLC Scholarly and Reference Division and Palgrave Publishers Ltd (formerly Macmillan Press Ltd). ISBN 0–333–77136–2 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Banking, capital markets, and corporate governance / edited by Hiroshi Osano and Toshiaki Tachibanaki. p. cm. “Preliminary versions of the papers in this book were

presented at the Conference, ‘Banking, capital markets, and

corporate governance’, held at Lake Biwa, in July 1998”—Pref.

Includes bibliographical references and index.

ISBN 0–333–77136–2

1. Banks and banking. 2. Capital markets. 3. Corporate governance. I. Osano, Hiroshi, 1955- II. Tachibanaki, Toshiaki, 1943– HG1521 .B365 2001

332.1—dc21

2001021886 10 10

9 09

8 08

7 07

6 06

5 05

4 04

3 03

2 02

1 01

Printed in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire

Contents Preface

vii

List of Contributors

viii

1

Introduction Hiroshi Osano and Toshiaki Tachibanaki

1

Part I Banking 2

3

4

5

6

Liquidity Demand of the Corporate Sector and Soft Budget Constraint Noriyuki Yanagawa Incentive Effects of Conditional Bank Recapitalisation: Lending and Disclosure of Non-Performing Loans Philippe Aghion, Patrick Bolton and Steven Fries

13

31

Injection of Public Funds into Banks under Deposit Insurance and Bank Regulation Hiroshi Osano

51

Governance Structure of Banks and Their Business Performance Toshiaki Tachibanaki and Hideo Okamura

85

A Vacuum of Governance in Japanese Bank Management Masaharu Hanazaki and Akiyoshi Horiuchi

Part II

133

Corporate Governance

7

Corporate Bankruptcy Michelle J. White

183

8

Executive Option Plans and Incentives to Take Risk in Levered Firms: Equity Value or Firm Value Maximisation? Gerald T. Garvey and Amin Mawani

v

204

vi Contents

9

Does the Decision to Retain Retiring Executives on the Board of Directors Help to Control Agency Problems in American and Japanese Firms? James A. Brickley, Jeffrey L. Coles and James S. Linck

Part III

233

Capital Market

10 Viable Design of a Security with a Pre-existing Market  Kazuhiko Ohashi

253

11 Investment, Security Design and Information Gabrielle Demange and Guy Laroque

272

Part IV

General Comments

12 General Comments Masahiro Okuno-Fujiwara

291

Index

297

Preface

Preliminary versions of the papers in this book were presented at the Conference, `Banking, Capital Markets and Corporate Governance', held at Lake Biwa in July 1998. The conference was organised by Hiroshi Osano and Toshiaki Tachibanaki, and ®nanced by the Kansai Economic Research Centre and the Suntory Foundation. We are indebted to these two institutions for their ®nancial support. This volume is the third English publication in the Biwako Conference series. The ®rst was published in 1994, Labour Market and Economic Performance, and the second in 1997, Internal Labour Markets, Incentives and Employment. Both were published by Macmillan Press. We are grateful to Macmillan (now Palgrave) for continuing to publish the Biwako Conference series. The Biwako Conference has been held every year for the more than thirty-®ve years, and is known as one of the most prestigious conferences for economists in Japan. Capital and ®nancial markets, the subjects of this book, are important because they collect funds from one place, and allocate them to another. The banking sector is an important participant in these markets. It is vital for any country that both the banking sector and capital market work ef®ciently so that capital and money are effectively utilised. Corporate governance is also crucial because it affects a ®rm's performance. Banking, capital market and corporate governance are all interrelated; we investigate them extensively. Finally, the editors would like to express their gratitude to the following individuals, Messrs Yasuo Shingu (the former President of the Kansai Economic Research Centre), Yoshihisa Akiyama (the President of the Centre), Katsutoshi Kojima (the Executive Director of the Centre), Shinichiro Torii (the President of the Suntory Foundation), Kenji Sugiya (the Executive Director of the Foundation), and Professors Chikashi Moriguchi (Tezukayama University) and Toshihisa Toyoda (Kobe University), for their advice and support. HI R O S H I OS A N O TO S H I A K I TA C H I B A N A K I

vii

List of Contributors

Philippe Aghion: Professor of Economics, Department of Economics, University College, London, UK Patrick Bolton: Professor of Economics, Department of Economics, Princeton University, USA James A. Brickley: Gleason Professor of Business Administration, Simon School of Business, University of Rochester, USA Jeffrey L. Coles: Professor of Finance, College of Business, Arizona State University, USA Gabrielle Demange: Professor of Economics, DELTA, Paris, France Steven Fries: Director of Policy Studies, European Bank for Reconstruction and Development, London, UK Gerald T. Garvey: Professor of Economics and Finance, Peter F. Drucker School of Management, Claremont Graduate University, USA Masaharu Hanazaki: Professor of Economics, Institute of Economic Research, Hitotsubashi University, Tokyo, Japan viii

List of Contributors ix

Akiyoshi Horiuchi: Professor of Economics, Faculty of Economics, University of Tokyo, Japan Guy Laroque: Directeur des Etudes et SyntheÁses Economiques INSEE, France James S. Linck: Assistant Professor of Business Administration, Terry College of Business Administration, University of Georgia, USA Amin Mawani: Professor of Accounting, School of Business Administration, York University, Canada  Kazuhiko Ohashi: Associate Professor of Finance, Graduate School of International Corporate Strategy, Hitotsubashi University, Japan Hideo Okamura: Lecturer in Finance, School of Business Administration, Kwansei Gakuin University, Idyogo, Japan Masahiro Okuno-Fujiwara: Professor of Economics, University of Tokyo, Japan Hiroshi Osano: Professor of Economics, Institute of Economic Research, Kyoto University, Japan Toshiaki Tachibanaki: Professor of Economics, Institute of Economic Research, Kyoto University, Japan Michelle J. White: Professor of Economics, Department of Economics, University of California at San Diego, USA

x List of Contributors

Noriyuki Yanagawa: Associate Professor, Department of Economics, University of Tokyo, Japan

1

Introduction Hiroshi Osano and Toshiaki Tachibanaki

1.1

Motivation for this book

Issues surrounding the fragility of the banking system, corporate governance and the increasing securitisation of corporate ®nancing have gained enormous practical signi®cance over recent decades. All are closely interrelated. For example, the fragility of the banking system can be caused by dysfunctional corporate governance mechanisms in banks or their borrowing ®rms, and can be aggravated by increasing reliance on securitisation in corporate ®nancing. On the other hand, the breakdown of the banking system can create a need for the replacement of the existing system of corporate governance, and for further increasing the securitisation of corporate ®nancing. The increasing securitisation of corporate ®nancing can also affect the existing system of corporate governance because it changes the traditional methods of corporate ®nancing through banks. To explore these complicated problems, this book has three aims. The ®rst is to consider and evaluate the role of banking policy in banking crises. Since the early 1980s, many countries ± including not only emerging countries in East Asia and South America, but also developed countries, such as Japan, Northern European countries and the United States ± have suffered from the fragility of their banking systems. In banking crises, policymakers face the serious problem of whether or not private assets provide suf®cient liquidity for ef®cient functioning of the productive sector. If private assets cannot provide enough liquidity, the government needs to play an active role in supplying the de®cient liquidity to the productive sector. To attain this objective, they also need to determine how the liquidity asset holdings of the private sector affect private sector activity. These questions are concerned with several

1

2 Introduction

recently controversial topics: the injection of cash funds into insolvent banks, credit crunch, in¯ation targeting in recessions, and extremely loose monetary policies to increase liquidity, such as the `zero interest rate' policy in Japan. Another serious problem in banking crises is that banks are likely to rollover loans in default, or to underwrite more risky activities, in order to hide their poor ®nancial condition and gamble on hopes of resurrection. The intuitive mechanism behind this moral hazard action is explained as follows. In banking crises, the regulator can choose a variety of policies, such as bank closures or bank rescues. A tough bailout policy is likely to demand that insolvent banks must dismiss their managers. This is so because a soft bailout policy under which managers of insolvent banks retain their management positions causes formidable political disputes and creates future moral hazard problems when bank managers take on risky investments. Thus, managers of insolvent banks are afraid of the tough bailout policy and may be induced to rollover loans in default, or gamble on resurrection by ®nancing more risky activities to hide the extent of their banks' loan losses. As a result, understanding how a banking policy in banking crises affects incentives for managers of failing banks to restructure the banks' assets is important in understanding the effectiveness of such policies in banking crises. The problem of corporate governance of ®nancial intermediaries is also relevant in preventing the breakdown of the banking system. This problem is particularly important in Japan and some continental European countries because banks play a major role in monitoring the management of their borrowing ®rms, and act as an important shareholder in these countries. If bank managers are not provided with proper incentives, they may be induced to allow their borrowing ®rms to make too many risky investments. Thus, it is often stated that lack of effective governance in bank management may have caused recent banking crises. In order to discuss the problem of banking crises, we, therefore, need to evaluate the corporate governance mechanisms of the banking industry, and ask who monitors the monitor, that is, the bank. This issue also needs to be investigated, together with the monitoring mechanisms used by public administration, because regulation mechanisms have an important effect on the strategies of banks. Beyond the investigation of corporate governance mechanisms of banks, there is a great deal of discussion about how good or bad the existing governance mechanisms are of non-®nancial ®rms in industrialised countries. These arguments are provoked by the fact that each industrialised country has a different system of corporate governance that

Hiroshi Osano and Toshiaki Tachibanaki 3

disciplines the management of ®rms: e.g., the United States and the United Kingdom have a market-oriented system of corporate governance, whereas Germany and Japan have a bank-oriented one. Indeed, corporate governance mechanisms consist of economic and legal institutions, such as the design of bankruptcy procedures, the allocation of control to the board of directors, policies for the retention of retiring chief executive of®cers (CEOs) and other executives, the design of managerial compensation contracts, and so on. Also relevant are economic and legal institutions concerning controlling rights and property rights. The second purpose of this book is to explore how these economic and legal institutions ± particularly bankruptcy laws, managerial compensation contracts, and policies for retaining retiring executives (on the board of directors) ± can regulate the management of non®nancial ®rms. The basic problem studied in this book is made up of the questions of how to allocate control rights among various stakeholders in the event of default, and how to design managerial compensation arrangements to deal with the corporate governance problem. The latter question is closely concerned with the effectiveness of different stock option arrangements and policies for retaining retiring executives (on the board of directors), as performance incentives for executives. Finally, the third purpose of this book is to discuss the effects of signi®cant increases in the securitisation of corporate ®nancing and the amount of ®nancial innovation, which have recently been observed in many developed countries. These two ®nancial phenomena bring a large number of various securities to the ®nancial market: asset backed securities (ABS), collateralised mortgage obligations (CMO), mortgage backed securities (MBS), ¯oating rate bonds, zero coupons, swaps, options, indexed securities, money market funds, and so on. These new securities, as well as traditional securities, such as equities, preferred stocks, bonds, convertible bonds, warrants, and commercial papers, are issued to satisfy different kinds of motives. In the presence of informational asymmetries between ®rms and suppliers of capital, ®rms face an adverse selection problem in that ®rms with highly productive investment opportunities may be imitated by ®rms with lower productive investment opportunities. In this situation, the ingenious design of securities can promote the dissemination of information about risky projects: ®rm managers issue securities to signal the ®rm's potential value and opportunities, or their efforts and abilities. The ingenious design of securities can also resolve the corporate governance problem by enhancing the role of the large investor (rather than small investors) as a delegated monitor who disciplines the performance of a company's

4 Introduction

management. Furthermore, the creation of new securities can provide investors with an opportunity to trade previously unavailable contingent claims. Hence, the problem of security design affects the behaviour not only of issuing ®rms but also of investors and ®nancial intermediaries, including venture capitalists. Some papers in this book examine how the security design can be organised to satisfy the needs of various participants in the ®nancial market under asymmetric information.

1.2

Content of the book

This book consists of three main parts. Part I examines the issue of banking both theoretically and empirically. Part II discusses corporate governance, including bankruptcy. Part III explores capital market issues, in particular security design. Although we have classi®ed all contributions into three parts, all are related to each other, for obvious reasons. Banking, capital markets, and corporate governance are interrelated issues. The classi®cation of papers into three parts is only formal, and none should be regarded as completely independent from the others. All papers deal with banking, capital markets and corporate governance. Part I consists of ®ve papers. Yanagawa (Chapter 2) investigates the relationship between liquidity demands by corporate sector and investment decisions. Since asset prices ¯uctuate, investment activity is affected by a change in liquidity demand. The demand for liquidity is discussed from the viewpoint of the agency problem between entrepreneurs and investors. In particular, the moral hazard problem created for entrepreneurs is analysed. Liquidity asset holdings impact on investments in two ways. The ®rst positive effect is that the liquidity holder can absorb a liquidity shock, and so can continue a necessary project. The second, a negative effect, is that the liquidity helps to bail out the project. This is similar to the soft budget constraint problem. The difference between the ex ante shock and the ex post shock makes determining the impact of different liquidity holding levels problematic. Yanagawa proposes that it is dif®cult to control the liquidity level when the value of liquidity ¯uctuates, despite the fact that ®rms want to keep the optimal level of liquidity at the time of shock. In such a case, it is dif®cult to obtain suf®cient funds to implement an initial project, and, thus, investment levels should decline. Then, Yanagawa examines the endogenous determination of the price of liquidity assets and investments. He obtains several interesting ®ndings regarding the relationship

Hiroshi Osano and Toshiaki Tachibanaki 5

between asset prices and investment levels. Finally, he examines the role of credit lines offered by ®nancial intermediaries. We have seen large scale banking crises over recent years, including in Japan. The paper by Aghion, Bolton and Fries (Chapter 3) examines the causes and outcomes of the Japanese banking crisis theoretically. The writers present their model, called the Aghion, Bolton and Fries (ABF) model, to help understand banking crises in transition economies. They claim here that their analysis is also applicable to Japan, with several minor modi®cations. They argue that banks in transition economies, as well as in Japan, are run by managers who are not the actual (or sole) owners of the bank, and so may run the bank in their own self-interest. Such an emphasis on self-interest and superior information (i.e., informational asymmetries) by these managers in determining the quality of loans is likely to produce unintended and undesirable effects: insolvent banks can respond by simply hiding their insolvency, thus prolonging and magnifying the crisis. We know well that this phenomenon can be found in both transitional economies and in Japan. Aghion, Bolton and Fries provide us with an elegant theoretical interpretation of actual stories from recent crises. The main reasoning behind their result may be summarised as follows. First, a tough recapitalisation policy in which bank managers are dismissed results in the bank managers rolling over bad loans in order to conceal loan losses, and, thus, in the softening of the ®rms' budget constraints. Second, however, a socially ef®cient outcome can generally be achieved through a recapitalisation policy combined with the carving out of bad loans at a suitable non-linear transfer price. Finally, it is noted that regulatory authorities in these countries are fragile for various reasons. Osano (Chapter 4) discusses another issue related to those canvassed in the paper by Aghion, Bolton and Fries, concerning the rescue of insolvent banks in banking crises. Osano considers theoretically the situation arising under a deposit insurance system when regulators inject public funds into a bank in trouble. Since it is assumed that the regulator is unable to observe the efforts of a bank to reduce liquidity shocks arising from the deterioration of the bank's assets, it is desirable to inject public funds into the bank in order to construct incentive mechanisms for overcoming the moral hazard problem. It is noted that the two papers (Chapter 3 and 4) both discuss the moral hazard problem faced by banks. Osano provides us with several reasons for injecting public funds into a bank, noting that the effect of such injections will depend on several

6 Introduction

factors. For example, different policies are recommended, depending upon the procedure used to implement the injection. There are several procedures that can be used, such as (1) the purchase of subordinated bonds, (2) the purchase of preferred stocks, and (3) the purchase of common stocks. Osano presents the optimal security design amongst these three procedures, noting that the choice will depend upon the degree of information about the bank held by the regulator. He, nevertheless, proposes that inef®cient banks should be closed, even if there remains the option of injecting public funds, because the injection policy cannot be independent of the bank closure policy. It is vital for the regulator to determine quickly whether a bank in trouble should be closed, or whether injection of public funds into the bank is desirable. The paper by Tachibanaki and Okamura (Chapter 5) investigates the role of shareholders in determining banks' corporate governance structures under the condition of intercorporate shareholding, which is common in Japan. In other words, they are interested in whether or not banking activity bene®ts from the intercorporate shareholding, which is, probably, the most distinguishing feature of capital markets in Japan. The other reason why intercorporate shareholding is crucial in the banking sector is that debt holders (i.e., individual savers) are not interested in governing or monitoring banks' behaviour. Shareholders are the only agency interested in governing banks. Tachibanaki and Okamura estimate the productivity function and the cost of capital for individual banks. The productivity function is explained by various explanatory variables including the structure of shareholding, asset values, employment, etc. It is estimated for panel data. The cost of capital is estimated under the condition of intercorporate shareholding. This is the ®rst attempt to conduct such an analysis for Japan. They obtain the following results. Intercorporate shareholding has had different impacts on productivity from time to time. The role of life insurance companies, which are the most important source of intercorporate shareholders, is peculiar. The reason for this peculiarity arises from the fact that the great majority of life insurance companies in Japan are mutual companies. The estimation of the cost of capital for banks suggests the following two observations. The ®rst is that differences in the cost of capital for various banks are growing because of widening differences in productivity. The second is that the overall cost of capital has been in a declining trend. Hanazaki and Horiuchi (Chapter 6) present an overview of the governance structure in the Japanese banking industry. They conclude

Hiroshi Osano and Toshiaki Tachibanaki 7

that the management has been independent of outsiders' control. Therefore, there was no effective governance. They even name it the `vacuum of governance'. Their judgement supports the views offered in the Aghion, Bolton and Fries' paper (Chapter 3), and partly supplements the Tachibanaki and Okamura paper. Nearly all countries, from developed countries to developing countries and transitory economies, experienced banking crises during the 1980s and 1990s. Hanazaki and Horiuchi, nevertheless, propose that the banking crisis in Japan is unique and more serious than those experienced in other countries, and attribute this to de®ciency of effective governance in the banking industry. More concretely, they propose that the following three disciplinary mechanisms have not worked effectively: (1) the capital market disciplinary function, (2) effective competition in the industry, and (3) supervision of the regulatory authorities. Empirical evidence is presented to demonstrate the ineffectiveness of these mechanisms, and the authors canvass a variety of factors, including the disclosure of nonperforming loans, the danger of the vicious circle, the safety net in the banking industry, the dangers of the forbearance policy, the insuf®ciency of deposit insurance, the limitations of the traditional rescue method, delayed deregulation, and conspiratorial relationships between the regulator and banks (i.e., Amakudari). Then, the authors conduct an international comparative study based on the regression analysis, using the ®gures obtained, and conclude that the behaviour of the Japanese banking industry did not change in the ®rst half of the 1990s, despite the depression. They also calculate the ®gure for the Japan premium. Part II discusses corporate governance, including bankruptcy. White (Chapter 7) presents a survey of bankruptcy procedures in the US and European countries, the UK, Germany, and France in particular, and canvasses several economic issues and tradeoffs in bankruptcy. She describes the bankruptcy liquidation and bankruptcy reorganisation procedures in these countries, and ®nds that they are basically similar, with some differences between the US and Europe. Even in the US, there are several procedures; most importantly, the differences between Chapter 7 and Chapter 11 are substantial. Since most large ®rms in the US adopt reorganisation rather than liquidation, there are many empirical studies on the former. She summarises these studies. White then discusses various economic issues and tradeoffs in bankruptcy, based on a large number of case studies and examples. The subject covered is wide. She looks at the race between creditors attempting to be the ®rst to collect, and shows that it is an example of the prisoner's

8 Introduction

dilemma. Managers of failing ®rms may have an incentive to delay bankruptcy. Moreover, if there is a possibility of reorganising in bankruptcy, an additional incentive appears to obtain partial forgiveness of debt under Chapter 11. In these circumstances, various stakeholders, such as managers, creditors, and equity holders, bargain openly, and a result is given based on the sequential bargaining model. White proposes that there is an ef®ciency justi®cation for having two separate bankruptcy procedures: liquidation and reorganisation. The decision between these two will be based on differences in ®rms' levels of economic ef®ciency and ®nancial distress, and on the difference between type I and type II errors. Filtering failure occurs when the bankruptcy procedure generates either type of error. She elegantly analyses the above tradeoffs for various countries, and provides us with several useful interpretations, covering topics such as non-bankruptcy workouts, costly bankruptcies, shirking, and under-investment. The paper by Garvey and Mawani (Chapter 8) examines the issue of executive stock options used as incentives in leveraged ®rms. Since risky investment decisions are made by managers not shareholders, it is important to examine managerial incentives, in the same way as the `asset substitution problem' has already been investigated in relation to shareholders and bondholders. The authors present a direct test, examining the relationship between capital structure and the risk incentives in CEO stock option contracts. Garvey and Mawani show that by adjusting the exercise price for the option, executive stock option plans can be designed to control the manager's incentive to take risks without diluting her incentives to exert effort or otherwise increase ®rm value. Their theoretical framework is an extension of the John and John analysis, and the Yermack analysis, and the results they obtain are consistent with these other analyses, although some differences remain. The authors' empirical tests show that shareholder and executive leverage levels were negatively related over the entire ®rm sample used, even in the 2SLS estimates that allowed for endogenous capital structure choices. This is veri®ed by examining the relationship between managerial compensation and equity risk using a variety of basic option-pricing techniques. The results support the hypothesis that incentive contracts mitigate risk-taking incentives in leveraged ®rms because a given change in equity risk implies a larger increase in underlying asset risk as ®rm leverage increases. As noted above, this result is obtained in the simultaneous equation model for endogenous capital structure choices. Finally, they consider a more complete measure of managerial risk

Hiroshi Osano and Toshiaki Tachibanaki 9

attitudes. They ®nd even stronger evidence that leverage does not increase managers' inclination to take on risky projects. Brickley, Coles and Linck (Chapter 9) examine the issue of agency problems of retiring executives who can be retained on the board of directors in both American and Japanese ®rms. Their study is interesting and useful in view of the many studies that have documented the similarities and differences between American and Japanese management styles, including the behaviour of top executives. There is an agency problem (i.e., potential con¯icts of interest) between top managers and shareholders. Similar agency problems and con¯icts of interest between external and internal corporate governance stakeholders also arise in areas such as corporate takeovers, executive compensation and turnover, bank and block ownership, and the board of directors. The three authors here concentrate on top executives who can be retained on the board of directors. There are two conditions for retaining top managers. The ®rst is that managers must value board seats. The second is that the retention decision must be based on past performance. The paper focuses on the second condition. Brickley, Coles and Linck ®rst examine existing studies for both Japan and the United States. They suggest that the board retention decision in Japan and the US is determined by past performance. Simply, there is a high correlation between retention rate and performance. Then, they conduct their own research, by estimating the logit model for the retention rate. They ®nd that absolute performance is more important in explaining the retention decision than relative performance, in contrast to the prediction made in standard agency theory. Part III looks at capital market issues; in particular, security design. Two papers in this part are theoretical and fairly rigorous. They, nevertheless, have several practical implications for venture capital and security design. Â hashi (Chapter 10) tackles the issue of ®nancial innovation in the O context of securitisation. He argues that the design of a new security is crucially affected by the availability of pre-existing securities, and the correlation between payoff levels for these securities and the value of the underlying assets backing the newly-created security. Next, he shows that the informed investor could trade the pre-existing securities freely to hedge that part of the risk in the created security about which he has no private information. This enables us to facilitate the innovation of new securities, and can provide another role for short-sales constraints to play in ®nancial innovation. Â hashi then takes the case of the pre-existing security being a TO bond and the created security an MBS. It is likely that the pre-existing

10 Introduction

security is traded competitively, while the created security is traded strategically. Finally, he discusses the implications of his analysis for the relationship between ®nancial innovation and investment decision making. It is possible to understand that the pre-existing security represents market index futures, such as S&P 500 or Nikkei 225 in the case of an IPO, while the created security represents the IPO stock. The result implies that the IPO stock cannot be sold if its payoff is not viable, and, thus, the entrepreneur can invest only in projects that generate viable stock returns. The paper by Demange and Laroque (Chapter 11) considers a venture capitalist who undertakes risky projects and ®nances these projects. They show how expected distortions in the functioning of the stock markets, and the nature of the risks to be hedged, in¯uence the ex ante investment and security-design decisions of venture capitalists. They consider a two-stage model for investigating the above problem. The ®rst stage evaluates different project choices, and looks at how different types of investment are ¯oated on the stock market. In the second stage after an interim stage, the securities are exchanged against the sure (numeraire) good on a competitive basis. They consider three groups of competitive agents who trade in the market, and two speci®cations for the information received by the risk adverse public, depending upon the degree of risk aversion and on information availability. Under the usual CARA gaussian model, which assumes that ®rms are rational decision-makers, Demange and Laroque show that the ex-ante utility is the product of two terms, namely speculative gains and insurance gains. The maximisation of the insurance gains suggests that the venture capitalist forms a single ®rm if risk adverse traders are informed, while their hedging needs must be taken into account if they are uninformed. The authors then analyse the optimal capital stock choice for the capitalist under various combinations of information signals available at trading, and the hedging needs of risk adverse traders. Finally, they evaluate their results in relation to other studies discussing security design, initial public offerings, and the incompleteness of markets. Although the theoretical results in this paper are reached through highly technical calculations, the practical implications are valuable in venture capital decisions, and, thus, for new ®rms. The last message prepared by Masahiro Okuno provides an overall discussion and evaluation of all papers and the conference in general. We invited him to attend the conference, and to provide us with his comments. His message summarises the conference elegantly, and presents an evaluation that should be useful in understanding the content of this book.

Part I

Banking

This page intentionally left blank

2

Liquidity Demand of the Corporate Sector and Soft Budget Constraint Noriyuki Yanagawa

This paper examines the relation between liquidity demands of the corporate sector and investment decisions. It shows that the liquidity asset holding has two impacts on investments. First, the liquidity asset can absorb a liquidity shock and helps to continue a pro®table project. Second, it helps to continue an undesirable project and decreases the incentive of ®rms. This negative aspect is similar to the soft budget constraint. According to these two impacts, ®rms must keep the optimal level of liquidity at the time of liquidity shock. If the value of liquidity asset ¯uctuates, however, it will become dif®cult to control the liquidity level. Hence the ¯uctuation of asset prices may decrease the aggregate investments even though all agents are risk neutral. The paper also examines the impacts of investments on liquidity asset prices, and endogenously determines both asset prices and investment levels. It shows that even if the asset prices are expected to rise perfectly, aggregate investments should decrease as long as the growth rate of the prices is too high. It also examines the role of the credit line offered by ®nancial intermediaries.

1

Introduction

This paper examines the relation between liquidity and investment decisions of the corporate sector. Keeping liquidity asset holdings appropriately is an important element for corporate ®nancial management, and the demand for liquidity assets in the corporate sector is important in ®nancial markets. In the economic literature, however, the liquidity demand of the corporate sector has not been examined very much. In the literature of the macro economy, much empirical evidence supports the hypothesis that liquidity is an important factor for investment decisions.1 In that literature, however, the theoretical

13

14 Liquidity Demand of the Corporate Sector

explanation about this relationship is not yet derived clearly. Furthermore, in that literature, the value of liquidity assets is given, and only the effects of liquidity level on investment decisions are explored. However, those asset values must be affected by the activity of the corporate sector. This paper will make clear, therefore, the interrelation between asset prices and investment decisions. We will show that the amount of liquidity holding affects the incentive of entrepreneurs and possible investment levels. The ¯uctuation of asset pricing should decrease the investment level even though all investors are risk neutral. It will be also shown that the liquidity holding of some ®rms may raise the asset prices too high and may cause to promote inef®cient investments of other ®rms. Recently, the importance of the liquidity demand of the corporate sector has been explored theoretically by Holmstrom and Tirole (1997, 1998). They have examined not only the liquidity demand of the corporate sector but also the supply side of liquidity. Then they have derived a new theory about asset pricing. In the traditional asset pricing models, the corporate sector is a black box and it is hard to examine the effects of government policies. In their model, however, government policies directly affect the price of assets. The spirit of this paper is close to the Holmstrom and Tirole model, although we will show more complicated interactions between asset pricing and investments, and will show a negative side of liquidity. Similar to the model of Holmstrom and Tirole, this paper assumes the liquidity demand of the corporate sector is coming from the agency problem between investors and entrepreneurs. More precisely, it assumes there is a moral hazard problem of entrepreneurs. In order to solve the moral hazard, investors must pay a rent to the ®rm. This means the payment from the ®rm to the investors is limited and this limitation generates an under-investment problem.2 If there is a possibility of liquidity shock and the ®rms have to make additional investments, this problem becomes signi®cant. Since the pro®t that can deliver to the investors is limited by the moral hazard problem, the additional investments may not be implemented since the investors cannot expect suf®cient return from the additional investments. To avoid such liquidity problems, ®rms have to retain liquidity assets to absorb the liquidity shock in the future. By holding liquidity assets, the ®rms can implement the initial projects. This is a positive aspect of the liquidity asset holding pointed out by Holmstrom and Tirole (1998). Keeping liquidity assets in a ®rm, however, also has a negative effect on the incentive of entrepreneurs. Even if a project becomes less productive

Noriyuki Yanagawa 15

and it should be shut down, entrepreneurs can continue the project by using the liquidity asset. In other words, the liquidity asset helps to bail out the project. This must decrease the incentive of entrepreneurs since it raises the gain from low effort choice. In order to clarify this point, we formulate two types of liquidity shocks. One is the ex ante shock that occurs before the choice of effort level and the other is ex post shock that occurs after the choice of effort level. At the ex post shock there is no incentive problem. The response to the ex post shock, however, affects the incentive of entrepreneurs. If the ®rm has suf®cient liquidity, the project is not terminated by the ex post shock and the incentive of entrepreneurs should be decreased. This negative aspect of liquidity asset holding is similar to soft budget constraint problems (Kornai, 1986). As Dewatripont and Maskin (1995) have pointed out, the soft budget constraint is an incentive problem which comes from the inability to commit not to bail out. In our situation, the liquidity asset holding makes it dif®cult to commit to terminate the project for the ex post shock and this generates the incentive problems. Moreover, this negative aspect is also related to the free-cash¯ow problem that was pointed out by Jensen (1986) and Stulz (1990). Jensen (1986) argued that managers have an incentive to invest in unpro®table projects if they have an excess cash ¯ow.3 Since the liquidity asset holding has a useful aspect for the ex ante liquidity shock, there is a trade-off. The liquidity asset holding has a negative and a positive impact on the realisation of the investment. Thus ®rms must choose a proper amount of liquidity asset to show that entrepreneurs have suf®cient incentive to choose high effort. In other words, there is an optimal level (or range) of liquidity asset holding. Too much or too little will decrease the incentive of entrepreneurs and thus decrease possible investments. We should point out that ®rms must keep this optimal level at the time of liquidity shocks. Firms must get the liquidity assets, however, before the liquidity shocks. If the value of the liquidity asset is constant over time, this requirement becomes less important. Firms only have to get a proper amount of the asset. If the value of liquidity asset ¯uctuates, however, it becomes a serious problem. Firms cannot keep the optimal amount of liquidity. Then it becomes impossible for the ®rms to get a suf®cient amount of the fund to implement the initial projects and as a result the investment level should be decreased. It should be noted that the same problem occurs even if the asset value is only expected to rise. For example, when there is a boom in the asset market and the price is expected to rise rapidly, the investment level

16 Liquidity Demand of the Corporate Sector

should be decreased. This could be one explanation for why a boom in ®nancial market suddenly generates depression. Next we examine the case where the price of liquidity assets is endogenously determined. If private claims of ®rms are mainly used for liquidity assets, liquidity shocks that change the value of ®rms affect the value of liquidity assets, and may become a reason for the ¯uctuation of liquidity assets. As a result, the possibility of a small liquidity shock may decrease investment levels signi®cantly. Moreover, we can show that the appropriate liquidity holding of some ®rms may raise the asset price too high and may cause to promote inef®cient investments of other ®rms. This paper also examines the credit line offered by ®nancial intermediaries. This credit line works just like the liquidity assets. One difference between the liquidity asset and the credit line is that the ®nancial intermediary controls the value of credit line. Thus the value of credit line can be constant over time and this is an advantage to its use. This advantage, however, will disappear if there is a ¯uctuation of other variables, for example the level of additional investment. There may be another advantage to the credit line. If the ®nancial intermediary can distinguish between the ex ante shock and ex post shock the intermediary allows ®rms to uses the credit line only for the ex ante shock. In Section 2, we will explain the basic model, and in Section 3, we explain the simple case in which there is only the ex ante shock. In Section 4, we introduce the possibility of ex post liquidity shock and examine the relation between the value of a liquidation asset and the optimal investment level. In Section 5, we shows that the ¯uctuation of asset value generates a problem and high variance of asset price decreases the investment level. In Section 6 we present a general equilibrium model to determine asset prices and investment levels endogenously. In Section 7 we examine the role of the credit line, and in Section 8 we present some discussions.

2

Model

We consider the following simple model where there are three types of agents, ®rms, investors and ®nancial intermediaries. All agents are risk neutral and we assume the investors have the following simple utility function: Uˆ

4 X iˆ1

ci

…1†

Noriyuki Yanagawa 17

By this assumption, we exclude the liquidity demand of investors. At each period, consumers receive suf®ciently large endowments to ®nance necessary investments but they cannot sell claims on their future endowments. This simpli®cation is used to highlight the key point of our argument, such that the possibility of moral hazard by ®rms (entrepreneurs) generates the liquidity demands. Each ®rm has a same investment opportunity at t = 0. It make an investment I at t = 0, considering the possibility of liquidity shocks at t = 1 or t = 3. The investment will generate R if the project succeeds and 0 if it fails. Since each ®rm only has A( < I), it has to gather at least (I ± A) for the investment. We assume there is a moral hazard problem at t = 2. An entrepreneur can choose high effort eH or low effort eL , but this effort choice is unobservable to investors. If an entrepreneur chooses eH , the probability of success will be PH and it will become PL if he/she chooses eL . On the other hand, the entrepreneur gets the private non-pecuniary bene®t B by choosing eL . We assume the following relation is satis®ed: PH R > PL R ‡ B

…2†

This means that eH is socially desirable. Entrepreneurs, however, may have an incentive to choose eL when he/she cannot get all of the realised gain R. The key point of our argument is the possibility of liquidity shock at t = 1 and t = 3. We assume that the ®rm that faces the liquidity shock has to add investments. If the ®rm cannot implement these additional investments, the project must be shut down and the pro®t becomes zero. At t = 1, the liquidity shock occurs with probability s and a ®rm facing this shock must additionally invest k (k is exogenous). We call this liquidity shock `ex ante liquidity shock'. At t = date 4, a ®rm which chose eL faces the `ex post liquidity shock' with probability u. The ®rm faces this shock must additionally invest f (f is also exogenous). The assumption that only the ®rms that choose eL will face the ex post liquidity shock is used for simplifying the explanation. Even if we change the assumption and assume that even the high effort ®rms will face the `ex post shock', our results are not affected. In Figure 2.1, we summarise the sequence of decisions. At t = 0 the initial investment is implemented and the contract between ®rms and investors are determined. At t = 1, the ex ante liquidity shock occurs with probability s and whether the project is shut down or not is determined. At t = 2, entrepreneurs choose eH or eL and this choice affects the probability that ®rms realise R. At t = 3, the ex post liquidity shock occurs

18 Liquidity Demand of the Corporate Sector Figure 2.1 t=0

t=1

Investment contracts

ex ante shock

t=2

continue

effort choice

terminate

t=3

t=0

R is

ex post continue shock realised terminate

with probability u only when the entrepreneur chose eL , and the project is shut down if the additional investment is not implemented. At t = 4, the pro®t is realised. To make our argument interesting, we assume that the project has a positive (expected) net present value if the ®rm chooses eH , but it has a negative (expected) net present value if it chooses eL . Assumption 1

P H R � I � sk > 0 > P L R � I � sk � uf

…3†

This assumption also means that the project is not realised if the entrepreneur cannot show credibly that he/she will choose the high effort level.

3

Simple case: no ex post liquidity shock

First, we examine the incentive of entrepreneurs when there is no possibility of ex post liquidity shock, that is u = 0. In order to check the incentive problem at t = 2, let us de®ne R as the payment from a ®rm to investors. If an entrepreneur chooses eH , R will be realised with probability PH . Therefore, the expected gain of the entrepreneur who has chosen eH is PH …R � R†

…4†

On the other hand, if the entrepreneur chooses eL , R will be realised only with probability PL , but she/he can get the private bene®t B. Thus the expected gain for the entrepreneur who has chosen eL is PL …R � R† ‡ B

…5†

From (4) and (5), R should satisfy the following condition in order to implement eH :

Noriyuki Yanagawa 19

RR�

B PH � PL

…6†

This upper bound of R means that the ®rm cannot commit by itself to pay out all possible gain to investors. This result generates at least two problems. First this upper bound might be too low to compensate the cost of initial investment. Let us de®ne R0 as the minimum necessary payment for the project, that is R0 ˆ

I � A ‡ sk PH

…7†

Then the project is not realised if R < R0 , since investors cannot expect suf®cient return from the project. A second problem arises when the ex ante liquidity shock occurs and the ®rm does not have suf®cient liquidity. Even if k < PH R (this means this additional investment is desirable), the additional investment only gives negative net present value to the investors as long as k > PH R. Thus this project should be terminated if the ®rm does not have suf®cient liquidity assets which can be used for the liquidity shock. This point has been explored by Holmstrom and Tirole (1998). When there is the ex ante liquidity shock, the ®rm has to invest k additionally, but it is only possible to pay out PH R to new investors.4 Hence the ®rm has to have liquidity asset at least (k � PH R) to implement the additional investment. This liquidity asset holding can be done by gathering …I � A† ‡ …k � P H R†

…8†

at t = 0. (I � A) is invested to the project and (k � PH R) is kept as the liquidity asset We assume here that the ®rm only uses this (k � PH R) for the liquidity and cannot use it for other investment opportunities or consumption. If the liquidity shock did not occur, the ®rm must return this (k � PH R) to the investors. If the liquidity shock occurs the ®rm uses (k � PH R) for the additional investment and gets PH R additionally from investors to implement the additional investment. We can easily show that investors who supply (I � A) + (k � PH R) can expect non-negative net return as long as R > R0 . Thus this asset holding is realised. Furthermore the ®rm can keep more than (k � PH R) as the liquidity asset. For example, even if the ®rm gathers (I � A) + k and keeps k as the liquidity asset, this project is realised. In this sense, it is not so dif®cult for a ®rm to keep appropriate liquidity value as long as the value of liquidity assets is exogenous and constant. If the value of liquidity assets is

20 Liquidity Demand of the Corporate Sector

endogenous or ¯uctuates, however, the problem of insuf®cient liquidity arises as explored by Holmstrom and Tirole (1998): private claims cannot work well to keep a suf®ciently high value of liquidity assets when there are macro shocks. Their argument is of course very important, but this paper stresses another important problem coming from the ¯uctuation of the liquidity asset value. If there is a possibility of ex post liquidity shocks too high value of liquidity assets decreases the incentive of entrepreneurs. Thus the ¯uctuation of liquidity value may become serious. In the next section we explore this point, and in Section 6, we examine the reason of the ¯uctuation by using of a general equilibrium framework.

4

Ex post liquidity shock and soft budget constraint

In this section, we introduce the ex post liquidity shock and examine the effects of the ex post liquidity shock on the incentive of entrepreneurs. Since the ex post shock occurs after the choice of entrepreneurs, the maximum possible payment after the ex post shock becomes R. Even so, however, the ex post shock affects the incentive of entrepreneurs' choice at t = 2. Since the ex post liquidity shock occurs only when the entrepreneur chose eL , the incentive to choose eL should be decreased if the additional investment for the ex post shock is not realised. The entrepreneur, however, tries to continue the project under the ex post liquidity shock, since he/she can get the private gain B. If the entrepreneur has suf®cient liquidity asset, therefore, he/she will continue the project even under the ex post shock and this has negative impact on the choice of eH . This situation is similar to the soft budget constraint problem. As Dewatripont and Maskin (1995) have pointed out, the soft budget constraint is an incentive problem and inability to commit to no bailout ex ante is the main reason. This situation is also related to the free cash ¯ow problem pointed out by Jensen (1986), which says too many cash ¯ow managers invest in unpro®table projects. In our situation, the liquidity asset holding makes it dif®cult to terminate the project for the ex post shock and this generates the incentive problem of entrepreneurs. The important point is that the liquidity asset holding is useful for the ex ante liquidity shock. Therefore there is a trade-off. The liquidity asset holding has a negative impact and a positive impact on the realisation of the investment. In order to highlight this incentive problem of entrepreneurs, we make the following assumption:

Noriyuki Yanagawa 21

Assumption 2

PL R < f < PL R ‡ B

…9†

This assumption means that outside investors do not have any incentive to make an additional investment for the ex post liquidity shock, but entrepreneurs do. Hence there is a soft budget constraint problem if the entrepreneur can continue the project even under the ex post liquidity shock. Although the continuation or termination of the project is endogenously determined, we ®rst check the incentive problem of entrepreneurs in the case where the continuation or termination is exogenously determined. 4.1

Ex post shock with continuation

The possibility of ex post liquidity shock affects the gain of entrepreneur who chose eL , and thus affects the maximum possible payment to implement the high effort level eH . First we derive the maximum possible payment R under the cases where the additional investment will be implemented under the ex post liquidity shock. To implement eH , R must satisfy the following condition: P H …R � R†  …1 � u†P L …R � R† ‡ B

…10†

The left-hand side is the expected gain when the entrepreneur chooses eH . The right-hand side is the expected gain when he/she chooses eL . If the ex post liquidity shock does not occur (this probability is 1 � u), he/she can get the promised share (R � R) and the private bene®t B. Under the ex post shock, however, he/she pays out all expected gain to implement the additional investment. Because, contrary to the cases of ex ante shock, there is no incentive problem any more at t = 3, the ®rm can promise to pay out all R. Thus the entrepreneur expects to get only the private bene®t B under the ex post shock. From this (10), we have the maximum possible payment RC : RC ˆ R �

B P H � …1 � u†P L

…11†

If the entrepreneur has suf®cient liquidity asset, he/she can use this asset for the additional investment and he/she may not have to pay out R. In such cases, however, the left-hand side of (10) becomes higher. Thus the maximum possible payment becomes lower than RC if entrepreneurs have liquidity assets.

22 Liquidity Demand of the Corporate Sector

4.2

Ex post shock with termination

Next we check the incentive problem of an entrepreneur when the project is terminated under the ex post liquidity shock. In this situation, the maximum possible payment R must satisfy the following condition to implement eH : PH …R � R†I  …1 � u†P L …R � R† ‡ …1 � u†B

…12†

This condition is similar to (10). In this case, however, the project is terminated when the ex post liquidity shock occurs. Thus the entrepreneur can get the private gain B only with probability (1 � u). From (12), we get the following maximum possible payment with termination, RS : RS ˆ R �

…1 � u†B PH � …1 � u†P L

…13†

By comparing (11) and (13), we can see that Rs > Rc . This means that the maximum possible payment increases by terminating the project under the ex post liquidity shock. In other words, an entrepreneur can promise to pay more if he/she can commit by himself/herself to terminate the project when the ex post liquidity shock occurs. In this model, the continuation decision is endogenous. If the entrepreneur has suf®cient liquidity asset, he/she can continue the project even under the ex post shock. Furthermore he/she has an incentive to continue the project because of the private gain B. Thus holding too much liquidity holding tends to continue the project even under the ex post shock and has a negative impact on the maximum possible payment. In the next sub-section, we will make clear this point. 4.3

Liquidity asset holding and optimal investment level

In this sub-section, we examine whether the project is terminated or not under the ex post liquidity shock. Since an entrepreneur can get the private gain B by continuing the project, he/she will choose the continuation as long as he/she can implement the additional investment. Thus the value of the liquidity asset which the ®rm has is important for the continuation decision and the maximum possible payoff. Moreover, whether the maximum possible payment is suf®cient for the initial investment cost or not is another important factor. In order to highlight on the soft budget constraint problem, we assume that

Noriyuki Yanagawa 23

RC < Ro < RS …1 � s†RS < R

0

…14† …15†

(14) means that if the project is not terminated under the ex post shock, the ®rm cannot promise suf®cient payment covering the investment costs.5 (15) means that the project does not give suf®cient return if the additional investment for the ex ante shock is not implemented. Under these assumptions, we will check the relation between the value of the liquidity asset and the realisation of the project. Let L be the value of liquidation asset value at t = 1. First we can easily show that the project is not realised if L < k � PH RS . The reason is simple. When there is the ex ante liquidity shock, the ®rm must implement the additional investment k at t = 1. The ®rm only has L so that it must gathers k � L from outside investors, who expect to get only PH RS . (Since all liquidity assets were invested at t = 1, this ®rm cannot continue the project under the ex post liquidity shock.) Thus this amount of liquidity asset is insuf®cient to implement the additional investment and this means the project is not realised. > Next we examine the case in which L ˆ k + f � PL R. In this case, the liquidity value is suf®cient for the additional investment for the ex post shock, even if the ®rm paid the additional investment for the ex ante shock.6 It follows that the ®rm cannot commit to terminate the project for the ex post shock. Thus the ®rm is only possible to pay out RC , which cannot over RO . If k � PH RS  L < k + f � PL R, the situation becomes different. As long as the ex ante shock occurs, a ®rm uses the liquidation asset suf®ciently for the ex ante shock. Thus the liquidation value becomes too low for the ex post shock and the project should be terminated under the ex post shock. If the ex ante shock does not occur, however, the ®rm has too much liquidity asset at t = 3. Then the ®rm will continue the project even under the ex post shock as long as L > f � PL R. Thus the ®rm cannot promise to pay out RS with probability (1 � u). To make clear this point we assume here that …1 � s†RC ‡ sRS < R0

…16†

This assumption is satis®ed as long as RC or s is suf®ciently low. Under this assumption, the ®rm cannot promise to pay out suf®cient return if L > f � PL R. Therefore, the project is realised only if k � PH RS < L < f � PL R. If f � PL R < k � PH RS , this project cannot be realised for any L. Figure 2.2 summarise this relation.

24 Liquidity Demand of the Corporate Sector Figure 2.2 (k � PH R s < f � PL R)

Investment

I

L

_ k–PHRS

f–PLR

Proposition 1 If f � PL R < k � PH RS , the project is not implemented.

If f � PL R < k � PH RS , the project is implemented only when the value of

the liquidation asset takes k � PH RS < L < f � PL R.

This proposition means that holding too much or too little liquidity asset

is bad for implementing the project. However, the best strategy for a ®rm

is simple, as long as there is no uncertainty about the value of liquidity

assets. The best strategy for a ®rm is to keep the liquidity asset in the range

k � PH RS < L < f � PL R. If the value of the liquidity asset ¯uctuates,

however, a ®rm cannot control the liquidation value at t = 1. Since the

value of the liquidation asset at t = 1 is important for implementing the

Noriyuki Yanagawa 25

project, this uncontrollability generates a serious problem. We will explain this problem in the next section.

5

Fluctuation of liquidity asset value

In this section we consider the cases where the value of liquidity asset ¯uctuates over time. In those cases, ®rms cannot control the value of holding liquidity asset perfectly. Thus there is a possibility that the liquidity asset at t = 1 does not take the proper value to realise the suf®cient payment. For the later explanation we de®ne [k � PH RS , f � PL R] as the L* range. Let suppose that the value of the liquidity asset ¯uctuates at t = 1 and if the value of the liquidity asset at t = 0 is L, the value at t = 1 will take ( L‡m with 0:5 L�m with 0:5 m is an exogenous parameter. For simplicity, we assume there is no ¯uctuation of liquidity value after t = 2. As long as m is small, the liquidation asset will be in the L* range and it works properly to responding the liquidity shocks. If m is suf®ciently large, however, the liquidation value cannot be in the range of L* and the ®rm is unable to promise to pay a suf®cient return. More precisely, we get the following proposition. Proposition 2 L

H

S

R If m > f �P R�k‡P ,the holding of a liquidity asset does not work well and 2 the ®rm cannot realise the project.

This proposition means the liquidity asset does not work properly if the disturbance of the asset value is suf®ciently high. In other words, the variance of asset price is important even though all investors are risk neutral. The total investment level should be decreased by an increase of the variance. This means that the policy to decrease the asset price should be meaningful. This point will be examined in the later section. Even if the asset price tends to go up, we get similar results. Let suppose the value of liquidation asset is as follows. If the value is L at t = 0, the asset value at t = 1 becomes  L with 0:5 L ‡  with 0:5

26 Liquidity Demand of the Corporate Sector

In this situation, the expected net capital gain is positive, but keeping this asset as the liquidity asset does not contribute to implement the project if  is suf®ciently large. This result has an important implication for macroeconomic problems. Suppose some ®rms have the equity of a ®rm as the liquidity asset. The above result shows that if the increase of the stock price is not so large, this stock works as the liquidity asset and the investments of those ®rms should be increased. If the stock price goes up so quickly, however, the stock does not work as the liquidity asset properly and macro investment levels go down. Thus this result is potentially useful to explain economic ¯uctuation problems. Furthermore, if there are several types of asset which can be used for the liquidity asset, ®rms tend to use a safe (i.e., low variance) asset. Thus the price of the safety asset tends to be high even if all investors are risk neutral.

6

General equilibrium analysis

In this section, we extend the previous arguments to a general equilibrium framework and examine how the ¯uctuation of liquidity asset value occurs. We assume here that only claims issued by ®rms can be used to transfer wealth from one period to next. In other words, the claims are only used for the liquidity demand. This means the ¯uctuation of ®rms' value generates the ¯uctuation of liquidity value. Thus aggregate shocks which affect the ®rms' value may generates the inef®ciency of investments. In this section we examine this problem carefully. In order to set up a general equilibrium model, we suppose that there is a continuum of ®rms with unit mass. Each ®rm has the same technology as assumed in the previous sections and faces the possibility of liquidity shocks as explained in the previous sections. We assume each ®rm has the market portfolio (that is, the symmetric combination of all equities) as the liquidity asset in order to decrease the variance of asset value. First we examine the total value of the market portfolio. At t = 0, the expected value of each ®rm is P H R � sk

…17†

If the stochastic variable s of each ®rm is independent, the total value of the market portfolio after the ex ante liquidity shock is also (17). If there are aggregate shocks, however, the total value of market portfolio must ¯uctuate. Let us de®ne the value of market portfolio L(!), L…!† ˆ PH RS � s…!†k

…18†

Noriyuki Yanagawa 27

! is the parameter of aggregate shock and it is distributed in the range of [!min , !max ] so as to satisfy dL/d! > 0. We assume here that each ®rm is symmetric and thus any other kinds of portfolio cannot decrease the variance of this market portfolio. In this situation, the value of the liquidity asset must ¯uctuates at t = 1 since the value of the market portfolio ¯uctuates. As pointed out by Holmstrom and Tirole (1998), this economy cannot supply for the total demand of liquidity if L(!min ) is suf®ciently small. Since the main point of this paper is the soft budget problem, however, we simply exclude this possibility and assume L…!min † > k

…19†

Under this assumption, the liquidity value is suf®ciently high to implement the additional investments for the ex ante shock. Even if the insuf®cient liquidity problem does not exist, the ¯uctuation of the liquidity value generates the soft budget problem. If the range of ¯uctuation [L(!min ), L(!max )] is not included in the range L*, each ®rm cannot control the liquidity value appropriately and the initial investments cannot be realised as explained in the previous section. This mean that private claims are insuf®cient for absorbing the liquidity shocks even if the total value of the market portfolio is always suf®ciently high. Rather, too high value generates the problem since it promotes the undesirable investment for the ex post shock. Moreover, we can see that the high value of the market portfolio is partially coming from the implementation of additional investment for the ex ante shock. Therefore the liquidity holding for the ex ante shock may harm the total economy, although such liquidity holding is meaningful for each individual ®rm. An intuitive reason is as follows. By holding the liquidity asset, each ®rm that faces the ex ante liquidity shock can accommodate the shock and implement the additional investments. As a result of the additional investment, the equity value of those ®rms does not decrease and the value of the market portfolio becomes high. In this sense the liquidity asset is working well. These high equity prices, however, generates too much liquidity holding and makes it possible to accommodate the ex post liquidity shock. Hence if some ®rms failed to accommodate the ex ante shock, it might be better for the economy. The value of the market portfolio does not go up so much and ®rms do not implement the inef®cient additional investment for the ex post shock. In summary, the liquidity asset holding may promote desirable investments, but it also generates the boom of ®nancial markets

28 Liquidity Demand of the Corporate Sector

and may allow inef®cient investments. If investors initially expect this scenario, they do not invest suf®ciently and the initial investment level must be decreased.

7

Provision of credit line

In the previous sections, we have assumed that a ®rm keeps a liquidity asset for the liquidity shocks. As explored by Holmstrom and Tirole (1998), however, ®rms need not have any asset if ®nancial intermediaries offers credit lines. Instead of keeping the liquidity asset, a ®rm makes a contract to a ®nancial intermediary and uses the credit line if there is a liquidity shock. Under the liquidity shock, the additional investment only generates a negative rate of return, since the ®rm cannot promise to pay out all possible gain. This means that the credit line generates negative pro®ts to the intermediary. Thus the ®rm has to pay some credit line fee at the date of contracts. There are several types of credit line contracts which can implement the project. A simple way is as follows. A ®rm makes a contract that allows the ®rm to use the credit line L. The ®rm need not pay any fee or interest rate when it uses the credit line, but initially the ®rm has to pay L to the ®nancial intermediary for the credit line fee. The ®rm gathers the fund for the fee from investors. In other words, the ®rm must gather just the same amount as the case of the liquidity asset holding from investors. In this case, however, the fund L goes to the ®nancial intermediary instead of purchasing the liquidity asset. One advantage of using the credit line is that the amount of credit line does not ¯uctuate over time. The maximum amount of credit line is usually determined by contracts, and it does not change over time. Hence, as long as the initial level of credit line is determined properly, the credit line helps to realise the project. If other valuables ¯uctuate over time, however, this advantage of the credit line may disappear. For example, let us suppose k will ¯uctuate at t = 1 and there is uncertainty about k at t = 0. In this case, the necessary amount of credit line also ¯uctuates. This means, the credit line which is usable for the ex post shock must ¯uctuate even though the contracted total credit line is constant. Hence the merit of the credit line explained above is not general. Financial intermediaries, however, may have informational advantage. In this paper we have assumed that investors cannot distinguish between the ex post shock and ex ante shock. Then ®rms can use the liquidity asset even for the ex post shock, and the soft budget constraint problem arises. If

Noriyuki Yanagawa 29

®nancial intermediaries can distinguish those two shocks, therefore, they can solve the soft budget constraint problem by prohibiting using the credit line for the ex post shock.

8

Discussion

The arguments in the previous sections have shown that the ¯uctuation of asset value has negative effects on investments. We have shown that even if the price is expected to rise, this negative effect works. One way to solve this problem may be government intervention. The stabilisation of the liquidity asset price can be obtained through market interventions or direct supply of liquidity assets (such as government bonds). By adjusting the supply level of liquidity, government can control the investment levels indirectly. In this paper we have assumed for simplicity the investment level I is exogenous. We can easily extend this model, however, to the cases where investment level is endogenously determined. In such cases the equilibrium investment level changes when the maximum possible payments ¯uctuate. Hence there exists an under-investment effect because of the ¯uctuation of asset prices. Moreover, this paper has used a simple four periods model. An extension of this model to more general versions of dynamic models would be interesting. By using an in®nite horizon model, for example, we can derive the dynamic of asset prices more rigorously. Then we can make clearer the dynamic interactions between the dynamics of asset prices and investment decisions.

Notes 1. See, for example, Fazzari, Hubbard and Petersen (1988) or Blanchard, Lopez-deSilanes and Shleifer (1994). 2. This logic is also similar to Greenwald, Stiglitz and Weiss (1984) or Myers and Majluf (1984). 3. Recently, several papers empirically and theoretically examined this negative aspect of liquidity. See, for example, Hadlock (1998), or Myers and Rajan (1998). 4. Since the project must be shut down and pro®t becomes zero if the additional investment is not implemented, the old investors agree that their gain becomes zero by paying out PH R to new investors. 5. More precisely, investors require more than RO , since the liquidity asset will be used for continuing the project even for the ex post shock. Thus (14) is a suf®cient condition. 6. We are assuming that the liquidity asset is unable to use for consumption.

30 Liquidity Demand of the Corporate Sector

References Blanchard, O.J., E. Lopez-de-Silanes and A. Shleifer (1994) `What Do Firms Do with Cash Windfalls?', Journal of Financial Economics, 36, pp. 337±60. Dewatripont, M. and E. Maskin (1995) `Credit and Ef®ciency in Centralized and Decentralized Economies', Review of Economic Studies, October. Fazzari, S.M., R.G. Hubbard and B.C. Petersen (1988) `Financial Constraints and Corporate Investment', Brookings Papers on Economic Activity, pp. 141±95. Greenwald, B., J.E. Stiglitz and A. Weiss (1984) `Informational Imperfections in the Capital Market and Macroeconomic Fluctuations', American Economic Review, 74, pp. 194±99. Hadlock, C. (1998) `Ownership, Liquidity, and Investment', RAND Journal of Economics, 29, no. 3, pp. 487±508. Holmstrom, B. and J. Tirole (1998) `Private and Public Supply of Liquidity', Journal of Political Economy, vol. 106, no. 1, pp. 1±40. Holmstrom, B. and J. Tirole (1997) `LAPM: A Liquidity-Based Asset Pricing Model', mimeo. Israel, R. (1991) `Capital Structure and the Market for Corporate Control: The Defensive Role of Debt Financing', Journal of Finance, 46, pp. 292±355. Jensen, M. (1986) `Agency Cost for Free Cash Flow, Corporate Finance, and the Takeovers', American Economic Review, 76, pp. 323±29. Kornai, J. (1986) `The Soft Budget Constraint', KYKLOS, 39 (1), pp. 3±30. Myers, S.C. and N.S. Majluf (1984) `Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have', Journal of Financial Economics, 13, pp. 187±221. Myers, S.C. and R.G. Rajan (1998) `The Paradox of Liquidity', Quarterly Journal of Economics, pp. 733±71. Stulz, R. (1990) `Managerial Discretion and Optimal Financing Policies', Journal of Financial Economics, 26, pp. 3±27.

3

Incentive Effects of Conditional Bank Recapitalisation: Lending and Disclosure of Non-Performing Loans1 Philippe Aghion, Patrick Bolton and Steven Fries

1

Introduction

It is widely believed that one of the main sources of the prolonged economic slump of the Japanese economy in the 1990 is the collapse of asset and real estate prices, which have led to a banking crisis of an unprecedented scale. It is in large part the failure promptly to resolve this banking crisis that has led to a prolonged and massive recession. With a large fraction of non-performing loans, a reduced capital base and substantially lower liquidity the banking sector has dramatically reduced new lending activity and provoked a huge credit crunch. Other countries have witnessed large scale banking crises in recent years, but what sets the Japanese experience apart is the problem of hidden loan losses and opaque accounting practices as well as the stunning complacency of regulatory authorities. More than in other countries banks have actively sought to hide the extent of their loan losses, so much so that every year ®nancial markets were discovering to their horror that loan losses were substantially higher than anticipated. More than in other countries, regulatory authorities have hesitated and wavered in determining a strategy towards the resolution of the banking crisis, thus prolonging and worsening the crisis. Some politicians favoured a tough policy towards insolvent banks, recommending that they be closed as soon as possible and their assets redeployed. Others, were mainly concerned about the systemic risks of closing the larger insolvent banks and favoured a softer strategy based on recapitalisations of distressed banks. From the perspective of bank managers there was substantial policy uncertainty until very recently. In particular they could not rule out the 31

32 Incentive Effects of Conditional Bank Recapitalisation

possibility that if their bank was found to be insolvent it would be closed down. Thus, in an attempt to save their institution they naturally chose to under-report loan losses and actively lobbied for new capital injections. They had every incentive to do so. In an earlier paper on Bank Restructuring in Transition Economies we emphasised that when regulators must rely on banks to truthfully report loan losses then the implementation of strict closure rules and prompt corrective action may have unintended and undesirable effects: insolvent banks would then respond by simply hiding their insolvency, thus prolonging and magnifying the crisis (see Aghion, Bolton and Fries, 1999 henceforth ABF). In that paper we also developed a simple model to address the question of the optimal bank restructuring policy in the face of informational asymmetries on the quality of outstanding bank loans. In this paper we argue that with some minor modi®cations the basic framework in ABF developed for transition economies is also relevant for Japan. The key common element is the basic observation that in each case banks are run by managers who are not the actual (or sole) owners of the bank. That these managers run the bank in their own self-interest and that they have superior information on the quality of the loans they make. Who actually owns the bank is not that relevant since bank owners are likely to be either too dispersed or too ignorant to be willing or able actively to monitor bank managers. A second objective of this paper is to explore several extensions of the basic model developed in ABF to test for the robustness of the main conclusions obtained in the simple ABF model. The basic set-up considered in ABF includes three types of agents: ®rms, banks and the regulators. Firms and banks are controlled by their managers who derive private bene®ts from their continued operations and the main source of discipline on their behaviour is the possibility of dismissal associated with insolvency. The regulators' objectives are to promote the ef®cient restructuring or liquidation of ®rms which have defaulted on their bank loans.2 As already emphasised, banks are assumed to have private information about the quality of their loan portfolio. The main conclusions reached in ABF are that: 1. a tough recapitalisation policy in which the bank manager is always dismissed results in the bank managers rolling over bad loans in order to conceal loan losses and therefore in the softening of the ®rms' budget constraints. However, 2. the socially ef®cient outcome can generally be achieved through a recapitalisation policy combined with the carving out of bad loans at a suitable non-linear transfer price.

Philippe Aghion, Patrick Bolton and Steven Fries 33

Other studies on bank restructuring in Transition Economies from the mid-1990s have reached similar conclusions. Most notably the work of Mitchell (1995), who also emphasises the problem that when a bank manager suffers in some way when the bank gets into trouble he will roll over loans in default in order to postpone facing the cost of ®nancial distress. Two other related papers are Suarez (1995) and Povel (1997). The former paper studies bank closure rules and recapitalisation in a dynamic complete information model. Given the informational assumptions stressed in this paper, it is not entirely surprising that it ®nds that the closure of insolvent banks has good ex ante incentive properties. The latter paper deals with bankruptcy of non-®nancial ®rms but emphasizes a similar tension between ex ante incentives to avoid bankruptcy and ex post incentives to ®le for bankruptcy in a timely fashion. The remainder of the paper is organised as follows. Section 2 describes a slightly modi®ed version of the ABF model. Section 3 summarises the main incentive effects of various standard bank restructuring policies including conditional bank recapitalisations. Section 4 then considers the ®rst of three extensions by examining the effects of high-powered incentives for bank managers. Section 5 explores the effects of less stringent informational assumptions. Section 6 brie¯y considers an extension where bank managers may have different managerial abilities. Finally, Section 7 provides a brief summary of the main lessons of our analysis.

2

The ABF model

The model considered in Aghion, Bolton and Fries (1999) has three types of agents interacting with each other: ®rms, banks and regulators. 2.1

Firms

Firms are run by self-interested managers, who control an asset, which yields a random return. In the ®rst period, the return is: ( )  with probability p …1† 0 with probability 1 � p where,  > 0 and p 2 (0, 1) is exogenously given. In the second period, the ®rm also has a random continuation value, which is the discounted stream of future returns. Each ®rm has an outstanding stock of bank debt and no other liabilities. This stock of debt imposes a repayment obligation on the ®rm of D 2 [0, ]. When a ®rm defaults, the bank can either

34 Incentive Effects of Conditional Bank Recapitalisation

liquidate the ®rm or allow it to continue. The liquidation value of the ®rm is L. The continuation value is random and given by: ( )  with probability 1 � …2† 0 with probability where, V > L > 0. In the event of default, the continuation value can be costlessly observed. The ®rm's manager always honours the debt repayment obligations if he can. This assumption rules out strategic defaults by ®rms. The model can accommodate strategic defaults as follows: suppose that, in addition to the ®rm continuation value, the ®rm's manager has a random private continuation value from retaining his position given by: ( ) V with probability 1 �  …3† 0 with probability  If the ®rm's manager decides to default strategically, he gets the foregone debt service payments. Thus, if one assumes that V > D the manager defaults strategically only when his expected continuation value is equal to zero. The probability of strategic default is then given by p. When one allows for strategic default the manager is able to divert resources from the ®rm. The only constraint that prevents him from diverting all the ®rm's resources is the threat of liquidation when there is a default on the ®rm's debt. 2.2

Banks

Banks are also run by self-interested managers. On the asset side of their balance sheets, banks have a portfolio of loans to ®rms, each of which has a scheduled debt service payment of D. In the event of a default, and in the absence of strategic behavior by the bank manager, the bank liquidates the ®rm with probability and obtains L. The alternative to liquidation is ®rm continuation with a realised return . To introduce the possibility of bank failures it is assumed that ®rms' returns are correlated so that the fraction of the bank's performing loans is a random variable which takes on a range of values, p1 > p2 > p3 > p4 > 0, with respective probabilities 1 , 2 , 3 , 4 > 0. The expected fraction of performing loans is then given by p = 4iˆ1 1 p1 . On the liability side of their balance sheet, banks issue deposits in the amount d to fund each loan. The bank's realised net worth is then given by: Wi ˆ …1 � pi †‰ L ‡ …1 � †Š ‡ pi D � d

…4†

Philippe Aghion, Patrick Bolton and Steven Fries 35

where W4 < W3 < W2 ˆ 0 < W1

…5†

That is, only banks in states 1 and 2 are solvent while banks in states 3 and 4 are insolvent.3 If one allows for strategic behaviour by ®rm managers as outlined above then the realised net worth of the bank becomes: Wi ˆ ‰1 � pi …1 � †Š‰ L ‡ …1 � †Š ‡ pi …1 � †D � d

…6†

That is, the probability of default on bank loans increases by pi . The ABF model captures ex ante moral hazard in bank lending to ®rms by introducing an effort variable which affects the ex ante probability of bank failure. The bank manager can choose between two effort levels e 2 {0, 1}. The cost of effort is c(e), with c(0) = 0 and c(1) = c, and when e = 1 the probability distribution i (1) (®rst-order) stochastically dominates the probability distribution when e = 0, i (0): j X iˆ1

i …1† >

j X

i …0†

…7†

iˆ1

for all j = 1, 2, 3. The main ingredient in the ABF model ± the ex post incentive problem of truthful reporting of non-performing loans ± is captured as follows: when a ®rm defaults, the sale of the ®rm's assets can be observed costlessly so that the liquidation decision is observable and veri®able. Loan continuation and provisioning decisions, on the other hand, are entirely at the discretion of the bank manager and cannot be veri®ed. Unless a nonperforming loan is actually liquidated it is not possible to verify whether the loan is performing or not. This admittedly strong assumption is not entirely unrealistic in the case of Japan, according to recent press reports. Indeed, the systematic deception of bank regulators seems to have gone as far as temporarily `parking' non-performing loans in reputable international banks. In the ABF model the deception of bank regulators takes a very simple form: the bank manager inef®ciently re®nances bad loans in order to hide (or understate) the overall extent of loan losses. Bank managers can also deceive regulators by overstating loan losses in order to elicit a greater recapitalisation. The core analysis of the ABF paper centres around these two forms of strategic behaviour by bank managers. A bank manager's objective function involves a monetary and a private bene®t component. The monetary component is the sum of a ®xed salary

36 Incentive Effects of Conditional Bank Recapitalisation

(which is normalised to zero) and a share, say equal to b 2 (0, 1), of the bank's (reported) net worth. The private bene®t component re¯ects the facts that: (i) bank managers like power and (ii) bank managers, like ®rm managers, would rather retain their job than be ®red. In addition, the bank manager's objective function includes the cost of effort, if any, that she expends in managing the bank's loan portfolio. Formally, the bank manager's objective boils down to: ci † ‡ e UB ˆ b max…0; W B‰1 ‡ max…0; Wi ‡ R†Š � c…e†

…8†

where ( e Bˆ

B if the bank manager remains in place 0 if not

)

ci is the reported net worth of the bank (absent recapitalisation). and W Any additional resources accruing to the bank in period one, in particular as a result of recapitalisation, is denoted by R. The analysis in ABF only considers the case where the bank manager has no monetary incentives, so that b = 0 and UB ˆ e B‰1 ‡ max…0; Wi ‡ R†Š � c…e† 2.3

…9†

The regulator

In ABF the regulator is pictured as a benevolent social planner with full commitment powers, whose objective is to: (i) maximise the expected social return of the underlying assets of ®rms, (ii) induce bank managers to make sound ex ante lending decisions, and (iii) to minimise the costs associated with any excessive recapitalisation of banks. The purpose of this unrealistic assumption is to isolate the constraints on bank closure policy imposed by banks' ability to hide loan losses and to abstract from political economy issues relating to the motives of the regulators' political constituency. These latter issues are clearly of fundamental importance in the case of Japan. Indeed, the ®nal bank restructuring package that has been put in place is as much the result of political compromise as of purely economic or ®nancial logic. In ABF the regulator's problem is straightforward when there is complete information about the true net worth of banks. Then the regulator would avoid excessive recapitalisations by simply transferring ± Wi (to cover insured deposits) to those banks in states i = 3, 4 and nothing to banks in states i = 1, 2. Note that such a policy ± which is essentially the

Philippe Aghion, Patrick Bolton and Steven Fries 37

received wisdom on bank closure policies ± would also best mitigate the ex ante moral hazard problem in bank lending. By dismissing the bank manager whenever the bank is insolvent this closure policy maximises incentives to make prudent lending decisions. When the regulator must instead rely on the bank to report its loan losses then dismissal can be avoided when the bank is insolvent by underreporting loan losses. Similarly, when the regulator decides to recapitalise a distressed bank, but does not know the bank's net worth, the bank manager may be able to increase the size of the recapitalisation by in¯ating loan losses and claiming to be in the worst possible state. Hence if the regulator wants to guarantee that all banks reach at least a minimum reported net worth of zero, it must be prepared to bailout banks up to an amount ± W4 , the worst possible net worth. Such misrepresentation by all bank managers would lead to an excessive recapitalisation with an ex ante dead-weight loss of ‰1 …W1 � W4 † ‡ 2 …W2 � W4 † ‡ 3 …W3 � W4 †Š ˆ E

…10†

Of course, the regulator has the option to limit the size of the recapitalisation to an amount less than ± W4 , but then it exposes itself to the possibility of inadequate recapitalisation of those banks in the worst state of nature. The expected social return of the underlying assets of ®rms is given by their expected ®rst period cash ¯ows, p (e) (where p (e) = 4iˆ1 i (e)pi ) plus their expected continuation values bi †Š ‡ bi Lg

i ˆ pi …1 � † ‡ …1 � pi †fmin‰…1 � †; …1 �

…11†

That is, for the proportion pi of ®rms with high cash ¯ows the expected continuation value is (1 � ), since these ®rms will never be liquidated by their managers. For the proportion (1 � pi ) of ®rms with low cash ¯ows the manager is forced to default and the average continuation value per loan is min[(1 � ), (1 � ^i )] + ^i L. Here ^i denotes the fraction of defaulting loans the bank manager chooses to liquidate in each state i = 1, . . ., 4. Formally, the regulator's objective in the ABF model takes the form: UG ˆ p …e† ‡

4 X

i i � E � c…e†

…12†

iˆ1

So that social ef®ciency requires ful®lment of three conditions: 1. a ®rm should be liquidated if, and only if, its liquidation value exceeds the continuation value ~, that is, ^i should be equal to for i = 1, . . ., 44

38 Incentive Effects of Conditional Bank Recapitalisation

2 only those banks with truly negative net worth should be recapitalised; that is, E should be equal to zero; 3. bank managers should choose effort e = 1 given that 4 X p…1† � p…0† ‡ ‰i …1† � i …0†Š i > c …13† iˆ1

3

Incentive effects of conditional bank recapitalisations

Using this model ABF makes two main points concerning bank recapitalisation policies. First, since bank managers can to a large extent neutralise the incentive effects of strict closure rules by misreporting their loan losses these rules cannot be assumed to have the desired incentive effects on ex ante lending decisions. In addition, these rules clearly create incentives to hide loan losses, thereby introducing massive distortions in loan renewal decisions. These distortions result in crowding out of sound new investments by non-performing old loans and magnify the size of ultimate loan losses. As a result of these distortions it is not always true that strict bank closure rules are a more ef®cient regulatory response than unconditional bailouts, since the latter eliminate incentives to hide loan losses and thus eliminate the distortion of keeping non-performing loans alive in order to hide losses. More strikingly, bailouts may in some special cases even provide better ex ante incentives to make sound lending decisions than strict closure rules. Second, the best regulatory response in the ABF model is neither strict closure rules nor unconditional bailouts, but instead conditional recapitalisation. That is, the recapitalisation should be conditional on liquidation, sale or carve-out of non-performing loans. This scheme provides the best incentives for truthful reporting of loan losses and replicates the ex ante incentive effects for prudent lending of strict closure rules. Before turning to the discussion of the robustness of this basic analysis in ABF we brie¯y summarise the analysis of these two main points. 3.1

Closure of insolvent banks vs bailouts

Consider ®rst the incentive effects of strict closure rules 3.1.1

Strict closure rules

Suppose that the manager of a bank which reports a negative net worth is automatically dismissed and the bank's assets redeployed. Then, the manager of a bank with realised p1 or p2 has no incentive to manipulate either the accounts of the bank or the decisions to liquidate ®rms or to

Philippe Aghion, Patrick Bolton and Steven Fries 39

write-down their loans. However, the bank would be insolvent if either p3 or p4 were realised. With such outcomes, the bank manager will act as if pk = p2 has occurred in order to preserve his job. Since the liquidation of ®rms is veri®able, the bank manager will pretend that pk = p2 by liquidating a fraction 2 of ®rms in its portfolio, where 2 is de®ned as the fraction of liquidated loans in the portfolio of a bank with realised p2 , (that is 2 = (1 � p2 ) ). That is, the bank manager liquidates a fraction >^ k of defaulting ®rms, where bk ˆ 2 ˆ …1 � p2 † …14† …1 � pk † So that, the proportion of defaulted loans that are actually liquidated by the bank manager in states p3 or p4 is less than the socially ef®cient bk < . The incentive of bank managers to maintain the proportion: appearance of bank solvency under a tough bailout policy, thus leads to a softening of ®rms' budget constraints. Strict closure rules thus result in an insuf®cient number of ®rm liquidations. The loss in social surplus due to the softness of banks on ®rms in default is the foregone liquidation value of those ®rms which are continued even though they have a zero continuation value.5 When one allows for strategic defaults these effects are ampli®ed, for then ®rm managers may choose to default strategically if they anticipate a lower probability of liquidation by their insolvent bank. Thus, with strategic defaults there will be an additional build up of non-performing loans in banks' portfolios. More formally, suppose that ‰1 � p2 …1 � †ŠV < D < V One can then verify that in the case of a solvent bank (in state 1 or 2), the pair of strategies (> ^i = , strategic default with probability ) is the unique Nash equilibrium. But in state 4, there exists a Nash equilibrium involving b4 < , strategic default a higher probability of strategic default, namely: ( b4 satis®es with probability one), where b4 ˆ 2 ˆ ‰1 � p2 …1 � †Š ‰1 � p4 …1 � 1†Š The ex ante incentive effects of strict closure rules, taking into account the bank manager's ex post misreporting of non-performing loans are as follows. The bank manager chooses e = 1 only if 4 X iˆ1

1 …1†B‰1 ‡ max…0; Wi ‡ R†Š � c >

4 X iˆ1

1 …0†B‰1 ‡ max…0; Wi ‡ R†Š

40 Incentive Effects of Conditional Bank Recapitalisation

or equivalently ‰1 …1† � 1 …0†ŠBW1 > c

…15†

No bank manager is dismissed in equilibrium because of the misrepresentation of the bank's net worth. Note that the value of private bene®ts equals B in all states of nature except state 1, when the value of private bene®ts equals B(1 + W1 ). The expected value of private bene®ts thus rises with managerial effort to the extent that this effort raises the probability that state 1 will occur. 3.1.2

Bank bailouts

Under an unconditional bailout policy a bank manager is immune from dismissal, regardless of reported net worth. This approach creates an incentive for bank managers to overstate their problem loans so as to increase the amount of recapitalisation. Bank managers can easily overstate anticipated losses by taking excessively high charges. The main bene®t of bailouts is that they restore bank manager's incentives to impose ®nancial discipline on the ®rms they lend to. In other words, with bailouts, the bank manager hardens the budget constraint on ®rm managers. There are at least two social costs resulting from a generous bailout policy. One is the dead-weight costs from excessive recapitalisation. The second concerns ex ante moral hazard in lending. Note, however, that in the ABF model the only incentive for bank managers to choose e = 1 under a bank bailout policy arises from the associated private bene®ts. The bank manager would has an incentive to choose e = 1 only if: 3 X ‰i …1† � i …0†ŠB‰max…0; Wi † � W4 Š > c

…16†

iˆ1

or, …1 …1† � 1 …0††B
max…0; W1 † ‡ …4 …1† � 4 …0††BW4 > c

…17†

Comparing equations (15) and (17) reveals that whenever 4 (1) ± 4 (0) < 0 the incentive-compatibility constraint on managerial effort is less tight under a strict bank closure policy than under a bailout policy provided that jW4 j is not too large. Thus, for ex ante incentive for prudent lending considerations the regulator may want to allow for some overstatement of bad loans ex post under a bailout policy.

Philippe Aghion, Patrick Bolton and Steven Fries 41

A comparison of the two policy regimes leads to the overall conclusion that while strict bank closure rules dominate when the banking system is believed to be basically sound, bailouts dominate when there are reasons to believe that banks are likely to be insolvent. If one limits the choice of policy to these two regimes a careful evaluation of the likely state of the banking system is required to determine which is the adequate policy response. Thus, if Japanese regulators had reasons to believe that the banking sector was reasonably healthy in the immediate aftermath of the ®nancial collapse in 1990 they might have been justi®ed to opt for a strict enforcement of bank closure rules. On the other hand, in light of new data over the following years of a deteriorating economic situation and a worsening of bank equity values they might have been led to change their mind about the severity of the banking crisis and thus to opt for a different policy response favouring bailouts in an attempt to clean up bank balance sheets. In addition, if the original reading of the overall health of the banking sector was mistaken then the initial commitment to enforcing strict bank closure rules may have exacerbated the non-performing loans problem by inducing banks to hide their loan losses. As the analysis in ABF highlights, if one limits the policy options to these two alternatives there is scope for massive policy errors. However, when one widens the set of policy alternatives to include conditional recapitalisations then the choice of policy may be less sensitive to a precise reading of the state of the banking sector. Indeed, conditional recapitalisations emerge as the favoured response irrespective of prior beliefs about the state of the banking sector for a wide range of parameter values in the ABF model. 3.2 Bank recapitalisations conditional on liquidation of non-performing loans As liquidation of non-performing loans is an observable and veri®able action it is natural to consider recapitalisation policies conditional on liquidation. In ABF it is shown that it is possible to achieve two of the three objectives of the regulator with such a policy, namely: 1. the ef®cient liquidation of ®rms in default so that ^i = , and 2. no excessive bank recapitalisation. To achieve the third objective of providing incentives to overcome moral hazard in lending, however, a complementary policy may be required. Ef®cient liquidation and no excessive recapitalisation can be obtained if the regulator introduces a non-linear transfer pricing scheme specifying

42 Incentive Effects of Conditional Bank Recapitalisation

the terms at which the regulator accepts to purchase loan contracts from banks. In the highly simpli®ed setting of the ABF model the non-linear transfer pricing scheme takes the following form: 1. a low transfer amount, tL , for loans in default which are liquidated, up   2 of the bank's portfolio and, to a threshold m 2. beyond that threshold transfers per liquidated loan are increased to t H > tL . It is shown in ABF that by adequately specifying the non-linear pricing schedule {tL , tH , m} the regulator can induce banks to truthfully disclose the state of their ®nances. The main result in that paper is as follows: Proposition: There exists an m  2 such that the two-part transfer scheme price (tL , tH , m): i) recapitalises only those banks which are truly insolvent, bi = ) if, and and ii) leads to the ef®cient liquidation of ®rms in default ( only if p4 D ‡ …1 � p4 †…1 � †V ‡ …4 � 2 †V ‡ 2 L  d

…18†

Proof: see ABF.& A non-linear tariff is needed mainly in order to discriminate between truly insolvent banks who require recapitalisation and others who may simply be looking for a cheap source of funds. A linear tariff would inevitably apply to all banks whether solvent or not. The solvent ones would take advantage of the rescue scheme to obtain cheap government funding. Such a recapitalisation scheme would be wasteful and would not provide the best incentives for prudent lending by banks. A conditional policy is required to make sure that banks do not take advantage of the recapitalisation scheme by overstating the extent of their losses. By making recapitalisation conditional on liquidation the regulator makes sure that the bank is recapitalised for `realised' as opposed to `reported' losses. Given the size of some recent bank recapitalisation programmes (e.g., in Argentina, Mexico, and also Japan) it should clearly be an overriding concern that the size of the recapitalisation is limited as much as possible. One way of meeting that objective is to make sure that only those banks that are most in need of recapitalisation get funding and that they get funding only for `realised' losses.

Philippe Aghion, Patrick Bolton and Steven Fries 43

It is important to clarify that liquidation does not need to take the form of an actual sale of a loan in the secondary market. It should be understood rather as a transfer of a bad loan to a government restructuring agency specially set up as part of the conditional recapitalisation scheme. The reason why one may want to avoid sale of loans in the secondary market is obvious. If there is a general liquidity shortage these loans would be sold below long-term value. Thus, in the case of real estate loans it would be highly inef®cient to sell such non-performing loans while the real estate market is depressed. On the other hand, a transfer of such loans to a specialised agency, who would hold on to them until the market recovers would serve the purpose of cleaning up bank balance sheets without at the same time worsening or prolonging the real estate crisis. Why does the conditional recapitalisation policy work best only if condition (18) is satis®ed? This condition states that under a non-linear tariff which perfectly discriminates solvent and insolvent banks the recapitalisation of the worst hit banks would be suf®cient. This is clearly not always possible. Some banks may require massive recapitalisations. These would be so large that healthier banks might be tempted to overstate their loan losses just to be able to obtain better terms for their liquidated bad loans. Thus, if condition (18) is not satis®ed, even this conditional recapitalisation policy would lead to excessive recapitalisation of solvent banks in state p2 (and/or of insolvent banks in state p3 ), with the associated dead-weight costs. However, it is probably asking for too much to be able to avoid any excessive recapitalisation and a conditional policy with a suitably designed non-linear tariff is likely to beat any other scheme of recapitalisation along this dimension even if it is not able to always achieve the ®rst-best. Two other potential problems that might arise with such a scheme. The ®rst is that banks might be encouraged to get rid of their worst loans if the transfer tariff does not adequately discriminate between different underlying qualities of loans. The second problem with this non-linear pricing scheme is that it may create incentives for solvent banks to sell their bad loans to insolvent banks. To prevent such pro®table arbitrage from taking place the regulator would need to monitor the secondary market for loans and scrutinise more closely net purchasing banks. Despite these dif®culties we believe that the basic economic principles underlying this scheme are suf®ciently sound that it is worth exploring further the feasibility and desirability of such a bank rescue plan. In particular, an important strength of this scheme is that it achieves ex post ef®ciency regardless of the regulator's knowledge (or beliefs) i (e) about the state of the overall banking system. In this important respect it

44 Incentive Effects of Conditional Bank Recapitalisation

dominates the strict closure rules and unconditional bailouts considered above, none of which would achieve these requirements for ex post ef®ciency, except perhaps for a small subset of parameter values for i (e). The analysis in ABF thus shows that conditioning bank recapitalisation on an observable and veri®able action of bank managers can increase the ex post ef®ciency of bank bailouts and, under certain circumstances, meet the two requirements for ex post ef®ciency. As for ex ante effort incentives, it turns out that the implementation of strict closure rules provide bank managers with precisely the same incentives as the conditional recapitalisation considered here. It is straightforward to show that the incentive-compatibility constraint for a bank manager under the conditional bank recapitalisation policy simpli®es to ‰1 …1† � 1 …0†ŠBWi > c

…19†

which is the same as under the strict closure rules policy. This equivalence comes from the fact that both policies give the bank manager the option of distorting his ex post report about loan losses (i.e., about i ), although our conditional scheme is designed in such a way that bank managers are indifferent between distorting (and announcing state 1) and not distorting. This explains why ex ante effort incentives are the same under the two policies, even though our scheme avoids the ex post inef®ciencies induced by the strict closure rules policy. Our scheme should thus be seen as a strict improvement over the standard recommended strict closure rules. This conclusion hinges however on the fact that the manager is never dismissed in equilibrium under any recapitalisation policy. In Section 7 we consider an extension of the model where the true state of nature is publicly revealed in period 2. To summarise the main lessons from the analysis in ABF are that by reducing the incentive of bank managers to exaggerate the extent of their bad loans, a suitably designed conditional recapitalisation scheme in which government transfers to insolvent banks are linked to `realised' as opposed to `reported' losses can achieve some of the ex post ef®ciency objectives. We believe that these lessons are relevant for the Japanese banking crisis and we now outline that they are also robust to a number of changes brought to the basic ABF model.

4

High-powered incentives

The analysis in ABF proceeds under the assumption that bank managers are not provided with high-powered monetary incentives linked to banks'

Philippe Aghion, Patrick Bolton and Steven Fries 45

reported net worth. We now consider the more general case where a bank manager's expected utility takes the form E‰UB …e†Š ˆ

4 X

ci † ‡ e i …e†fb maxf0; W Bi …1 ‡ max…0; Wi ‡ R†Šg � c…e†

…20†

iˆ1

c i is the reported net worth per loan of the bank, and R is where b > 0, and W the government-funded recapitalisation. The ®rst term in equation (20) depicts utility derived by the bank manager from monetary compensation and the second term utility obtained from private bene®ts. The introduction of a high-powered monetary incentive scheme affects the analysis of the various bank recapitalisation policies ®rst through its impact on the incentive to misreport banks' net worth ex post. Now, bank managers have an additional incentive to hide loan losses, since by boosting reported net worth (and hiding loan losses) they not only avoid losing their job but also increase their bonus. In particular, a bank manager may now want to misreport net worth of W1 instead of only W2 as argued in Section 3.1. However, with a `soft' recapitalisation policy, a high-powered incentive scheme raises the cost of overstating the size of the non-performing loans problem. The trade-off between monetary and private bene®ts in this case can lead bank managers to misreport net worth of W1 instead of W4 . While the introduction of a high-powered incentive scheme alters the incentive to overstate banks' net worth, it provides no incremental incentive to supply effort ex ante beyond that provided by private bene®ts with either a strict closure or a soft recapitalisation policy. This result follows immediately from equation (20) and the fact that reported net worths are independent of the realised state of nature. Expected utility from monetary bene®ts does not depend on the probabilities of various states of nature which a bank manager can in¯uence by choosing effort ex ante. Under a conditional bank recapitalisation policy linked to liquidations of ®rms in default, high-powered incentives schemes for bank managers have yet another effect. The introduction of such incentives can undermine the conditional bank recapitalisation scheme's effectiveness by creating a monetary incentive for managers of insolvent banks not to participate in the scheme and to misreport net worth. In fact, given the bank manager's objective of the form given by equation (20), no bank manager would have an incentive to participate in the conditional recapitalisation scheme. Consider in particular a bank in state W4 . The c1 + B utility of its manager from misreporting a net worth of W1 is bW

46 Incentive Effects of Conditional Bank Recapitalisation

while her utility from participating in the conditional recapitalisation scheme is only B. Thus all bank managers would report a net worth of W1 under a conditional bank recapitalisation scheme with a high-powered incentive. Finally, contrasting the various recapitalisation policies from the point of view of ex ante effort incentives, we ®nd that, since reported net worth under a strict closure policy is independent of the state of nature, highpowered incentives provide no additional ex ante effort incentive beyond that associated with a bank manager's private bene®ts. Therefore, the ex ante incentive constraint remains: ‰1 …1† � 1 …0†ŠBW1  c

…21†

However, the ex ante incentive constraint under a soft recapitalisation is now changed to: d1 ‡ ‰4 …1† � 4 …0†ŠbW4  c ‰1 …1† � 1 …0†ŠB:W

…22†

In particular, for given private bene®ts B, there exists a cut-off level of b, say bs , such that for b < bs (respectively b > bs ) the strict closure policy and/ or our conditional scheme provide better (respectively worse) ex ante incentives than a soft recapitalisation policy. In other words, since the bene®t of overstating loan losses under a soft recapitalisation increases with the manager's share of monetary bene®ts, the larger b the bigger the ex ante incentive effects under a soft recapitalisation as compared to a strict closure policy or our conditional recapitalisation scheme. A high-powered incentive contract based on a parameter that can be manipulated by bank managers thus not only fails to strengthen incentives for bank managers to perform, but it also introduces new ± or exacerbates existing ± distortions of bank recapitalisation schemes. If banks are state-owned, a possible strengthening of managerial incentives might be to privatise banks so that measures of performance can be based on an observable and veri®able measure of banks' net worth. In particular, privatisation may introduce the possibility of basing highpowered incentives on the equity market's assessment of banks' net worth. In effect, privatisation could be one (admittedly imperfect) way of reducing the information asymmetry between bank managers and the government. Equity market measures of private banks' net worth, however, re¯ect both their true net worth and expectations of any government bailouts. The careful design of bank recapitalisation and closure rules for private banks is thus of paramount importance if this measure of performance is

Philippe Aghion, Patrick Bolton and Steven Fries 47

to prove reliable. These are, of course, important policy issues in their own right, but ones which go beyond the scope of this particular paper.

5 Partial information asymmetry between bank managers and the government In this section, we relax the assumption that a bank manager can misreport the net worth of the bank without limit by introducing a constraint on the scope for misrepresentation. We shall assume that a bank manager can only misreport the net worth of the bank by at most an increment of one state of nature. Therefore, a bank in state 4 can only safely claim to be in state 3 without the misrepresentation being detected by government auditors. Similarly, a bank in state 1 can only safely claim to be in state 2. In easing the extent of the information asymmetry between bank managers and the government, the distortions introduced by banks' strategic behaviour are naturally reduced. The results regarding the social costs associated with an excessive recapitalisation and with inef®cient liquidation/continuation decisions for ®rms in default are stated here without proof. First, the social inef®ciency under the `tough' recapitalisation policy associated with the excessive continuation of ®rms in default by bank managers seeking to hide their bad loans, declines. This social cost falls because banks in state 4 can no longer misrepresent their net worth as W1 or W2 . Second, the easing of the information asymmetry reduces the social inef®ciencies associated with the excessive recapitalisation and liquidations under the `soft' recapitalisation policy, because banks in state 1 can no longer misrepresent their net worth as W3 or W4 . The conditional recapitalisation policy, however, is unaffected by this change. Reducing the information asymmetry between bank managers and the government can either increase or decrease the incentive for bank managers to supply effort ex ante, depending on the government's recapitalisation policy. Under the `tough' recapitalisation policy, the incentive±compatibility constraint now becomes ‰1 …1† � 1 …0†ŠBW1 � ‰4 …1† � 4 …0†ŠB > c

…23†

Because a bank manager in the worst state of nature can no longer misrepresent the bank's net worth to preserve her job, there is an increased incentive for supplying effort with the reduced information asymmetry.

48 Incentive Effects of Conditional Bank Recapitalisation

Under the `soft' recapitalisation policy, however, the incentive for managerial effort can be diminished by the reduction in the information asymmetry because banks in the two better states of nature are less able to attract excessive recapitalisations, with banks in state 1 being unable to misreport and banks in state 2 being only able to misreport that they are in state 3. Thus, easing of the information asymmetry between bank managers and the government improves unambiguously the social ef®ciency of the `tough' recapitalisation policy. Not only is there less misrepresentation of loan quality and fewer excessive recapitalisations of banks, there is also greater incentive for bank managers to supply effort compared with the case of a complete information asymmetry. However, for the `soft' and `inbetween' policies, the gains from less misrepresentation of loan quality and fewer excessive recapitalisations are offset by diminished incentives for managerial effort. This adverse effect of the improved information arises from the fact that banks in better states of nature receive excessive recapitalisation under the `soft' and `in-between' policies with a complete information asymmetry. This feature of the recapitalisation policies provides bank managers with an incentive to exert effort to increase the likelihood of these states occurring, an incentive which is curtailed by reducing the extent of the information asymmetry.

6

Bank managers of heterogeneous quality

In this section we explore another extension where bank managers differ in their abilities in managing a bank pro®tably. We shall suppose that managers can be either good (G) or bad (B). If they are bad they end up in state W4 for sure, while if they are good their bank never ends up in that state. Thus, the realisation of W4 identi®es the manager's type perfectly. Also, bad managers are not able to generate as high liquidation and/or continuation values per loan as good managers. We shall assume that: G ˆ B where > 1 LG ˆ LB There is thus potentially a substantial ef®ciency gain in removing bad managers in state W4 . Presumably, an advantage of `tough' bailout policies is then to facilitate the replacement of bad bank managers. Note, however, that in our model such a policy would not have the desired effect since bad bank managers can conceal the true state of the bank. But, consider a small departure from our model where the true state of the

Philippe Aghion, Patrick Bolton and Steven Fries 49

bank would become public with a small probability  > 0 despite the manager's efforts to conceal the bank's insolvency. There is then the following trade-off in imposing a `tough' bailout policy: the policy brings about an ex post improvement from replacing a bad manager when he or she is discovered, but it leads to a higher cost of misrepresentation. If we denote by  the prior probability that a manager is good and by v ˆ … ‡ 1 � †B L ˆ … ‡ 1 � †LB the expected continuation/liquidation values under a new bank manager, then the expected gain per loan from replacing a bad manager is b4 …L � LB † ‡ …1 � b4 †…v � vB †† …1 � p4 †… while the expected misrepresentation cost is b4 †L …1 � p4 †… � Thus, it pays to impose a `tough' bailout policy only if the probability of detection of insolvency is suf®ciently high: >

b4 †L … � b4 †… � 1† b4 … � 1†LB ‡ …1 �

This observation underscores again the main underlying theme of this paper that `tough' bailout policies are ef®cient only if a powerful auditing and monitoring system is in place which limits bank managers' discretion in writing down non-performing loans. By all appearances, such a system is unlikely to be in place today in Japan. The evidence of banking crises in other advanced economies suggests that such a system was also missing in these economies.

7

Conclusion

The BIS regulations have almost entirely focused on the problem of how to avoid ®nancial distress and how to give banks incentives for prudent lending. In contrast, little attention has been devoted to the problem of how to close or restructure a distressed bank. This problem is complex mainly because banks have considerable private information about loan losses and a lot of discretion in deciding what to write down and what loans to de®ne as non-performing. This paper makes a small step in

50 Incentive Effects of Conditional Bank Recapitalisation

analysing this issue and proposes some general principles which should underlie a bankruptcy institution for banks. These principles have been derived in a highly stylised model and can only serve as a broad framework to organise our analysis of bank recapitalisations. Much additional research is required to be able to specify a more systematic policy towards bank failures.

Notes 1. The research reported in this paper is part of a wider research project on Private Sector Development initiated by the European Bank for Reconstruction and Development (EBRD). We are grateful to the EBRD for ®nancial and intellectual support. 2. Another objective we consider in an extension is to replace inef®cient bank managers. 3. At least four different states of nature are needed in order to compare alternative bank bailout policies. 4. The total continuation value should include the ®rm manager's private value V. This value is assumed to be large enough that the manager would never want to shut down the ®rm willingly. Moreover it is not worth bribing the manager to close down the ®rm whenever the lender's continuation value is equal to zero since V > L. This latter assumption implies that the total continuation value is always higher than the liquidation value (at least in the case where there are no strategic defaults), so that it would be ex post ef®cient not to liquidate. However, since it is clearly ex ante ef®cient to commit to liquidation (so as to provide incentives to repay the debt) the regulator's objective should be to enforce liquidation whenever the bank's continuation value is equal to zero. 5. The loss in social surplus also includes the misallocation of funds which could have been directed to better investments. An important limitation of the ABF model is that it does not account for that cost.

References Aghion, P., P. Bolton and S. Fries (1999) `Optimal Design of Bank Bailouts: The Case of Transition Economies', Journal of Institutional and Theoretical Economics, 155 (1), pp. 51±70. Bolton, P. and D. Scharfstein (1990) `A theory of predation based on agency problems in ®nancial contracting', American Economic Review, vol. 80, no. 1. Mitchell, J. (1995) `Cancelling, transferring or repaying bad debts: cleaning banks' balance sheets in economies in transition', Cornell University mimeograph. Povel, P. (1997) `Optimal ``soft'' or ``tough'' bankruptcy procedures', Financial Markets Group Discussion Paper No. 240, London School of Economics. Suarez, J. (1995) `Closure Rules and the Prudential Regulation of Banks', mimeo, CEMFI, Madrid.

4

Injection of Public Funds into Banks under Deposit Insurance and Bank Regulation Hiroshi Osano

We discuss the optimality of the regulator injecting public funds into a bank in the presence of deposit insurance, and characterise an optimal injection policy to prevent the bank from taking moral hazard action. We show that under certain conditions, the regulator's optimal policy is to inject new cash funds into the bank. Furthermore, if the regulator does not have enough information on the bank, the regulator can inject public funds into the bank through the purchase of subordinated bonds by setting the interest rate equal to zero; on the other hand, if the regulator has enough information on the bank, the regulator may inject public funds into the bank through the purchase of preferred stocks. However, we also indicate that this kind of injection policy cannot be independent of the bank closure policy of the regulator: inef®cient banks should be closed. The author is grateful to Patrick Bolton, Ichiro Ide and Shuji Kobayakawa for helpful comments.

1

Introduction

In March 1998, the Deposit Insurance Corporation injected cash funds to support the liquidity of major Japanese banks. In Japan, the policy of cash injection into major banks is carried out through the purchase of subordinated bonds (loans) and preferred stocks (see Table 4.1). The most troubling aspect of this procedure is the concern that the injection of public funds causes insolvent banks to continue their operations. In such a case, the greater is the possibility of fresh capital injection, the greater the incentive for the bank to gamble with deposit insurance funds. Moreover, although the regulator can inject public funds into banks through the purchase of alternative securities issued by banks such as subordinated bonds, preferred stocks and common stocks, there has been no rigorous analysis of the optimal security design for fresh capital

51

52 Injection of Public Funds into Banks Table 4.1 Injection of public funds in Japan in March, 1998 Ratinga

Financial method

Injected amount (billion yen)

Underwriting condition

Tokyo Mitsubishi

Aa2

100

Libor + 0.90%b

Daiichi Kangyo Sakura

A1 A3

99 100

0.75%c Libor + 1.20%b

Sumitomo

A1

100

Libor + 0.90%b

Fuji

A3

100

Libor + 1.10%b

Sanwa Tokai

A1 A2

100 100

Libor + 0.55%b Libor + 0.90%b

Asahi

A2

100

Libor + 1.00%b

Daiwa

Baa3

100

Libor + 2.70%b

Industrial Bank of Japan Long-term Trust Bank of Japan Nippon Credit Bank Mitsubishi Trust Bank Sumitomo Trust Bank Mitsui Trust Bank Yasuda Trust Bank Toyo Trust Bank Chuo Trust Bank

A2

Permanent subordinated bond Preferred stock Permanent subordinated bond Permanent subordinated bond Permanent subordinated bond Subordinated bond Permanent subordinated loan Permanent subordinated loan Permanent subordinated loan Subordinated bond

100

Libor + 0.55%b

Preferred stock Permanent subordinated bond Preferred stock

130

1.00%c

46.6 60

Libor + 2.45%b 3.00%c

Permanent subordinated bond Permanent subordinated bond Permanent subordinated bond Permanent subordinated bond Permanent subordinated bond Preferred stock

50

Libor + 1.10%b

100

Libor + 1.10%b

100

Libor + 1.45%b

150

Libor + 2.45%b

50

Libor + 1.10%b

32

2.50%c

28 20

Libor + 2.45%b Libor + 1.10%b

20

Libor + 2.45%b

30

Libor + 2.95%b

Baa2

Ba1 Baa1 Baa1 Baa2 Baa2 Baa1 Baa3

Yokohama

A3

Hokuriku

Baa3

Ashikaga

Baa3

Total amount

Permanent subordinated Permanent subordinated Permanent subordinated Permanent subordinated

loan loan loan bond

1,815.6

Notes: a) The rating of each bank is taken from that of Moody's Investors Service. b) The interest rate of bonds or loans is represented by (London Inter-Bank Offered Rate (six month)) + . c) Dividend rate of preferred stocks.

Hiroshi Osano 53

injection into banks. The purpose of this paper is to explore the optimality of the regulator injecting public funds into banks, and to examine the optimal design of the injection policy to avert moral hazard behaviour by banks. These problems are discussed in the situation where the regulator is unable to monitor the extent to which bank resources are devoted to activities that lower the liquidity risks of bank assets. Under the current deposit insurance system, the effect of the injection of public funds depends on several factors. First, if the insurance premium is independent of bank risks,1 an increase in the level of injected capital funds reduces deposit insurance payments to be covered by the regulator in the case of bank insolvency. This bene®t results from the avoidance of early liquidation of banks because one way to keep the banks in operation requires the regulator to provide the banks with additional cash funds. Second, a rise in the level of injected capital funds may induce banks to take moral hazard action and increase the possibility of bank insolvency. This possibility arises if banks are induced to choose a lower effort to reduce the liquidity risk of their assets, because even failing banks may be allowed to continue in business. However, to resolve the moral hazard problem, the regulator can design an optimal scheme for injection of public funds, consisting of the bank closure rule, the level of injection of public funds, and the choice of security purchased by the regulator in the injection of public funds into banks. The combination of bank closure policy and design of security in the injection of public funds is particularly important for the regulator in order to control the problem of moral hazard by banks, because the policy combination affects the moral hazard incentive for banks through a change in the expected value of the residual claims of banks.2 As a result, we can even take account of the possibility that the injection of public funds into banks mitigates the incentive for banks to take moral hazard action. On the other hand, if the regulator does not inject public funds into banks, the regulator uses only the combination of bank closure policy and compensation policy for depositors through deposit insurance payment. The basic assumption in this paper is that the regulator is unable to observe the effort of a bank in reducing a liquidity shock arising from the bank's asset.3 The liquidity shock not only affects the ex post revenues or the liquidation value of the bank but also produces an additional interim liquidity need for the bank to be re®nanced through new deposit funds collected by the bank or through public funds injected by the regulator into the bank. If the regulator injects public funds into the bank, this procedure may reduce the incentive for the bank to choose a higher level

54 Injection of Public Funds into Banks

of effort to reduce the liquidity shock. Hence, the regulator needs to construct an incentive compatible mechanism of cash injection to overcome the moral hazard problem. We focus on the problem of optimal regulatory design by taking as given the regulator's role as a provider of deposit insurance. Thus, we have not directly addressed the question of whether or not government-sponsored deposit insurance should be provided. Our main results are summarised as follows: (1) In the absence of moral hazard action by the bank, the regulator need not inject public funds into the bank. (2) Under moral hazard action by the bank, the regulator needs to inject new cash funds into the bank under certain conditions. (i) If the regulator injects public funds through the purchase of subordinated bonds, then the regulator can set the interest rate of subordinated bonds at zero and set the level of injection of public funds at any level below or equal to the total liquidity shock as long as the bank is kept open; in contrast, the regulator closes the bank and need not inject additional cash funds into the bank if the total liquidity need is higher than a threshold point. (ii) If the regulator injects public funds through the purchase of preferred stocks, then the regulator sets the optimal injection level equal to (a) zero for a lower level of the total liquidity shock, (b) the total liquidity shock itself for intermediate levels of the total liquidity shock; and (c) again to zero for a higher level of the total liquidity shock at which the bank should be closed in period 1. (iii) If the regulator injects public funds through the purchase of common stocks, then an unambiguous rule cannot be obtained without further restriction of the parameter con®guration of the model. (iv) The optimal security design in the injection of public funds depends on the information that the regulator obtains. If the regulator does not have enough information on the bank, the regulator can inject public funds into the bank through the purchase of subordinated bonds into the bank by setting the interest rate equal to zero. If the regulator has enough information on the bank, the regulator may inject cash funds through the purchase of preferred stocks. There is a considerable literature on deposit insurance and bank capital regulation. One strand of the literature focuses on capital requirements in an asymmetric information environment (see, for example, Campbell, Chan and Marino, 1992; Giammarino, Lewis and Sappington, 1993; and

Hiroshi Osano 55

Besanko and Kanatas, 1997). These studies examine the normative issue of deposit insurance and capital requirements. However, they do not analyse a regulatory forbearance policy in the context of optimal bank regulation under asymmetric information Dreyfus, Saunders and Allen (1994), Nagarajan and Sealey (1995) and Rochet and Tirole (1996) develop a rigorous model of forbearance in the context of optimal bank regulation and investigate the endogenous determination of an optimal closure policy. Dreyfus, Saunders and Allen study the role of coverage caps on the scope of insured deposits as an alternative to a regulatory forbearance policy backed by cash infusions, and show that the optimal level of the ceiling for insured deposit coverage is beyond the minimum level consistent with the existence of a feasible ®nancing plan. Nagarajan and Sealey also suggest that regulatory forbearance is optimal under moral hazard when the bank's insolvency is due to factors beyond its control. Rochet and Tirole indicate that the ¯exibility afforded by decentralised interbank loan transactions ± for example, interbank peer monitoring ± does not necessarily make it possible for the regulator to avoid undesirable rescue operations. Aghion, Bolton and Fries (1999) and Mitchell (2001) also discuss how the bank bailout scheme can mitigate or overcome bank managers' incentives to misreport or hide loan losses. What distinguishes our model from those mentioned above is that we explicitly consider the design of security in the injection of public funds by the regulator. To the best of our knowledge, this is the ®rst attempt to analyse rigorously some policy questions regarding the optimal choice of securities purchased by the regulator in the injection of public funds into banks (an exception is Osano (2001), who studies this problem with a managerial compensation contract in a different context). The rest of this paper is organised as follows. Section 2 sets out the basic model, and describes the regulator's optimisation problem. Section 3 analyses the optimal regulatory design problem by taking a choice of security in the regulator's injection policy as given, and then proceed to discuss the problem of security design in the injection policy. Section 4 extends the basic model and investigates how the results obtained in Section 3 are modi®ed. The ®nal section is devoted to conclusions.

2 2.1

The model The basic environment

We consider a three-period model in which a representative bank can issue deposit accounts to depositors under the auspices of a regulator

56 Injection of Public Funds into Banks

offering deposit insurance covering all deposits. All agents are risk neutral. The equilibrium risk-free interest rate is zero; thus, all cash ¯ows are discounted at this rate. In period 0, the bank takes equity capital E and deposit funds D0 , and invests these funds in a portfolio of risky assets K (E + D0 = K). To focus on the effect of the policy for injection of public funds into the bank, we assume that not only the total risky investment level but also the leverage ratio of the bank in the initial period are taken to be ®xed.4 The amount of equity E may then be regarded as ®xed by a capital standard imposed on the bank by the regulator. By assumption, bank deposits are fully insured by the regulator.5 The regulator charges a premium p per insured deposit. The deposit insurance premium is paid by bank owners in period 0. Since this paper is not concerned with the pricing of the `fair' risk adjusted insurance premium, the deposit insurance premium does not necessarily re¯ect the `fair' risk of deposit funds (see Chan, Greenbaum and Thakor, 1992). The bank owners take actions in period 0 to affect the distribution of the additional amount of ®nancing need in period 1 for the operating expenditures of the bank's risky investment, the distribution of the liquidity value of the bank's assets in period 1, and the distribution of the returns realised in period 2 from the bank's risky investment.6 The actions taken by the bank may be interpreted as the bank's monitoring of its borrowing ®rms. We assume that the bank owners can choose a higher  or a lower monitoring level (m). Since the monitoring level (m) monitoring level can be thought of as corporate resources over which the bank owners have discretion, the monitoring activity would be costly to the bank owners. This cost factor is captured by cm, where c is constant. The objective of the bank owners is thus to maximise the value of the bank's initial equity minus the monitoring cost incurred upon them. In period 1, the bank neither engages in any further investments nor receives any cash ¯ows accruing from the borrowing ®rms. Instead, the bank must ®nance an additional amount x for the borrowing ®rms so that x ˆ  ‡ ;

…1†

where  represents the idiosyncratic shock of risky investment projects (borrowing ®rms) and  the aggregate shock of the economy.7 The idiosyncratic liquidity shock  is distributed according to a conditional cumulative distribution function G( j m), with a conditional density function g( j m) on the interval [, ] for each m; whereas the aggregate liquidity shock  is distributed according to a cumulative distribution  These two function F(), with a density function f() on the interval [, ].

Hiroshi Osano 57

liquidity shocks are assumed to be distributed independently. Given the relation (1), the total liquidity shock x is distributed according to a conditional cumulative distribution function G(x �  j m) = G(x j m, ), with a conditional density function g(x �  j m) = g(x j m, ) on the interval  for each m and . [ + ,  + ] We assume that the lower level of monitoring shifts the distribution of the total liquidity shock in the sense of ®rst-order stochastic dominance (that is, it increases the likelihood of the higher level of x realising for given ). This implies that the distribution G(x j m, ) ®rst-order <  ): G(x j m, ) ˆ stochastically dominates the distribution G(x j m, G(x j  Furthermore, we also assume  ) for all x 2 [ + ,  + ) and all  2 [, ]. m,  ) > g(x j m, ) for x 2 that there exists a threshold level xo such that g(x j m,  ) < g(x j m, ) for x 2 (x8,  + ]. This assumption is [ + , xo ) and g(x j m, satis®ed for all unimodal distributions under ®rst-order stochastic dominance. From now on, we use the distribution of x instead of the distribution of  for the convenience of analysis. After the observation of the total liquidity shock x but before the realisation of the bank's returns from risky investment, the regulator can choose either to keep the bank open or to close the bank immediately. We assume that the regulator cannot observe the monitoring level of the bank, m, nor the idiosyncratic liquidity shock, , although she can observe the other liquidity shocks and know the distribution functions. Given the observed total liquidity shock x and the observed aggregate liquidity shock , the regulator then decides whether or not to allow the bank to operate until period 2. This choice is captured as a threshold point of the total liquidity shock xB () for each : the bank is allowed to continue until < period 2 if x ˆ xB (); and the bank is closed in period 1 if x > xB (). Since xB () depends on the observed aggregate liquidity risk , forbearance policies adopted by the regulator are state-contingent.8;9 If the regulator allows the bank to operate until period 2, it may inject fresh capital into the bank to avoid the current or future liquidation of the bank's portfolio. Let I(x, ) denote the level of the injection of cash funds in period 1. Because bank deposits are fully insured by the regulator, the remaining liquidity need is re®nanced by additional new deposit funds D1 (x, ). Thus, I(x, ) + D1 (x, ) = x.10 Indeed, the injection of public funds does not let the bank have free additional funds because the regulator can inject public funds through the purchase of securities issued by the bank. Here, we focus on the following three kinds of security: subordinated bonds, preferred stocks, or common stocks. If public funds are injected through the purchase of subordinated bonds, the regulator is senior to the bank owners in the event of bank

58 Injection of Public Funds into Banks

insolvency in period 2 after total deposits are repaid to depositors. Thus, residual claims after the repayment to depositors are devoted ®rst to the repayment of subordinated bonds owned by the regulator. Similarly, if public funds are injected through the purchase of preferred stocks, we assume that the regulator is senior to the bank owners with respect to the bank's residual assets in the event of bank insolvency in period 2 if the other claimholders are fully repaid from the bank's ®nal returns. Thus, the main difference between subordinated bonds and preferred stocks in this paper is that the regulator can also claim dividends in period 2 under preferred stocks if the bank's returns from risky investment are larger than the sum of total deposits and outstanding preferred stock book values.11 Finally, if public funds are injected through the purchase of common stocks, both the regulator and the bank owners have an equal right to receive residual claims in period 2 after depositors are fully compensated. Since the price of common stocks to the regulator is the same as that to the bank owners, the situation still holds true even though the bank buys back its common stocks from the regulator. Now, suppose that the bank is allowed to operate until period 2. Then, the total liquidity need x is re®nanced; and the bank's ®nal returns from the maturity of risky investment, R � R , are realised in period 2. The variable R is modelled as a random variable according to a cumulative distribution function H(R), with a density function h(R) on the interval  We assume that the random variable R is distributed independently [R, R]. of x and . However, the bank's returns from risky investment decrease with an increase in  through the additional term �R . Depositors are repaid from the bank's ex post revenues. If the bank is insolvent, the regulator covers all the deposit funds by tax proceeds collected from other sectors of the economy. The bank owners may receive some part of the residual asset value of the bank according to the seniority rule of the securities purchased by the regulator for the injection of cash funds into the bank. On the other hand, if the bank is closed in period 1, then the total liquidity need x is not re®nanced; and the bank's assets are liquidated. Depositors are repaid only from the liquidation value L � L  in period 1. For simplicity, we assume that the variable L is deterministic. The liquidation value decreases with an increase in  through the additional term �L . Again, if the bank is insolvent, the regulator covers all the deposit funds by tax proceeds collected from other sectors of the economy. The bank owners can receive the residual asset value only if the bank's liquidation value is large enough to pay off the deposit claims.

Hiroshi Osano 59 Figure 4.1 Timing of the regulator's and bank's decision processes given the security choice in the injection of public funds

t = 0

bank owners’ effort _ decision m∈ {m, _ m}

idiosyncratic liquidity shock _ ∈ [_, ] ∋ ∋

total liquidity shock x

t=1

aggregate liquidity _ shock ξ∈ [_ ξ, ξ]

∋

regulator’s forbearance decision

[continue] x ≤ xB(ξ)

regulator’s injection decision I(x, ξ) newly collected deposit funds D1(x, ξ) = x - I(x, ξ)

t = 2

[close] x > xB(ξ)

realisation of the bank’s liquidity value L - p ξ L

distribution of the residual claims

realisation of the bank’s revenues R -pRξ

distribution of the residual claims

The timing of the regulator's and bank's decision processes is now illustrated in Figure 4.1.

60 Injection of Public Funds into Banks

2.2

The regulator's problem

Given the model structure presented in the preceding subsection, the regulator now designs an optimal regulatory mechanism {p, xB (), I(x, ),  for each choice of securities purchased D1 (x, ) j (x, ) 2 [ + ,  + ]  [, ]} by the regulator in the injection of public funds into the bank: subordinated bonds, preferred stocks, and common stocks. For the resulting optimal regulatory mechanism under each ®nancial scheme, the regulator then determines optimal security design in the injection of public funds by comparing the payoffs attained under the three different ®nancial schemes. For simplicity, we assume that the regulator purchases the same security in the injection policy for all the realised total liquidity shock x and the realised aggregate shock . We will brie¯y discuss the effect of this assumption in section 4.4. To formulate the regulator's problem, we provide several notations. Let  Since the bank y 2 �  {xB (), I(x, ), D1 (x, ) j (x, ) 2 [ + ,  + ]  [, ]}. owners pay off in period 0 the insurance premium p that the regulator receives, let SB (m, y) � p � cm,PB (m, y) � p � cm, and BC (m, y) � p � cm(SR (m, y) + p, PR (m, y) + p, and CR (m, y) + p) denote the ex ante net payoff of the bank (regulator) evaluated in period 0 under the injection of public funds through the purchase of subordinated bonds, preferred stocks, and common stocks, respectively. The ex ante gross payoffs of the j bank and the regulator, i (m, y) (i = B, R; j = S, P, C), will be speci®ed in the next section. We begin by discussing the constraints that the regulator must face. Suppose that the lower monitoring level m is more likely to yield a larger  so that the regulator liquidity shock than the higher monitoring level m always prefers the higher monitoring level m. Then, for each choice of securities purchased by the regulator, the incentive compatibility constraint for the bank is described by > j j B …m; y† � p � cm ˆ B …m; y† � p � cm;

…2†

where j = S, P and C. The bank must also attain at least the zero net expected payoff because the bank owners are risk neutral and the risk-free interest rate is zero. For each choice of securities purchased by the regulator, the individual rationality constraint for the bank is expressed by > j B …m; y† � p � cm ˆ 0; where j = S, P and C.

…3†

Hiroshi Osano 61

The feasibility constraint ensures that the sum of injected public funds I(x, ) and newly collected deposit funds D1 (x, ) ®nances the total liquidity need x: I…x; † ‡ D1 …x; † ˆ x;

for …x; † 2 ‰ ‡ ;  ‡ Š  ‰; Š:

…4†

The regulator also impose the non-negative constraints of injected public and newly collected deposit funds such that > > x ˆ I…x; † ˆ 0;

for …x; † 2 ‰ ‡ ;  ‡ Š  ‰; Š:

…5†

For each choice of securities purchased by the regulator, the regulator's problem is now characterised as follows: j

Maxp;y R …m; y† ‡ p subject to …2† � …5†:

…6†

j

where j = S, P and C. Let VR * denote the optimal value of the maximization problem (6) for j = S, P and C. Then, the regulator selects the optimal security design in the injection of public funds that gives it the maximum of the three values: VSR *, VPR *, and VCR *. We are now in a position to solve the maximisation problem (6). Since the individual rationality constraint (3) is always satis®ed with equality, j  y) � cm.  Substituting p = the insurance premium p is represented by B (m, j   B (m, y) � cm into the objective function of (6), we obtain the following ®rst-order conditions with respect to I(x, ) and xB () for each choice of securities purchased by the regulator (j = S, P, C):12 " j # j j j @R …m; y† @B …m; y† @B …m; y† @B …m; y† ‡ ‡ 1 � ˆ 2 …x; † � 3 …x; †; @I…x; † @I…x; † @I…x; † @I…x; † j @R …m; y†

@xB …†

‡

j @B …m; y†

@xB …†

" j # j @B …m; y† @B …m; y† ˆ 0; ‡ 1 � @xB …† @xB …†

…7† …8†

where 1 , 2 (x, ), and 3 (x, ) are the nonnegative multipliers associated > > with the incentive compatibility constraint (2), x ˆ I(x, ), and I(x, ) ˆ 0, j j @R …m;y† @B …m;y† respectively, and @I…x;† and @I…x;† are evaluated under the condition D1 (x, ) = x � I(x, ) from (4). We summarise the de®nitions of the derivatives of (7) and (8) in the Appendix. If the bank's monitoring level is perfectly observed by the regulator, the incentive compatibility constraint is not required. Then, the optimal level of injection of public funds and the optimal threshold level for bank

62 Injection of Public Funds into Banks

closure are determined from (7) and (8) by setting 1 = 0. Given the choice of securities purchased by the regulator, this corresponds to the ®rst-best solution to the regulator's problem as a benchmark against which the solution under moral hazard can be compared. In fact, since the bank's monitoring level cannot be observed by the regulator, the problem of moral hazard action by the bank cannot be neglected. Then, we can suppose 1 >j 0 so that the incentive compatibility constraint is strictly j j j @B …m† @B …m† @B …m† @B …m† binding. If @I…x;† � @I…x;† > 0 …or @I…x;† � @I…x;† < 0† the bank has more (or less) incentive to choose the higher monitoring effort as the level of injection of public funds is greater. This is so because the incentive compatibility constraint for the j bank isj more (or less) relaxed with an j j @ …m† @ …m† @ …m† @ …m† increase in I(x, ). Similarly, if @xBB …† � @xBB …† > 0…or @xBB …† � @xBB …† < 0† the bank has more (or less) incentive to choose the higher monitoring effort as the threshold point for bank closure is greater.

3

The optimal regulatory policy

To evaluate the ®rst-order conditions of (7) and (8), we need to specify j i (m, y) (i = B, R; j = S, P, C). Thus, for each choice of securities purchased by the regulator, we begin by describing explicitly the bank's ex ante gross j payoff prior to the insurance premium and monitoring payment B (m, y) (j = S, P, C) and the regulator's ex ante gross payoff prior to the insurance j premium receipt R (m, y) (j = S, P, C). Then, we examine the optimality conditions on the regulator's problem under each choice of securities purchased by the regulator. Finally, we compute the regulator's optimal net payoff for each choice of securities purchased by the regulator and determine the optimal security design in the injection of public funds. 3.1

Subordinated bonds

Let us begin with the case of subordinated bonds. For simplicity, we ignore the interest rate of subordinated bonds. In a following section (subsection 4.2), we allow for a non-zero interest rate of subordinated bonds and discuss how our main results in this section are modi®ed. Given D0 = K � E and D1 (x, ) = x � I(x, ) from (4), the bank's ex ante gross payoff prior to the insurance premium and monitoring payment is represented by Z  Z xB …† Z R SB …m; y† ˆ ‰R � R  � K ‡ E � xŠdH…R†dG…xjm; †dF…† Z ‡



‡

L�K‡E L



Z

R ‡K�E‡x

‡

xB …†

‰L � L  � K ‡ EŠdG…xjm; †dF…†:

…9†

Hiroshi Osano 63

The ®rst integral indicates the expected value of residual claims retained by the bank owners in period 2 if the bank is allowed to continue until period 2. The second integral denotes the expected value of residual claims retained by the bank owners in period 1 if the bank is closed in period 1. Several remarks on this representation are in order. First, the regulator closes the bank in period 1 as long as xB () < x. Second, we should also notice that the bank's ex post residual claims are positive as long as the bank's ex post asset proceeds are larger than the sum of deposit and > subordinated bond claims (R � R  ˆ K � E + x � I(x, ) + I(x, ) if the bank is allowed to continue until period 2, and L � L  > K � E if the bank is closed in period 1). Similarly, using D0 = K � E and D1 (x, ) = x � I(x, ) from (4), the regulator's ex ante gross payoff prior to the insurance premium receipt is described by SR …m; y† ˆ ‡

Z Z 

Z Z 

xB …† ‡ xB …†

Z Z

R

R ‡K�E‡x

I…x; †dH…R†dG…xjm; †dF…†

R ‡K�E‡x

R ‡K�E‡x�I…x;†

‡

‰R � R  � K ‡ E � x

‡ I…x; †dH…R†dG…xjm; †dF…† Z  Z xB …† Z R ‡K�E‡x�I…x;† ‰K � E ‡ x � I…x; † � 

‡

R

� R ‡ R ŠdH…R†dG…xjm; †dF…† Z  Z xB …† I…x; †dG…xjm; †dF…† � Z �



‡

 L�K‡E L

Z

‡

xB …†

‰K � E � L ‡ L ŠdG…xjm; †dF…†:

…10†

Here, the ®rst and second integrals are the expected repayments of subordinated bonds from the bank to the regulator in period 2 if the bank is allowed to operate until period 2. The third integral represents the expected deposit insurance payment to be covered by the regulator in period 2 if the bank is allowed to continue until period 2 and if the bank's assets are insuf®cient. The fourth integral indicates the amount of cash funds injected into the bank in period 1. The ®nal integral expresses the expected deposit insurance payment to be covered by the regulator in period 1 if the bank is closed in period 1 and if the bank's assets are insuf®cient.

64 Injection of Public Funds into Banks

Several comments on the representation (10) should be made. First, the regulator obtains from the bank the full amount of the promised > repayment of subordinated bonds as long as R ˆ R  + K ± E + x. The regulator can also receive a partial amount of the promised repayment if > > R  + K � E + x ˆ R ˆ R  + K � E + x � I(x, ). Second, if the bank is closed in period 1, the regulator need not inject fresh capital into the bank. Thus, the regulator does not have the possibility of obtaining the ex post residual > claims of the bank if x ˆ xB (). We next proceed to derive the partial differentiation results of SB (m, y) and SR (m, y) with respect to I(x, ) and xB (). Given I(x, ) = 0 in the case of bank closure in period 1 (that is, x 2 (xB (),  + ]), partial differentiation of (9) and (10) with respect to I(x, ) on x 2 [ + , xB ()] shows that @SB …m; y† ˆ 0; @I…x; †

for …m; x; † 2 fm; mg  ‰ ‡ ; xB …†Š  ‰; Š;

…11†

@SR …m; y†

ˆ 0; @I…x; †

for …m; x; † 2 fm; mg  ‰ ‡ ; xB …†Š  ‰; Š:

…12†

Equations (11) and (12) imply that an increase in the amount of public funds injected through the purchase of subordinated bonds has no effects on the ex ante gross payoff of either the bank or the regulator. The neutral effect on the ex ante gross payoff of the bank is due to the consideration that the bank's ex post residual claims are independent of the ®nancial method, whether the total liquidity need is ®nanced by newly collected deposit funds or subordinated bonds under the zero interest rate of subordinated bonds. Thus, an increase in the amount of public funds injected with subordinated bonds does not induce the bank to take the lower monitoring level. Furthermore, the additional cash injection by the regulator through the purchase of subordinated bonds simply cancels out the sum of the increasing expected repayment of subordinated bonds from the bank to the regulator and the decreasing expected payment of deposit insurance to be covered by the regulator in the event of bank insolvency. These arguments indicate that the injection of public funds with subordinated bonds has no effect on the ex ante gross payoff of the regulator. Given (5) and (7), these ®ndings lead us to state that the level of injection of public funds with subordinated bonds is indeterminate in the range of [0, x] if the bank is kept open. We also partially differentiate (9) and (10) with respect to xB () for (m, )  and obtain 2 {m, m}  [, ] ZR @SB …m; y† ˆ g…xB …†jm; †f …†f ‰R � R  � K ‡ E @xB …† R ‡K�E‡xB …† � xB …†ŠdH…R† ‡ …†…K � E � L ‡ L †g;

…13†

Hiroshi Osano 65

Z R ‡K�E‡xB …† @SR …m; y† ˆ g…xB …†jm; †f …†f ‰R � R  � K ‡ E @xB …† R � xB …†ŠdH…R† ‡ ‰1 � …†Š‰K � E � L ‡ L †g; ZR @SB …m; y† @SR …m; y† ‡ ˆ g…xB …†jm; †f …†f ‰R � R  � K ‡ E @xB …† @xB …† R � xB …†ŠdH…R† ‡ ‰K � E � L ‡ L Šg;

Z f

…14†

…15†

@SB …m; y† @SB …m; y† � ˆ ‰g…xB …†jm; † � g…xB …†jm; †Šf …† @xB …† @xB …† R

R ‡K�E‡xB …†

‰R � R  � K ‡ E � xB …†ŠdH…R† ‡ …†…K � E � L ‡ L †g;

…16† < < < L�K ‡E where …† ˆ 1 if  ˆ  ˆ L�K‡E ; and …† ˆ 0 if <  ˆ : L L Equation (13) implies that an increase in the threshold point for bank closure raises the ex ante gross payoff of the bank prior to the insurance premium and monitoring payment for  2 …L�KL‡E ; Š, but that such an increase has ambiguous effects for  2 ‰; L�KL‡EŠ: This is so because the higher threshold point for bank closure leads to the larger expected residual claims of the bank in the case of bank forbearance in period 1 for all , whereas it also forces the bank to lose the possibility of receiving the residual claims in the case of bank closure in period 1 for  2 ‰; L�KL‡EŠ: Equation (14) suggests that a rise in the threshold point for bank closure reduces the ex ante gross payoff of the regulator prior to the insurance premium receipt for  2 ‰; L�KL‡EŠ; but that its effect is ambiguous for  2 …L�K‡E L ; Š: The reason is that the higher threshold point for bank closure increases not only the expected payments of deposit insurance but also the expected losses due to the insolvency of subordinated bonds in the case of bank forbearance in period 1 for all , while it reduces the expected payments of deposit insurance in the case of bank closure in period 1 if  2 …L�KL‡E ; Š: Equation (15) represents the effect of an increase in the threshold point for bank closure on the sum of the ex ante gross payoffs of the bank and the regulator, SB (m, y) + SR (m, y). Equation (16) also shows that the effect of an increase in the threshold point for bank closure on the incentive for the bank to take the lower monitoring level depends on several factors:   L�K‡E <   < L�K‡E and g…xB …†jm; † < g…xB …†jm; † …that is;  < < L and xB …† > x †. L As indicated in (8), the threshold point for bank closure is now determined so as to balance these effects. Since these effects and the multiplier 1 are independent of I(x, ) (as shown in (11), (12), (15), and (16)), the equilibrium threshold point for bank closure in the presence of

66 Injection of Public Funds into Banks

cash injection through the purchase of subordinated bonds is equal to that in its absence. We now characterise an optimal solution to the benchmark case of no moral hazard, that is, 1 = 0. Given (7), (8), (11), (12), and (15) with I(x, ) = 0 for x > xB (), we obtain the following proposition. Proposition 1 Suppose that the regulator injects public funds into the bank through the purchase of subordinated bonds. In the absence of moral hazard action by the bank, the optimal level of the injection of public funds I* (x, ) is given by I  …x; † 2 ‰0; xŠ; I  …x; † ˆ 0;

< for x ˆ xB …†; for x > xB …†:

…17a† …17b†

The optimal threshold level for bank closure x*B () is determined so that (8) holds with 1 = 0 and (15). This threshold point is the same as that obtained in the absence of the injection of public funds. Now, we proceed to examine an optimal solution under moral hazard, 1 > 0. Given (7), (8), (11), (12), (15) and (16) with I(x, ) = 0 if x > xB (), we have the following proposition. Proposition 2 Suppose that the regulator injects public funds into the bank through the purchase of subordinated bonds. In the presence of moral hazard action by the bank, the optimal level of the injection of public funds I*(x, ) is still given by I  …x; † 2 ‰0; xŠ; 

I …x; † ˆ 0;

< for x ˆ xB …†; for x >

xB …†:

…18a† …18b†

The optimal threshold level for bank closure xB () is determined so that (8) holds with (15) and (16). This threshold point is the same as that obtained in the absence of the injection of public funds. Proposition 2 suggests that if the regulator injects public funds into the bank through the purchase of subordinated bonds under moral hazard, the regulator can set the optimal level of injection of public funds at any level below or equal to the total liquidity need as long as the bank continues; in contrast, the regulator need not inject additional cash funds into the bank if the total liquidity need is higher than a threshold point xB (). Comparing this result with that of Proposition 1, we see that the optimal rule for injection of public funds in the presence of moral hazard is identical with that in the absence of moral hazard, in which the

Hiroshi Osano 67

incentive compatibility constraint plays no role in the ®rst-order conditions with respect to the level of injection of public funds. Thus, if the bank closure rule is taken as given, the injection decision of the regulator is ef®cient even in the presence of moral hazard. Furthermore, under moral hazard, the bank is closed on a criterion that is independent of whether or not public funds are injected. However, since the incentive compatibility constraint plays a key role in the ®rst-order conditions with respect to the threshold point for bank closure, the optimal closure rule under moral hazard is different from the ef®cient one without moral hazard. The intuition behind Proposition 2 can be clari®ed as follows. If the bank is kept open, an increase in the amount of public funds injected through the purchase of subordinated bonds reduces the regulator's expected liabilities of deposit insurance payments because the possibility of insolvency of bank deposits decreases. In fact, this gain from the reduction of the regulator's expected liabilities of deposit insurance payments is offset by the cost from the increased possibility of insolvency of subordinated bonds. On the other hand, the bank's ex post residual claims are independent of whether the total liquidity need is ®nanced by new deposit funds or subordinated bonds. Thus, the injection of capital funds through the purchase of subordinated bonds does not affect the incentive for the bank to choose the lower monitoring effort. As a result, the regulator is indifferent to any level of injection of public funds: the level of injection of public funds is indeterminate. This neutralises the effect of an increase in the cash injection on the equilibrium threshold point for bank closure. The equilibrium threshold point for bank closure, therefore, does not depend on whether or not public funds are injected. 3.2.

Preferred stocks

We next discuss the case of preferred stocks. Given D0 = K � E and D1 (x, ) = x � I(x, ) from (4), the bank's ex ante gross payoff prior to the insurance premium and monitoring payment is represented by PB …m; y† ˆ

Z Z 

xB …†

‡

Z

R

E ‰R � R  � K ‡ E E ‡ I…x; † R ‡K�E‡x

�xŠdH…R†dG…xjm; †dF…† Z L�K‡E Z ‡

L

‰L � L  � K ‡ EŠdG…xjm; †dF…†: ‡ 

xB …†

…19†

The ®rst integral expresses the expected value of residual claims retained by the bank owners in period 2 if the bank is allowed to continue until

68 Injection of Public Funds into Banks

period 2. Note that the bank owners receive the ratio E‡IE…x;† of the residual claims if the regulator injects public funds into the bank through the purchase of preferred stocks. The second integral represents the expected value of residual claims retained by the bank owners in period 1 if the bank is closed in period 1. The regulator's ex ante gross payoff prior to the insurance premium receipt is described by PR …m; y† ˆ

Z Z 

xB …†

Z

R

I…x; † f ‰R � R  � K ‡ E � xŠ E ‡ I…x; † R ‡K�E‡x

‡

‡ I…x; †gdH…R†dG…xjm; †dF…† Z  Z xB …† Z R ‡K�E‡x ‡ ‰R � R  � K ‡ E � x 

R ‡K�E‡x�I…x;†

‡

‡ I…x; †ŠdH…R†dG…xjm; †dF…† Z  Z xB …† Z R ‡K�E‡x�I…x;† ‰K � E ‡ x � I…x; † � R � 

‡

R

‡ R ŠdH…R†dG…xjm; †dF…† Z  Z xB …† I…x; †dG…xjm; †dF…† � Z �



‡

 L�K‡E L

Z

‡ xB …†

‰K � E � L ‡ L ŠdG…xjm; †dF…†:

…20†

Here, the ®rst and second integrals are the expected repayments of preferred stocks from the bank to the regulator in period 2 if the bank is allowed to continue until period 2. Note that the regulator receives not I…x;† of the residual claims but also the only the dividends at the ratio E‡I…x;† > revenues from the sale of preferred stocks to the bank if R � R  ˆ K � E + x � I(x, ) + I(x,  + I(x, )).13 The third integral indicates the expected deposit insurance payments to be covered by the regulator in period 2 if the bank is allowed to continue until period 2 and if the bank's assets are insuf®cient. The fourth integral expresses the amount of preferred stocks injected into the bank in period 1. The last integral denotes the expected deposit insurance payments to be covered by the regulator in period 1 if the bank is closed in period 1 and if the bank's assets are insuf®cient. Given I(x, ) = 0 in the case of bank closure in period 1 (x 2 (xB (),  + ]), partial differentiation of (19) and (20) with respect to I(x, ) on x 2 [ + ,  xB ()] shows that for (m, x, ) 2 {m, m}  [ + , xB ()]  [, ],

Hiroshi Osano 69

@PB …m; y† ˆ �g…xjm; †f …† @I…x; † @PR …m; y† ˆ g…xjm; †f …† @I…x; †

Z

Z

R

E‰R � R  � K ‡ E � xŠ dH…R† < 0; ‰E ‡ I…x; †Š2 R ‡K�E‡x …21†

R

E‰R � R  � K ‡ E � xŠ dH…R† > 0; ‰E ‡ I…x; †Š2 R ‡K�E‡x …22†

@PB …m; y† @RP …m; y† ‡ ˆ 0; @I…x; † @I…x; † @PB …m; y† @BP …m; y† � ˆ �‰g…xjm; † � g…xjm; †Šf …† @I…x; † @I…x; † ZR E‰R � R  � K ‡ E � xŠ < dH…R† 0; 2 > ‰E ‡ I…x; †Š R ‡K�E‡x < if and if x x : >

…23†

…24†

The inequality (24) is immediate from the assumption g(x j m, ) � (g(x j m, ) > 0 (or g(x j m, ) � g(x j m ) < 0) if and only if x < xo (x > xo ). Equation (21) implies that an increase in the amount of public funds injected through the purchase of preferred stocks reduces the ex ante gross payoff of the bank prior to the insurance premium and monitoring payment. The reason is that the expected value of the residual claims of the bank decreases as the ratio of preferred stocks to total outstanding stocks is larger. If the liquidity need is re®nanced by injected cash funds instead of newly collected deposit funds, the expected value of the residual claims of the bank decreases. Equation (22) reveals that the additional injection of cash funds through the purchase of preferred stocks raises the ex ante gross payoff of the regulator prior to the insurance premium receipt. This is because the greater ratio of preferred stocks to total outstanding stocks leads to the greater expected value of the residual claims of the regulator if the bank's assets are greater than its liabilities. @PB …m;y† Indeed, Equation (23) indicates that these two effects, @I…x;† and P @B …m;y† , cancel each other out. @I…x;† Inequality (24) shows that the greater amount of public funds injected through the purchase of preferred stocks strengthens or weakens the incentive for the bank to take the lower monitoring action according as x < xo or x > xo . As indicated in (21), a rise in the amount of public funds injected reduces PB (m, y) regardless of the monitoring level chosen by the bank. Thus, the incentive for the bank to choose the lower monitoring

70 Injection of Public Funds into Banks

 y) is level is strengthened or weakened according as this impact on PB (m, larger or smaller than that on PB (m, y). More speci®cally, under our assumption, the effect on moral hazard action by the bank depends on the characteristics of the density function of x, g(x j m, ). Because of the   assumption g(x j m, ) � g(x j m, ) < 0 for x > xo , the moral hazard incentive for the bank is strengthened or weakened according to which x is smaller or larger than xo . Given (23) and (24), we now determine the level of I(x, ) by evaluating (7). In the absence of moral hazard 1 = 0, it follows from (7) and (23) that 2 (x, ) � 3 (x, ) = 0 for all x and . This implies that I(x, ) is indeterminate in the range of [0, x] for all . On the other hand, in the presence of moral hazard 1 > 0, it is found from (7), (23) and (24) that   > 2 (x, ) � 3 (x, ) > 0 according as x > xo: Because of i (x, ) ˆ 0 for i = 2, 3   [,  this ®nding shows that and I(x, ) = 0 for all (x, ) 2 (xB (),  + ] 3 (x, ) > 0 (that is, I(x, ) = 0) if x < min(xB (), xo ); 2 (x, ) > 0 (that is, I(x, ) < = x) if min(xB (), xo ) < x ˆ xB (); and I(x, ) = 0 if x > xB (). We next partially differentiate (19) and (20) with respect to xB () for  (m, ) 2 {m, m}  [, ]: ZR @PB …m; y† E ‰R � R  � K ˆ g…xB …† j m; †f …†f @xB …† E ‡ I…x B …†; † R ‡K�E‡xB …† ‡ E � xB …†ŠdH…R† ‡ …†…K � E � L ‡ L †g; @PR …m; y† @xB …†

Z ˆ g…xB …† j m; †f …†f

…25† R

I…xB …†; † ‰R � R  � K I…xB …†; †

R ‡K�E‡xB …† E ‡ Z R ‡K�E‡xB …†

‡ E � xB …†ŠdH…R† ‡

‰R � R  � K ‡ E

R

� xB …†ŠdH…R† ‡ ‰1 � …†Š‰K � E � L ‡ L Šg; @PB …m; y† @xB …†

‡

@RP …m; y† @xB …†

ZR ˆ g…xB …† j m; †f …†f ‰R � R  � K ‡ E R

…26† …27†

� xB …†ŠdH…R† ‡ ‰K � E � L ‡ L Šg; @PB …m; y† @xB …†

�

@PB …m; y† @xB …†

ˆ ‰g…xB …† j m; † � g…xB …† j m; †Šf …† Z f

R

R ‡K�E‡xB …†

E ‰R � R  � K E ‡ I…xB …†; †

‡ E � xB …†ŠdH…R† ‡ …†…K � E � L ‡ L †g; < < where …† ˆ 1 if  ˆ  ˆ L�K‡E L ; and …† ˆ 0 if

L�K ‡E L

< <  ˆ :

…28†

Hiroshi Osano 71

Given (25)±(28), we can repeat arguments on the optimal threshold point for bank closure that are similar to those in the subordinated bond case. Thus, in the subsequent analysis, we only brie¯y characterise the optimal threshold point for bank closure in the preferred stock case. In the absence of moral hazard, it is apparent from (8), (15) and (27) that the optimal threshold point for bank closure in the preferred stock case is the same as that in the subordinated bond case. On the other hand, under moral hazard, we can ®nd one important difference between the optimal threshold points for bank closure in these two cases: in the presence of cash injection through the purchase of preferred stocks, the optimal threshold point for bank closure is not the same as that in its > absence if xB () ˆ xo . This is because the regulator is not indifferent to any injection level in the preferred stock case: the regulator uses the injection > > ®nancing up to the maximal level it can set, I (x, ) = x, if xB () ˆ x ˆ o min (xB (),x †. The foregoing arguments lead us to show the following propositions: Proposition 3 Suppose that the regulator injects public funds into the bank through the purchase of preferred stocks. In the absence of moral hazard, both the optimal rule for injection of public funds I* (x, ) and the optimal threshold point for bank closure xB () are the same as those determined in the case of the injection of public funds through the purchase of subordinated bonds. Proposition 4 Suppose that the regulator injects public funds into the bank through the purchase of preferred stocks. In the presence of moral hazard, the optimal level of the injection of public funds I* (x, ) is given by14 I  …x; † ˆ 0; 

I …x; † ˆ x; 

f or …x; † 2 ‰" ‡ ; min…xB …†; x ††  ‰; Š; 

f or …x; † 2 ‰min…xB …†; x †; xB …†Š  ‰; Š;

I …x; † ˆ 0;

f or …x; † 2 …xB …†;  ‡ Š  ‰; Š:

…29a† …29b† …29c†

The optimal threshold point for bank closure xB () is determined so that (8) holds under (27) and (28). This optimal threshold point xB () is not the same as that obtained in the absence of the injection of public funds. Proposition 3 suggests that under no moral hazard, not only the optimal rule for injection of public funds but also the optimal threshold point for bank closure are independent of the variety of securities purchased by the regulator in the injection of public funds. In this sense, this proposition can be interpreted as a restatement of the classical Modigliani±Miller theorem in the event of fresh capital injection into the bank.

72 Injection of Public Funds into Banks

Under moral hazard action by the bank, Proposition 4 shows that the optimal level of injection of additional cash funds into the bank through the purchase of preferred stocks is equal to (i) zero for the lower level of the total liquidity shock, (ii) the total liquidity shock itself for an intermediate level of the total liquidity shock, and (iii) zero again for the higher level of the total liquidity shock at which the bank should be closed in period 1. Thus, the optimal threshold point for bank closure in the presence of the injection of public funds is not the same as that in the absence of their injection. The rational behind Proposition 4 can be clari®ed as follows. If the bank is kept open, an increase in the amount of public funds injected through the purchase of preferred stocks reduces the expected liabilities of the regulator covering deposit insurance payments because the possibility of insolvency of bank deposits declines. Furthermore, if the bank's assets are larger than its deposit claims, the additional injection of public funds enables the regulator to receive more claims (up to the injection amount of preferred stocks) as a preferred stockholder before the bank owners are paid the residual claims. However, these gains are exactly counterbalanced by the injection payments incurred by the regulator through the purchase of preferred stocks. On the other hand, since the regulator also > shares the residual claims with the bank owners if R ˆ R  + K � E + x, an increase in the amount of public funds injected raises the ex ante gross payoff of the regulator prior to the insurance premium receipt. Indeed, this decreases the ex ante gross payoff of the bank prior to the insurance premium and monitoring payment. These two effects cancel each other  As a result, the out so that (23) holds for all (x, ) 2 [ + ,  +  ]  [, ]. remaining effect of the additional injection of public funds through the purchase of preferred stocks is only the effect on the incentive for the bank to choose the higher monitoring effort. This effect is captured by a  y) � PB (m, y) for the lower and higher levels of the total decline in PB (m,  [ [xB (),  + ]  [, ],  liquidity shock, (x, ) 2 [ + , min (xB (), (xo )]  [, ] P P and by a rise in B (m, y) � B (m,  y) for the intermediate level of the  total liquidity shock, (x, ) 2 [min (xB (), xo ), xB ()]  [, ]. For the lower and higher levels of the total liquidity shock, the remaining incentive effect of the injection of public funds strengthens the incentive compatibility constraint by inducing the bank to take the lower monitoring level. The injection of public funds through the purchase of preferred stocks, therefore, is not optimal. The total liquidity need is re®nanced through the collection of insured new deposits for a lower level of the total liquidity shock, but is not re®nanced for a higher level of the total liquidity shock at which the bank should be closed in period 1. On

Hiroshi Osano 73

the other hand, for an intermediate level of the total liquidity shock, the remaining incentive effect mitigates the incentive compatibility constraint by motivating the bank to take the higher monitoring level. Because the remaining incentive effect is effective for any level of I (x, ) 2 [0, x], the optimal injection level increases to the total liquidity shock x. 3.3

Common stocks

Since the analysis of the injection of public funds through the purchase of common stocks is similar to that through the purchase of preferred stocks, we only brie¯y discuss an optimal regulatory policy in this case. A detailed analysis is presented in the Appendix. In the absence of moral hazard, we can set the optimal rule for injection of public funds to be identical with that described in Proposition 1 and 3. In contrast, under moral hazard, we cannot obtain an unambiguous optimal rule for the injection level. However, if the injection of public funds with common stocks induces the bank to choose the lower monitoring level, then such an injection is not optimal. With regard to the optimal threshold point for bank closure, we can repeat arguments similar to those in the case of subordinated bonds or preferred stocks. In the absence of moral hazard, the optimal threshold point for bank closure in this case is again the same as in the case of subordinated bonds or preferred stocks. Under moral hazard, the optimal threshold point for bank closure in the presence of the injection of public funds with common stocks is not the same as that in the absence of their injection because the optimal level of injection of public funds is not necessarily equal to zero. 3.4

Optimal security design in the injection of public funds

In subsections 3.1±3.3, we have studied the regulator's problem by taking as given the security purchased by the regulator for the injection of public funds. We now consider the problem of optimal security design for the regulator. For simplicity, the regulator is assumed to select the same security for all the realisations of the observed total and aggregate liquidity shocks, x and . Without moral hazard, the discussions in subsections 3.1±3.3 show that both the optimal level of injection of public funds and the optimal threshold point for bank closure are independent of the choice of the security purchased by the regulator in the injection policy. Thus, the welfare level of the regulator does not depend on its choice of security in the injection policy. This implies that the security choice in the injection of public funds is a matter of indifference for the regulator. The resulting

74 Injection of Public Funds into Banks

combination of the level of injection of public funds and the threshold point for bank closure is the ®rst-best. Under moral hazard, if the regulator does not have enough information on the effect of the injection policy and if it has some good reason to avoid ®nancial fragility, Proposition 2 suggests that it should use subordinated bonds that have a neutral effect on the ef®ciency of the banking system. In this case, the regulator may inject additional capital funds up to the total liquidity shock in¯icted upon each bank and may set the interest rate of subordinated bonds equal to zero, although a bank with high total liquidity shock should be closed. However, if the regulator has more information on the effect of the injection policy, Proposition 4 suggests that the regulator should use preferred stocks for intermediate levels of the total liquidity shock. Since the injection policy with preferred stocks can motivate the bank to choose the higher monitoring level for that range of the total liquidity shock, it would be better than the injection policy with subordinated bonds. In contrast, for the lower or higher level of the total liquidity shock, the regulator should not inject fresh capital into the bank through the purchase of preferred stocks because this procedure induces the bank to choose the lower monitoring effort. In fact, the regulator has an alternative option of injecting public funds through the purchase of common stocks. However, as indicated in subsection 3.3, the effect of the injection policy with common stocks is indeterminate. Furthermore, if the bank is more likely to be induced to take moral hazard action, this scheme is detrimental to the ef®ciency of the banking system. Several remarks on this policy conclusion should be added. First, we presume that the regulator requires the bank to repay the injecting subordinated bonds or buy back the injecting preferred stocks. If this requirement is broken or not imposed, our results do not hold. Second, our regulatory procedure demands that the regulator should close banks with high levels of total liquidity shock. This requirement is one of the most critical points of our regulatory scheme.

4

Possible extensions of the basic model

In this section, we relax several restrictive assumptions in the basic model and discuss how the results obtained in the preceding section are thus modi®ed. 4.1

Ex post moral hazard

Up to this point, we have assumed that the bank takes its moral hazard action before the regulator injects cash funds into the bank. However, one

Hiroshi Osano 75

can envisage circumstances in which the bank takes its moral hazard action after the regulator injects public funds into the bank. The timing of the regulator's and bank's decision processes is then the following: (i) In period 0, the bank takes equity capital and deposit funds, and invests these funds in a portfolio of risky assets. Again, we assume that not only the total risky investment level but also the leverage ratio of the bank in the initial period are taken as ®xed. The regulator chooses a regulatory scheme consisting of the bank closure policy, the injection amount and the choice of securities. (ii) In period 1, the liquidity shock is realised. Then, the regulator chooses either to keep the bank open or to close the bank immediately. (a) If the regulator allows the bank to continue until period 2, the regulator injects public funds into the bank through the purchase of securities. Given the amount of public funds injected, the bank chooses a monitoring level that affects the distribution of the returns realised in period 2 from the bank's risky investment. (b) If the regulator decides to close the bank, the bank is liquidated. (iii) In period 2, the ®nal returns of the bank from the maturity of risky investment are realised if the bank is allowed to continue until period 2. We assume that the higher level of monitoring shifts the distribution of the bank's ex post returns from its risky investment in the sense of ®rstorder stochastic dominance (that is, it increases the likelihood that the higher level of the bank's ex post returns is realised). Let m denote the monitoring level of the bank, H (R j m) the distribution function of the bank's ex post returns, and h (R j m) the density function of the bank's ex post returns. Suppose that the bank can choose a higher monitoring level (m) or a lower monitoring level (m). Then, the above assumption implies that the distribution H (R j m) ®rst-order stochastically dominates <  the distribution H (R j m): H (R j m) ˆ H (R j m) for all R 2 [R, R]. Under these assumptions, we obtain the following results: (1) In the absence of the moral hazard incentive for the bank, the regulator need not inject public funds into the bank. (2) In the presence of the moral hazard incentive for the bank, the regulator needs to inject new cash funds into the bank through the purchase of securities under certain conditions. (i) If the regulator injects cash funds into the bank through the purchase of subordinated bonds, then an optimal injection level is set (a) at any

76 Injection of Public Funds into Banks

level below or equal to the liquidity shock itself if the bank is permitted to continue in period 2, and (b) equal to zero if the bank should be closed in period 1. (ii) If the regulator injects cash funds into the bank through the purchase of preferred or common stocks, then we cannot establish an unambiguous rule even though we assume that the high monitoring level shifts the distribution of the bank's ex post returns in the sense of ®rst-order stochastic dominance. The main feature that distinguishes the results obtained in the case of the ex post moral hazard incentive for the bank from those obtained in the ex ante case is that the policy of injection through the purchase of preferred stocks has unambiguous effects in the case of the ex ante moral hazard incentive for the bank, but not in the case of the ex post moral hazard incentive. 4.2

Interest rate of subordinated bonds

We next allow for the interest rate of subordinated bonds and investigate whether or not our results will continue to hold. Let r denote the interest rate of subordinated bonds. For simplicity, we assume that the interest rate of subordinated bonds is exogenously ®xed before the resolution of the total liquidity shock. Then, as argued in the Appendix, we see that the injection policy with subordinated bonds does not have neutrality. However, if the regulator can freely adjust the interest rate of subordinated bonds, then it can neutralise the effect of the injection policy by setting the interest rate equal to zero. Indeed, if the regulator has enough information, it may demand a positive interest rate of subordinated bonds and motivate the bank to choose the higher monitoring effort. 4.3

Variety of preferred stocks

In Section 3, we considered only one type of preferred stock, wherein preferred stock holders are senior to common stockholders with respect to the residual claims of the bank after depositors (and, implicitly, senior debt holders) are fully compensated. In fact, there are several kinds of preferred stock. In one, preferred stock holders are senior to common stockholders with respect to the dividend payments of the bank, while in another preferred stock holders are senior to common stockholders with respect to both the residual claims and dividend payments of the bank. The former type is closer to common stocks, whereas the latter type can still be dealt with a framework similar to that in subsection 3.2.

Hiroshi Osano 77

4.4 A more general class of forms of security in the injection of public funds In addition to the problems mentioned in subsections 4.2 and 4.3, we can take account of more general forms of security. In Section 3, for all pairs of the realised total and aggregate liquidity shocks, (x, ), the regulator is assumed to inject public funds through the purchase of the same security. If the regulator can make the choice of securities in the injection of public funds contingent on (x, ), we need to modify the ®rst-order conditions of (7) and (8) by allowing for this contingent choice. Since the regulator gains more freedom to adjust the injection policy, the state-contingent security design raises the threshold point for bank closure and increases the possibility that the injection of public funds is optimal. The more important restriction is that in the preceding section, we focus on only three kinds of security: subordinated bonds, preferred stocks and common stocks. Thus, we need to consider a more general class of state-contingent forms of security that is not restricted to the above three kinds. In other work (H. Osano, 1998), we study the more general class of forms of security contingent on both the total and aggregate liquidity shocks. Osano (2001) also examines the more general class of forms of security with a managerial compensation contract. 4.5

Endogeneity of initial capital requirement

Even though initial deposit funds D0 and equity capital E are endogenously determined, the main results of the preceding section still hold. Since the initial capital requirement of the bankis mainly concerned with the ex ante optimality conditions, the endogenous determination of the initial capital requirement does not essentially modify the ex post optimality conditions with respect to the capital injection level or the threshold level for bank closure. 4.6

Social cost of public funds

If we introduce a cost of government involvement in bank regulation, we may assume that there exists a social cost of using public funds to ®nance the deposit insurance programme. The additional cost includes not only the social loss of deposit insurance payments but also the social loss of injected cash funds. The former effect enhances the necessity of the injection of cash funds into the bank, whereas the latter effect erodes it. Thus, a rise in the social cost of public funds does not necessarily result in unambiguous effects on the threshold point for bank closure or on the possibility that the injection of public funds is optimal. 4.7

Trustworthiness of the deposit insurance system

In Section 3, we assume that deposits are fully insured by the regulator. If depositors do not believe this promise, banks with a bad reputation

78 Injection of Public Funds into Banks

cannot collect new deposit funds for re®nancing the total liquidity need. In this case, the injection of public funds can be rationalised to prevent the breakdown of the ®nancial system. The crucial point is that the regulator is committed to the policy of closing banks with a liquidity shock greater than the threshold point for bank closure. 4.8

Caps on the coverage of insured deposits

A problem related to the previous issue is that in many countries the regulator usually sets a ceiling on the amount of insurance per account. This modi®cation affects the ®nancing constraint for the bank that is represented by (4). Under the coverage cap on insured deposits, in order to allow a bank to re®nance the total liquidity need, the regulator can inject cash funds into the bank or relax any existing ceiling (cap) on the use of insured deposits. Since uninsured deposit funds require a market insurance premium that re¯ects the risk of bank insolvency, Dreyfus, Saunders and Allen (1994) argue that a rise in the coverage cap on insured deposits is an alternative to the policy of cash injection by the regulator. In addition, our model explicitly supposes that the regulator injects cash funds into the bank through the purchase of securities, whereas the model of Dreyfus, Saunders and Allen does not. Thus, the differences of these two policies are larger in our paper than in Dreyfus, Saunders and Allen (1994). The presence of the coverage cap on the use of insured deposits, therefore, may reduce the possibility that the optimal level of injection of public funds is positive. 4.9

Interbank loan transactions

Another modi®cation that affects the ®nancing constraint for the bank, (4), is to introduce the possibility of interbank loan transactions. The consideration of this possibility enables us to study the systemic risk problem, that is, the propagation of a bank's distress to other banks linked to that bank through ®nancial transactions. Here, we can only mention a short survey of this promising ®eld.15 Rochet and Tirole (1996) argue that the current existence of substantial interbank lending implies that banks do a considerable amount of monitoring of each other. Nevertheless, they show that the interbank peer monitoring does not necessarily lead the regulator to avoid undesirable rescue operations. If there exists a possibility of the closure of the entire banking system, they suggest that the central bank should provide enough liquidity to avert such systemic risk problems. They also suggest that the central bank should directly save a solvent but failing lending bank rather than indirectly bail out a borrowing bank. Thus, a `soft budget constraint' does not imply `too big to fail' in their model.

Hiroshi Osano 79

Kahn and Roberds (1998) present a simple model of interbank clearing and settlement that offers an explanation for the National-Banking-Era pattern of mutual insurance against liquidity risks. This mutual insurance pattern was not constructed against liquidity risks during normal times; rather, it was constructed as collateralised lending against liquidity risks during panics. Kahn and Roberts argue that panic-induced shifts in information structure could have motivated shifts in the settlement medium of clearinghouses, from reserve assets during normal times to interbank debt during panics. Then, they suggest that if clearinghouse banks have the ability to commit ex ante to debt clearing at a penalty rate, such precommitment can reduce the banks' incentives to gamble. However, since the aggregate liquidity shock is not taken into account in their model, we cannot exclude the case in which central bank assistance is required if a large aggregate shock hits the economy.

5

Conclusion

We have examined the optimality of the regulator injecting public funds into a bank in the presence of deposit insurance, and have characterised an optimal injection policy to avert moral hazard action by the bank. We have shown that under certain conditions, the regulator's optimal policy is to inject new cash funds into the bank. Furthermore, if the regulator does not have enough information on the bank, it can inject public funds into the bank through the purchase of subordinated bonds by setting the interest rate equal to zero; on the other hand, if the regulator has enough information on the bank, the regulator may inject public funds into the bank through the purchase of preferred stocks. However, we have also indicated that this kind of injection policy cannot be independent of the bank closure policy: inef®cient banks should be closed.

Appendix De®nitions of the derivatives in the ®rst-order conditions: let us introduce Z B …m; y† ˆ Z R …m; y† ˆ



 



B …m; y…†; †dF…† ˆ

R …m; y…†; †dF…† ˆ

Z Z

‡



‡

‡



‡

Z Z

B …m; y…x; †; x; †dG…x j m; †dF…†; R …m; y…x; †; x; †dG…x j m; †dF…†;

where y…† 2 �…†  fxB …†; I…x; †; D1 …x; †jx 2 ‰ ‡ ;  ‡ Šg and y…x; † 2 �…x; †  fxB …†; I…x; †; D1 …x; †g:

80 Injection of Public Funds into Banks Then, de®ne @i …m; y† @i …m; y…x; †; x; †  g…x j m; †f …†; @I…x; † @I…x; † @i …m; y† @i …m; y…†; †  f …†; @xB …† @xB …†

i ˆ B; R;

i ˆ B; R:

…A1† …A2†

The derivatives in the ®rst-order conditions (7) and (8) are given by (A1) and (A2). Common stocks: We ®rst examine the regulator's problem in the injection of public funds through the purchase of common stocks. The bank's ex ante gross payoff prior to the insurance premium and monitoring payment is represented by C

B …m; y† ˆ

Z Z 

xB …†

Z

‡

R

R ‡K�E‡x�I…x;†

E ‰R � R  � K ‡ E E ‡ I…x; †

� x ‡ I…x; †ŠdH…R†dG…x j m; †dF…† Z L�K‡E Z ‡ L ‡ ‰L � L  � K ‡ EŠdG…x j m; †dF…†: 

xB …†

…A3†

The regulator's ex ante gross payoff prior to the insurance premium receipt is similarly described by Z  Z xB …† Z R

I…x; † C

‰R � R  � K ‡ E � x R …m; y† ˆ  ‡ R ‡K�E‡x�I…x;† E ‡ I…x; † ‡ I…x; †ŠdH…R†dG…x j m; †dF…† Z  Z xB …† Z R ‡K�E‡x�I…x;† � ‰K � E ‡ x � I …x; † 

‡

R

� R ‡ R Šd H…R†dG…x j m; †dF…† Z  Z xB …† I…x; †dG…x j m; †dF…† � 

Z �

‡



L�K‡E L

Z

‡

xB …†

‰K � E � L ‡ L ŠdG…x j m; †dF…†:

…A4†

Note that (A3) and (A4) continue to hold true irrespective of whether the bank buys its common stocks back from the regulator. Partially differentiating (A3) and (A4)  we obtain with respect to I(x, ) for (m, x, ) E {m, m}  [ + , xB ()]  [, ], @CB …m; y† ˆ g…x j m; †f …† @I …x; †

Z

R

R ‡K�E‡x�I…x;†

@CR …m; y† ˆ g…x j m; †f …† @I…x; †

Z

E‰�R ‡ R  ‡ K ‡ xŠ dH…R†; ‰E ‡ I…x; †Š2

R

E‰R � R  � K � xŠ dH…R†: ‰E ‡ I…x; †Š2 R ‡K�E‡x�I…x;†

…A5† …A6†

It is obvious from (A5) and (A6) that @CR …m; y† @BC …m; y† ‡ ˆ 0; @I…x; † @I…x; †

…A7†

Hiroshi Osano 81 @C …m;y†

@C …m;y†

B B However, the sign of @I…x;† � @I…x;† is indeterminate. Thus, under moral hazard (1 > 0), we cannot explicitly derive the rule for injection of capital funds from (7) and (A5)±(A7) alone.

Subordinated bonds with a positive interest rate: If we take account of the interest rate of subordinated bonds, then the bank's ex ante gross payoff prior to the insurance premium and monitoring payment, (9), is rewritten by Z Z BS …m; y† ˆ 

xB …†

‡

Z

R

R ‡K�E‡x‡rI…x;†

‰R � R  � K ‡ E � x

� rI…x; †ŠdH…R†dG…x j m; †dF…† Z L�K‡E Z ‡ L ‰L � L  � K ‡ EŠdG…x j m; †dF…†: ‡ 

xB …†

…A8†

The regulator's ex ante gross payoff prior to the insurance premium receipt, (10), is also reformulated by Z Z …m; y† ˆ S

R 

Z

xB …†

‡

Z Z ‡ 

R

R ‡K�E‡x‡rI…x;† xB …†

Z

‡

…1 ‡ r†I…x; †dH…R†dG…x j m; †dF…†

R ‡K�E‡x‡rI…x;†

R ‡K�E‡x�I…x;†

‰R � R  � K ‡ E � x

‡ I…x; †ŠdH…R†dG…x j m; †dF…† Z  Z xB …† Z R ‡K�E‡x�I…x;† ‰K � E ‡ x � I…x; † � 

‡

R

� R ‡ R ŠdH…R†dG…x j m; †dF…† Z  Z xB …† � I…x; †dG…x j m; †dF…† 

Z �

‡



L�K‡E L

Z

‡

xB …†

‰K � E � L ‡ L ŠdG…x j m; †dF…†:

…A9†

Partially differentiating (A8) and (A9) with respect to I(x, ) allows us to show that the injection of public funds with subordinated bonds does not have neutrality.

Notes 1. Chan, Greenbaum and Thakor (1992) show that fairly priced deposit insurance and incentive compatibility may not be consistent, and that the optimal scheme may involve the intertemporal mispricing of deposit insurance under asymmetric information. Furthermore, in most countries, the insurance premium of deposit insurance is determined independently of bank risks.

82 Injection of Public Funds into Banks 2. The regulator's problem studied here is in some ways similar to the security design problem that determines an optimal ®nancial contract to resolve an agency problem between ®rm insiders and outside investors. See Townsend È f and von Thadden (1979), Diamond (1984), Gale and Hellwig (1985), Berglo (1994) and Hart and Moore (1998). Dewatripont and Tirole (1994) apply the ®nancial structure of ®rms to the regulation of banks, and show that a convenient combination of debt and equity can provide outside investors with the appropriate incentives to implement the second-best decision rule. However, they focus on the determination of the initial debt-equity ratio to develop a framework for the study of prudential regulation. 3. In the main part of this paper, we con®ne our attention to the ex ante moral hazard incentive for the bank that exists before the injection of public funds. In subsection 4.1, we will brie¯y discuss the ex post moral hazard incentive for the bank that is caused after the injection of public funds. 4. In subsection 4.5, we will brie¯y discuss the relaxation of this assumption. 5. In subsection 4.8, we will touch on the rule of caps on the coverage of insured deposits. In Japan, all banking deposits are fully insured by March, 2001. 6. See note 3. 7. Holmstro È m and Tirole (1997) consider a liquidity shock problem at the interim period in a different context. 8. In fact, we may not be able to exclude the possibility that there exist multiple threshold points of the total liquidity shock at which the regulator is indifferent as to whether it closes the bank in period 1 or allows the bank to remain open until period 2. However, barring other special reasons, it would be politically dif®cult to justify `good banks' with a lower total liquidity shock being closed in period 1 while `bad banks' with a higher total liquidity shock are kept open until period 2. We therefore assume that the threshold point for bank closure is uniquely determined. 9. Nagarajan and Sealey (1995) also study a contingent forbearance policy. 10. This ®nancing constraint for the bank in period 1 would be rather restrictive because the bank could obtain new cash funds by making interbank loan transactions or because depositors would not wish to provide additional new deposit funds due to the presence of a ceiling on the coverage of insured deposits. Hence, we will brie¯y discuss the effects of changes in this ®nancing constraint on the optimal injection scheme in subsections 4.8 and 4.9. 11. In fact, the terminology of preferred stocks includes several variations: in one type, preferred stockholders are senior to common stockholders with respect to only the dividend payment, while in another type, preferred stockholders are senior to common stockholders with respect to both the bank's dividend and residual asset payments. In subsection 4.3, we will examine the problem arising from the variety of preferred stocks. 12. Throughout, we assume that the second-order conditions for a maximum are satis®ed, and that the optimal threshold point for bank closure is interior. The latter assumption implies that a bank with the lower total liquidity shock is kept open, whereas a bank with the higher total liquidity shock is closed. 13. The Japanese banks into which public funds were injected through the purchase of preferred stocks in 1998 are required to buy back the issued preferred stocks from the Deposit Insurance Organisation in several years. See Okina (1998).

Hiroshi Osano 83 14. We assume that I*(x, ) = x for x = min (xB (), x8). 15. For an introduction and overview of this ®eld, see Berger, Hancock and Marquardt (1996).

References Aghion, P., P. Bolton and S. Fries (1999) `Optimal Design of Bank Bailouts: The Case of Transition Economics', Journal of Institutional and Theoretical Economics, vol. 155, pp. 51±70. Berger, A.N., D. Hancock and J.C. Marquardt (1996) `A Framework for Analyzing Ef®ciency, Risks, Costs, and Innovations in the Payment System', Journal of Money, Credit, and Banking, vol. 28, pp. 696±732. Berglo È f, E. and E.-L. von Thadden (1994) `Short-term versus Long-term Interests: Capital Structure with Multiple Investors', Quarterly Journal of Economics, vol. 109, pp. 1055±84. Besanko, D. and G. Kanatas (1996) `The Regulation of Bank Capital: Do Capital Standards Promote Bank Safety?', Journal of Financial Intermediation, vol. 5, pp. 160±83. Campbell, T.S., Y.-S. Chan and A.M. Marino (1992) `An Incentive-Based Theory of Bank Regulation', Journal of Financial Intermediation, vol. 2, pp. 255±76. Chan, Y.-S., S.I. Greenbaum and A.V. Thakor (1992) `Is Fairly Priced Deposit Insurance Possible?', Journal of Finance, vol. 47, pp. 227±45. Dewatripont, M. and J. Tirole (1994) The Prudential Regulation of Banks (Cambridge, MA: The MIT Press). Diamond, D. (1984) `Financial Intermediaries and Delegated Monitoring', Review of Economic Studies, vol. 51, pp. 393±414. Dreyfus, J.-F., A. Saunders and L. Allen (1994) `Deposit Insurance and Regulatory Forbearance: Are Caps on Insured Deposits Optimal?', Journal of Money, Credit, and Banking, vol. 26, pp. 412±38. Gale, D. and M. Hellwig (1985) `Incentive Compatible Debt Contracts: The Oneperiod Problem', Review of Economic Studies, vol. 52, pp. 647±63. Giammarino, R.M., T.R. Lewis and D.E.M. Sappington (1993) `An Incentive Approach to Banking Regulation', Journal of Finance, vol. 48, pp. 1523±42. Hart, O. and J. Moore (1998) `Default and Renegotiation: A Dynamic Model of Debt', Quarterly Journal of Economics, vol. 113, pp. 1±41. Holmstro È m, B. and J. Tirole (1998) `Private and Public Supply of Liquidity', Journal of Political Economy, vol. 106, pp. 1±40. Kahn, C.M. and W. Roberds (1998) `On the Role of Bank Coalitions in the Provision of Liquidity', Federal Reserve Bank of Atlanta, mimeo. Mitchell, J. (2001) `Bad Debts and the Clearing of Banks' Balance Sheets: An Application to Transition Economies', Journal of Financial Intermediation, vol. 10, pp. 1±27. Nagarajan, S. and C.W. Sealey (1995) `Forbearance, Deposit Insurance Pricing, and Incentive Compatible Bank Regulation', Journal of Banking and Finance, vol. 19, pp. 1109±30. Okina, Y. (1998) Information Disclosure and Japanese Financial System, Tokyo, Toyokeizai Shinpo-sha (in Japanese). Osano, H. (1998) `Optimal Regulatory Design in the Injection of Public Funds into Banks under Deposit Insurance', Kyoto Institute of Economic Research, mimeo.

84 Injection of Public Funds into Banks Osano, H. (2001) `Managerial Compensation Contract and Bank Bailout Policy', forthcoming in Journal of Banking and Finance. Rochet, J.-C. and J. Tirole (1996) `Financial Crises, Payment System Problems, and Discount Window Lending', Journal of Money, Credit and Banking, vol. 28, pp. 804±24. Townsend, R. (1979) `Optimal Contracts and Competitive Markets with Costly State Veri®cation', Journal of Economic Theory, vol. 21, pp. 265±93.

5

Governance Structure of Banks and Their Business Performance Toshiaki Tachibanaki and Hideo Okamura

1

Introduction

Banks in Japan played an important role in the so-called indirect ®nancing system which implies that banks collect funds from individual investors as a form of deposits and lend them to non-®nancial ®rms. Banks, in particular main banks, have paid special attention to monitoring the performance of their borrowers, and have acted as an important stakeholder in the Japanese corporate ®nance system. Banks, however, have not been monitored adequately by public administration, namely the Ministry of Finance (MOF) and/or the Bank of Japan (BOJ). Although the MOF was a regulator to private banks, it has not made suf®cient effort to administrate and monitor them. Some of the regulations by the MOF contributed, implicitly or explicitly, to preventing any bankruptcies of private banks. Mergers and acquisitions, which were initiated and recommended by the MOF, of ®nancially troubled banks by larger banks were quite common in the past. It should be recognised that these larger banks were strong enough ®nancially to be able to acquire banks in trouble because they could enjoy extra pro®t and rent due to such mergers which were assured by the regulations. See, for example, Tachibanaki (1996) about the effect of regulation on banks in Japan. The situation changed totally after the end of the so-called `bubble economy' because many of banks had serious ®nancial troubles. The reason is that they accumulated huge bad debts, and in fact some of them went bankrupt. An important issue emerged: `Who monitors the monitor (i.e., the bank)?' The banks used to monitor non-®nancial ®rms through their lending activity. `Who monitors banks?' is the central concern of this study. It is possible, however, that banks were monitored marginally

85

86 Governance Structure of Banks and Their Business Performance

by other agents. Therefore, it is interesting to inquire whether banks were monitored by somebody, and to seek for any difference in productivity or ef®ciency among various banks which are monitored differently. Put simply, does the difference in corporate governance structure for banks affect productivity among banks? We employ two methods to investigate this subject. The ®rst is to estimate the productivity function for a panel data of banks, and the second is to estimate the cost of capital for banks. There is growing concern about the effect of corporate governance on ®rms' performance in Japan. Such a concern has been concentrated on non-®nancial ®rms, namely the effect of banks' monitoring activity, and of the voice of shareholders on them. No serious attention has been paid to ®nancial ®rms. In particular, no serious investigations have been presented to show some quantitative effect of corporate governance on ®nancial ®rms' performance except for several casual studies. As noted, there are two institutions which monitor, govern or even control ®rms' management and performance. They are the suppliers providing ®rms with ®nancial resources. The ®rst is shareholders who own the capital of the ®rm, and the second debt-holders who lend funds. Shareholders are able to appoint managers or executives of the ®rm through the shareholders' voting power, and thus provide the ®rm with considerable in¯uence on the ®rm's management. Other signals, which can evaluate the performance of the ®rm or can be used for corporate governance, are the effect of the movement in equity prices, and a threat of hostile takeovers. The second is debt ®nance. One popular idea regarding the effect of debt on corporate governance is based on the theory of incomplete contract, see Aghion and Bolton (1992), Hart (1995) and others. The essence of this idea can be described as follows. When the performance of ®rms is normal, i.e., a certain level of sales and/or pro®t is assured, a debtholder will receive the predetermined level of cash ¯ow. When the performance is bad, however, or the ®rm goes bankrupt, debt-holders go into action. The management authority is transferred from the managers (i.e., the chief executives) to the debt-holders at the time of bankruptcy. The possibility of such a transfer can work as the corporate governance mechanism for the ®rm. One important feature which has a great in¯uence regarding the above corporate governance structure in Japan is that intercorporate shareholding is common, namely the majority of equities and shares are held by each other among friendly ®rms in a group, such as Mitsubishi, Sumitomo, and others, or in a framework or relationship among parents±

Toshiaki Tachibanaki and Hideo Okamura 87

subsidiary ®rms. In particular, the role of ®nancial ®rms is important for intercorporate shareholding. The most notable characteristic of the intercorporate shareholding system in Japan is that a ®rm which holds other ®rms' shares is usually quiet or silent about other ®rms' management policy. In other words, neither strong control nor ef®cient monitoring from shareholders is seen. Needless to say, this quiet or silent relationship is symmetrical among ®rms who commit to intercorporate shareholding. This suggests to us to imply that the degree of control is weaker as the rate of intercorporate shareholding increases. Although the intercorporate shareholding system lowers the degree of corporate governance or control, there is one theory which stresses the merit of the system: it does not urge managers to produce higher short-run pro®t or to raise share prices, but to raise long-run interests in the ®rm because the long-run motive rather than the short-run one helps to increase the relative position of the ®rm in the industry, or to raise the long-run pro®t of the ®rm. Another interesting feature of intercorporate shareholding is that it tends to lower the face value of the cost of capital due to the accounting rule. The reason for this underestimation is that the part of the pro®t which was retained in the other ®rm (i.e., the counter-part of shareholding) cannot be included in the pro®t of the ®rm, although the share price of the ®rm is in¯uenced by the increase in the share price of the other ®rm. Thus, unless all that part of the pro®t in the other ®rm is distributed as dividends rather than as retained earnings, the cost of capital is underestimated. If the manager in the ®rm determined the amount of investment or the amount of ®nancing from outside based on the underestimated cost of capital, over-investment or over-®nancing would be likely to occur, and thus it would imply less ef®ciency in management. In sum, intercorporate shareholding is likely to produce a discrepancy between the pro®t and the share price in the ®rm, and can be a source of a bias in the estimation of the cost of capital. Debt ®nancing may reduce the bargaining power, or the monitoring ability of debt-holders as Ikeo (1994) pointed out. We can raise such evidence by showing the case in which the main bank in Japan lost its monitoring ability during the period of the post-bubble economy. It is an interesting subject to investigate whether banks, in particular Japanese banks, were able to keep their monitoring capability or controlling activity. It is crucial to consider the peculiar role of debt-holders in the determination of banks' behaviour. Debt-holders of a bank are savers who deposit their funds at the bank, while debt-holders of a non-®nancial

88 Governance Structure of Banks and Their Business Performance

®rm are banks who lend their funds to the ®rm. The distinction regarding debt-holders between a bank and a non-®nancial ®rm is very important because debt-holders of a bank are normally individual households and ordinary people who are neither interested in monitoring their bank nor capable of examining the performance of the bank. More importantly, the incapability of an extremely large number of depositors at a bank who have only an extremely small amount of deposit regarding their monitoring activity must be recognised. In other words, neither power nor interest is kept by debt-holders of a bank (i.e., depositors) unlike debtholders of a non-®nancial ®rm (i.e., banks). Since the amount of deposit is guaranteed by the deposit insurance system, it is nearly impossible to expect the interest in monitoring activity by depositors. Then, `who monitors banks?' becomes an important issue. According to Dewatripont and Tirole (1994) it is the government, normally the Treasury Department or the Ministry of Finance, which monitors or governs the activity of banks. Ef®cient monitoring activity is expected to assure sound banking business activity in a country. Shimizu and Horiuchi (1997), and Tachibanaki (1998a), however, concluded that the monitoring activity of the Ministry of Finance in Japan was quite insuf®cient. Thus, it is possible to guess that no ef®cient management was observed for the Japanese banks due to such insuf®cient monitoring activity by the government. Shimizu and Horiuchi (1997) were interested in the role of `descending from heavens', i.e., bankers (some of them are board members of a bank) who started to work at a bank after retiring from the MOF. Many banks accept retired MOF people in exchange for the regulation rents enjoyed by banks. Horiuchi and Shimizu showed that no governance role of these bankers (i.e. ex-MOF people) was observed. Another important tendency is deregulation in the ®nancial industry taking place in Japan. Since the degree of freedom in the business activity of banks is increasing currently, it is necessary to expect more adequate and sound monitoring activity of the MOF. One important role of such activity would be to reduce the possibility of bank's bankruptcy, or to prepare an ef®cient and cost-minimised rule dealing with a ®nancially troubled bank, or a bank in bankruptcy. Since the discussion on the monitoring activity of debt-holders and on the governance for a bank requires various analyses, we concentrate on the two subjects, namely the direct governance role of shareholders and the indirect governance role through the market mechanism. Part 2 examines various subjects related to the management of banks such as the role of shareholders and the particularity of the banking business. Part 3 shows an empirical analysis of the estimation of productivity in banks in

Toshiaki Tachibanaki and Hideo Okamura 89

order to reveal the effect of the structure of shareholders. It uses a panel data, and aims at examining the effect of shareholders on management ef®ciency in the banking industry. Part 4 presents the estimated cost of capital for banks to examine the governance role through the market mechanism. One contribution of this study is to adjust for the effect of intercorporate shareholding, when the cost of capital is estimated. Finally, Part 5 provides a conclusion and some future issues.

2

Governance structure and ®rms' management

As Berle and Means (1932) argued in their classical work, the separation between holding a ®rm and managing a ®rm was one of the important issues. In particular, the con¯ict of interests between managers and owners (i.e., shareholders) was their concern during and after the 1930s. One of the most useful approaches dealing with this issue is the development of economics of information, which provided us with the so-called agency approach. It discussed the effect of capital structure, namely equity versus debt, and of monetary compensations to top executives on management ef®ciency. The rising interest in this ®eld is the effect of incomplete contract on ef®ciency. In the literature of corporate governance attention is paid to examining the role of each stakeholder, such as shareholders, debt-holders, managers, employees, customers, etc, and their relationships in addition to the examination of the direct controlling role of shareholders. 2.1 The direct role of shareholders and the effect of intercorporate shareholding The separated system in corporated ®rms implies that professional managers, who are not the main shareholders of a ®rm, manage it, while its equities are held by a large number of shareholders. Managers attempt to maximise the ®rm's value by minimising the con¯ict of interests among various stakeholders. In turn, managers also are governed or monitored by shareholders whether they perform well or not. There are various ways to monitor and control managers from the shareholders' point of view. The direct control is to use the voting power at the shareholders' meeting, and to nominate and/or discharge managers, while the indirect control through the market mechanism is to use the possibility of hostile takeovers. Large shareholders are interested in monitoring a ®rm, because they have invested a lot. They are able to monitor or control managers of a

90 Governance Structure of Banks and Their Business Performance

®rm, or the performance of a ®rm more ef®ciently. At the same time, they can enjoy a higher return to ef®cient management with a smaller cost. Among various large shareholders, institutional investors, such as banks, pension funds, trust funds, etc, who are professional in these business activities can be good candidates for ef®cient monitoring (see Shleifer and Vishny, 1997). The indirect control through the market mechanism typically has two channels. One is to use information about the decrease in the share price of a ®rm. This indicates the possibility of bad performance in business activity, and can be transmitted to the shareholders' meeting which may dismiss the ®rm's managers. The second is to use the possibility of hostile takeover, and again to dismiss the ®rm's managers and appoint new ones. The Japanese case needs more explanation when we discuss the direct role of shareholders. The reason is the effect of the intercorporate shareholding system. As Kurasawa (1993), Yonezawa (1995), and Tachibanaki and Nagakubo (1997) have pointed out, the intercorporate shareholding system reduces, probably, the possibility of hostile takeovers, because the majority of such shareholders are silent. They are called `stable and silent shareholders' in Japan and they neither commit to frequent selling and buying activities of shares nor address their hostile opinion to the ®rm. This obviously raises the degree of autonomy of managers, and thus is likely to increase the productivity of a ®rm because of no serious outside interference and intervention. Put simply, managers have freedom. There are two reasons why we can propose such an increase in productivity under the intercorporate shareholding system. One is that it raises the long-run interest in managing a ®rm without worrying about the short-run strong demand for higher dividends by shareholders. The second is that the feature of a labour-managed ®rm in interpreting the Japanese ®rm in general, as proposed by, for example Tachibanaki (1998a and b), can work well if their is no serious interferences from shareholders come. The intercorporate shareholding system can help in functioning the function of the above two reasons. It is important, however, to recognise that intercorporate shareholding is likely to reduce the face value of the cost of capital of a ®rm as noted previously, and to induce overinvestment for a ®rm. This overinvestment causes a problem of inef®ciency in management. Consequently, it is worth attempiting to estimate the cost of capital adjusted for the intercorporate shareholding system in order to know a more realistic story in corporate ®nance.

Toshiaki Tachibanaki and Hideo Okamura 91

2.2

Governance by debt and ef®ciency in banking activity

Aghion and Bolton (1992) proposed elegantly the usefulness of debt ®nancing in controlling and monitoring managers. A bankruptcy transfers the management authority from managers to debt-holders. Hirota and Ikeo (1996) presented empirical evidence, proposing that the possibility of re-organisation due to debt ®nancing such as a possible transfer of the management authority worked well in Japan to keep a higher level of corporate governance. We have to keep in mind, however, the special role of debt-holders (i.e., depositors) at banks. In other words, the major part of debts in the balance sheet of a bank are depositors, and at the same time there are so many depositors. Such depositors are not interested in monitoring activity of a bank, nor having any capability of monitoring. In sum, debt-holders are unable to take action in useful monitoring activity of a bank. Another reason for the lack of monitoring activity by debt-holders for a bank is that the deposit insurance system reduces the depositors' incentive of monitoring. Moreover, the linear and ®xed rate of insurance premium in the deposit insurance system is likely to induce moral hazard of managers in the banking industry, as proposed by Ikeo (1991). Moral hazard was also observed in Japan in view of the fact that the MOF heavily regulated the banking industry, and provided a rescue action for a ®nancially troubled bank. If banks knew that the MOF would help them in an emergency, they might not have a strong incentive of ef®cient management.

3 3.1

Productivity in banks and the structure of shareholders Purpose of the analysis

The purpose of this part is to estimate the determination of productivity in the banking industry, and to investigate the effect of the shareholders' structure on productivity. There are several studies which investigated these subjects for non-®nancial ®rms. Hirota (1996) estimated the effect of not only the structure of shareholders, but also the rate of debt over total ®nance, the main bank system, and some other variables, on ef®ciency in management. He concluded that the Japanese type of corporate ®nance was effective for raising ef®ciency, in particular ef®ciency in ®rms which had relatively low growth rates. It is noted, however, that no signi®cant in¯uence of the shareholders' structure was observed. Yonezawa and Miyazaki (1996) discovered that a high degree of intercorporate shareholding would raise average productivity. Moreover,

92 Governance Structure of Banks and Their Business Performance

they examined the effect of the labour share over total cost in a ®rm, and proposed the following observation; The behaviour of ®nancial ®rms, and of ®rms which are owned by foreigners is similar to the behaviour of the neo-classical type of a ®rm, which aims at maximising the share price. In other words, average productivity in these ®rms is normally higher than that in other ®rms. At the same time, a high rate of shareholding by executives over total shares raises productivity. Although the major concern in these studies was addressed to non®nancial ®rms, Saito (1997) examined regional banks, and estimated the effect of their lending behaviour and the shareholders' structure for a panel data. He concluded that the effect of the shareholders' structure on regional banks' lending activity was evident. Additionally, he proposed that the equity market would watch the lending activity and the shareholders' structure of regional banks carefully. As pointed out earlier, depositors at banks, who are supposed to be debtholders, cannot or are not interested in monitoring the banking activity. Therefore, it is possible to guess that shareholders of a bank wish to monitor it more strongly than depositors to compensate for the lack of monitoring activity by depositors, and thus it is likely that large shareholders want to monitor the banking activity more actively. One important contribution of this study is to investigate the role of life insurance companies as the most important shareholders in Japan's corporate ®nance. Since a life insurance company is able to hold shares of other ®rms, the rate up to ten per cent of total shares of a ®rm, it can be the largest shareholder of a ®rm. Incidentally, a bank can hold at the rate only up-to ®ve per cent of total shares of a ®rm. The Japanese life insurance companies have extremely large assets because life insurances are quite popular. It should be an interesting subject to inquire whether or not life insurance companies played an important role as a governing or monitoring institution for corporate ®nance and/or corporate governance. Life insurance companies are different from other corporated ®rms because they are normally mutual companies which do not issue any shares and stocks, although they are able to hold shares of other ®rms. Needless to say, nobody can hold any shares of a mutually owned life insurance company because no shares are issued. This asymmetry nature of a life insurance company regarding its shareholding structure gives several clues to understand the intrinsic nature of a life insurance company. First, life insurance companies are interested mainly in the highest monetary return to shareholding of other ®rms, and are not interested in acting as normal shareholders who are expected to govern or monitor a

Toshiaki Tachibanaki and Hideo Okamura 93

company. Second, life insurance holders, who are called owners of a life insurance company, do not pay close attention to the performance of either their life insurance company or other ®rms. Thus, the managers of a life insurance company also are not interested in monitoring other nonlife insurance companies whose large part of shares are held by the life insurance company. It is an attractive subject to examine which possibility is empirically plausible among the above two. In sum, we are going to investigate the role of life insurance companies in corporate governance in Japan. The ®nal subject is to analyse the effect of the labour side, i.e., workers who receive wages, by examining the role of the labour share over total cost on productivity. 3.2

Data

We use the data prepared by the Nomura Research Institute, which publishes ®nancial statements of various ®nancial ®rms. We apply a ®xed model because the data are basically panel data from 1981 to 1995, a total of 15 years. There are 10 city banks, three long-term credit banks, and seven trust banks. One city bank, Hokkaido Takushoku Bank, and two long-term credit banks, The Long-term Credit Bank of Japan and Japan Bond and Credit Bank, went bankrupt and declared their closure or nationalisation in 1997 and 1998. Since the data for these banks are available, we include them statistically. It should be a useful attempt, if we examine both the case of all samples (i.e., twenty banks), and the case of divided samples by its industrial feature (i.e., city banks, long-term credit banks, and trust banks, respectively). We adopted per capita (i.e., per employee) valued-added as a symbol of productivity (or ef®ciency) of a bank. It is the sum of business pro®t, total compensation (both wages and non-wage payments), and depreciation. The reason why we include total compensation for employees is that employees play an important role in the determination of productivity, and act as an important intermediary among various stakeholders. This is, in particular, true when we are concerned with corporate governance of a bank. One idea is to adopt total factor productivity (TFP) as given by, for example, Hirota (1996), and Yonezawa and Miyazaki (1996). Another possibility is to consider per capita after-tax pro®t as a symbol of productivity. We understand that per capita value-added is the most appropriate variable to indicate ef®ciency of a bank because it can take into account various aspects particularly when we are concerned with corporate governance.

94 Governance Structure of Banks and Their Business Performance

It is unfortunate that the data on business pro®t are available only after 1988. Therefore, we examine for eight years (i.e., from 1988 to 1995), by taking the sum of business pro®t, total compensation and depreciation. This total value-added is divided by the number of employees in a bank to measure productivity. If we examined only for the period after 1988, it would imply that we ignored the period when there was drastic change in the Japanese economy in the 1980s, i.e., the birth and death of the so-called bubble economy. It is quite likely that management behaviour of a bank was greatly in¯uenced by this bubble economy. Therefore, we examined also for the two periods from 1981 to 1985, and from 1986 to 1990, separately in order to investigate the effect of the bubble economy on banks' behaviour. We use per capita after-tax pro®t for the dependent variable for these periods because of the lack of data on business pro®t. We use the following independent variables for each bank; the number of employees, total asset value, the rate of the shares held by top ten shareholders over total shares (called simply top ten shares), and the rate of the shares held by life insurance companies at top ten shareholders over total shares (called simply life insurance shares). We include each individual bank dummy variable to control for the particular nature of each bank. We estimate the following two equations: Yit ˆ b1 X1it ‡ b2 X2it ‡ b3 X3it Yit ˆ b0 1 X1it ‡ b0 2 X2it

‡ b5 X5it ‡ Di ‡ eit

‡ b0 4 X4it ‡ b5 X5it ‡ Di ‡ eit

…1† …2†

where i denotes individual bank, and, t denotes year. Y is the logarithm of per capita value-added, or per capita after-tax pro®t. X1 is the logarithm of the number of employees, X2 is the logarithm of total asset, X3 is the ratio of shares held by top ten shareholders over total share values, X4 is the ratio of shares held by life insurance companies, Di is the i-th individual bank's dummy, and eit is the error term peculiar to i-th bank and t-th period. eit is independent of each explanatory variable. bi is the parameter. See, for example, Baltagi (1995) about the estimation method. There are two important notes for equation (1) or (2). One is that we apply the instrumental variable method in estimating (1) or (2) in order to take into consideration a possibility that some explanatory variables are determined endogenously. Second is that we add one explanatory variable in equations (1) and (2), that is X5 which signi®es the labour share. X5 is measured by ratio of total compensation over total value-added when per capita value-added is used as the dependent variable, while X5 is measured by the ratio of total

Toshiaki Tachibanaki and Hideo Okamura 95

compensation over the sum of total after-tax pro®t and total compensation when per capita after-tax pro®t is used. The labour share variable was included to examine whether there exists some in¯uence of the distribution of value-added to employees on productivity. In other words, we are interested in investigating the relationship between employees' contribution and productivity, which sheds light on the understanding of corporate governance of a bank. Table 5.1 shows the summary statistics for data from 1988 to 1995. The dependent variable is per capita value-added, Table 5.2 is for data from 1981 to 1985. The dependent variable is per capita after-tax pro®t. Table 5.3 is for data from 1986 to 1990. The dependent variable is per capita after-tax pro®t. 3.3

Estimated results

Table 5.4 presents the estimated result for 1988±95, where per capita value-added is used as the dependent variable. As noted, per capita valueadded is measured by the sum of business pro®t, total compensation and depreciation. The bottom line shows the case in which the labour share variable is added as an explanatory variable. First, the effect of top ten shares on per capita value-added is evaluated. We ®nd the positive and statistically signi®cant coef®cient for the observations of total twenty banks regardless of whether or not the labour share variable is added. The result for city banks is the same as for total twenty banks, while it is the same for trust banks only for the case in which the labour share is not added. It would be desirable to understand that the separated samples, i.e., city banks, trust banks and long-term credit banks, trust banks, respectively, provide us with more reliable and robust results in view of much different result between total twenty banks and separated samples. The effect of life insurance shares is negative with statistical signi®cance for the observations of total twenty banks. The coef®cient is negative signi®cantly for city banks when the labour share variable is excluded, while it is negative signi®cantly for long-term credit banks and trust banks regardless of the labour share variable. The number of employees is positive and statistically signi®cant for all twenty banks, while the sign is mixed for separated samples. We ®nd the negative and statistically signi®cant coef®cient for the total asset variable for total twenty banks and the similar result for city banks when the labour share is excluded, and in some cases of long-term credit banks and trust banks. The case of trust banks is positive in some cases. The labour share coef®cient is negative and signi®cant statistically in many cases,

value-added (one million yen)/number of employees

number of employees

value-added (one million yen)

business pro®t (one million yen)

depreciation (one million yen)

wages and non-wage payments (one million yen)

total asset (one million yen)

labour share (%)

top ten shares (%)

life insurance shares (%)

96

Table 5.1 Summary statistics (1988±95) *Dependent Variable: Per capita Value-added Total 20 Banks average

standard error

minimum

25.86 9,355.88 244,460.88 132,921.08 15,986.63 95,553.17 28,687,820.00 44.43 25.01 9.13

11.08 5,593.98 171,694.44 109,349.66 12,933.66 60,627.60 19,408,220.00 15.38 7.60 5.02

average

standard error

minimum

maximum

25.00 13,850.36 359,241.06 191,512.51 25,439.81 142,288.74 41,132,100.00 43.16 24.16 12.73

7.46 4,392.85 166,077.18 117,770.80 11,851.92 51,725.35 17,811,900.00 10.43 2.32 3.85

14.22 5,907.00

89,316.00

11,378.00

6,568.00 49,980.00 10,361,400.00 23.31 20.36 2.69

47.96 22,919.00 748,712.00 519,054.00 47,675.00 247,713.00 66,590,800.00 80.94 29.01 19.67

9.74

1,739.00

18,600.00

�4,168.00

677.00

19,088.00 1,334,930.00 15.49 17.16 1.64

maximum 77.67 22,919.00 748,712.00 519,054.00 47,675.00 247,713.00 66,590,840.00 111.12 75.91 19.67

City Banks

value-added (one million yen)/number of employees

number of employees

value-added (one million yen)

business pro®t (one million yen)

depreciation (one million yen)

wages and non-wage payments (one million yen)

total asset (one million yen)

labour share (%)

top ten shares (%)

life insurance shares (%)

Table 5.1 Summary statistics (1988±95) (continued ) Long-term Credit Banks, and Trust Banks

value-added (one million yen)/number of employees

number of employees

value-added (one million yen)

business pro®t (one million yen)

depreciation (one million yen)

wages and non-wage payments (one million yen)

total asset (one million yen)

labour share (%)

top ten shares (%)

life insurance shares (%)

average

standard error

26.71 4,861.39 129,680.70 74,329.65 6,533.45 48,817.60 16,243,600.00 45.69 25.85 5.53

13.78 1,664.89 71,182.66 56,949.74 3,901.42 17,283.05 11,279,300.00 19.08 10.45 3.10

average

standard error

minimum 9.74 1,739.00 18,600.00 �4,168.00 677.00 19,088.00 1,334,930.00 15.49 17.16 1.64

maximum 77.67 7,708.00 330,287.00 248,555.00 17,815.00 74,013.00 43,602,700.00 111.12 75.91 13.49

Trust Banks

value-added (one million yen)/number of employees number of employees value-added (one million yen) business pro®t (one million yen) depreciation (one million yen) wages and non-wage payments (one million yen)

20.64 5,318.02 117,501.93 60,595.55 6,938.98 49,967.39

7.90 1,651.56 66,765.85 50,434.30 4,355.22 17,203.91

minimum 9.74 1,739.00 18,600.00 �4,168.00 677.00 19,088.00

maximum 45.47 7,708.00 285,532.00 204,309.00 17,815.00 73,627.00

97

Summary statistics (1988±95) (continued )

total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

98

Table 5.1

10,644,700.00 51.66 28.34 4.34

5,856,600.00 19.26 11.60 1.77

1,334,930.00 25.01 19.56 1.64

Notes: (1) value-added: (business pro®t + depreciation) + (wages + non-wages). (2) labour share: [(wages + non-wages)/(value-added)]  100. (3) top ten shares: the rate of the shares held by top ten shareholders over total shares (%). (4) life insurance shares: the rate of the shares held by life insurance companies at top ten shareholders over total shares (%)

Sources:

Financial statements prepared by Nikkei Needs, and Nomura Research.

Survey on Dividends, Survey on Shareholders, Survey on Firm Groups.

21,500,700.00 111.12 75.91 9.22

Table 5.2 Summary statistics (1981±85) *Dependent Variable: Per capita after-tax Total 20 Banks average after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

3.35 9,081.68 27,814.75 65,377.16 13,630,700.00 71.53 27.38 9.37

standard error 2.57 5,594.31 21,684.87 41,739.62 9,373,970.00 13.47 8.88 5.43

minimum 0.58 1,664.00 1,423.00 11,620.00 706,720.00 41.09 16.99 0.00

maximum 10.01 22,850.00 81,855.00 164,869.00 36,369,400.00 90.79 63.12 19.94

City Banks average after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

2.71 13,732.86 39,216.78 99,917.18 19,352,800.00 74.11 26.29 13.09

standard error 1.29 4,097.45 22,879.87 30,692.41 8,629,130.00 8.21 2.83 4.21

minimum 0.99 6,513.00 8,638.00 43,491.00 5,313,683.00 58.40 19.99 2.14

maximum 5.96 22,850.00 81,855.00 164,869.00 36,369,400.00 84.83 30.75 19.94 99

Table 5.2 Summary Statistics (1981±85) (continued ) Long-term Credit Banks, and Trust Banks

after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

4.00 4,430.50 16,412.72 30,837.14 7,908,620.00 68.94 28.47 5.65

standard error 3.29 1,514.04 12,698.69 11,966.72 6,021,030.00 16.89 12.20 3.71

minimum 0.58 1,664.00 1,423.00 11,620.00 706,720.00 41.09 16.99 0.00

maximum 10.01 6,528.00 49,067.00 49,781.00 24,767,700.00 90.79 63.12 13.98

Trust Banks average after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

2.00 4,835.54 10,743.49 32,835.23 4,761,290.00 78.74 32.02 4.26

standard error 1.30 1,384.01 8,357.90 11,482.21 2,796,110.00 8.91 12.99 2.74

minimum 0.58 2,152.00 1,423.00 12,310.00 706,720.00 56.99 21.51 0.00

Notes: (1) labour share: [(wages + non-wages)/(after-tax pro®t + wages + non-wages)]  100. (2) top ten shares: the rate of the shares held by top ten shareholders over total shares (%). (3) life insurance shares: the rate of the shares held by life insurance companies at top ten shareholders over total shares (%).

Sources:

Financial statements prepared by Nikkei Needs, and Nomura Research.

Survey on Dividends, Survey on Shareholders, Survey on Firm Groups.

maximum 5.60 6,528.00 34,021.00 49,781.00 10,062,600.00 90.79 63.12 9.22

100

average

Table 5.3 Summary statistics (1986±90) *Dependent Variable: Per capita after-tax Pro®t Total 20 Banks

after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

average

standard error

7.80 8,645.16 64,689.68 79,149.74 25,597,800.00 57.00 25.43 9.50

4.38 5,020.23 46,183.59 48,655.09 18,297,600.00 12.86 5.64 5.21

average

standard error

6.31 12,593.08 85,984.74 116,202.02 36,160,500.00 61.04 25.45 13.18

2.84 4,065.78 53,134.67 41,630.86 18,466,600.00 10.22 2.76 4.11

minimum 1.74 2,092.00 3,665.00 16,596.00 1,552,029.00 29.71 17.16 1.64

maximum 20.29 22,919.00 198,314.00 238,136.00 66,590,800.00 83.60 50.16 19.94

City Banks

after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

minimum 2.18 6,282.00 13,780.00 49,727.00 9,034,756.00 41.66 20.36 2.59

maximum 12.37 22,919.00 198,314.00 238,136.00 66,590,800.00 78.30 30.17 19.94 101

Table 5.3 Summary statistics (1986±90) (continued )

after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

average

standard error

9.29 4,697.24 43,394.62 42,097.46 15,035,200.00 52.96 25.41 5.82

5.11 1,606.48 23,677.31 15,749.34 10,379,100.00 14.02 7.52 3.21

average

standard error

minimum 1.74 2,092.00 3,665.00 16,596.00 1,552,029.00 29.71 17.16 1.64

maximum 20.29 7,451.00 93,520.00 68,453.00 43,602,700.00 83.60 50.16 13.50

Trust Banks

after-tax pro®t (one million yen)/ number of employees number of employees after-tax pro®t (one million yen) wages and non-wage payments (one million yen) total asset (one million yen) labour share (%) top ten shares (%) life insurance shares (%)

6.62 5,182.94 38,094.37 44,622.69 10,158,000.00 58.76 27.73 4.61

3.10 1,519.74 22,681.01 15,451.65 5,757,660.00 12.58 7.81 2.06

minimum 1.74 2,110.00 3,665.00 16,596.00 1,552,029.00 43.20 20.75 1.64

Notes: (1) labour share: [(wages + non-wages)/(after-tax pro®t + wages + non-wages)]  100. (2) top ten shares: the rate of the shares held by top ten shareholders over total shares (%). (3) life insurance shares: the rate of the shares held by life insurance companies at top ten shareholders over total shares (%).

Sources:

Financial statements prepared by Nikkei Needs, and Nomura Research. Survey on Dividends,

Survey on Shareholders, Survey on Firm Groups.

maximum 12.33 7,451.00 77,556.00 66,453.00 21,500,700.00 83.60 50.16 9.23

102

Long-term Credit Banks, and Trust Banks

Toshiaki Tachibanaki and Hideo Okamura 103

implying that ef®ciency in management decreases as the distribution to employees increases. The overall result in Table 5.4 suggests the following conclusions. The shareholders' governance for banks, in particular city banks, was ineffective from 1988 to 1995 because we observed the following: the higher the ratio of intercorporate shareholding, the lower the ef®ciency in management. The similar proposition is possible for life insurance companies' shareholding. Life insurance companies did not govern banks effectively. Another interesting result is that no positive in¯uence on productivity is observed as the total asset value of a bank increases. We observe even a decrease in productivity, implying that a bigger bank is not necessarily ef®cient in management. An exception is trust banks which show a small positive correlation between total asset and productivity. Finally, the negative coef®cient of labour shares signi®es that the wage payment does not correspond to productivity. It is likely to suggest that banks collude with their employees in the determination of wages, and the ef®ciency wage hypothesis is not supported in the banking industry. Table 5.5 shows the result for the period of 1981±85 where per capita after-tax pro®t was used as the dependent variable. The bottom line in this table includes the labour share variable as an additional explanatory variable. We ®nd that both top ten shares and life insurance shares are statistically signi®cant for city banks in the case in which the labour share variable is added. Although many coef®cients of the number of employees are negative with statistical signi®cance, the majority of total asset are statistically signi®cant and positive. All the labour shares' coef®cient are statistically signi®cant and negative. From Table 5.5 we can conclude that the governance capability of shareholders was effective for city banks from 1981 to 1985: the higher the rate of intercorporate shareholding, the higher the ef®ciency of city banks, management. The life insurance companies' governance capability for city banks was also effective, although it was weaker. No governance capability of shareholders, however, was observed for long-term credit banks and trust banks. Unlike the result in Table 5.4, there was a positive relationship between productivity and total asset value for all types of banks. The coef®cients, however, of the labour share variable were negative as in Table 5.4. Table 5.6 shows the result for the period from 1986 to 1990, where the dependent variable is measured by per capita after-tax pro®t. We can observe the following results based on Table 5.6.

Table 5.4

The structure of shareholdings for banks and management ef®ciency (1988±95) Total 20 Banks

life insurance shares log (employees) log (asset) Adjusted R-squared

3.01** (2.01) 5.49** (2.42) �3.70** (�2.55)

�12.60*** (�4.13) 2.47*** (3.70) �2.01*** (�4.20)

0.30

0.61

Total 20 Banks labour share top ten shares life insurance shares log (employees) log (asset) Adjusted R-squared

�1.90*** (�6.61) 1.92** (2.29)

�19.80*** (�3.36) �0.74 (�1.12) �0.98*** (�2.98)

�13.53*** (�3.25) 1.07*** (3.41) �1.15*** (�3.84)

0.58

0.65

2.93** (2.15) �1.85** (�2.08)

�6.12*** (�4.24) 0.82** (2.42) �0.61** (�2.47)

0.90

0.92

�2.41*** (�12.57) �5.96** (�2.06) �0.53 (�1.57) 0.10 (0.48) 0.89

�2.29 (�0.68) �0.24 (�0.30) 0.64

�2.34*** (�12.71) �3.46 (�1.52) 0.04 (0.23) 0.03 (0.14) 0.91

�2.60*** (�6.29) 2.39 (1.64) 5.03 (1.44) �1.19 (�1.60) 0.75

�2.90*** (�2.91)

�15.35** (�2.43) 3.44* (1.71) �1.49*** (�2.68)

�7.43*** (�3.26) 1.32* (1.91)

0.61

0.67

Long-term Credit and Trust Banks

City Banks

�2.06*** (�16.10)

�0.80 (�0.57)

Trust Banks 104

top ten shares

Long-term Credit and Trust Banks

City Banks

10.65 (1.37) �4.31 (�1.92) 0.50 (0.73) 0.73

Trust Banks �2.19*** (�13.64) �8.89*** (�3.29) 1.74** (2.04) �0.60** (�2.51) 0.92

�0.55 (�0.38) �2.13 (�0.77) �5.69 (�0.89) 0.95 (0.64) 0.78

Dependent Variable: Per capita Value-added.

Fixed Model and Instrumental Variable Method.

Figures in parentheses are the ratio of the coef®cient over the standard errors: *** signi®cant at 1% level ** at 5% level; and * at 10% level.

�2.00*** (�6.16) �7.31 (�1.35) 1.25 (0.79) �0.68 (�1.60) 0.89

Table 5.5 The structure of shareholdings for banks and management ef®ciency (1981±85) Total 20 Banks top ten shares life insurance shares log (employees) log (asset) Adjusted R-squared

�0.61 (�0.23) �1.49 (�1.32) 1.00*** (10.91) 0.96

102.08 (0.29) 22.89 (0.27) 1.43 (0.96) 0.02

Total 20 Banks labour share top ten shares life insurance shares log (employees) log (asset)

9.50 (1.61) �7.21** (�2.70) �0.75 (�0.96)

�1.62 (�0.22) �7.11** (�2.29) �0.51 (�0.63)

�1.40 (�1.16) 0.96*** (8.19)

0.90

0.89

0.96

�1.22*** (�3.81) 0.44*** (14.02)

�0.15 (�0.05) �1.44* (�1.99) 0.45*** (8.61)

0.99

0.99

�4.63*** (�18.90) 2.55** (2.22) �0.92 (�1.59) 0.37** (2.45) 0.99

�132.53 (�1.40) 5.16 (0.93) 0.96*** (4.09) 0.84

Long-term Credit and Trust Banks

City Banks �4.11*** (�9.75)

�4.03 (�0.907)

Trust Banks

�4.87*** (�18.11) 2.78* (1.83) 0.05 (0.07) 0.59*** (3.35) 0.99

�4.36*** (�18.13) 1.13 (1.06)

�0.93 (�0.20)

�425.96 (�1.39) �1.05 9.06 (0.64) (1.15) 1.07*** 1.21*** (7.94) (4.74) 0.96

0.68

Trust Banks �4.35*** (�17.70)

�0.85*** (�2.74) 0.43*** (9.23)

�0.84 (�0.12) �0.74* (�1.78) 0.42*** (9.13)

0.99

0.99

�4.50*** �4.48*** (�19.66) (�18.41) 0.81 (0.73) 15.83 (0.52) �1.29*** �1.60** (�3.47) (�2.17) 0.40*** 0.39*** (8.48) (7.86) 0.99

Dependent Variable: Per capita After-tax Pro®t.

Fixed Model and Instrumental Variable Method.

Figures in parentheses are the ratio of the coef®cient over the standard errors: *** signi®cant at 1% level; ** at 5% level; and * at 10% level.

0.99

105

Adjusted R-squared

�4.26*** (�18.24) 1.36 (1.66)

Long-term Credit and Trust Banks

City Banks

Table 5.6

The structure of shareholdings for banks and management ef®ciency (1986±90)

top ten shares life insurance shares log (employees) log (asset) Adjusted R-squared

�1.27 (�0.47)

top ten shares life insurance shares log (employees) log (asset) Adjusted R-squared

0.46 (0.07)

�1.28* (�1.74) 0.48*** (2.98)

�4.58 (�0.85) �0.93 (�1.65) 0.40** (2.11)

�1.25** (�2.27) 0.71** (2.21)

�0.03 (�0.00) �1.28*** (�3.33) 0.69*** (3.08)

0.90

0.89

0.82

0.82

Total 20 Banks labour share

Long-term Credit and Trust Banks

City Banks

�3.77*** (�23.49) 1.31* (1.71) �0.30 (�1.36) 0.45*** (10.49)

2.27 (1.56) �0.61*** (�3.81) 0.48*** (9.44)

0.99

0.99

�3.93*** (�33.55) 2.26** (2.12)

�3.03 (�0.97) 0.58 (1.01) 0.92

�4.43 (�0.52) �1.33 (�0.64) 0.29 (0.64) 0.92

Long-term Credit and Trust Banks

City Banks

�3.64*** (�25.80)

�2.42 (�0.75)

Trust Banks

�3.86*** (�34.67)

�0.21** (�2.06) 0.47*** (9.12)

2.60** (2.26) �0.40*** (�5.54) 0.43*** (11.52)

0.99

0.99

�3.49*** (�11.73) 0.32 (0.26)

�2.61 (�0.97) �1.68 (�0.54) 0.06 (0.08) 0.90

106

Total 20 Banks

�8.68 (�0.91) 0.86 (0.29) �0.53 (�0.64) 0.89

Trust Banks �3.45*** (�13.29)

�1.23 (�1.03) 0.59*** (2.91)

0.78 (0.24) �1.46* (�1.84) 0.62*** (3.71)

0.99

0.99

�3.44*** �3.49*** (�14.99) (�17.76) �0.43 (�0.55) �1.05 (�0.45) �0.64 �0.26 (�0.72) (�0.37) 0.28 0.20 (1.40) (1.08) 0.99

Dependent Variable: Per capita After-tax Pro®t.

Fixed Model and Instrumental Variable Method.

Figures in parentheses are the ratio of the coef®cient over the standard errors: *** signi®cant at 1% level; ** at 5% level; and * at 10% level.

0.99

Toshiaki Tachibanaki and Hideo Okamura 107

The coef®cients of both top ten shares and life insurance shares are statistically signi®cant and negative for city banks in the case in which the labour share variable is added. The effect of the number of employees is all negative for city banks, and it is also negative largely for other types of banks. The effect of total asset values is positive for both total twenty banks and city banks. This is true also for long-term credit banks and trust banks in the case in which the labour share variable is added. Finally, all the samples show the negative coef®cients for the labour share variables. The results in Table 5.6 enable us to propose the following conclusions. The governance capability of both shareholders and life insurance companies for city banks was effective from 1986 to 1990. It was not, however, effective for long-term credit banks and trust banks. The positive effect of total asset values on productivity can be explained by the speci®c feature of the bubble economy in the late 1980s. The similar result, i.e., the negative coef®cient of the labour share variable, was obtained also for this period as the results in Tables 5.4 and 5.5 showed. It should be useful to summarise and discuss the overall conclusions on the basis of Tables 5.4, 5.5 and 5.6 One interesting observation is that the higher the degree of intercorporate shareholding in the 1990s, the lower the ef®ciency in management in the banking industry, although the positive effect was observed in the 1980s. One reason for this shift of shareholders' governancing role may be summarised as follows. The problem of bad debts and long-run recessions after the bubble economy in the 1990s changed the course of the corporate governance system in the banking industry, or invalidated the traditional intercorporate governance system which worked relatively well in the 1980s. An alternative interpretation may run as follows: although the traditional governance role which was expected from shareholders did not work well even in the 1980s, it raised the internal governance capability. `Internal' here means both managers who climbed to the top by internal promotion ladder, and employees who are potential future managers. Both managers and employees worked hard to raise productivity of their ®rm (i.e., bank) because it was not necessary for them to worry about the intervention of shareholders who would be sometimes interested only in the short-run dividends motive. In other words, shareholders were largely silent, and the notion of labour managed ®rms was relevant in the 1980s. See, for example, Tachibanaki (1998b) about the interpretations of these results. The similar argument is possible for the role of life insurance companies. They were marginally effective in the 1980s as shareholders because life insurance companies could enjoy high monetary returns to

108 Governance Structure of Banks and Their Business Performance

shareholding due to the increase in the stock price in the 1980s. It was not necessary for them to express their voice to the company whose shares are held by them because the amount of capital gains due to stockholding was large. Since life insurance companies were quiet as shareholders, managers in other ®rms were able to manage their own ®rms without worrying about shareholders' (i.e., life insurance companies's) intervention. This is the role of corporate governance performed by life insurance companies. The above mechanism could not work in the 1990s because the stock price started to decline considerably. Life insurance companies were expected to play an important role in the 1990s as institutional shareholders who govern and express their opinion to ®rms whose shares are held by them. They, however, were unable to do so, or did not have a strong intention to control ®rms for the following reasons. First, they were not accustomed to act as an active shareholder like other shareholders in Japan. Second, intercorporate shareholding implicitly assumed that shareholders who are managers of the ®rm and at the same time who hold shares of their counter-part were quiet and silent about management of their counter-part. Third, life insurance companies did not have a strong incentive as being an active governance agency for corporated ®rms because they do not issue their stocks. They are mutual companies rather than stock companies. In other words, since life insurance companies are mutual companies, they are not controlled or governed ef®ciently by anybody. Put simply, their performance in management was not so ef®cient in general. It is desirable that life insurance companies are concerned more seriously with the management of other companies. A similar statement can be addressed to life insurance holders who should be concerned with the management in their life insurance company. Finally, it is important to point out one institutional background which governed the whole structure of the ®nancial industry in Japan. That is the heavy regulation on the industry by the Ministry of Finance and the Bank of Japan. This heavy regulation, which does not encourage severe competition, protected the industry strongly, and thus produced a fairly large amount of regulatory rents. See, for example, Tachibanaki (1996) on the effect of the regulation on the ®nancial industry. The lack of severe competition in the ®nancial industry and the extra rents enjoyed by the industry led many banks and life insurance companies to be less interested in increasing their productivity. Thus, they did not commit to strong control and governance activity for other ®rms. The relationship between total asset value and management ef®ciency is interesting because it was positive in the 1980s, while it was negative in

Toshiaki Tachibanaki and Hideo Okamura 109

the 1990s. The heavy regulation on the ®nancial industry in the 1980s is responsible for the former positive relationship, because a larger total asset value of a bank can help this relationship. A large amount of bad debts, however, which were accumulated by a bank whose total asset value is large in the 1990s, is responsible for the latter negative relationship, since such a bank committed to the excessive lending activity during the bubble economy in the late 1980s. It is desirable for banks to have a better risk management policy. The negative coef®cient of the labour share variable can be interpreted in different ways. Incidentally, these different interpretations can provide us with a clue to understand the Japanese way of corporate ®nance and/or corporate governance. First, the marginal productivity principle in the determination of wages is not supported in the ®nancial industry. Somewhat related to this point, it is possible to propose that the ef®ciency wage hypothesis, which postulates that higher productivity or the hard work of employees is expected by higher wage payments, does not work. Second, as pointed out earlier, there has been a heavy regulation on the ®nancial industry. This regulation is responsible for the excessive wage payments to employees in banks. The lack of competition, moreover, has not encouraged banks to take an aggressive business activity in the ®eld of new ®nancial services and new business innovation, and to be conscious with cost minimisation. These observations imply the negative effect of labour shares on productivity. There are a plenty of documents which describe that productivity in the Japanese banking industry is much inferior to that in the US banking industry. Third, it is possible to guess that the rate of return to capital (i.e., equity) is lower because of the high monetary rewards to employees. This possible low rate of return to capital does not give a strong motivation of monitoring or governing the management of a bank severely to capital holders. The low incentive and thus the lack of governance of capital holders is likely to lower productivity of a bank. The above mechanism is intensi®ed by the Japanese characteristic of intercorporate shareholding in general. Intercorporate shareholding implicitly assumes quiet and silent shareholders regarding the management of a ®rm. While in the case of the manufacturing industry it worked positively and favourably to raise productivity, it worked negatively in the case of the ®nancial industry. The distinction appears due largely to severe competition in the former and heavy regulation in the latter.

110 Governance Structure of Banks and Their Business Performance

4 4.1

The cost of capital at banks Motivation

The purpose of this part is to estimate the cost of equity capital at major banks, and presents our interpretation of the monitoring activity and the performance of each bank based on the estimated cost of capital. It would be interesting to compare the estimated result for the ®nancial industries and the one for non-®nancial industries which was estimated by Ando and Auerbach (1988a, 1988b, 1990), One important feature of this exercise is to take into account the effect of intercorporate shareholding in view of the fact that intercorporate shareholding provides us with some underestimation of the cost of capital. In other words, we attempt to adjust for a bias caused by intercorporate shareholding. The above task is important because the unadjusted cost of capital may misguide the manager to commit to overinvestment, and is likely to lead the ®rm to inef®cient management. Therefore, we are going to estimate the cost of capital for each individual bank whose shares are held by other ®rms, and assess each bank's performance based on the adjusted cost of capital. We employ two methods for the adjustment. It is necessary to explain the reason why we do not take into consideration the role of debt. Since a bank collects ®nancial funds from the two sources, broadly speaking, debt (i.e., deposits and other forms of debt) and equity, debt can be included in addition to equity. The cost of capital is normally calculated based on debt, equity and sometimes retained earnings for non-®nancial ®rms. Since debt for ®nancial ®rms is different from debt for non-®nancial ®rms, it is risky to treat it for the former in the same way as for the latter. Thus, we ignore debt to calculate the cost of capital for ®nancial ®rms. 4.2

Estimation methods

We estimate the cost of capital for 10 city banks, three long-term credit banks, and seven trust banks. The estimation period is from 1988 to 1994. Intercorporate shareholding overevaluates the price-earnings ratio, and thus underevaluates the cost of capital according to current accounting rule, as Kobayashi (1990), Kurasawa, Taki and Okazaki (1992), Yonezawa (1995) pointed out. It is important to treat the pro®t and its evaluation in the equity market correctly. Moreover, we have to apply the ex ante, or expected rate of return to equity to calculate the cost of capital. The data, however, are available only on the basis of the ex post sense. We have to take into account this fact.

Toshiaki Tachibanaki and Hideo Okamura 111

There are several methods for calculating the cost of capital such as the rate of return to equity method, the discounted dividends method, the inverse of the price-earnings ratio, etc. We use the method based on the price-earnings ratio, and the price-cash ¯ow ratio. The inverse of the estimated value can be regarded as the cost of capital. The calculation method is explained here with considerable depth. One important element of intercorporate shareholding is that the amount of dividends which are received from the counter-part of intercorporate shareholding is included in the pro®t of the ®rm, while the amount of capital gains is not included. In other words, the share price of the ®rm can be re¯ected by the amount of retained earnings of the counter-part through the increase in the share price of the counter-part. However, the amount of retained earnings of the counter-part is not included in the pro®t of the ®rm. Equation (3) is used to adjust for intercorporate shareholding in the entire equity market. It is not, however, legitimate to apply equation (3) for an individual ®rm's equity market where the propensity to dividends payout differs from ®rm to ®rm. Kurasawa, Taki and Okazaki (1992) proposed a relevant method for an each individual ®rm in the case in which the equity market is rational and perfect, and thus the determination of equity price is ef®cient. The adjusted PER is in¯uenced by both the rate of shareholding and the propensity to dividends payout in the following way: …The adjusted PER† ˆ …Unadjusted PER†  …1 � †=…1 � d†

…3†

where  is the rate of shareholding and d is the propensity to dividends payout. Tables 5.7, 5.8 and 5.9 present the unadjusted PER (A1 and B1), and the adjusted PER (A2 and B2). `A' signi®es that the sum of after-tax pro®ts and depreciations is used as the index of pro®ts, while `B' signi®es that only after-tax pro®ts is used as the index of pro®ts. Each table shows the percentage value of the inverse of the PCFR and the PER, respectively. The numerical rate of intercorporate shareholding from 1988 to 1992 is available in Takano (1993), and its extension to 1994 was calculated by us by using the Survey on Distribution on Shareholder published by the Tokyo Stock Exchange. Suppose we consider two ®rms (®rm 1 and ®rm 2) which hold the other ®rm's shares each other. The observed pro®t (E1 *, E2 *) and equity price (P1 *, P2 *) are given by the following formulations, E1  ˆ E1 ‡ d2
2
E2  P1  ˆ P1 ‡ 2
P2 

E2  ˆ E2 ‡ d1
1
E1  P2  ˆ P2 ‡ 1
P1 

…4†

112

Table 5.7 Equity cost of capital (city banks) Daiichi-kangyo year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

1988 1989 1990 1991 1992 1993 1994

2.095 2.448 1.722 1.543 1.256 0.906 1.015

3.224 3.734 2.543 2.184 1.605 1.113 1.133

1.128 1.286 0.821 0.641 0.349 0.207 0.118

year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

1988 1989 1990 1991 1992 1993 1994

2.206 2.091 2.080 3.546 1.928 1.767 1.707

3.393 3.189 3.071 5.019 2.464 2.171 1.906

1.188 1.098 0.991 1.473 0.536 0.404 0.199

4.837 7.945 3.833 2.644 1.797 1.087 1.223

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

(B3±B1)

1.899 2.104 1.354 1.283 0.950 0.581 0.591

2.921 3.209 2.000 1.816 1.214 0.713 0.659

1.022 1.105 0.645 0.533 0.264 0.133 0.069

4.303 6.589 2.687 1.860 0.873 ± ±

2.404 4.485 1.333 0.577 �0.077 ± ±

(A3±A1)

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

(B3±B1)

25.006 11.257 26.867 ± ± ± ±

1.868 1.602 1.447 2.449 1.038 0.855 0.696

2.873 2.443 2.137 3.467 1.326 1.051 0.777

1.006 0.841 0.689 1.018 0.288 0.195 0.081

(A3±A1) 2.742 5.497 2.111 1.101 0.541 0.181 0.208

Sakura

27.212 13.348 28.947 ± ± ± ±

21.791 9.101 15.775 ± ± ± ±

19.923 7.499 14.328 ± ± ± ±

Table 5.7

Equity cost of capital (city banks) (continued )

Fuji year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

1988 1989 1990 1991 1992 1993 1994

2.078 2.058 2.067 1.311 1.405 1.189 1.510

3.197 3.139 3.051 1.855 1.795 1.461 1.686

1.119 1.081 0.985 0.545 0.390 0.272 0.176

4.151 3.905 4.081 1.717 2.008 1.683 2.201

(A3±A1) 2.073 1.847 2.014 0.406 0.603 0.494 0.691

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

1.857 1.730 1.583 0.544 0.556 0.427 0.641

2.857 2.639 2.337 0.770 0.710 0.525 0.715

1.000 0.909 0.754 0.226 0.154 0.098 0.075

3.603 3.063 2.839 ± ± ± ±

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

(B3±B1)

2.016 2.086 1.336 1.505 0.693 0.375 0.607

3.101 3.181 1.972 2.130 0.886 0.461 0.678

1.085 1.096 0.636 0.625 0.193 0.086 0.071

4.797 6.019 2.574 2.362 0.391 ± 0.165

2.781 3.933 1.238 0.858 �0.302 ± �0.442

(B3±B1) 1.746 1.333 1.256 ± ± ± ±

Tokyo-Mitsubishi year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

1988 1989 1990 1991 1992 1993 1994

2.339 2.500 1.812 2.191 1.305 0.966 1.333

3.598 3.813 2.675 3.101 1.668 1.187 1.488

1.259 1.313 0.863 0.910 0.363 0.221 0.155

5.717 7.503 3.959 4.183 1.663 1.082 1.979

(A3±A1) 3.379 5.003 2.147 1.992 0.358 0.116 0.646

113

114

Table 5.7 Equity cost of capital (city banks) (continued ) Asahi year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

(A3±A1)

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

(B3±B1)

1988 1989 1990 1991 1992 1993 1994

3.022 2.622 2.775 3.091 2.203 2.139 2.095

4.649 4.000 4.097 4.375 2.816 2.628 2.339

1.627 1.377 1.322 1.284 0.612 0.488 0.244

± ± ± ± 11.326 75.829 9.481

2.417 1.977 1.982 1.736 0.863 0.836 0.810

3.718 3.015 2.926 2.458 1.102 1.026 0.905

1.301 1.038 0.944 0.721 0.240 0.191 0.094

± ± ± ± ± 2.470 0.333

± ± ± ± ± 1.635 �0.478

year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

(A3±A1)

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

(B3±B1)

1988 1989 1990 1991 1992 1993 1994

2.425 2.390 2.344 3.124 2.020 1.451 1.574

3.730 3.646 3.461 4.422 2.582 1.782 1.757

1.306 1.256 1.117 1.298 0.561 0.331 0.183

2.162 2.064 1.898 2.306 1.273 0.828 0.764

3.325 3.148 2.801 3.265 1.627 1.017 0.853

1.164 1.084 0.904 0.958 0.354 0.189 0.089

6.428 5.579 4.901 6.646 1.801 0.818 0.455

4.266 3.515 3.003 4.340 0.528 �0.010 �0.309

± ± ± ± 13.529 77.968 11.576

Sanwa

7.376 6.679 6.479 10.201 3.961 2.425 3.052

4.951 4.289 4.135 7.077 1.941 0.974 1.478

Table 5.7 Equity cost of capital (city banks) (continued ) Sumitomo year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

1988 1989 1990 1991 1992 1993 1994

2.197 2.461 2.241 2.645 0.779 0.990 ±

3.379 3.754 3.309 3.745 0.995 1.216 ±

1.183 1.293 1.068 1.099 0.216 0.226 ±

year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

1988 1989 1990 1991 1992 1993 1994

2.003 1.932 2.244 3.219 2.263 1.993 2.606

3.082 2.947 3.314 4.556 2.892 2.448 2.909

1.079 187.673 1.015 ± 1.069 ± 1.337 ± 0.629 ± 0.455 ± 0.303 ±

3.490 4.875 4.505 4.549 ± ± ±

(A3±A1) 1.293 2.414 2.264 1.903 ± ± ±

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

1.962 2.182 1.893 2.069 0.237 0.528 ±

3.019 3.328 2.794 2.929 0.303 0.649 ±

1.057 1.146 0.902 0.860 0.066 0.121 ±

3.039 4.180 3.600 3.058 ± ± ±

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

1.709 1.575 1.771 2.223 1.203 0.890 1.096

2.630 2.403 2.614 3.147 1.537 1.093 1.224

0.920 150.265 0.827 ± 0.844 ± 0.924 ± 0.334 ± 0.203 ± 0.128 ±

(B3±B1) 1.077 1.998 1.707 0.989 ± ± ±

Daiwa (A3±A1) 185.670 ± ± ± ± ± ±

(B3±B1) 148.556 ± ± ± ± ± ±

115

Equity cost of capital (city banks) (continued )

116

Table 5.7 Tokai year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

(A3±A1)

Cost of Capital B1

Cost of Capital B2

Cost of (B2±B1) Capital B3

(B3±B1)

1988 1989 1990 1991 1992 1993 1994

1.689 2.061 2.198 2.000 2.054 1.852 1.819

2.598 3.143 3.245 2.832 2.624 2.275 2.030

0.909 1.082 1.047 0.831 0.571 0.423 0.212

4.833 ± 41.841 ± ± ± ±

1.329 1.552 1.636 0.882 0.979 0.913 0.976

2.045 2.367 2.415 1.249 1.251 1.121 1.089

0.716 0.815 0.779 0.367 0.272 0.208 0.114

3.105 ± 24.628 ± ± ± ±

6.521 ± 44.039 ± ± ± ±

4.434 ± 26.264 ± ± ± ±

Hokkaido-Takushoku year

Cost of Capital A1

Cost of Capital A2

Cost of (A2±A1) Capital A3

(A3±A1)

Cost of Capital B1

Cost of Capital B2

Cost of (B2±B1) Capital B3

(B3±B1)

1988 1989 1990 1991 1992 1993 1994

2.616 2.120 3.204 4.620 3.138 2.746 4.334

4.025 3.234 4.731 6.539 4.010 3.373 4.838

1.409 1.114 1.526 1.920 0.872 0.627 0.504

± 91.480 ± ± ± ± ±

1.896 1.479 2.121 3.065 1.592 0.964 1.827

2.917 2.256 3.132 4.338 2.035 1.184 2.040

1.021 0.777 1.011 1.273 0.442 0.220 0.213

± 53.772 ± ± ± ± ±

± 93.600 ± ± ± ± ±

± 55.251 ± ± ± ± ±

Table 5.7

Equity cost of capital (city banks) (continued )

Notes: (1) Figures are percentages. (2) The method for calculation is explained in the main text. (3) The degree of intercorporate shareholding and the propensity to dividends payout are as follows in. year intercorporate shareholding propensity to dividends payout degree of adjustment

1988

1989

1990

1991

1992

1993

1994

42.83

42.06

40.60

40.15

39.92

40.37

40.50

28.13

27.64

30.30

38.06

58.18

66.27

82.91

53.84

52.52

47.64

41.55

27.79

22.84

11.63

(4) The degree of adjustment is calculated by equation (4.1) in the main text.

Sources: Financial Statements prepared by Nikkei Needs, and Nomura Research Institute, and the same as in

Table 5.1±5.6.

117

118 Governance Structure of Banks and Their Business Performance Table 5.8 Equity cost of capital (trust banks) Mitsui Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1988 1989 1990 1991 1992 1993 1994

2.772 3.201 2.445 2.432 2.167 1.481 1.667

4.265 4.883 3.610 3.443 2.769 1.820 1.861

2.542 2.854 1.995 1.766 1.151 0.878 1.061

3.911 4.354 2.946 2.500 1.470 1.078 1.185

1.369 1.499 0.951 0.734 0.320 0.200 0.123

1.493 1.682 1.165 1.011 0.602 0.338 0.194

Mitsubishi Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1988 1989 1990 1991 1992 1993 1994

2.389 3.414 2.608 3.243 2.320 1.495 1.423

3.675 5.208 3.851 4.590 2.965 1.836 1.589

2.171 2.916 2.052 2.010 1.165 0.780 0.749

3.340 4.448 3.030 2.846 1.488 0.959 0.836

1.169 1.532 0.978 0.835 0.324 0.178 0.087

1.286 1.793 1.243 1.347 0.645 0.341 0.166

Yasuda Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1988 1989 1990 1991 1992 1993 1994

3.160 3.090 2.400 2.874 1.943 1.680 2.038

4.862 4.714 3.544 4.069 2.483 2.064 2.275

2.833 2.682 1.928 2.083 1.061 1.037 1.322

4.358 4.090 2.846 2.949 1.356 1.274 1.476

1.525 1.409 0.918 0.866 0.295 0.237 0.154

1.702 1.623 1.144 1.194 0.540 0.384 0.237

Toyo Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1988 1989 1990 1991 1992 1993 1994

2.939 2.945 2.090 2.845 1.917 1.538 1.854

4.522 4.492 3.086 4.028 2.450 1.889 2.070

2.728 2.617 1.678 1.863 0.992 0.852 1.125

4.198 3.992 2.477 2.638 1.268 1.046 1.256

1.469 1.375 0.799 0.774 0.276 0.194 0.131

1.583 1.547 0.996 1.182 0.533 0.351 0.216

Toshiaki Tachibanaki and Hideo Okamura 119 Table 5.8 Equity cost of capital (trust banks) (continued ) Chuo Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

1988 1989 1990 1991 1992 1993 1994

(Unlisted) 2.566 2.329 1.760 1.660 1.663 1.800

(Unlisted) 3.913 3.438 2.491 2.122 2.043 2.009

(Unlisted) 2.239 1.803 1.218 0.935 0.937 1.007

(Unlisted) 3.415 2.662 1.725 1.195 1.151 1.124

1.348 1.110 0.731 0.461 0.380 0.209

(B2±B1)

1.176 0.859 0.506 0.260 0.214 0.117

Nihon Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1988 1989 1990 1991 1992 1993 1994

1.610 1.416 3.176 4.264 4.266 4.050 ±

2.476 2.159 4.689 6.036 5.451 4.975 ±

1.417 1.143 2.078 1.911 1.560 1.584 ±

2.179 1.744 3.069 2.705 1.993 1.945 ±

0.763 0.601 0.990 0.794 0.433 0.362 ±

0.867 0.744 1.513 1.772 1.185 0.925 ±

Sumitomo Trust year

Cost of Capital A1

Cost of (A2±A1) Capital A2

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1988 1989 1990 1991 1992 1993 1994

2.464 3.276 3.129 3.391 2.761 1.566 1.697

3.791 4.996 4.620 4.800 3.528 1.924 1.894

2.244 2.797 2.462 2.381 1.550 0.751 0.859

3.453 4.266 3.635 3.370 1.981 0.922 0.959

1.208 1.469 1.173 0.989 0.431 0.171 0.100

1.327 1.721 1.491 1.409 0.767 0.358 0.197

where Ei (i = 1, 2) is the pro®t in the case in which there is no intercorporate shareholding, i is the rate of i-th ®rm's shares which are held by the other ®rm, di is the propensity to dividends payout, and Pi is the equity price in the case in which there is no intercorporate shareholding. The above formulations imply that the part of pro®t of the other ®rm, i.e., the rate multiplied by di * i , is added to the pro®t of the ®rm, while

120

Table 5.9 Equity cost of capital (long-term credit banks) Industrial Bank of Japan year

Cost of Capital A1

Cost of Capital A2

(A2±A1) Cost of Capital A3

1988 1989 1990 1991 1992 1993 1994

1.025 1.023 1.017 1.453 0.699 0.554 0.648

1.577 1.560 1.501 2.057 0.894 0.680 0.724

0.552 0.537 0.484 0.604 0.194 0.127 0.075

2.077 1.869 1.311 2.540 ± 0.058 ±

(A3±A1) 1.052 0.846 0.295 1.087 ± �0.496 ±

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

0.954 0.942 0.934 1.307 0.537 0.407 0.474

1.467 1.438 1.380 1.850 0.686 0.500 0.530

0.514 0.495 0.445 0.543 0.149 0.093 0.055

1.853 1.621 1.102 1.944 ± ± ±

Cost of Capital B1

Cost of Capital B2

(B2±B1)

Cost of Capital B3

1.564 1.440 1.571 2.380 1.231 1.170 1.111

2.405 2.197 2.319 3.370 1.573 1.437 1.241

0.842 0.757 0.748 0.989 0.342 0.267 0.129

(B3±B1) 0.900 0.678 0.168 0.637 ± ± ±

Long-term Credit Bank of Japan year

Cost of Capital A1

Cost of Capital A2

(A2±A1) Cost of Capital A3

1988 1989 1990 1991 1992 1993 1994

1.614 1.508 1.655 2.543 1.506 1.378 1.432

2.484 2.300 2.443 3.600 1.924 1.692 1.599

0.869 0.792 0.788 1.057 0.418 0.315 0.167

± 78.937 73.696 ± 0.762 0.819 5.420

(A3±A1) ± 77.430 72.042 ± �0.744 �0.559 3.988

± 72.585 65.778 ± 1.454 1.268 6.115

(B3±B1) ± 71.144 64.207 ± 0.223 0.098 5.004

Table 5.9

Equity cost of capital (long-term credit banks) (continued )

Japan Bond and Credit Bank year

Cost of Capital A1

Cost of Capital A2

(A2±A1) Cost of Capital A3

1988 1989 1990 1991 1992 1993 1994

1.704 1.301 1.751 2.467 1.812 1.826 2.240

2.621 1.985 2.586 3.492 2.316 2.243 2.500

0.917 0.684 0.834 1.025 0.504 0.417 0.261

± 4.620 23.162 ± ± 0.245 ±

(A3±A1) ± 3.318 21.411 ± ± �1.581 ±

Cost of Capital B1

Cost of Capital B2

(B2±B1)

1.596 1.210 1.608 2.232 1.403 1.351 1.529

2.455 1.846 2.374 3.160 1.793 1.660 1.707

0.859 0.636 0.766 0.928 0.390 0.309 0.178

Cost of Capital B3 ± 4.117 19.745 ± 0.634 1.037 0.184

(B3±B1) ± 2.907 18.137 ± �0.769 �0.314 �1.345

121

122 Governance Structure of Banks and Their Business Performance

the equity price of the ®rm is added only by the rate of corporate shareholding. There are two approaches in order to adjust for the in¯uence of intercorporate shareholding; the ®rst is to presume that all pro®ts are distributed as dividends, and the second is to consider the case in which there is no intercorporate shareholding. We take the second approach. The price-earnings ratio at the second approach is written as follows: P1 = E1 ˆ …P1  � 2
P2  †=…E1  � d2
2
E2  †

…5†

When we consider more than three ®rms, the adjusted PER is written as follows: PER ˆ …P �  i
Pi †=…E �  Di †

…6†

where P* is the current share value of the ®rm,  i * Pi is the current share value of equities which the ®rm holds, E* is the accounting pro®t of the ®rm, and  Di is the dividends received. The most dif®cult task is to estimate  i * Pi accurately because their values for a bank were not published of®cially, but only its accrued capital gains of holding equities. We used the sum of the latter and the book value of holding stocks as a substitute for the current value of holding stocks. One de®ciency of this method is that the accrued capital gains contain not only ones for equity but also for corporate bond. In view of the fact that the share of corporate bonds in the Japanese ®rms is small the bias due to this method is not large. We present two cases, namely PCFR where E* is given by sum of aftertax pro®t and depreciations, and PER where E* is given only by after-tax pro®t. The cost of capital is obtained by taking the inverse of PCFR and PER, respectively. It is called A3 and B3, respectively. It is noted that the numerator in equation (6) turned out being negative in several cases. In such a case no numerical value is written (i.e., blank). Unfortunately, nearly all trust banks gave negative values. Thus, the result for them has been eliminated. 4.3

Estimated results

Tables 5.7, 5.8 and 5.9 show the estimated results on the cost of capital, and Figures 5.1, 5.2, 5.3, 5.4 and 5.5 are their graphic representation for A1. The unadjusted costs of capital give the values about 2.03.0% for city banks in 1988. Although they increased somewhat right-after the bubble economy, say in 1991, the majority are between 1.52.5% in 1994. The difference in the cost of capital among city banks has expanded since

Toshiaki Tachibanaki and Hideo Okamura 123

1991. The values for trust banks were 2.53.0% except for one in 1988. They declined since 1992, and reached 1.52.0% levels in 1994. Regarding long-term credit banks the difference between Industrial Bank of Japan (IBJ) and other banks was considerable, and the difference between Long-term Credit Bank of Japan and Japan Bond and Credit Bank has been widening since 1992. We examined two methods to adjust for intercorporate shareholding. The ®rst is based on equation (3), and the second is based on equation (6). We compare, ®rst, the unadjusted cost of capital and the adjusted one based on equation (3). It should be noted that the common adjustment is made each year for all banks. The result suggests that the degree of adjustment has declined from 53.84% in 1988 to 11.63% in 1994. This decline re¯ects the gradual decrease in the propensity to dividends payout after 1988 in many ®rms, and equation (3) is able to absorb it. The difference between the unadjusted cost capital and the adjusted one was about 1.0% for city banks, and trust banks in 1988, while it was 0.5%1.0% for long-term credit banks. The difference declined since then, and it is about 0.10.3% for many banks in 1994. A3 and B3 in Tables 5.7, 5.8 and 5.9, which are calculated based on equation (6), are probably the most desirable ®gures at least in the conceptual sense. It was unfortunate, however, that the empirical results were less impressive because some of the ®gures in numerators and/or denominators were negative, and the estimated costs of capital ¯uctuated considerably. The result for trust banks was particularly bad. There are two main reasons for this. First, the estimation method for current share values is not appropriate. We used the sum of the book share value and the accrued capital gain of all securities to represent the current share values. Since both the average equity price and the general interest rates decreased in the 1990s, the contribution of accrued capital gains in bonds other than stocks increased relatively. Second, many banks had a large amount of bad debts in the 1990s, and some of them committed to cancel these bad debts. This obviously provides banks with negative pro®ts. The reliable result, nevertheless, is available for several city banks such as Daiichi-Kangyo, Fuji, Tokyo-Mitsubishi and Sanwa banks. The empirical result is presented for these banks. It is found that the cost of capital in 1988 was around 47%, and it declined signi®cantly to 1.03.0% in 1994. The difference between the unadjusted cost of capital and the adjusted one in 1988 and 1989 was about 2.02.5% points. Therefore, the degree of underestimation was very large, say between 100 and 200%. In 1993 and 1994 the difference, however, declined to

Equity cost of capital (city banks 1)

124

Figure 5.1 5.0 4.5 4.0 3.5 3.0

Tokyo-Mitsubishi Sumitomo

2.5

Fuji 2.0 Sakura 1.5

Sanwa

1.0 0.5

0

1988

1989

1990

1991

1992

1993

1994

Figure 5.2

Equity cost of capital (city banks 2)

5.0 4.5 4.0 3.5 3.0 Daiichi-kangyo 2.5

Asahi Daiwa

2.0 Tokai 1.5

Hokkaido-Takushoku

1.0 0.5

0

1988

1989

1990

1991

1992

1993

1994

125

Equity cost of capital (trust banks 1)

126

Figure 5.3 5.0 4.5 4.0 3.5 3.0

Mitsui Trust 2.5 2.0

Mitsubishi Trust Yasuda Trust

1.5

Sumitomo Trust

1.0 0.5 0

1988

1989

1990

1991

1992

1993

1994

Figure 5.4

Equity cost of capital (trust banks 2)

5.0 4.5 4.0 3.5 3.0 Tokyo-Mitsubishi

2.5

Chuo Trust

2.0

Nihon Trust 1.5 1.0 0.5 0 1988

1989

1990

1991

1992

1993

1994

127

Equity cost of capital (long-term credit banks)

128

Figure 5.5

5.0 4.5 4.0 3.5 3.0 Industrial Bank of Japan Long-term Credit Bank of Japan Japan Boad and Credit Bank

2.5 2.0 1.5 1.0 0.5 0

1988

1989

1990

1991

1992

1993

1994

Toshiaki Tachibanaki and Hideo Okamura 129

0.11.0% points, implying that the degree of underestimation is about 20100%. In some cases the overestimation appeared. The above result suggests the following conclusions. The cost of capital for banks has been in a decreasing trend since 1988. However, the difference among banks has been widening. The principal cause of the declining trend is that after-tax pro®t of each bank decreased signi®cantly due partly to the problem of bad debts in the 1990s. Also, a general background such that the Japanese economy was under a serious recession, and thus the growth rate and the interest rate were low, may be suggested. It is important, however, to recognise that the difference in the performance of each bank was greater in the 1990s than the past. Another important conclusion is that the degree of underestimation caused by intercorporate shareholding has declined signi®cantly since 1988. The degree of underestimation is affected by the following two factors: the rate of intercorporate shareholding, and the propensity for dividend payouts. Since the rate of intercorporate shareholding has been almost constant with a very minor decrease, it is quite likely that the increase in the propensity to dividends payout is responsible for the decrease in the underestimation of the cost of capital. We should not forget, however, the fact that the market force (i.e., the business performance of each bank) differentiated the equity price of each bank considerably, and thus that the difference in the cost of capital is considerably large among banks, sometimes it is more than a double.

5

Concluding remarks

This study investigated the role of shareholders in the determination of banks' corporate governance under the condition that intercorporate shareholding is common in Japan. One important aspect, which cannot be ignored, is that no debt-holders (i.e., individual savers) are interested in governing or monitoring banks' behaviour or management. It would be appropriate to state that individual savers and depositors are unable to monitor banks' activity. Therefore, shareholders of banks' stocks are supposed to be crucial in the governance role. Several important conclusions arise from this study. First, the role of intercorporate shareholding differs from period to period, if we are concerned with the role of a bank. In the 1980s it worked fairly well to raise management ef®ciency. It did not, however, work well in the 1990s after the Japanese economy experienced the bubble economy of the late

130 Governance Structure of Banks and Their Business Performance

1980s. We presented several speci®c reasons why such a change in the role of shareholders occurred during the bubble economy. Second, we discussed in detail the peculiar role of life insurance companies which are, probably, the most important institutional shareholders. Life insurance companies played an important role in raising productivity of a bank in some cases, while they did not in other cases. One reason for this mixed outcome stems from the fact that life insurance companies are mutual companies. Third, the relationship between total asset value of a bank and its productivity also differs from time to time. It was positive in the 1980s, and negative in the 1990s. The huge number of bad debts accumulated in the early 1990s, and the heavy regulation on the ®nancial industry (i.e., the lack of competition in the industry) are responsible for this shift in the relationship between total asset and productivity. Fourth, the negative effect of the labour share variable on productivity was observed. We presented several causes to explain it, such as the inappropriateness of the ef®ciency wage hypothesis, the heavy regulation on the industry, the lower rate of return to capital, etc. Fifth, the cost of equity capital for banks has been in a decreasing trend since 1988. We found about 50100% underestimation of the cost of capital in 1988, and 1050% underestimation in 1994. The task of estimation for the cost of capital is dif®cult and complicated. More speci®cally, we require more data to obtain a reliable result under the condition of intercorporate shareholding, and thus hope more disclosure of the data in the ®nancial industry. Sixth, the difference in the cost of capital among banks has been widening. In some cases the difference is over twice between the highest cost of capital and the lowest one. It re¯ects the fact that the difference in management ef®ciency among banks appears to be large. Seventh, the effect of intercorporate shareholding on the cost of capital has been declining since 1988. There are two reasons for this trend. One is a change in the rate of intercorporate shareholding, and the other is a change in the propensity to dividends payment. We found that the latter is largely responsible. There have been no serious studies on such subjects as `Who governs and control a bank?' `Does a bank play an important role as a monitor?' `What is the role of life insurance companies in the governance structure?' We hope that this study sheds light on the understanding of these issues and problems.

Toshiaki Tachibanaki and Hideo Okamura 131

References Aghion, P. and P. Bolton (1992) `An Incomplete Contract Approach to Financial Contracting', Review of Economic Studies, 59, pp. 473±94. Ando, A. and A. Auerbach (1988a) `The Cost of Capital in the United States and Japan: A Comparison', Journal of the Japanese and International Economies, 2, pp. 134±58. Ando, A. and A. Auerbach (1988b) `The Corporate Cost of Capital in Japan and the United States: A Comparison', in J. Shoven (ed.), Government Policy towards Industry in the United States and Japan, Cambridge University Press. Ando, A. and A. Auerbach (1990) `The Cost of Capital in Japan: Recent Evidence and Further Results', Journal of the Japanese and International Economies, 4, pp. 323±50. Baltagi, B. (1995) Econometric Analysis of Panel Data, John Wiley and Sons. Berle, A.A. and G.C. Means (1932) The Modern Corporation and Private Property, Macmillan Press. Dewatripont, M. and J. Tirole (1994) The Prudential Regulation of Banks, Bonton, MA: MIT Press. Hart, O. (1995) Firms, Contracts, and Financial Structure, Oxford: Clarendon Press. Hirota, S. and K. Ikeo (1996) `Corporate Finance and Ef®ciency in Management', in H. Ito (ed.) Japan's Firm System, chapter 2, University of Tokyo Press (in Japanese). Hirota, S. (1996) `Japan's Financial and Equity Markets, and Corporate Governance', in T. Tachibanaki and Y. Tsutsui (eds.), Japan's Capital Markets, chapter 10, Nihon-hyoronsha (in Japanese). Horiuchi, A. and K. Shimizu (1997) `An Analysis of the Issue Who Monitors the Monitor: The Impact of Amakudari on Bank Performance', mimeo. Ikeo, K. (1991) `Economics of BIS Rule', Keizai Kenkyu, vol. 42, no. 3 (in Japanese). Ikeo, K. (1994) `Comment on Horiuchi Paper', Bank of Japan: Kinyu Kenkyu, vol. 13, no. 3 (in Japanese). Kobayashi, T. (1990) `Fundamental Value of Stock Price', in K. Nishimura and Y. Miwa (eds.), Japan's Stock and Land Prices, chapter 12, University of Tokyo Press (in Japanese). Kurasawa, T. (1993) `Financial Effect of Intercorporate Shareholding', Nihonkeizai Kenkyu (in Japanese). Kurasawa, T., A. Taki and T. Okazaki (1992) `Cost of Capital in Banks', Bank of Japan: Kinyu Kenkyu, vol. 11, no. 1 (in Japanese). Saito, T. (1997) `Evaluation of Regional Banks in Equity Markets', mimeo (in Japanese). Shimizu, K. and A. Horiuchi (1997) `Safety Net and Stability of Financial Systems in Japan', in K. Asako, S. Fukuda and N. Yoshino (eds.), Modern Macroeconomic Analysis, University of Tokyo Press, pp. 85±122 (in Japanese). Shleifer, A. and W. Vishny (1997) `A Survey of Corporate Governance', The Journal of Finance, vol L I I , no. 2, pp. 737±83. Tachibanaki, T. (1996) Public Policies and the Japanese Economy, London: Macmillan. Tachibanaki, T. (1998a) `Capital and Labour in Corporate Governance', in FAIR (eds.), Policy Reforms in Japan, pp. 211±17 (in Japanese). Tachibanaki, T. (ed.) (1998b) Who Runs Japanese Business?, London: Edward Elgar. Tachibanaki, T. and R. Nagakubo (1997) `Intercorporate Shareholding and Firm's Behaviour', Financial Review, Ministry of Finance, October (in Japanese).

132 Governance Structure of Banks and Their Business Performance Takano, M. (1993) `Intercorporate Shareholding in Listed Firms and Determination of Stock Prices', Daiwa Toshi Shiryo, November (in Japanese). Yonezawa, Y. (1995) Economics of Stock Market, Nihonkeizai-shinbunsha (in Japanese). Yonezawa, Y. and M. Miyazaki (1996) `Japan's Corporate Governance and Productivity', in T. Tachibanaki and Y. Tsutsui (eds.), Japan's Capital Markets, chapter 11, Nihon-hyoronsha (in Japanese).

6

A Vacuum of Governance in Japanese Bank Management Masaharu Hanazaki and Akiyoshi Horiuchi

This paper overviews the bank crisis Japan has faced since the early 1990s. Undeniably, the current bank crisis is the aftermath of the ®nancial overexpansion that occurred in the late 1980s. However, this paper focuses on the issues related to managerial governance in the banking sector. We propose a hypothesis that there has existed a vacuum of governance in Japanese bank management in the sense that bank managers have enjoyed wide latitude. This hypothesis of the governance vacuum will explain why the bank crisis has been so serious and so protracted.

1

Introduction

The period of the 1980s and the early 1990s was characterised by the global bank crisis. Not only many industrialised countries such as the United States and Japan, but also most developing countries and the economies transiting from the central planning to the market-oriented system experienced more or less bank crisis. Lindgren et al. (1996) describe `[a] review of the experience since 1980 of the 181 current Fund member countries reveals that 133 have experienced signi®cant banking sector problems at some stage during the past ®fteen years' (p. 20). After aggressively expanding their credit to risky projects like real estate developments, many banks were found trapped in the dif®culty of a large amount of non-performing loans in those countries. The government had to step in to bail out heavily damaged banks by pouring in public money in some cases. It may be a comfort for Japanese people to hear that the bank crisis is not peculiar to Japan.1 However, the Japanese bank crisis seems to be unique to its long duration and seriousness of its bad in¯uence on the 133

134 A Vacuum of Governance in Japanese Bank Management

macroeconomy. Japan has taken half a decade to deal with the bad loan problem in the banking sector without notable success. The bad loan problem has grown so serious as to endanger the viability of the current ®nancial system in the late 1997. We think there are common factors which can explain the banking sector problem in South East Asian countries including Japan. This paper tries to give an answer to the question of why Japan has suffered from so serious a bank crisis from the perspective of corporate governance. We propose a hypothesis that the current bank crisis in Japan has been caused by de®ciency of effective governance in bank management. Needless to say, the bank is a corporation whose management must be disciplined based on effective monitoring in order to keep its managerial ef®ciency. Generally speaking, bank management could be disciplined by three means: (1) the capital markets where either investors (including depositors) would monitor performance of individual banks, or the threat of hostile takeovers would discipline bank managers for their bad performance, (2) the competition in the banking industry which would weed out inef®cient banks, and (3) supervision of the regulatory authorities which would either prevent banks from taking excessive risk, or force managers of distressed banks to restructure their businesses. We explain how these disciplinary mechanisms have not effectively worked in Japan. The de®ciency of effective governance, i.e., a vacuum of managerial governance, has brought forth inef®cient management, and particularly delayed responses to the management restructuring necessitated by the increasing non-performing loans since the beginning of the 1990s. This paper is organised as follows. Section 2 gives an overview of the current banking crisis in Japan. It is pointed out that there exists the danger of a vicious circle between the decrease in bank capital and business setback. Section 3 examines how the comprehensive safety net implemented by the government undermined the capital market mechanism of disciplining bank management (the disciplinary channel (1)). In Section 4, we argue that the Japanese government has rigidly controlled the deregulation process in the ®nancial markets so that the market competition (the disciplinary channel (2)) was unable to in¯uence bank management. Section 5 investigates the disciplinary

Masaharu Hanazaki and Akiyoshi Horiuchi 135

channel (3) for bank management. Speci®cally, we discuss how the Japanese government disciplined bank management from the prudence perspective. We argue that the pervasive relationship between the regulatory authority and private banks (so-called amakudari) increased fragility of the banking industry. Our argument from Sections 3 to 5 in this paper suggests there has existed `a vacuum' in the governance structure in Japanese bank management in the sense that the managers are immune from external discipline. Section 6 provides some tests to show that the `vacuum' of management governance in the banking industry delayed necessary structural readjustments responding to serious crisis. The vacuum of management governance in the banking industry was viable only if investors in the capital market trusted the comprehensive safety net implemented by the government. However, the prolonged bank crisis has undermined investors' trust in the government's capability to manage the traditional safety net. Thus, the capital market started to ®ll the vacuum of managerial governance and bank managers were threatened by the harsh evaluation of the capital market. We discuss this market process of ®lling the vacuum in bank management governance in Section 7. Finally, Section 8 summarises the discussions in this paper, and draws policy implications.

2

The current crisis in the Japanese banking sector

The current bank crisis in Japan was caused by the shortage of bank capital brought forth by a huge amount of non-performing loans. However, the information on non-performing loans is very imperfect and fragmented. Moreover, the government and banks have repeated mistakes of not disclosing full-scale information about non-performing loans. This negative attitude of the government toward disclosure has made the non-performing loan problem messy in Japan and increased the public's disbelief in any information about non-performing loans disseminated by the government. This section ®rst explains the present situation of nonperforming loans in the Japanese banking sector, then explains how the delayed settlement of the non-performing loan problem is likely to produce a vicious circle between the weakened banking sector and economic setback in Japan. 2.1

The disclosure of non-performing loans

Since the de®nition of non-performing loans is by nature elusive, it is dif®cult to grasp the exact situation of the bad loan problem. However,

136 A Vacuum of Governance in Japanese Bank Management

the dif®culty comes mainly from the fact that the information about nonperforming loans remained less comprehensive at the earlier stage of the bad loan problem and is only partially disclosed even now. For example, the Major Banks2 and the regional banks started to disclose only the amount of narrowly de®ned non-performing loans in March 1996. The non-performing loans contained (a) loans to bankrupted borrowers, (b) loans with interest payment overdue for longer than 180 days, and (c) loans with interest reduction to borrowers in trouble.3 Miscellaneous cooperative banks such as shinkin banks, credit cooperatives and agricultural cooperatives, have not yet disclosed any ®gures for nonperforming loans. We can obtain only the aggregated ®gures of nonperforming loans for those cooperative banks. It is noteworthy that the amount of their deposits accounts for nearly 30 per cent of total bank deposits in Japan. Table 6.1 summarises the of®cial annual ®gures of non-performing loans from March 1996 to March 1998. According to this table, the average of non-performing loan ratio (the ratio of non-performing loans over the total loans) in the banking sector was around 3.6 per cent as of March 1998. This is 1.2 per cent point lower compared with the ®gure of March 1996. More than four ®fths of the non-performing loan was covered by the provision for loan losses (i.e., the provision ratio was 3.63 per cent at March 1998). Thus, Table 6.1 appears to show that, except for the cooperative credit banks, the uncovered non-performing loan ratio of which still stayed at high level, the problem of non-performing loans has already been reduced to a minor policy problem in Japan. However, the Japanese banking sector came to a serious crisis immediately after September 1997. We cannot perceive the threatening bank crisis from Table 6.1.4 2.2

Extended de®nition of non-performing loans

The of®cial de®nition of non-performing loans adopted in Table 6.1 is not suf®ciently comprehensive compared with the US standard prescribed by the SEC. Thus, the Major Banks and the regional banks extended the de®nition following the US criteria at the end of March 1998. Due to this extension, the amount of non-performing loans reportedly jumped up by around 40 per cent from ¥25 trillion to ¥35.2 trillion for those banks. Nevertheless, many people do not believe that the disclosed ®gures of non-performing loans show the real dif®culty the Japanese banking sector has been facing. Actually, following guidance by the Ministry of Finance (MOF), individual banks assess the amount of the `problematic loans'

Masaharu Hanazaki and Akiyoshi Horiuchi 137 Table 6.1 Non-performing loans (NPL) in the banking sector (¥ 100 billion)

Major banks (a) Total loans (b) NPLs (b/a: %) (c) Provision (c/a: %) Regional banks (a) Total loans (b) NPLs (b/a: %) (c) Provision (c/a: %) Total cooperatives (a) Total loans (b) NPLs (b/a: %) (c) Provision (c/a: %) Shinkin banks (a) Total loans (b) NPLs (b/a: %) (c) Provision (c/a: %) Credit cooperatives (a) Total loans (b) NPLs (b/a: %) (c) Provision (c/a: %) Total (a) Total loans (b) NPLs (b/a: %) (c) Provision (c/a: %)

March 1996

March 1997

March 1998

3,918.5 218.7 (5.58) 103.5 (2.64)

3,953.1 164.4 (4.16) 93.9 (2.38)

3,658.7 145.2 (3.97) 136.0 (3.72)

1,896.8 66.4 (3.50) 29.5 (1.56)

1,902.9 53.5 (2.81) 29.5 (1.55)

1,872.6 50.1 (2.68) 42.1 (2.25)

1,312.1 63.0 (4.80) 17.6 (1.34)

1,285.4 61.1 (4.75) 26.6 (2.07)

1,353.4 54.4 (4.02) 40.9 (3.02)

696.0 32.0 (4.60) 10.3 (1.48)

702.0 32.4 (4.62) 16.2 (2.31)

704.1 32.4 (4.60) 26.8 (3.81)

173.7 20.5 (11.80) 1.8 (1.04)

172.1 21.2 (12.32) 3.0 (1.74)

150.9 12.0 (7.95) 4.1 (2.72)

7,127.4 348.0 (4.88) 150.5 (2.11)

7,141.4 279.0 (3.91) 149.9 (2.10)

6,884.7 249.8 (3.63) 219.0 (3.18)

Note: In this table, the non-performing loans are de®ned by the old standards. Those banks

which went down during the sample period are excluded from the table.

Source: Federation of Bankers Associations of Japan, Analysis of Financial Statements of All Banks.

which would be more or less dif®cult for the banks to collect. The ®gures of the `problematic loans' are not disclosed because bankers think they are too comprehensive and too ambiguous. They include a large amount of loans bankers believe to be collectable without great dif®culties. However,

138 A Vacuum of Governance in Japanese Bank Management Figure 6.1 Several de®nitions of non-performing loans

¥100 billion 16 14 12

Covered Uncovered

10 8 6 4 2 0 March 96

March 97 Official figures

March 98

March 98 US Standard

March 96 Problematic loans self-assessed by banks

outsiders think the amount of `problematic loans' indicates the actual situation of bad loans for each bank. The MOF reported a survey of the amount of `problematic loans' assessed by banks themselves in January 1998. According to this report, the total of loans which is either impossible or very dif®cult to collect was ¥11.4 trillion for the Major and the regional banks (i.e., just 1.8 per cent of the total loan of these banks). In addition to this, however, those banks held the problematic loans that must be treated carefully in order to collect amounting to ¥65.3 trillion. In total, the problematic loan was higher than 12 per cent of the total loans for these banks. This is substantially higher than the non-performing loans ratios of®cially disclosed. Figure 6.1 compares the ®gure of narrowly de®ned non-performing loans with the ®gures of both the non-performing loans de®ned by the US standard and the problematic loans self-assessed by banks as of March 1998, and gives the impression that the Japanese non-performing loan

Masaharu Hanazaki and Akiyoshi Horiuchi 139

problem is far from being settled. Some of the problematic loans are likely to change into non-performing ones in the near future. In particular, the Japanese economy has been suffering from sluggishness since early 1997 partly due to the turmoil in the ®nancial system. The business setback would further increase the amount of non-performing loans. Thus, we should not be optimistic about the capability of the banking sector to recover from its deteriorated balance sheet.5 Moreover, both the banks and regulatory authorities showed negative attitudes toward the disclosure of bad loan information. At ®rst, they disclosed only the narrowly de®ned non-performing loans of the restricted section of the banking sector. Then, they gradually extended the de®nition of bad loans. The disclosed amount of non-performing loans has increased accordingly. This disclosure policy produced an unfortunate consequence of depriving the of®cial ®gures of credibility. 2.3

The danger of a vicious circle

In October 1995, Japanese banks faced `the Japanese premium' in the international money market for the ®rst time since the ®rst oil crisis in 1974. (The Japanese premium is de®ned by the difference between LIBOR and TIBOR for US dollar.) In the summer of 1995, some Japanese ®nancial institutions went down due to the huge amounts of non-performing loans. After abruptly jumping up to higher than 30 basis point (in terms of 3 month US dollar), the Japan premium remained at around 10 basis point until the beginning of November 1997. In late November, the Japan premium went up to 100 basis point re¯ecting the turmoil in the domestic money market.6 The development of the Japan premium suggests that the international money market had already started to give an alarming signal to the Japanese banking system in the autumn of 1995. However, the Japanese government belatedly started to force the banks to recapitalise by announcing that the `prompt corrective action rules' would be introduced in April 1998. Both this announcement and the prolonged sluggishness of stock prices compelled banks to reduce credit supply in 1997. This is a `credit crunch phenomenon'. The requirement of more comprehensive disclosure of non-performing loans seems to have made the credit crunch more serious. Combined with the impact of the tax increases in April 1997, the credit crunch promoted the business setback that increased the amount of non-performing loans in the banking sector. Obviously, this is a vicious circle between de®ciency of bank capital and the macroeconomic slowdown.7 The fall in stock prices caused by the economic setback has decreased the accumulated amount of unrealised capital gain on share-holdings in

140 A Vacuum of Governance in Japanese Bank Management

the banking sector. Since the unrealised capital gain has been an important ingredient of bank capital (tier II), the fall in stock prices makes the de®ciency of bank capital worse leading to another round of credit crunch. This is also a sort of vicious circle. Thus, in order to revitalise the Japanese economy, we urgently need to reconstruct the banking sector by speci®c policy measures to suppress the vicious circles which have their origin in the non-performing loans in the banking sector.8

3

The safety net in Japan

Why have we suffered so serious a bank crisis? This paper proposes the hypothesis that the bank crisis in Japan is an issue of managerial governance. In this section, we investigate whether or not the capital market disciplined bank management effectively in Japan. 3.1

Lack of capital market discipline

From the viewpoint of the standard theory of corporate ®nance, the degree of concentrated ownership of ®rms is important as an effective device of capital market control of corporate management (Prowse, 1992). In reality, Japanese banks are more diffusely held than non-®nancial companies are. According to Kim and Rhee (1997), the top six shareholders of banks hold on the average 18.4 per cent of the total shares outstanding. In contrast, Prowse (1992) ®nds that the top ®ve shareholders for the Japanese mining and manufacturing companies hold 33.1 per cent of the total shares outstanding. In this sense, the Japanese capital market is not so powerful in monitoring bank management. We should also note that insurance companies have often been the largest shareholders of banks. The insurance companies were quite helpful to incumbent bank managers when they were required to strengthen their capital since the end of the 1980s. Speci®cally, Japanese banks issued a large amount of subordinate debt (or subordinate loans) to increase their equity capital following the BIS capital adequacy requirement. Insurance companies actively bought most of the debt to help bank management. The main object of the insurance companies in buying subordinated debt issued by banks was not to monitor bank management more strongly, but to keep business relationships with the banks. The insurance industry is most heavily protected in the Japanese ®nancial industries. The government often tried to make use of the rent accumulated in the insurance industry for the `public purpose', and the insurance companies tended faithfully to obey the government policy. For example, many market observers in the Tokyo Stock Exchange said that the Japanese government

Masaharu Hanazaki and Akiyoshi Horiuchi 141

implicitly guided insurance companies and other institutional investors to keep their positions in the stock market in order to counterbalance the downward pressures on stock price levels. Thus, it was a plausible story that the government permitted banks to issue subordinate debts to increase their capital at the end of the 1980s, immediately after the BIS capital adequacy regulation became effective, then implicitly order (or recommend?) insurance companies to support banks by buying most of the debts. If so, the insurance companies have been far from a reliable monitor of bank management. 3.2

The mechanisms of the safety net in Japan

We de®ne the ®nancial safety net as a social system of dealing with distressed banks and of distributing social costs associated with bank failures among related parties. The government provides a ®nancial safety net in order to minimise the spillover effects of failures of banks and other ®nancial institutions on the ®nancial system as a whole. The safety net also has important implications for risk sharing in the ®nancial system. Speci®cally, the operation of the safety net changes the ex post distribution of social costs associated with bank failures. The safety net decreases monitoring incentives of depositors and other investors either explicitly or implicitly protected from bank failure losses. Thus, appropriate incentive mechanisms are required to reinforce monitoring of bank management to keep the safety net system viable. The wider the scope of the ®nancial safety net, the stronger are the moral hazard incentives given to bank management, and thus, the more energetically the regulatory authorities must monitor banks to prevent excessive risk-taking in place of depositors and investors.9 The Japanese ®nancial system operates under an extensive safety net implemented by the regulatory authorities. The MOF has executed programmes to rescue distressed ®nancial institutions in tight collaboration with the Bank of Japan (BOJ) and private ®nancial institutions, particularly major banks. Before 1990 there occurred some bank failures though the number was quite small.10 The MOF guided (more precisely ordered) private banks to rescue their distressed peers. Probably, the most important rescue programme implemented by the MOF before 1990 was the case of merger of Heiwa-Sogo Bank by Sumitomo Bank in October 1986. Heiwa-Sogo got into dif®culty during the ®rst half of the 1980s. In 1985, the MOF made a bailout plan for this bank to prevent the crisis of Heiwa-Sogo from destabilising the Japanese banking industry as a whole. Finally, in 1986, the MOF succeeded in persuading Sumitomo to absorb Heiwa-Sogo. Despite de facto bankruptcy, the closure of Heiwa-Sogo did

142 A Vacuum of Governance in Japanese Bank Management

not cause damage to depositors and holders of other debt issued by this bank. Sumitomo bore the cost of dealing with the distressed bank. On the other hand, Sumitomo was able to expand its branch network at once by absorbing Heiwa-Sogo's branches. In other cases, the MOF often placed its of®cers on the board of the distressed bank with a view to reorganising its management. Dispatching of®cials to a distressed bank may be an effective signal to inform the public that the government has made a commitment to rescue the bank at any cost. This signal might have helped the MOF to persuade other banks to collaborate with the bailing out programme.11 In reality, however, this signaling does not seem to be always successful. One of the most recent cases was Hyogo Bank, to which the late chief of the Banking Bureau of the MOF was sent to reorganise its management. Despite this intervention, Hyogo ®nally went bankrupt in October 1995. This paper will examine how the human relationship between regulatory authorities and private banks, which is called amakudari in Japanese, in¯uences the stability of the banking sector. 3.3

Comprehensive safety net

Since the actions taken by the authorities to rescue troubled banks have been covert, it is dif®cult to estimate the social costs of the safety net and the exact distribution of the burden among the various agents. However, the safety net was comprehensive in the sense that not only depositors but also almost all other debt-holders (except for some ®nancial institutions) were exempted from the burdens associated with bank failures. In most cases, even shareholders of failed banks seemed to be rescued from bank failures. For example, in the case of failures of credit cooperative banks, their equity holders were not required to share the costs of failures.12 The costs of preserving ®nancial stability have fallen disproportionately on sound private banks, particularly major banks. Until the early 1990s, the ®nancial authorities rarely paid the costs of the bailout procedure, con®ning their role to coordinating the rescue programme endured by private banks and other ®nancial institutions. In some cases, the BOJ may have extended loans to distressed banks at the of®cial discount rate, which was substantially lower than money market interest rates, but it is impossible to obtain any information about these unof®cial rescue programmes. After the rescue of Yamaichi in 1965, the BOJ utilised emergency loans (authorised by Article 25 of the BOJ Act) for the ®rst time to support the Tokyo Kyodo Bank, newly established in 1995 to take over two failed credit cooperatives in the Tokyo metropolitan area. The amount of the BOJ's emergency loans increased abruptly during 1995 due to

Masaharu Hanazaki and Akiyoshi Horiuchi 143

managerial crises in several small and medium scale banks (including Hyogo Bank) and reportedly reached a little more than ¥1.0 trillion. 3.4

Danger of forbearance policy

The MOF's implementation of the safety net was essentially covert. There were no explicit rules that the authorities should obey in implementing the safety net. Therefore, it was almost impossible for outsiders to evaluate the MOF's performance in operating the safety net. Herein lies a danger of the forbearance policy in the sense that the authorities postpone taking determined actions to liquidate de facto insolvent banks. The bureaucrats in charge of monitoring the management of individual banks have signi®cant incentives to postpone any de®nite policy decision which would reveal their incompetence or failures to the public. It is well known that the forbearance policy is likely to incur large social losses when the troubled banks ®nally fail after remaining in business for a long time due to this policy (Kane, 1985, 1993). It is easy to understand why the forbearance policy tends to increase the social cost of dealing with bank failures. A bank at the brink of bankruptcy has a particularly strong incentive to take extreme risk because it stands to lose almost nothing when it fails. On the other hand, depositors and most other investors are not cautious about the soundness of individual banks' management under the comprehensive safety net. The insuf®cient disclosure is likely to worsen the situation. Therefore, unless the authority stops its operation, the distressed bank continues to increase liability, most of which will ®nally be transferred to the safety net.13 3.5

Deposit insurance in Japan

The experience of the US ®nancial system suggests that deposit insurance should be an important element of the safety net. However, this was not the case in postwar Japan. At the end of the 1960s, some of the MOF of®cials were seriously concerned that the coming deregulation in the ®nancial service industry would increase the number of banks and other ®nancial institutions suffering from serious distress. The system of deposit insurance was introduced in 1971 in order to keep ®nancial stability in the face of ®nancial deregulation. However, the facility of the deposit insurance system was not actually utilised until 1992. The MOF continued to implement the traditional safety net to avoid the straightforward bankruptcy of depository ®nancial institutions. The MOF gave priority to the protection of weak (and therefore inef®cient) banks over the promotion of competition in the Japanese ®nancial industry, even after the introduction of deposit insurance. The Deposit Insurance Corporation

144 A Vacuum of Governance in Japanese Bank Management

(DIC) remained nominal for a long time. Its functions were limited compared with those of its US counterpart (the FDIC), being con®ned to paying off insured deposits in cases of bank failure, although the DIC has never resorted to paying off these deposits. In 1986, the Law of Deposit Insurance was amended to strengthen the DIC's competence. For example, the amended law allows the DIC to support schemes of rescuing or disposing of distressed banks by giving the necessary funds to private agents involved in the schemes. The DIC functioned for the ®rst time only in April 1992, when it supplied ¥8.0 billion to help Iyo Bank, a regional bank, absorb Toho Sogo Bank. The DIC has been equipped with a means of paying off insured deposits of failed banks from the time of its establishment. However, the government announced in December 1995 that they were not prepared to exercise it, although a quarter century had passed since the start of deposit insurance. In December 1997, the government declared that all investments into deposits and other bank debts such as bank debentures would be protected from bank failures. The purpose of this policy is to calm down people's concern with the danger of bank failures caused by the ®nancial crisis following the bankruptcy of Hokkaido-Takushoku Bank and failures of a few major securities companies including Yamaichi in the end of 1997. Of course, this commitment by the government is likely to produce further moral hazard on the side of bank management by weakening the incentive of depositors and investors to monitor bank management. However, the long-standing implementation of the comprehensive safety net has produced among depositors and other investors a perception that they will never be required to share the burden should their banks go bankrupt. Because of this widespread perception, the government adoption of paying off insured deposits without rescuing investors of bank debts other than insured deposits would result in an unexpected shock to the ®nancial system, thereby making the dif®culty more serious. Thus, at the end of 1997, the Japanese government could not but make a commitment to ensure that the widespread perception about the safety net was valid. As Ueda (1996) describes, `the most important safety net in this country has not been the deposit insurance system, but the public's con®dence in the MOF and the BOJ's ability to avoid a major instability in the ®nancial system.' This safety net may have had the merit of freeing people from the need to bother with the soundness of individual banks' management. However, it has also deprived investors of incentives to monitor the performance of individual banks and hindered the development of market mechanisms to discipline bank management. The lack of market

Masaharu Hanazaki and Akiyoshi Horiuchi 145

mechanisms, in turn, has made it quite dif®cult for the government to abandon the traditional safety net. We learn from this experience how dangerous it is for the authorities to have people believe in effectiveness of too comprehensive safety net.14 3.6

Limitation of the traditional rescue method

Since the beginning of the 1990s, when the so-called `bubble economy' burst, it has become increasingly dif®cult for the MOF to maintain the traditional procedure of bailing out bank failures. This is re¯ected in the utilisation since 1992 of the deposit insurance system to cope with the ®nancial distress of individual banks, although, as we have explained, the paying off of insured deposits has never been exercised. The scale of the DIC is as yet limited, but its increasing use marks a signi®cant change in the operation of the Japanese safety net. One of the reasons for this shift is that, with structural changes in ®nancial markets, there are fewer rents in banking for the MOF to use in in¯uencing banks. With ®nancial deregulation, it has become dif®cult for the authorities to manipulate regulatory means to favour some ®nancial institutions over others. For example, interest rate deregulation has reduced the meaning of branch of®ces for individual banks, making the MOF's administration with respect to the branch network less important.15 Since the early 1990s, it has become more and more dif®cult for the government to obtain cooperation from private banks in implementing the traditional safety net.16 Thus, it is a natural that the DIC has started to play a signi®cant role in dealing with troubled banks since April 1992, when the DIC supported Iyo Bank to absorb of the failed Toho Sogo Bank after more than twenty years inactivity. From April 1992 to May 1998, the DIC intervened in 25 cases of bailing out troubled banks and provided the banks cooperating in the bailout schemes with subsidies of more than ¥2.4 trillion (see Figure 6.2). In addition, the DIC is to be mobilised in 35 cases of bank failures including Hokkaido-Takushoku in the near future. Since the early 1990s, the traditional methods of dealing with bank failures have not yet disappeared, and many private banks are still playing an important role through collaboration with the regulators. However, the role of the DIC in the process appears to have become increasingly important, and it is likely that the deposit insurance system will be utilised substantially in the future. Use of the deposit insurance system to facilitate reorganisation did not, however, imply that banks would undergo formal bankruptcy procedures. Even after 1990, the government continued to avoid explicit bank failures. The government has provided sound banks with incentives either to merge

146 A Vacuum of Governance in Japanese Bank Management Figure 6.2 Financial support by the DIC ¥ billion

16,000

14,000

12,000 Debt undertaking Purchase of assets

10,000

Injection of money 8,000

6,000

4,000

2,000

0 1992

1993

1994

1995

1996

1997

1998

Notes: The ®gure in 1998 is the sum until the end of May. Source: Federation of Bankers Associations of Japan.

with insolvent ones or to collaborate with the government's policy of restructuring troubled banks by using the deposit insurance system intensively, rather than by preferential regulatory treatment. This implies a slow reorganisation of the ®nancial system and a marked increase in the burden borne by the DIC. The regulator needs to have its monitoring power strengthened, and thereby to keep the cost of bailing-out policy within manageable bounds. In particular, the regulator should be able to order banks in distress to cease operations before the negative value of their net wealth becomes too great. In fact, the New Law for Strengthening the DIC, instituted in 1996, authorises the regulator to take a policy of `prompt corrective action (PCA)'. Such a policy was started in April 1998. The Financial Supervision Agency, newly established in June 1998, will require individual banks to strengthen their capital bases following an explicit rule

Masaharu Hanazaki and Akiyoshi Horiuchi 147

so that the decreases in the equity capital of a bank will promptly induce the authority to strengthen its monitoring of the bank management. The purpose of the PCA is to prevent excessive risk-taking by banks and to encourage the regulator to intervene in bank management following explicit rules prescribed by the capital adequacy standard. The latter is expected to reduce the danger of `forbearance policy' on the side of the regulator. Unfortunately, the PCA was started belatedly after almost every Japanese bank had been deeply involved in the serious problem of nonperforming loans. The explicit involvement of the DIC in the operation of the safety net should be bene®cial to the Japanese economy. The social costs of bailing out distressed banks will become more transparent, and this will facilitate assessment of the ef®ciency of the current ®nancial safety net. It will also make the regulatory authorities' administration more accountable, as suggested by the recent experience of dealing with the two cooperative credit banks in Tokyo. The improvement of accountability is desirable because it will discipline the regulatory authorities, thereby preventing the forbearance policy, and will rationalise the safety net mechanisms.

4

Disciplinary in¯uence of market competition

The comprehensive safety net deprived the capital market of incentives to monitor and discipline bank management in Japan. Then, what about the disciplinary in¯uence of market competition on bank management? As Nickell, Nicolitsas and Dryden (1997) show with regard to manufacturing industries, we may expect full-scale market competition to exert a strong disciplinary in¯uence on corporate management by weeding out the inef®ciently managed ®rms. Regardless of its speci®c ownership structure or any other ®nancial governance structure, the corporate management would be disciplined by ®erce market competition. Many Japanese believe that the Japanese manufacturing ®rms have achieved excellent performance because they have long faced ®erce competition in the global market. At present, this belief remains a conventional view that must be empirically tested. However, it seems fairly well-grounded. In contrast, the Japanese ®nancial services industries including the banking have been protected from full-scale competition by the competition restricting regulation. Thus, the market competition has not worked to discipline management in the banking and other ®nancial services industries in Japan. 4.1

Role of competition-restricting regulations

The competition-restricting regulations, such as interest rate controls and restriction on new entry into banking and other ®nancial business

148 A Vacuum of Governance in Japanese Bank Management

through the system of compartmentalisation, conferred a handsome amount of rents on existing banks and other ®nancial institutions. The primary purpose of the MOF's administrative guidance was to suppress full-scale competition in each of the compartmentalised ®nancial business, thereby protecting the less competitive small-scale banks such as sogo banks, shinkin banks and credit cooperatives. The MOF's policy stance was often called the `convoy administration'.17 The rents created by the competition-restricting regulations might have contributed to stabilising the banking system under the Japanese comprehensive safety net in two ways. First, as economic theory shows, the existence of rents could provide private banks with incentives to refrain from excessive risk-taking in order to continue enjoying handsome rents, even without effective prudential regulations (Hellman, Murdock and Stiglitz, 1997). Furthermore, thanks to protection offered by the competition-restricting regulations, even inef®cient banks rarely went to the brink of managerial dif®culty that is particularly likely to induce moral hazard behaviour.18 Second, the monetary authorities were able to utilise the rents accumulated in the banking sector as a means of dealing with banks in ®nancial distress. Speci®cally, the regulators relied on the collaboration of private banks in implementing the safety net, and major banks faithfully bore a disproportionate share of the costs involved. This mechanism would not have worked had the major banks not enjoyed the rents stemming from the competition-restricting regulations. The MOF also utilised the competition-restricting regulations to give private banks an incentive to accept its initiatives in the process of dealing with bank failures. The MOF manipulated the regulatory means to do favours for those banks who toed the line and to penalise those who failed to heed their guidance. In other words, speci®c administrative guidance based on the competition-restricting regulations was an instrument for the MOF to determine the distribution of rents among banks. Thus, the competitionrestricting regulation was strategically important for the MOF in order to maintain the viability of the comprehensive safety net.19 4.2

Delayed deregulation in the ®nancial markets

However, we should note the competition-restricting regulation has gradually weakened the capability of the Japanese banks and other ®nancial institutions to adapt themselves to environmental changes since the mid-1970s. We may say that in practice ®nancial deregulation has been tightly controlled by the government (more speci®cally by the MOF). The Japanese government took the policy of gradualism for the purpose of

Masaharu Hanazaki and Akiyoshi Horiuchi 149

preventing `unduly destabilising' impacts of ®nancial deregulation. In reality, this gradualism was synonymous with the policy of protecting vested interests existing in the ®nancial industries, thereby suppressing the disciplinary effects the ®nancial deregulation was expected to exert on management in the ®nancial industries including the banking. Rather, the ®nancial deregulation was promoted by pressures from abroad, particularly from the US, than on the government initiative. For example, the ad hoc Yen/Dollar agreement between US and Japan realised through the strong requirement by the Reagan administration in 1984 compelled the Japanese government to provide an explicit timetable of liberalising ®nancial markets.20 Compared to liberalisation in international capital market, the Japanese ®nancial markets have been belatedly deregulated. The so-called `big bang' proposed by former Prime Minister Ryutaro Hashimoto in November 1996 was a sign of the government commitment of abandoning the policy of gradualism. Of course, we should not totally deny the impact of ®nancial deregulation on domestic ®nancial markets during the 1980s. In particular, major companies reduced their dependence on bank borrowing by issuing a large amount of corporate bonds in international markets. This `internationalisation' of corporate ®nance induced deregulation of domestic corporate bond markets since the mid-1980s (Horiuchi, 1996). However, generally speaking, the Japanese banks and other ®nancial institutions were able to base their business on the huge amount of wealth accumulated by households. The gross amount of ®nancial assets held by households reportedly amounted to ¥1,200 trillion as of the mid-1990s. Thus, it would be an exaggeration to say that the internationalisation of corporate ®nance exerted substantial in¯uence on their way of business.

5

The role of government in bank managerial governance

The previous sections stressed that neither the capital market nor market competition was effective in disciplining bank management in Japan, mainly because the intervention of government (the MOF) into the ®nancial markets through the comprehensive safety net and control of the deregulation process suppressed those disciplinary in¯uences. This is to some extent a natural outcome of the current legal framework which assigns great responsibility of monitoring bank management to the Ministry of Finance and the Bank of Japan. The Banking Law authorises the MOF to intervene into management of banks for purpose of prudential regulation. The BOJ is also in charge of monitoring bank management particularly from the viewpoint of money market adjust-

150 A Vacuum of Governance in Japanese Bank Management

ment. Thus, the current prolonged turmoil of the banking industries is mainly responsible for the ®nancial authorities. In the following section, we will ®rst examine how the Japanese government has implemented the prudential regulations, and then provide evidence to show the weakness of monitoring by the regulatory authority which led to fragility of the banking industry. 5.1

Capital adequacy regulations

Capital adequacy requirements, accompanied with rigorous monitoring by regulators, are a typical means of prudential regulation. During the period of economic reconstruction immediately after World War Two, the MOF was seriously concerned about the prudence of bank management, because banks' equity capital per deposit had fallen sharply from 29.9 per cent in 1930 to only 5.6 per cent by 1953. With a view to strengthening banks' capital bases, the MOF started in 1953 instructing banks to reduce current expenses to 78 per cent or less of current revenues. This administrative guidance continued until 1973. In 1954, the MOF introduced the capital adequacy regulation, which required banks to increase broadly de®ned capital to more than 10 per cent of total deposits.21 This could be regarded as a forerunner of the capital adequacy regulation introduced by the Bank for International Settlements (BIS) in 1987. However, some depository ®nancial institutions were not covered by this capital adequacy regulation. For example, the sogo banks were required only to maintain more than the prescribed minimum amount of equity capital (book value). Thus, they could have increased their leverage ratio without limit had they wished to do so. When the sogo banks converted to regional banks in February 1989, the MOF started to impose the same minimum capital adequacy ratio on the sogo banks (now called the regional banks of tier two) as for the city banks and the other regional banks. Shinkin banks, which are cooperative ®nancial institutions, had been free from the capital adequacy regulation until May 1986, when the MOF introduced administrative guidance in the form of a minimum capital adequacy ratio. Thus, until the late 1980s, the capital adequacy regulation did not cover the whole range of depository ®nancial institutions. Moreover, the regulation seemed to be ineffective. Figure 6.3 shows that, from 1960 to the mid-1970s, the average of the (broadly de®ned) capital/deposits ratio for the banking sector, which is comprised of city banks and regional banks, remained almost constant at 6 per cent, far below the MOF's requirement of 10 per cent. Furthermore, the average capital/deposit ratio dropped abruptly to below 4 per cent during the 1980.22

Masaharu Hanazaki and Akiyoshi Horiuchi 151 Figure 6.3 Capital/deposit ratio of Japanese commercial banks 1960±85

% 8

6

4

2

0 1960

1965

1970

1975

1980

1985

Source: Federation of Bankers Associations of Japan, Analysis of Financial Statements of All Banks.

5.2

Bank capital and amakudari

We have stressed that bank management has been disciplined neither by the capital market nor by market competition. In the previous subsection, we explained that the MOF was not serious in implementing prudential regulation. If our explanation is true, there existed a vacuum in the governance of bank management. This vacuum should have made the Japanese banking sector potentially fragile. However, Aoki, Patrick and Sheard (1994) argue that the ®nancial authority has been disciplined to monitor bank management through the so-called amakudari system; i.e., the system prevailing among private banks (and other ®rms) of accepting post-retirement of®cials to their managerial boards.23 According to their argument, the amakudari system has given regulatory of®cers incentives to monitor bank management rigorously and faithfully. If they fail to achieve a good performance as a

152 A Vacuum of Governance in Japanese Bank Management

monitor, they will lose the chance of obtaining a good job in a private bank after retirement. Thus, following Aoki, Patrick and Sheard (1994), the bank performance in terms of soundness will be positively in¯uenced by amakudari. However, this amakudari system is accompanied by the danger of an agency problem, because the bureaucrats are to supervise the management of the banks that are likely to employ them after their retirement.24 If the ®nancial authority and private banks bargain with each other through manipulating monitoring effectiveness and accepting amakudari of®cials, the amakudari system would undermine effectiveness of monitoring by the ®nancial authority and allow banks to engage in unsound management at the expense of depositors and/or taxpayers (Horiuchi and Shimizu, 2001). This agency problem hypothesis predicts that the banks accepting amakudari of®cials from the ®nancial authority will show poor performance in terms of soundness. This is in sharp contrast with the hypothesis advocated by Aoki, Patrick and Sheard (1994). We test the hypothesis of whether or not the amakudari system undermines the prudence of the Japanese banking sector by a simple statistical method. Here, we take a sample of 125 regional banks existing as of March 1996. We classify those sampled banks into four categories according to whether or not they accept amakudari of®cers from the regulatory authorities. The ®rst group (MOF&BOJ) contains the banks which accept amakudari of®cers from both the MOF and the BOJ. The second (MOF) consists of the banks accepting of®cers only from the MOF. The third (BOJ) is the group of banks accepting amakudari only from the BOJ. Finally, the fourth (NON) consists of the banks that do not accept amakudari of®cers at all. Table 6.2 compares performances the respective categories of regional banks achieved from 1985 to 1989. Here, the banks are subdivided according to their amakudari status as of 1985. The bank performance is measured in four terms: capital/asset ratio EQT; the annual growth of total assets GAS; and the current pro®ts per equity capital PRO; and the nonperforming loan/total loan ratio BAD measured at March 1996. Except for BAD, these performance variables are the averages of the sample period. In Table 6.2, the capital/asset ratio (EQT) is signi®cantly lower for both categories MOF&BOJ and MOF than for category NON, while we ®nd no signi®cant difference between the respective categories as for asset growth (GAS) and pro®tability (PRO). For example, the capital/asset ratio (EQT) for category MOF&BOJ banks, which accepted amakudari of®cials from both the MOF and BOJ as of 1985, was on average 0.562 per cent point lower than that of the category NON banks. The differences are

Masaharu Hanazaki and Akiyoshi Horiuchi 153 Table 6.2 Amakudari and performance of regional banks, 1985±89

EQT GAS PRO BAD

MOF&BOJ (41)

MOF (43)

BOJ (21)

NON (20)

2.849*** 10.945 8.913 4.145***

3.008*** 9.927 9.087 4.145***

3.390 10.526 8.641 2.205

3.411 9.815 8.610 2.200

Note: The asterisks ***, ** and * indicate the ®gures are different from the those of `NON' signi®cantly at 1%, 2.5% and 5% respectively. Panel A and B delete Daiko Bank because of its abnormal performances during the 1980s. The ®gures in parentheses are the numbers of banks belonging to respective categories.

statistically signi®cant at the 1 per cent level. As Keeley (1990) argues, the lower level of capital/asset ratio implies the higher level of risk. Thus, Table 6.2 suggests that the banks accepting amakudari of®cials from the MOF tend to take higher level of risk.25 While the equity/asset ratio (EQT) could be regarded as an ex ante measure of the risk-taking by banks, the bad loan ratio (BAD) could be an ex post measure of risk undertaken by banks. Thus, it would make sense to compare the bad loan ratio of the respective bank groups in order to examine how the amakudari relationship in¯uenced the soundness of bank management. The Japanese banks started to disclose the amount of comprehensively de®ned non-performing loans for the ®rst time in March 1996. Figure 6.4 presents the distribution of sampled banks according to their non-performing loan ratios (BAD). We may interpret the ®gures of non-performing loans as of March 1996 as an indication of the degree of risk the banks took during the latter half of the 1980s and the early 1990s. The BAD row in Table 6.2 presents the bad loan ratio for each category of amakudari status. The two groups of banks accepting amakudari of®cials from the MOF (i.e., MOF&BOJ and MOF) as of 1985 had a bad loan ratio (4.145) amount twice higher than the banks totally independent from the amakudari relationship (i.e., NON). These differences are statistically signi®cant at the 1 per cent level. In contrast, the average level of bad loan ratio for the banks accepting amakudari from the BOJ (i.e., banks of BOJ status) is not signi®cantly different from that of the NON. If we measure (ex post) risk by the bad loan ratio, the results are consistent with the hypothesis that the amakudari relationship undermines monitoring by the MOF. Some may doubt the causality between amakudari and bank performance by stressing the fact that the MOF more often than not dispatches of®cials to the banks in ®nancial distress in order to rehabilitate them. Certainly, there

154 A Vacuum of Governance in Japanese Bank Management Figure 6.4 The distribution of regional banks in terms of non-performing loan ratios 50

of banks the number

40

30

20

0

0

1.00

5.00

10.00

15.00

20.00

%

Source: Federation of Bankers Associations of Japan, Analysis of Financial Statements of All Banks.

were some cases of MOF of®cials being dispatched to private banks for the purpose of strengthening their management. However, almost all banks in distress had accepted amakudari of®cials long before they faced managerial dif®culties. It would be fair to say that, contrary to what Aoki et al. (1994) argued, the amakudari was not effective in improving performances of the banks accepting it. The regulator tends to help incumbent managers of distressed banks to continue their operation. We conclude that the lack of effective monitoring by the outsiders is the most conspicuous feature of governance in Japanese bank management. This feature has produced in¯exibility of bank management when confronted by the serious crisis of non-performing loans since the early 1990s.26

Masaharu Hanazaki and Akiyoshi Horiuchi 155

6

The vacuum of governance in bank management

This paper has stressed that there existed a vacuum of governance in bank management. In other words, the hypothesis of managerial entrenchment was applicable to the Japanese banking industry. This section provides some evidence showing what has resulted from the independence. Generally speaking, when they manage their ®rms independently from outsiders' control, corporate managers would (1) engage in expansionism to display their managerial capability (Gorton and Rosen, 1995), and (2) delay structural changes after their policy is found to fail (Boot, 1992). We should take this second point into consideration in order to explain why the Japanese banking crisis since the early 1990s is so serious. In our opinion, the seriousness of the banking crisis has come from the delayed response on the side of bank management to the accumulated nonperforming loans, rather than from the accumulation of non-performing loans in the banking sector. As has been pointed out by Lindgren, Garcia and Saal (1996), the bank crisis is not peculiar to Japan. Many countries have experienced larger or smaller bank crisises since 1980. However, Japan has taken too long to deal with this problem without remarkable success. The main reason for this failure is the delayed response of bank management to the crisis. More speci®cally, the Japanese banks have hesitated to take the drastic restructuring policy necessary to deal with the dif®culty of non-performing loans. The government was unable to take policy measures to force banks quickly to reinforce their capital in the increasing non-performing loans. 6.1

Some international comparisons

Figure 6.5 presents an international comparison of banking restructuring during the ®rst half of the 1990s based on the BIS Annual Report (1996). This ®gure shows that, except for the US, the pro®tability of commercial banks decreased in the ®rst half of the 1990s compared with the later half of the 1980s in all major industrial countries, including Japan. When we look at (1) the growth rate in the number of bank branches, (2) the growth rate in the total number of employees, and (3) the changes in wage index, Japan was unique in the sense that none of these measures decreases during the 1990s compared with the later half of the 1980s. In other words, the commercial banks in the other major industrialised countries downsized or reduced their scale of business after recognising a fall in pro®tability during the 1990s. Thus, Figure 6.5 shows how the Japanese banks were hesitant to restructure their business in spite of decreasing

156 A Vacuum of Governance in Japanese Bank Management Figure 6.5 Restructuring in the banking industry international comparison 10 Profit Branch Employment 5

Wages

Per cent

0

–5

–10

–15 Japan

US

Germany

France

Italy

UK

Notes: Pro®t (total pro®t per total assets), the difference between the average of 1986±88 and 1992±94; number of branches, the growth rate in the total number of branches from 1990 to 1995; number of employees, the growth rate in the total number of employees from 1990 to 1994; wage index (the ratio of wage payment over total revenue), the difference between the average of 1986±88 and 1992±94. Source: The BIS 66th Annual Report.

pro®tability after 1990 in comparison with their rivals in most industrialised countries. The second test examines to what extent salaries and wages (the staff costs) were responsive to ¯uctuations in pro®ts in the banking industry. We suppose that, if managers are entrenched, they can stabilise their salaries regardless of the pro®tability of their ®rms. Thus, the lower responsiveness of staff costs to pro®ts is regarded as the higher degree of managerial entrenchment. Our statistical test is based on the data from

Masaharu Hanazaki and Akiyoshi Horiuchi 157 Table 6.3 The relationship between staff cost and pro®t Dependent variable: staff cost Independent Variable

Equation (1)

Equation (2)

Const. PR PR  JPD R 2

25,778.4 (5.04)** 0.514 (20.83)**

25,963.7 (5.12)** 0.520 (20.95)** �0.328 (�1.78)* 0.9792

0.9791

Notes: PR is the one-year lag of pro®t before tax; JPN is a dummy variable taking 1 for Japan and 0 for other countries. The ®gures in parentheses are t-statistics. Staff cost and pro®t before tax are the real term. The asterisks ** and* indicate the levels of signi®cance at 5% and 10% respectively. Sources: OECD, Bank Pro®tability±Financial Statements of Banks; IMF, International Financial Statistics.

the OECD's Bank Pro®tability. We picked up the time series data of major 23 countries, including Japan, and tried panel analysis (the random effects method). The estimated equations are as follows: SCit ˆ Ci ‡ PRit�1

…1†

SCit ˆ Ci ‡  PRit�1 ‡ PRit�1 JPD;

…2†

where SC, C, PR, and JPD are respectively the staff cost of the banking sector, the constant term, the pro®t before tax of the banking sector, and the dummy variable taking 1 for Japan and 0 for other countries. The subscript i and t represent the cross-country element and the time-series element respectively. Using the GDP de¯ator of each country, the staff cost and the pro®t before tax are changed to the constant price basis. The data period for most countries, including Japan, is from 1979 to 1995, but for countries such as Korea and Mexico, which became OECD member countries quite recently, the data period is much shorter. The results are summarised in Table 6.3. As Equation (1) shows, the staff cost was positively correlated with the one-year lag of pro®t before tax. However, when introducing the cross term of PR and the Japan dummy variable JPD (Equation (2)), we found this cross term was signi®cantly offsetting the positive in¯uence of PR on the staff cost (SC). This result shows that the correlation between the staff cost and pro®tability was uniquely low in Japan compared to the international standard. In fact, uniqueness of the Japanese banking sector is observed in the relationship between Figure 6.6 and Figure 6.7. As seen in Figure 6.7, pro®tability of the Japanese banking sector has been declining steadily in the 1990s. However, its staff costs have taken an upward turn in the same

158 A Vacuum of Governance in Japanese Bank Management Figure 6.6 Staff cost of banking sector (1986 = 00) 140 130 Japan

120

USA 110

Germany

100

UK

90 80 year

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

Sources: OECD Bank Pro®tability±Financial Statements of Banks; IMF, International Financial Statistics.

Figure 6.7 Pro®t before tax of banking sector (1986 = 00)

250

200 150 100

Japan USA Germany UK

50 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 year Sources: OECD, Bank Pro®tability±Financial Statements of Banks; IMF, International Financial Statistics.

period. This phenomenon suggests delayed restructuring of the Japanese banking sector. The third international comparison is the relationship between gross income, which is the sum of interest and non-interest incomes, and operating expenses. It would be reasonable to assume that the gross income determines the operating expenses such as staff costs and property costs. However, under entrenched management, operating expenses are determined independently from the gross income situation.

Masaharu Hanazaki and Akiyoshi Horiuchi 159

The data source of each country's banking sector and the estimation method are the same as the second test above. The basic equation is represented by the following equation: OPit ˆ Ci ‡  GIit�1

…3†

where OP, C, and GI are the operating expenses, the constant term, and the gross income respectively. We add the cross terms of each country dummy with the gross income GI to the basic equation (3). Thus, OPit ˆ Ci ‡  GIit�1 ‡ GIt�1 DUM

…4†

where , GI, and DUM are scalars of coef®cients, gross incomes of respective countries, and country dummy variables. Among 23 sampled countries, the United States is dummied out in Equation (4). This means that the parameter  represents the coef®cient of GI for the United States and the parameter represents the differences in sensitivity of OP with GI between the United States and the other countries. The estimation results are found in Table 6.4. In Equation (3) we observe that the correlation between the operating expenses and the one-year lag of the gross income is signi®cantly positive. Equation (4), however, shows that among 22 countries the estimated coef®cients of gross income of 12 countries including Japan are signi®cantly different from that of the United States. And in most countries, the coef®cients are smaller than that of the United States. This suggests that in most countries the bank operating expenses were less sensitive to changes in the gross incomes than was the case in the United States. Notably, Japan's coef®cient is much smaller compared with that of the United States, although the magnitude of Japan's own coef®cient turns out to be signi®cantly negative. This relationship between the operating expenses and the gross income in the Japanese banking sector is also observed in Figure 6.8. According to this ®gure, the operating expenses have been steadily increasing in spite of the declining income stream in the 1990s. The international comparisons in this subsection suggest that the behaviour of the Japanese banking sector is unique in the sense that business expansion was not ceased in spite of the structurally depressed banking business situations in the ®rst half of the 1990s. 6.2

Adjustment speed of capital stock

Finally, we examine how sensitively the ®nancial sector in Japan has adjusted its physical capital stock responding to changes in the real output, in comparison with other sectors in Japan, particularly with the

160 A Vacuum of Governance in Japanese Bank Management Table 6.4 The relationship between operating expenses and gross income Dependent variable: operating expenses Independent Variable

Equation (3)

Equation (4)

Const. GI GI  JPD GI  GED GI  FRD GI  UKD GI  CAD GI  AUD GI  TUD GI  BED GI  POD GI  DED GI  SED GI  FID GI  SWD GI  GRD GI  SPD GI  ITD GI  KOD GI  NED GI  MED GI  LUD GI  NOD GI  ICD

0.010 (5.12)** 0.479 (10.69)**

0.017 (7.71)** 0.383 (5.22)** �0.691 (�3.02)** �0.224 (�1.87)** �0.402 (�2.19)** �0.039 (�0.42) �0.133 (�1.35) �0.336 (�2.20)** 0.282 (3.61)** �0.492 (�2.70)** �0.148 (�1.58) �0.176 (�2.10)** �0.006 (�0.06) 0.091 (0.84) �0.388 (�2.99)** �0.128 (�1.13) �0.051 (�0.56) �0.125 (�1.21) �0.120 (�0.96) �0.234 (�1.85)* 0.182 (2.24)** �1.331 (�4.69)** 0.008 (0.09) 0.242 (3.15)**

R 2

0.8260

0.8289

Notes: GI is the one-year lag of gross income; JPD, GED, FRD, UKD, CAD, AUD, TUD, BED,

POD, DED, SED, FID, SWD, GRD, SPD, ITD, KOD, NED, MED, LUD, NOD, ICD are dummy

variables taking 1 for Japan, Germany, France, United Kingdom, Canada, Austria, Turkey,

Belgium, Portugal, Denmark, Sweden, Finland, Switzerland, Greece, Spain, Italy, Korea,

Netherlands, Mexico, Luxembourg, Norway and Iceland respectively and 0 for other countries.

Gross income and Operating expenses are the real term and are normalized by the year-average

total assets. The ®gures in parentheses are t-statistics. The asterisks ** and * indicate the levels of

signi®cance at 5% and 10% respectively.

Source: OECD, Bank Pro®tability±Financial Statements of Banks; IMF, International Financial

Statistics.

manufacturing sectors. Our basic assumption is that in each industry's net investment expenditure IN, is determined by a simple stock adjustment principle: INt ˆ …Kt  � Kt�1 †;

< 0 > ;  ˆ 1

…5†

Figure 6.8

Operating expenses and gross income of the Japanese banking sector (1986 = 00)

170

160

150

140

130

Gross income

120

Operating expenses

120

100

90 80 1986

year 1987

1988

1989

1990

1991

1992

1993

1994 1995

Sources: OECD, Bank Pro®tability±Financial Statements of Banks; IMF, International Financial Statistics.

161

162 A Vacuum of Governance in Japanese Bank Management

where  and  are the speed of adjustment and the optimal operating rate, respectively, and Kt * is the desired level of capital stock at the period t determined by the history of output level Qt ±1 in the following way: Kt  ˆ V

n X

i Qt�i ;

iˆ0

n X

i ˆ 1

…6†

iˆ0

where V is the capital/output coef®cient. We can derive the adjustment speed in capital stock  for each industry by assuming V is equal to an average of the actual capital/output coef®cient from the time series regression with speci®c structure of lag pattern. Adding the depreciation term to the above equations, the equation for gross investment It is obtained: It ˆ …V ˆ V

n X iˆ0 n X

i Qt�1 � Kt�1 † ‡  Kt�1

i Qt�i ‡ … �  †Kt�1

…7†

iˆ0

Finally we specify the estimation model as follows: It ˆ C ‡

n X

i Qt�1 ‡ Kt�1 ;

…8†

iˆ0

where …9†  V i ˆ i ;  �   ˆ n P To estimate the parameter i , the sum of the lag coef®cients, our iˆ0

regressions are run utilising two distributed lag models such as Almon lags and Shiller lags. The data is quarterly basis from 1974. Q4 to 1997.Q1 for each industry including Finance and Insurance consisting of commercial banks, insurance, and others. We divide the sample period into two parts; one is from 1974.Q4 to 1989.Q4, the other from 1990.Q1 to 1997.Q1 because we are interested in changes in stock adjustment speed between the two time periods. Until 1989 most industries experienced relatively steady growth in output levels, while since 1990 the output growth has been either stagnant or declining as a result of the `bubble burst'. We estimated the above stock adjustment equation for 14 industries including the aggregated manufacturing sector to obtain the speci®c value of adjustment speed . Table 6.5 and 6.6 respectively summarise the estimated results based on the Almon lag method27 and the Shiller lag technique.28

Masaharu Hanazaki and Akiyoshi Horiuchi 163 Table 6.5 Estimation of the stock-adjustment type of investment function (Almon lag method) (1) Period: 1974.Q4±1989.Q4 n P Const. i V  R 2 iˆ0

Manufacturing Food Products

�2,380,720 (�5.45) ** �125,910 (�1.39)

Textiles

�34,653 (�0.30)

Paper and Pulp

�108,006 (�4.81) ** �38,689 (�0.78)

Chemicals Metal Products

�98,196 (�2.50) ** General Machinery �201,968 (�11.30) ** Electric Machinery �52,912 (�2.10) ** Transport. Equipment �478,448 (�11.08) ** Electricity, Gas and Water 153,307 (0.38) Finance and Insurance 300,983 (3.41) ** Real Estate 305,495 (2.47) ** Transport and Communi. 199,109 (1.54) Services

�653,259 (�1.51)

0.0800 (3.15) ** 0.0202 (1.43)

0.00101 (1.49)

0.923 2.698

0.030

0.02067 0.902 1.584 (10.89) ** �0.0657 0.03633 0.801 3.939 (�1.39) (10.26) ** 0.1820 �0.00425 0.798 4.066 (4.86) (�0.66) ** 0.1268 0.00059 0.789 4.807 (2.78) (0.66) ** 0.0254 0.02686 0.936 2.607 (1.78) (19.34) * ** 0.0581 0.01596 0.961 1.931 (6.42) (4.16) ** ** 0.0804 0.00225 0.983 1.975 (6.99) (0.28) ** 0.0593 0.02706 0.977 2.102 (6.27) (13.41) ** **

0.013

0.2534 (0.91)

0.00236 (0.20)

0.1610 (6.31) ** �0.0841 (�4.26) **

�0.07433 0.882 2.222 (�3.61) ** 0.07558 0.903 1.044 (9.95) **

�0.0085 0.02241 (�0.38) (15.95) ** 0.0403 0.04386 (1.32) (7.44) **

0.633 12.545

�0.017 0.045 0.026 0.010 0.030 0.041 0.028

0.020 0.072 �0.081

0.964 3.589

0.002

0.981 1.450

0.028

164 A Vacuum of Governance in Japanese Bank Management Table 6.5 Estimation of the stock-adjustment type of investment function (Almon lag method) (continued ) (2) Period: 1990.Q1±1997.Q1 n P Const. R 2 V  i iˆ0

Manufacturing Food Products Textiles Paper and Pulp Chemicals Metal Products General Machinery Electric Machinery

�11,947,800

(�7.32)

**

3,558,740

(3.19)

**

�260,338

(�1.22)

�362,075 (�3.49) ** 594,843 (1.78)

*

�1,202,830

(�6.88)

**

�782,192

(�3.96)

**

77,537

(0.24)

Transport Equipment �627,568 (�1.14) Electricity, Gas and Water Finance and Insurance Real Estate Transport and Communi.

0.2838 (17.34) ** �0.3036 (�2.26) ** 0.2019 (4.29) ** 0.5801 (11.87) ** 0.4652 (5.14) ** 0.4554 (12.48) ** 0.1332 (12.87) ** 0.1726 (6.64) ** 0.1611 (4.84) **

�0.01459 0.965 3.813 (�9.27) ** �0.01820 0.649 2.806 (�3.02) ** 0.01072 0.850 7.011 (0.87) �0.06586 (�22.65) ** �0.08716 (�8.74) ** �0.01057 (�4.57) ** 0.01109 (2.44)

**

�0.04474

(�11.28)

**

�0.00328

(�0.42)

0.029 0.117

0.828 4.956

0.094

0.953 4.800

0.095

0.954 2.996

0.044

0.851 2.201

0.078

0.806 3.495

0.046

0.691 15.711

0.015

0.2327 (0.41)

0.01600 (0.78)

385,865 (1.34)

0.1087 (1.71) ** �0.2023 (�4.94) **

�0.04303 0.203 1.868 (�2.23) ** 0.01908 0.722 1.890 (1.57)

0.2789 (2.46) **

�0.00949 0.541 6.908 (�1.09)

�485,048 (�0.79)

�0.108

0.962 4.953

�706,658 (�0.55)

2,939,190 (9.52) **

0.074

0.058 �0.107

0.040

Masaharu Hanazaki and Akiyoshi Horiuchi 165 Table 6.5 Estimation of the stock-adjustment type of investment function (Almon lag method) (continued ) Services

�17,282,800 0.9271 (�8.34) (10.38) ** **

�0.06240 0.831 3.612 (�11.07) **

0.257

Notes: The estimated equation is It ˆ C ‡

n X

i Qt � I ‡ Kt � 1;

iˆ0

where I is business investment, v capital-output ratio, C a constant term,  the adjustment speed of capital stock, Q output, and K capital stock. The ®gures in parentheses are t-statistics. The asterisks ** and * indicate the levels of signi®cance at 5% and 10% respectively.

The overall performance of then estimated functions varies among P industries. As far as the parameter i is concerned, we were unable to iˆ0

obtain a signi®cantly positive sum of the lag coef®cients for some industries. Fortunately, the regression result for Finance and Insurance is statistically signi®cant for three cases among four regressions. Moreover, as for the manufacturing sectors, the sums of the lag coef®cients are signi®cantly positive especially in the case of the Almon lag method. Figure 6.9 presents those industries whose sums of the lag coef®cients are signi®cantly positive in both periods as a result of the Almon lag estimation. In this ®gure the left and right columns for each industry present the adjustment speed during the period from 1974.Q4 to 1989.Q4 and the period from 1990.Q1 to 1997.Q1 respectively. It is noteworthy that all sectors in Figure 6.9 but Finance and Insurance increased their adjustment speed during the period after 1990 when they experienced either declines or stagnation in output levels. In contrast with this, Finance and Insurance decreases the adjustment speed during the period after 1990. These results suggest that the ®nancial services industry was less sensitive to the necessity of downsizing their production capacity in the face of stagnation in output levels than the manufacturing industries that have continued to compete with foreign rivals in the global markets since the 1960s. We claim that this hesitation of the ®nancial services industry including banking in restructuring their business comes from the vacuum of managerial governance we have explained in the previous sections.29

7

Revelation of vulnerability ± nature abhors a vacuum

As far as investors in the ®nancial markets believe in the government's capability of implementing the traditional safety net, the vulnerability of

166 A Vacuum of Governance in Japanese Bank Management Table 6.6 Estimation of the stock-adjustment type of investment function (Shiller lag method) (1) Period: 1974.Q4±1989.Q4 n P Const. i V  R 2 iˆ0

Manufacturing Food Products

�1,969,520 (�3.48) ** �111,785 (�1.18)

Textiles

123,732 (0.80)

Paper and Pulp

�138,051 (�5.47) ** �51,732 (�0.97)

Chemicals Metal Products

�104,844 (�2.23) ** General Machinery �208,861 (�9.93) ** Electric Machinery �93,668 (�4.41) ** Transport. Equipment �478,860 (�8.89) ** Electricity, Gas and Water 60,233 (0.13) Finance and Insurance 310,434 (3.48) ** Real Estate 340,808 (2.64) ** Transport and Communi. 147,855 (1.11) Services

�596,090 (�1.32)

0.0547 (1.64)

0.01657 (1.92) * 0.0179 0.02086 (1.21) (10.51) ** �0.1288 0.03240 (�2.07) (7.54) ** ** 0.2341 �0.01202 (5.48) (�1.66) ** 0.1031 0.00604 (1.83) (0.47) * 0.0277 0.02683 (1.63) (17.84) ** 0.0625 0.01454 (5.06) (2.87) ** ** 0.0534 0.01893 (5.22) (2.74) ** ** 0.0593 0.02703 (4.92) (10.78) ** **

0.918 2.698

0.020

0.894 1.584

0.011

0.798 3.939

�0.033

0.801 4.066

0.058

0.771 4.807

0.021

0.934 2.607

0.011

0.958 1.931

0.032

0.990 1.975

0.027

0.975 2.102

0.028

0.3206 (1.03)

�0.00046 0.595 12.545 (�0.03)

0.026

0.1714 (6.51) ** �0.0906 (�4.39) **

�0.07865 0.880 2.222 (�3.77) ** 0.07808 0.898 1.044 (9.81) **

0.077

0.0001 (0.00)

0.02225 (15.61) ** 0.04416 (7.17) **

0.0363 (1.13)

�0.087

0.963 3.589

0.000

0.980 1.450

0.025

Masaharu Hanazaki and Akiyoshi Horiuchi 167 Table 6.6 Estimation of the stock-adjustment type of investment function (Shiller lag method) (continued) (2) Period: 1990.Q11997.Q1 n P Const. R 2 V  i iˆ0

Manufacturing Food Products Textiles Paper and Pulp Chemicals Metal Products General Machinery Electric Machinery

�11,607,600 (�6.12) ** 5,348,530 (4.20) ** �214,854 (�0.92) �447,672 (�3.49) ** 323,465 (0.63) �1,062,430 (�5.20) ** �890,073 (�3.99) ** �178,740 (�0.52)

Transport. Equipment �738,689 (�1.21) Electricity, Gas and Water Finance and Insurance Real Estate Transport and Communi. Services

0.2801 (13.86) ** �0.5298 (�3.41) ** 0.1864 (3.58) ** 0.6351 (9.86) ** 0.5544 (3.69) ** 0.4202 (9.50) ** 0.1435 (11.15) ** 0.2017 (6.61) ** 0.1590 (4.29) **

�0.01469 0.963 3.813 (�8.97) ** �0.01149 0.676 2.806 (�1.78) ** 0.00920 0.831 7.011 (0.68) �0.06985 (�18.19) ** �0.09655 (�6.09) ** �0.01071 (�4.47) ** 0.01176 (2.49) ** �0.04849 (�13.37) ** �0.00046 (0.05)

0.027 0.128

0.790 4.956

0.112

0.952 4.800

0.088

0.964 2.996

0.048

0.914 2.201

0.092

0.775 3.495

0.045

0.594 15.711

0.015

0.2318 (0.36)

0.01603 (0.68)

517,961 (1.40) 3,131,500 (8.60) **

0.0694 (0.78) �0.2311 (�4.69) **

�0.02890 0.000 1.868 (�1.02) 0.02678 0.676 1.890 (1.87) **

0.2306 (1.85) * �15,055,200 0.8288 (�4.74) (5.99) ** **

�0.189

0.960 4.953

�703,230 (�0.48)

�252,251 (�0.38)

0.073

0.037 �0.122

�0.00586 0.485 6.908 (�0.61)

0.033

�0.05612 0.803 3.612 (�6.41) **

0.229

168 A Vacuum of Governance in Japanese Bank Management Table 6.6 Estimation of the stock-adjustment type of investment function (Shiller lag method) (continued ) Notes: The estimated equation is

It ˆ C ‡

n X

i Qt � I ‡ Kt � 1;

iˆ0

where I is business investment, v capital-output ratio, C a constant term,  the adjustment speed of capital stock, Q output, and K capital stock. The ®gures in parentheses are t-statistics. For sub-sectors within manufacturing, the estimation period ends in 1996. Q1 instead of in 1997. Q1. The asterisks ** and * indicate the levels of signi®cance at 5% and 10% respectively.

the banking sector which the vacuum of management governance has fostered would not reveal itself. Although investors had recognised deterioration of bank performance due to rapid increases in nonperforming loans, they trusted that in the end the traditional safety net Figure 6.9 The adjustment speed of capital stock among industries 0.12

1974.Q4 ~1989.Q4

0.10

1990.Q1~1997.Q1

0.08

0.06

0.04

0.02

0.00 Manufacturing Paper and Pulp

Chemicals

Metal Products

General Machinery

Electric Transportation Finance and Machinery Equipment Insurance

Notes: Estimated results of Almon lag method. For sub-sectors within manufacturing, the estimation period ends in 1996. Q1 instead of in 1997. Q1.

Masaharu Hanazaki and Akiyoshi Horiuchi 169

would protect them from losses associated with bank failures. Thus, they did not need to differentiate good banks from bad ones in the capital markets. However, as the non-performing loan problem dragged on in the banking sector, the traditional safety net apparently reached a dead end incurring the distrust of investors about government capability to bail out distressed banks. Still the government continues to assign major banks an important role in bailing out weakened peers following the traditional safety net in Japan. However, as we have pointed out, the DIC, the semipublic organisation, has signi®cantly increased its role in the framework of the Japanese safety net. This fact signals that the traditional safety net which major banks bear the burden of bailing out distressed banks no longer works smoothly. The Japan premium shows to what extent Japanese banks pay higher interest rate in the international inter-bank money markets than foreign banks do. Thus, this premium re¯ects investors' evaluation of Japanese banks relative to their foreign rivals. The higher Japan premium suggests investors are more seriously concerned with capacity of Japanese banks to repay their debt. Figure 6.10 presents movements of the Japan premium which is de®ned by subtracting the London inter-bank offered rate (LIBOR) from the Tokyo inter-bank offered rate (TIBOR) with respect to 3 month dollar. The TIBOR is the average of inter-bank money market rates for 16 Japanese and two foreign banks (Barclays and Citi) surveyed by the Federation of Bankers Associations of Japan, the two highest and the two lowest banks being excluded. Since the two foreign banks have enjoyed lower interest rate than Japanese banks during the last several years, TIBOR can be regarded as the average offered rate for Japanese banks in Tokyo. On the other hand, LIBOR is the average of London inter-bank money market rate for major 16 banks including three Japanese banks (Tokyo-Mitsubishi, Fuji and Sumitomo Trust) cutting off the highest four and the lowest four from the average. Nowadays, the three banks are excluded from the LIBOR because the offered rates for them are substantially higher than the rates for the foreign banks in London. Thus, we may interpret the difference between TIBOR and LIBOR as a measure to what extent Japanese banks are negatively evaluated compared with their foreign rivals.30 According to Figure 10, the positive Japan premium was not observed until the end of September 1995. The Japan premium at the end of September (September 29) was only 1.042 basis point. However, the premium made a jump to 20.313 basis point on October 2. This abrupt jump was caused by the announcement on September 29 that the US

170

Figure 6.10 Japan premium (from the end of February 1991 to the beginning of July 1998)

% 1.5

1

0.5

0

–0.5

–1 1992

1993

1994

1995

1996

1997

1998

Masaharu Hanazaki and Akiyoshi Horiuchi 171

authorities discovered Daiwa Bank's wrongdoing in New York. The MOF was found to be awkward at dealing with this Daiwa case. This fact also contributed to the market turbulence.31 Associated with the increasing number of bank failures in the summer of 1995, this scandal triggered the scepticism in ®nancial markets of the government's capability to stabilise the banking system by means of the traditional safety net. The abrupt jump of the Japan premium re¯ected the widespread scepticism among investors. The scepticism of investors about the traditional safety net led to the start of the capital market mechanism of disciplining bank management. Once the government lost investors' con®dence in its capability of implementing the traditional safety net, investors were naturally motivated to monitor and discipline bank management severely. In short, the capital market started to ®ll the vacuum that existed in the framework of governance in bank management. In order to `calm down' the capital market, the government should have quickly strengthened the monitoring and disciplining of bank management. Unfortunately, the Japanese government did not recognise this development in the capital market neglecting to introduce effective measures to force banks to a quick recapitalisation. Thus, the disciplining mechanism of the capital market started at the worst time when most of banks were suffering from a large amount of non-performing loans. It is noteworthy that the ®nancial turmoil triggered by the failures of Yamaichi, Hokkaido-Takushoku not only caused resurgence of signi®cant Japan premium but, with this turmoil, the investors started to differentiate between individual banks according to their respective performances. Table 6.7 compares LIBOR (3-month US dollar) for TokyoMitsubishi Bank with those of Sumitomo Trust Bank and Fuji Bank during the two time periods before and after November 1997. During the period 1 September 1995 to 31 October 1997, the market scarcely differentiated offering interest rates for these three Japanese banks. However, since the beginning of November 1997, the inter-bank money market has differentiated between banks. Speci®cally, we observed signi®cant divergence between LIBOR for Tokyo-Mitsubishi and those for the other two banks. This change in the market attitude toward Japanese banks suggests that since late 1997, investors have no further con®dence in the effectiveness of the traditional safety net which used to make it unnecessary for them to evaluate the soundness of individual banks based on their respective performance. It is natural that the disciplinary mechanism of the capital market was severe and rather destructive in this situation. Some people criticised the

172 A Vacuum of Governance in Japanese Bank Management Table 6.7 Divergences of LIBOR (3 month US dollar) between Japanese major banks (%) Period

SUMI-BOTM

FUJI-BOTM

JPN premium

1 Sep 95± 31 Oct 97 3 Nov 97± 6 Oct 98

0.00838 (0.1985) 0.10000 (0.07342)

0.00446 (0.01564) 0.10040 (0.07370)

0.10436 (0.06733) 0.35279 (0.16539)

Notes: SUMI, FUJI, and BOTM are Sumitomo Trust Bank, Fuji Bank and Bank of TokyoMitsubishi respectively. JPN premium is the Japan premium de®ned by subtracting LIBOR for Citi Bank from that for Bank of Tokyo-Mitsubishi. Figures in parentheses indicate standard deviations.

capital market for the brutal manner with which it deals with distressed banks. Some go so far as to argue for suppressing the capital market to prevent a destructive impact on the banking system. However, they should note that the government has for long neglected to ®ll the vacuum in the governance of bank management, and that the capital market started to ®ll the vacuum just at the worst moment. In order to avoid the destructive working of the capital market, the government should have committed itself to ®ll the vacuum in place of the capital market. After jumping to a signi®cantly positive level at the beginning of October 1995, the Japan premium remained positive until November 1997, when the Japanese banking sector faced a second attack as shown by Figure 6.10. The dragging forbearance policy of the government accounted for this Japan premium phenomenon. After the ®nancial turmoil at the end of 1997, the government took some emergency policy measures to regain ®nancial stability. In particular, in March 1998 the government injected `public funds' of ¥1.8 trillion into 21 major banks' capital further to the Emergency Law for Stabilising Financial Functions which became effective in February 1998. Despite these emergency measures taken by the government, the Japan premium did not disappear immediately. Rather, the premium increased after the capital injection in March 1998.32 Obviously, the Japanese government failed in bringing out a positive response from the capital market. This is because the emergency policy measures since the beginning of 1998 were unable to convince investors that the government would truly part with the forbearance policy to ®ll the governance vacuum. The capital market requires the Japanese banking system to be more rationalised through drastic restructuring. However, from the investors' viewpoint, the emergency policy of

Masaharu Hanazaki and Akiyoshi Horiuchi 173

injecting public funds into bank capital without properly considering the true performance of individual banks was nothing but the policy of protecting inef®ciently managed banks. Any policy to cope with the current bank crisis would not be successful without a positive response from the capital market. The government and the capital market are struggling with each other to ®ll the governance vacuum in the bank management. If the government wins, the market will be calmed. However, if the government loses this struggle, the market will become more cruel for the time being. This is a totally new situation that the government has never experienced.

8

Concluding remarks

This paper is an overview of the governance structure in the Japanese banking industry. We have stressed that bank management has been independent from outsiders' control. Even the Ministry of Finance has not effectively monitored and disciplined bank management from the viewpoint of taxpayers. Thus, we have not resolved the issue `Who monitors the monitor?' in the Japanese ®nancial system. We may say there exists a vacuum in the governance of bank management. The vacuum of governance in the banking sector is responsible for the delayed restructuring in the banking industry which has been suffering from the bad loan problem since the beginning of the 1990s. In April 1998 the Japanese government instituted a policy of introducing the prompt corrective action rule and of ordering banks to submit explicit time schedule of managerial restructuring under which condition the government injects public funds into banks capital. These policy measures seem to have at last induced hesitant banks to start restructuring their business. This fact in itself tells us that Japanese banks lacked a strong incentive to drastically reform their way of business on their own initiative.

Notes * An earlier version of this paper was presented at the NBER Japan Project Meeting held on 17±18 April 1998, in Cambridge, Massachusetts, and the Econometrics Conference, Capital Market and Corporate Governance held on 9±11 July 1998, in Shiga, Japan. The authors wish to thank Patrick Bolton, Michael Gibson, Shin-ichi Hirota, Anil Kashyap, Hugh Patrick, Eric Rosengren and Yoshiro Tsutsui for their constructive comments on the earlier version. Financial support from the Japan Ministry of Education Scienti®c Research Grant is gratefully acknowledged.

174 A Vacuum of Governance in Japanese Bank Management 1. We de®ne all depository ®nancial institutions as banks including not only the city banks and regional banks, but also various cooperative credit banks. 2. The group of Major Banks consisted of city banks, three long-term credit banks, and seven trust banks. Since Mitsubishi Bank and Tokyo Bank merged in April 1996, and Hokkaido-Takushoku Bank went bankrupt in November 1997, the number of city banks was nine as of July 1998. 3. The Major Banks and the regional banks began to disclose ®gures for nonperforming loans in 1993. However, until March 1996, the Major Banks did not disclose the non-performing loans belonging to category (c), and the regional banks disclosed only category (a). 4. The credit cooperatives seem to be particularly fragile. According to Table 6.1, the bad loans/total loans ratio is 12 per cent for the credit cooperatives, or nearly three times as high as that for the Major Banks, which is estimated at 4 per cent. As for shinkin banks, it was reported that if they were to subtract nonperforming loans from their equity capital, almost 90 per cent of these banks would be unable to satisfy the domestic standard of capital adequacy requirement (4 per cent) imposed on commercial banks in Japan (Nihon Keizai Shimbun, 16 May 1996). This newspaper report suggests the serious dif®culty of non-performing loans for the cooperative banks. 5. Under the MOF's guidance, the banks formed the Cooperative Credit Purchasing Corporation (CCPC) to help themselves write off bad loans. The banks sell bad loans to the CCPC at discount prices. Banks are required to fund the purchase of the loans that they bring to the CCPC. The essential purpose of the scheme is to make losses on non-performing loans explicit so that banks obtain tax relief. When a bank sells non-performing loans to the CCPC at discount prices, the amount of bad loans is reduced by that amount in the bank's balance sheet. However, if the CCPC is unable to collect the loans within several years, the selling banks must take them back. The total amount of non-performing loans sold by the banks to the CCPC from the second half of ®scal year 1992 to the ®rst half of ®scal year 1997 is ¥4.0 trillion in book value and ¥5.5 trillion in actual prices. However, more than 80 per cent of the loans sold to the CCPC were not collected by September 1997. Thus, it would be safe to consider the amount of bad loans sold to the CCPC as still remaining on the balance sheet of the banking sector, although the whole amount of the loans sold to the CCPC is deleted from the non-performing loans in Table 6.1. 6. A note `The Japan Premium: Work in Progress' presented by Joe Peek and Eric S. Rosengren to the NBER±Japan Project on 17±18 April 1998 gives information about changes in the Japan premium. 7. It should be noted, as Gibson (1995) points out, that the deterioration of bank performance would weaken competitiveness of industrial ®rms, particularly those heavily depending on bank credit, by increasing the cost of capital for them. This bad in¯uence of the bank crisis may endanger the long-run growth capability of the Japanese economy. However, we may be optimistic about the bad in¯uence on the major companies, because they have substantially reduced their dependence on bank credit since the early 1980s. According to the BOJ statistics, the average of blue chip companies' dependence on bank credit in their total ®nance was just 6 per cent and 5 per cent respectively during the second half of the 1980s and the ®rst half of the 1990s, whereas

Masaharu Hanazaki and Akiyoshi Horiuchi 175

8.

9.

10.

11.

12.

their dependence on bank credit was higher than 30 per cent during the high growth era until the mid-1970s (The Bank of Japan, Analysis of Major Companies Management). The major companies would be able to raise funds in international capital markets independently from the intermediation capability of Japanese banks. The Japanese government injected public funds of ¥1.8 trillion into 21 major banks (nine city banks, three long-term credit banks, six trust banks, and three big regional banks) by buying either preferred stocks or perpetual subordinated debt at March 1998 with a view to mitigating the credit crunch. This injection is estimated to have increased equity capital of those banks by 5.14 per cent. It is extremely unclear whether this policy was really effective in mitigating the credit crunch. Furthermore, it remains to be investigated whether or not the direct injection of public funds into bank capital would be consistent with the long-term objectives of restructuring the Japanese banking industry. Total abolition of the ®nancial safety net would strengthen the incentives of depositors and investors to monitor and discipline bank management. However, since most depositors are small-sized wealth-holders enjoying no economy of scale in collecting and analysing information about bank management and since there exists a `free-riders' problem to hinder ef®cient information production, it would be unrealistic to depend totally on market discipline to keep stability of the banking system. As Dewatripont and Tirole (1994) argue, we need to have a sort of ®nancial safety net in order to protect small-sized investors in the banking sector. Until the end of the 1980s, the number of banks that came close to failing was small, with the largest rescue programme involving not a bank but Yamaichi Securities Company in 1965. In this rescue, coordinated by the MOF, the BOJ provided emergency loans of ¥28.2 billion to Fuji Bank and two other banks which functioned as conduits supplying ®nancial support to Yamaichi. In 1965, for example, Kawachi Bank, a small regional bank in ®nancial distress, was absorbed by Sumitomo Bank, while in 1978 Mitsui Bank absorbed Toto Bank, which had suffered from stagnant performance for a long time. In both cases, the rescue programmes were implemented under the administrative guidance of the MOF. Actually, until the late 1960s, there were a few cases in which depositors were forced to bear some part of losses associated with bank failures. See Yamawaki (1996). We may regard the protection given to shareholders as compensation for their silence on bank management. In reality, the shareholders have been rather similar to debt-holders in the governance structure of bank management. This is evidenced by the fact that a dividend on bank shares has been extremely stable regardless of bank performance. For example, the pro®ts of city banks were either very small or negative during the ®ve years from 1993 to 1997 mainly due to large loan loss provisions. Nevertheless, the city banks continued to pay almost constant amount of dividends to their shareholders. The total amount of pro®ts for the city banks was less than minus ¥1.8 trillion for the ®ve years. On the other hand, the total amount of dividend paid out by the city banks was a little larger than ¥1.0 trillion for the same ®ve years. If they had not paid the dividend at all, the total amount of capital would have been larger by 10 per cent for those banks than the actual amount in March 1998.

176 A Vacuum of Governance in Japanese Bank Management 13. Unfortunately, we have observed a number of cases that suggest a forbearance policy on the part of the authorities during the early 1990s. For example, we may cite the case of Cosmo Credit Cooperative, which failed and was taken over by Tokyo Kyodo Bank in March 1996. Although Cosmo had already fallen into serious dif®culty, with negative pro®ts in early 1992, the Tokyo metropolitan government, responsible for monitoring management of credit cooperatives located in Tokyo, allowed it to conceal its actual bad situation by manipulating accounts (Nihon Keizai Shimbun, 7 May 1996). The failure of Musashino Shinkin Bank is also an example of the forbearance policy that came to light in 1996. Musashino Shinkin had been in trouble since 1993 and the MOF was in charge of examining the bank's account statements before publication. The MOF reportedly allowed the bank to engage in window dressing to record positive pro®ts even as of March 1996, when the estimated amount of problem loans was nearly 70 per cent of total loans. The MOF guided the bank to conceal its dif®culties by allowing managers to manipulate ®nancial statements. In September 1996, the MOF decided to introduce an explicit system of ordering banks in trouble to improve their management based upon of®cially announced criteria (Nihon Keizai Shimbun, 11 October 1996). 14. Needless to say, before adopting the policy of paying off deposits, the MOF should introduce more perfect disclosure of individual banks' bad loans to help investors outside the deposit insurance coverage to select sound banks. 15. With regard to the branch administration implemented by the MOF, see footnote 17 below. 16. A few recent cases exemplify the dif®culties the MOF faces in using traditional bailing-out policies. In the summer of 1992, Toyo Shinkin Bank, located in Osaka, was broken up because of insolvency due to bad loans. The MOF reportedly wanted Sanwa Bank, a leading city bank, to absorb it in the traditional fashion, but was unable to persuade Sanwa to do this. Instead, Toyo Shinkin was broken up into a number of pieces, each of which was absorbed by a different ®nancial institutions. In the process, the Deposit Insurance Corporation paid ¥20 billion to Sanwa, which absorbed the largest part and played a major role in the reorganisation. Another event signaling that traditional methods are running into trouble occurred early in 1994. Three local banks in the Tohoku area jointly announced a plan for a merger, another typical MOF bail out method. One of the banks had a serious bad loan problem, and many parties, including the MOF, were pessimistic about its future viability. The merger plan, which was undoubtedly the result of MOF administrative guidance, had to be abandoned following ®erce resistance from employees of the relatively sound banks involved. Some of the banks' managers also reportedly argued against the merger. 17. The MOF's administration of branch of®ces was another signi®cant area of regulation. During the high growth period, when almost all deposit interest rates were under regulation, branch of®ces were an important means of nonprice competition for banks and essentially the vehicle by which they competed for deposit funds. Under the MOF's administration, banks were not free either to expand or to change the location of their branch networks. In permitting new branches, the MOF reportedly gave preferential treatment to small banks. The number of branches of small-scale banks increased more

Masaharu Hanazaki and Akiyoshi Horiuchi 177

18.

19.

20. 21. 22.

23.

rapidly than did that of city banks, both during and after the high growth period (Horiuchi, 1984). The MOF partially abandoned branch administration by allowing regional banks and shinkin banks freely to increase the number of branch of®ces in May 1993. At that time, the MOF announced that the branch regulation for city banks would be gradually liberalised while taking into account the in¯uence on small and medium sized ®nancial institutions. In May 1995, the MOF totally liberalised the regulation regarding the number of branch of®ces for all banks. Aoki (1994) argues, by assuming asymmetric information about banks' monitoring activities, that the rent was necessary to motivate private banks to monitor their borrowers faithfully and ef®ciently. He suggests that the longterm relationship between major banks and borrower ®rms in Japan, called the `main bank relationship', was crucially dependent on the competitionrestricting regulations. However, the restricting full-scale competition was not always necessary to motivate banks to supply a `high quality' level of monitoring. The laissez-faire market would be able to motivate banks to conduct good monitoring. See Klein and Lef¯er (1981). Even during the 1990s, the MOF manipulated its administrative guidance with a view to induce private banks to collaborate with its rescue program. In 1994, for example, Mitsubishi Bank obtained preferential treatment from the MOF in exchange for rescuing Nippon Trust Bank, which had been seriously damaged by the accumulation of a huge amount of bad loans since the early 1990s. Mitsubishi Bank was `rewarded' by being allowed to pursue a full complement of trust banking business through Nippon Trust, which is now its subsidiary. Other banks were prohibited by the MOF from engaging in fullline trust banking business through their trust bank subsidiaries. The same story is true of the case in which Daiwa Bank ®nancially supported Cosmo Securities Company, which was seriously damaged by the depression in the securities market after the `bubble' burst at the beginning of the 1990s. Cosmo has been a subsidiary of Daiwa Bank. However, Cosmo retained its stock brokerage business. Frankel (1984) explains the process of the Yen/Dollar agreement. Takeda and Turner (1992) discusses the relationship between the internationalisation of Japanese ®nancial markets and domestic ®nancial deregulation in great detail. Broadly de®ned capital includes not only equity capital (book value), but also some reserve items. The MOF amended the capital adequacy regulation in 1986 when the accounting rules governing bank ®nancial statements were changed. Through this amendment, the MOF probably intended to make the capital adequacy regulation more realistic. It is unclear whether the MOF recognised the increasing need for prudential regulations in banking as of the mid-1980s. The new capital adequacy rule required banks' broadly de®ned capital to be at least 4 per cent of total assets, hardly a stringent requirement. Since 1987, banks with branches or of®ces in foreign countries have been subject to the BIS capital adequacy rule, but other banks continue to face only this domestic capital adequacy requirement of 4 per cent. There are a number of hypotheses to explain why the Japanese ®nancial system has accepted the amakudari system. Rixtel (1994) provides a useful overview of these hypotheses. Neglecting all other hypotheses, this paper

178 A Vacuum of Governance in Japanese Bank Management

24. 25. 26.

27. 28. 29.

30.

31.

32.

concentrates on analysing amakudari from the viewpoint of the effectiveness of the ®nancial safety net. Kane (1989) points out there exists a similar agency problem in the US banking system. This problem was responsible for the S&L failures during the 1980s in the United States. Since the BOJ has not played a signi®cant role with respect to prudential regulation, this result is plausible (Horiuchi and Shimizu, 2001). Since 1997, the bank crisis in Japan has been centred on the major bank group consisting of big city banks. The most emergent policy agenda for the government is how to strengthen capital basis of those big banks suffering from the serious non-performing loan problem. As is well-known, big banks have been mostly independent from the amakudari relationship with the regulators. In contrast, the group of regional banks appears to be relatively sound partly because the weakest ones had disappeared until the end of 1997. Thus, the negative in¯uence of amakudari on soundness of bank management does not explain the whole story of the current bank crisis in Japan. The Almon lag speci®cation used for estimation is as follows: the number of terms in the polynomial is 2 degrees, the number of distributed lags is 8 quarters, and the endpoint constraint is FAR. For the Shiller lag estimation, we use a second degree of differencing, 8 quarter-lags, a FAR endpoint constraint and a prior variance on the differenced coef®cients equal to 0.1. Peek and Rosengren (1998) point out that some Japanese big banks restructure their business in inef®cient ways. According to their analysis, Japanese banks increased their lending to the US real estate sector in the early 1990s. At the peak in 1992, their US. subsidiaries held around 20% of all commercial real estate loans in the US banking sector. However, they cut back their lending in the U.S. in response to a sharp decline in real estate prices in Japan even though the U.S. real estate prices were rising. At the same time, Japanese banks expanded their lending to the domestic market where prices were plummeting. Thus, they transferred their loans from more pro®table sections to less pro®table, and much more risky sections. Both LIBOR and TIBOR do not show the offered rates for individual banks. But apparently Japanese banks are not greatly differentiated from each other. For example, as on 22 June 1998, the Japan premium de®ned by (TIBOR±LIBOR) was 20.834 basis point. On the same day, the inter-bank offered rate was 5.875 per cent for Tokyo-Mitsubishi, and 5.9375 per cent for both Fuji and Sumitomo Trust respectively in London. The difference was only 6.22 basis point between Tokyo-Mitsubishi and the other two banks. Meanwhile, the offered rate for all the foreign banks was the same at 5.687 per cent on 22 June 1998. On 16 October 1995, the public hearing with regard to Japan's ®nancial system was held at the House of Representative in Washington, DC. This public hearing seemed to promote scepticism of markets against the capability of Japanese ®nancial authorities. The Japan premium went up further immediately after the public hearing, reaching the unprecedented high of 52.605 basis point on October 25. Interestingly, from 13 March to 2 April 1998, the divergences of LIBOR between Tokyo-Mitsubishi and the other two banks disappeared while the Japan premium remained signi®cantly positive. This suggests that the inter-

Masaharu Hanazaki and Akiyoshi Horiuchi 179 bank money market regarded the capital injection by the government in March 1998 as a partial revival of the traditional safety net without the de®nite resolution of the banking problem as a whole in Japan.

References Aoki, Masahiko (1994) `Monitoring characteristics of the main bank system: An analytical and developmental view', in Masahiko Aoki and Hugh Patrick (eds.), The Japanese Main Bank System: Its Relevancy for Developing and Transforming Economies, New York: Oxford University Press, pp. 109±41. Aoki, Masahiko, Hugh Patrick and Paul Sheard (1994) `The Japanese main bank system: An introductory overview', in Masahiko Aoki and Hugh Patrick (eds.), The Japanese Main Bank System, New York: Oxford University Press, pp. 3±50. Bank for International Settlements (1996) 66th Annual Report. Boot, Arnoud W.A. (1992) `Why Hang on to Losers? Divestitures and Takeovers', Journal of Finance, 47, pp. 1401±23. Dewatripont, Mathias and Jean Tirole (1994) The Prudential Regulation of Banks, Cambridge, MA: MIT Press. Frankel, Jeffrey A. (1984) The Yen±Dollar Agreement: Liberalizing Japanese Capital Markets, Policy Analyses in International Economics 9, Washington, DC: Institute for International Economics. Gibson, Michael S. (1995) `Can bank health affect investment?: Evidence from Japan', Journal of Business, 68, pp. 281±308. Gorton, Gary and Richard Rosen (1995) `Corporate control, portfolio choice, and the decline of Banking,' Journal of Finance, 50 (5), pp. 1377±420. Hellman, Thomas, Kevin Murdock and Joseph E. Stiglitz (1997) `Financial restraint: Toward a new paradigm', in Masahiko Aoki, Hyung-Ki Kim and Masahiro OkunoFujiwara (eds.), The Role of Government in East Asian Economic Development: Comparative Institutional Analysis, New York: Oxford University Press, pp. 163± 207. Horiuchi, Akiyoshi (1984) `Economic growth and ®nancial allocation in postwar Japan,' Brooking Discussion Papers in International Economic, 18. Horiuchi, Akiyoshi (1996) `An Evaluation of Japanese Financial Liberalization: A Case Study of Corporate Bond Markets', in Takatoshi Ito and Anne O. Krueger (eds.), Financial Deregulation and Integration in East Asia, Chicago: University of Chicago Press, pp. 167±91. Horiuchi, Akiyoshi and Katsutoshi Shimizu (1998) `The deterioration of banks' balance sheets in Japan: Risk-taking and recapitalization', Paci®c Basin Finance Journal, 6, pp. 1±26. Horiuchi, Akiyoshi and Katsutoshi Shimizu (2001) `Did amakudari undermine the effectiveness of regulatory monitoring in Japan?', Journal of Banking and Finance, 25, pp. 573±96. Kane, Edward J. (1985) The Gathering Crisis in Deposit Insurance, Cambridge, MA: MIT Press. Kane, Edward J. (1989) `Changing incentives facing ®nancial-services regulators', Journal of Financial Services Research 2 (3), pp. 265±74.

180 A Vacuum of Governance in Japanese Bank Management Kane, Edward J. (1993) `What lessons should Japan learn from the US depositinsurance mess?', Journal of the Japanese and International Economies, 7, pp. 329±55. Keeley, Michael C. (1990) `Deposit insurance, risk, and market power in banking', American Economic Review, 80, pp. 1183±200. Kim, K.A. and S.G. Rhee (1997) `Large shareholders of banks: Shareholder activism and the impact of the regulatory environment', unpublished manuscript. Klein, Benjamin and Keith B. Lef¯er (1981) `The role of market forces in assuring contractual performance', Journal of Political Economy, 89, pp. 615±41. Lindgren, Carl-Johan, Gillian Garcia and Matthew I. Saal (1996) Bank Soundness and Macroeconomic Policy, International Monetary Fund. Washington, DC: Marsh, Terry A. and Jean-Michel Paul (1996) `Japanese banks' bad loans: What happened?', paper presented at the Conference on Emerging Trends in Japanese Financial Markets. Nickell, S., D. Nicolitsas and N. Dryden (1997) `What makes ®rms perform well?', European Economic Review, 41, pp. 783±96. Peek, Joe and Eric Rosengren (1998) `The international transmission of ®nancial shocks: The case of Japan,' American Economic Review, 87, pp. 495±505. Prowse, Stephen D. (1992) `The structure of corporate ownership in Japan', Journal of Finance, 47, pp. 1121±40. Rixtel, Adrian (1994) `The change and continuity of amakudari in the private banking industry: Patterned equalization, bureaucratic intervention and career management', a paper presented at the 7th Conference of the European Association of Japanese Studies on The Change and Continuity of the Japanese Political Economy. Takeda, Masahiko and Phillip Turner (1992) `The liberalization of Japan's ®nancial markets: some major themes', BIS Economic Papers, 34. Ueda, Kazuo (1996) `Causes of the Japanese banking instability in the 1990s', paper presented at the FAIR±Wharton Conference in Tokyo. Weisbrod, Steven R., Howard Lee and Lilian Rojas-Suarez (1992) `Bank risk and the declining franchise value of the banking system in the United States and Japan', IMF Working Paper 92/45. Yamawaki, Takeshi (1995) `The forbearance policy: Japanese ®nancial regulation', Green College, Oxford, Reuter Foundation Paper 56.

Part II

Corporate Governance

This page intentionally left blank

7

Corporate Bankruptcy Michelle J. White

The ®rst section of this article brie¯y describes bankruptcy procedures in the United States and other countries. The second section describes some of the important economic issues and tradeoffs in bankruptcy.

1

Bankruptcy law

Most countries have two bankruptcy procedures, one for liquidating the assets of failing ®rms and the other for reorganising failing ®rms. Bankruptcy liquidation The bankruptcy liquidation procedure in the United States is called Chapter 7. When a ®rm ®les under Chapter 7, the bankruptcy court appoints a trustee who shuts the ®rm down and sells its assets. The bankruptcy priority rule that determines how the proceeds of the sale are divided is called the `absolute priority rule'. It speci®es that claims are paid in full in the following order: ®rst, administrative expenses of the bankruptcy process itself; second, claims taking statutory priority, such as tax claims, rent claims, and unpaid wages and bene®ts; and, third, unsecured creditors' claims, including those of trade creditors, long-term bondholders, and holders of damage claims against the ®rm. Subordination agreements which specify that certain unsecured creditors rank above others in priority are followed; otherwise all unsecured creditors' claims have equal priority. Equity holders receive the remainder, if any. Secured creditors are outside the priority ordering. They have bargained with the ®rm for the right to claim a particular asset or its value if the ®rm defaults and/or ®les for bankruptcy. Secured creditors thus may receive a payoff in bankruptcy even when all other creditors receive nothing.

183

184 Corporate Bankruptcy

The bankruptcy liquidation procedures in the United Kingdom, Germany and France are similar, but are less frequently used. In Germany, the high costs of ®ling for bankruptcy discourage ®rms from ®ling. In the UK, a particular type of creditor ± called a ¯oating charge creditor ± has a blanket security interest in all of the ®rm's assets not subject to other secured creditors' liens. If the ®rm defaults, this creditor has the right to begin liquidating any assets of the ®rm that are not subject to secured creditors' liens. Bankruptcy liquidation begins only after the ¯oating charge creditor has been repaid. Bankruptcy reorganisation The bankruptcy reorganisation procedure in the United States is called Chapter 11. Managers of ®rms in ®nancial distress have the right to choose between ®ling for bankruptcy under Chapter 7 or Chapter 11. Under Chapter 11, existing managers of ®rms usually remain in control as `debtors-in-possession', they have the exclusive right for at least six months (and normally much longer) to propose a reorganisation plan for the ®rm and they have substantial control over the bargaining process generally. The reorganisation plan speci®es how much each creditor will receive in cash or new claims on the ®rm. There are two separate procedures for adopting a plan. Under the most commonly used of the two, all classes of creditors and equity as a class must vote in favour of the plan. Creditors must receive at least what they would be entitled to if the ®rm liquidated and the proceeds were distributed according to the absolute priority rule. For each class of creditors, the required voting margin in favour of the plan is at least two-thirds in amount and one-half in number of claimants and, for equity, the voting margin is at least two-thirds of all shares. The payoff patterns under Chapters 7 versus 11 differ strongly. Under Chapter 7, high priority and secured creditors tend to receive full payoff and low priority creditors and equity receive nothing at all, while under Chapter 11, each class of creditor and equity receives partial payment. Assuming that reorganising the ®rm causes it to be worth more than its assets would bring in liquidation, Chapter 11 provides a framework within which creditors and managers bargain over the distribution of the extra value. There is also a procedure ± called `cramdown' ± under which a Chapter 11 reorganisation can be adopted by the bankruptcy judge despite being voted down by one or more class of creditors. This procedure has higher costs and is much less frequently used.

Michelle J. White 185

If the parties fail to adopt a reorganisation plan proposed by managers, then creditors are eventually allowed to offer their own plans. If no plan proposed by creditors is adopted, then eventually either the bankruptcy judge orders that the ®rm be liquidated under Chapter 7 or the ®rm may be offered for sale as a going concern under Chapter 11. In the latter case, the proceeds are distributed according to the absolute priority rule. Firms in Chapter 11 bene®t from a number of provisions intended to increase the probability that they successfully reorganise. Managers are allowed to retain pre-bankruptcy contracts that are pro®table and reject pre-bankruptcy contracts that are unpro®table (while non-bankrupt ®rms are required to complete all their contracts). They are allowed to terminate underfunded pension plans and the government picks up the uncovered pension costs. Firms' obligation to pay interest to most creditors ceases when they ®le for bankruptcy. With the bankruptcy court's approval, ®rms in bankruptcy may give highest priority to creditors who provide post-bankruptcy loans, even though much of the payoff to these new creditors will come at the expense of pre-bankruptcy creditors. Also, the ®rm escapes any obligation to pay taxes on debt forgiveness under the reorganisation plan until it becomes pro®table. The United Kingdom, France and Germany have all recently adopted new bankruptcy procedures intended to encourage reorganisation of ®rms in ®nancial distress. However, these procedures differ substantially from Chapter 11 and also differ among themselves. In all three countries, prebankruptcy managers are given much less power over the reorganisation process than under Chapter 11. Instead of managers choosing between liquidation and reorganisation procedures in bankruptcy, the bankruptcy judge or an of®cial appointed by the bankruptcy court makes the decision and, if reorganisation is chosen, the of®cial formulates the reorganisation plan. In France, bankruptcy of®cials appointed to decide whether ®rms in bankruptcy will be liquidated or reorganised have as their primary objective `safeguarding the business'. In the United Kingdom, managers may petition the bankruptcy court to initiate a reorganisation procedure called an `administration order'. However, ¯oating charge creditors have the right to block administration orders, so that the procedure is infrequently used. In Germany, the bankruptcy reorganisation procedure has also been infrequently used, because of high costs and high minimum payoff requirements. As a result, most reorganisations of ®nancially distressed ®rms have tended to occur outside of any formal bankruptcy procedure. (See White, 1996, for a description and comparison of bankruptcy procedures in the United States, the United Kingdom, France and Germany.)

186 Corporate Bankruptcy

Payoff patterns in bankruptcy For reasons of data availability, most of the empirical research on bankruptcy has concentrated on large American ®rms that have publicly traded debt or equity. Most large ®rms that are ®nancially distressed reorganise rather than liquidate in bankruptcy, so that there is little data available on payoff rates to creditors in Chapter 7 bankruptcies. What data are available suggest that payoff rates under Chapter 7 tend to be very low. Most of the empirical research has concentrated on measuring how often Chapter 11 reorganisation plans deviate from the absolute priority rule by making positive payoffs to junior creditors and/or equity when senior creditors receive less than full payoff. Studies by Franks and Torous (1989), Weiss (1990), LoPucki and Whitford (1990), and Eberhart et al. (1990) ®nd that violations of the absolute priority rule occur in half to three-quarters of all large ®rm reorganisations. Unsecured creditors typically receive payoff rates of between .50 and .70 in Chapter 11 reorganisations. Also, the average time from ®ling to approval of a reorganisation plan is 1±2 years. Additional evidence suggests that Chapter 11 is not a cure for ®nancially ailing ®rms. Hotchkiss (1995) found that about one-third of ®rms that reorganised under Chapter 11 underwent further ®nancial restructuring within a few years. Violations of the absolute priority rule in large ®rm reorganisations under Chapter 11 can be quanti®ed by regressing the amount paid to equity as a fraction of unsecured creditors' claims on the amount paid to unsecured creditors as a fraction of their claims (i.e., the payoff rate to unsecured creditors). Using data from studies by LoPucki and Whitford (1990) and Eberhart et al. (1990), the results show that equityholders receive a minimum payoff of about 5 per cent of the amount owed to creditors, regardless of how much creditors receive. Thus in return for agreeing to a reorganisation plan, equityholders demand and get a minimum of 5 per cent of creditors' claims even when ®rms are clearly insolvent. As creditors' payoff rate increases, the payoff to equity as a proportion of creditors' claims rises at an increasing rate. When creditors' payoff rate is around .50 ± a common ®gure ± equity receives about 15 per cent of creditors' claims. When creditors' payoff rate is around .90, equity receives about 40 per cent of creditors claims. The fact that the payoff rate to equity rises at an increasing rate is consistent with equityholders being compensated for giving up their option on the ®rm's future earnings (see below for discussion).

2

Economic issues and tradeoffs in bankruptcy

In this section, I discuss some of the important economic issues and tradeoffs in bankruptcy, using a series of examples.

Michelle J. White 187 Table 7.1

Creditors' incentive to race to be ®rst

C1 doesn't race C1 races

C2 doesn't race

C2 races

3.9, 3.1 5,0

1, 4 3, 2

Creditors' incentive to race to be ®rst to liquidate failing ®rms' assets By de®nition, an insolvent ®rm has assets whose value is less than the total amount of creditors' claims. If unsecured creditors perceive that a ®rm is or may be insolvent, they anticipate that it will not be able to repay all its creditors in full and ± as in a bank run ± they have an incentive to race against each other to be ®rst to collect from the ®rm. Unsecured creditors race to collect by declaring their loans in default and suing the ®rm for repayment. (Secured creditors have less incentive to race against each other since they can foreclose on their collateral if default occurs.) When unsecured creditors win their lawsuits, they have the right to liquidate any assets of the ®rm not subject to secured creditors' liens. But when individual creditors liquidate assets piecemeal, they disrupt the ®rm's operations and may force it to shut down even when the best use of its assets is continued operation. Also individual creditors are concerned only with repaying their own claims, not with selling the ®rm's assets for their maximum value (Jackson, 1986). The race by creditors to be ®rst to collect from failing ®rms is an example of the prisoner's dilemma and may lead to inef®cient use of failing ®rms' assets (Webb, 1991). Suppose a ®rm has two creditors of equal priority, C1 owed 5 and C2 owed 4. If one or both of them races to be ®rst to collect, then suppose the ®rm's assets will be liquidated piecemeal and will be worth 5 in total. As shown in Table 7.1, if C1 races to collect but C2 does not, then C1 gets full repayment of 5 and C2 gets zero; while if C2 races to collect and C1 does not, then C2 gets full repayment of 4 and C1 gets 1. If they both race to collect, then suppose the expected outcome is that C1 gets 3 and C2 gets 2. Alternatively if the two creditors agree not to race against each other, then suppose the ®rm's assets would be worth 7 rather than 5. The assets are worth more when creditors do not race to collect because they are sold all at once rather than piecemeal and/or because they are sold when they have maximum value. Then both creditors would be paid 7/9 ths of their claims, so that C1 receives 3.9 and C2 receives 3.1. Although both creditors are better off in the no race outcome, the only equilibrium of the game occurs where they both race to be ®rst.

188 Corporate Bankruptcy

In this context we can interpret bankruptcy liquidation as a mandatory procedure which imposes the ef®cient, no race outcome. Suppose when a ®rm ®les for bankruptcy, all lawsuits by creditors are suspended and a bankruptcy trustee is appointed to sell all of the ®rm's assets in the most ef®cient way. The resulting increase in value of the assets is the justi®cation for imposing bankruptcy liquidation as a mandatory collective procedure. This consideration is not just theoretical. LoPucki (1983) presented evidence that many bankruptcy ®lings in the United States occur because creditors were about to foreclose on valuable assets and force ®rms to shut down. In Germany and the United Kingdom, liquidations of failing ®rms' assets tend to be private rather than collective and failing ®rms are thought to shut down prematurely as a result. Investment incentives, priority rules, and the bankruptcy decision While creditors have an incentive to shut down failing ®rms prematurely by racing to be ®rst to collect, managers of these ®rms may have an incentive to delay bankruptcy as long as possible, because they lose their jobs and equity loses its value when the ®rm is liquidated. This can also give rise to inef®cient bankruptcy decisions. The bankruptcy decision has often been modeled by assuming that equityholders and a working capital lender form a coalition which ignores the interests of other creditors (Bulow and Shoven, 1978; White, 1980 and 1989; and Gertner and Scharfstein, 1991). Suppose bankruptcy ®lings must be voluntary and assume that the only bankruptcy procedure is liquidation. A failing ®rm has debt due immediately of 2 and debt due next period of 5. It has no cash. The liquidation value of the ®rm's assets is 7, but if it continues to operate until next period, it will earn either 5 or 8, each with .5 probability. In this example liquidation is economically ef®cient since the ®rm's expected future earnings of 6.5 are less than the liquidation value of its assets. However managers and equity wish to avoid liquidation since it wipes out the value of equity. Because the ®rm must repay debt of 2 immediately and has no cash, it must ®nd a lender who will lend it 2 in order to avoid liquidation. Suppose a new lender is considering making the loan, which could rank either higher or lower in priority than the old loan of 5. If the new loan is made, the ®rm will continue to operate for one more period and then will be liquidated. The combined return to the coalition of equity and the new lender if the new loan ranks lower than the old is .5(5 � 5) + .5(8 � 5) � 2 = �.5; while the coalition's return if the new loan ranks higher than the old is .5(2) + .5(3) �2 = .5. (In the latter case, the coalition receives the ®rst 2 of earnings and also receives the last 1 of earnings if the ®rm earns 8.) Thus the coalition

Michelle J. White 189

prefers liquidation if the new loan would rank below the old but prefers continuation if the new loan would rank above the old. Since liquidation is the economically ef®cient outcome, the result is that continuation occurs too often. The economically inef®cient decision for the ®rm to continue operating in this example results from a combination of the coalition bene®ting by shifting priority so that the last lender ranks above the earlier lender and the coalition bene®ting by using the earlier lender's funds to gamble on the risky investment of continuing the ®rm's operations. There are many ways for managers of failing ®rms to shift priority so that late lenders take priority; examples include giving the late lender a security interest in some asset of the ®rm not already subject to a creditor's lien or the ®rm ®ling to reorganise under Chapter 11 and giving the later lender superpriority as a post-bankruptcy administrative expense. Equity's incentive for the ®rm to engage in risky investments because equity bene®ts from the up-side risk while creditors bear the down-side risk applies also to the risky investment of continuing the ®rm's operations. The riskier the ®rm's future earnings, the more strongly the coalition prefers that the ®rm continue operating, regardless of whether doing so is economically ef®cient. When reorganising in bankruptcy is a possibility, the coalition has an additional incentive to keep the ®rm operating ± that of obtaining partial forgiveness of debt under a Chapter 11 reorganisation plan. In the example just discussed, suppose the earlier lender is willing to accept a reorganisation plan that calls for 50 per cent debt forgiveness. In that case the coalition's expected return from ®ling under Chapter 11 becomes .5(2.5) + .5(5.5) = 4. (Here the ®rst 2 of the ®rm's earnings goes to the new lender, the next 2.5 goes to the old lender, and the equity receives the remainder, so that the coalition receives 2.5 if the ®rm earns 5 and 5.5 if the ®rm earns 8.) The next two sections give reasons why creditors may be willing to agree to reorganisation plans that provide for debt forgiveness. Because managers and equityholders have an incentive to avoid or delay liquidation of failing ®rms even when liquidation is the most economically ef®cient outcome, it should not be surprising that payoff rates to creditors in Chapter 7 are very low. There would be an ef®ciency gain from measures that either reduced managers' incentive to delay ®ling for bankruptcy or provided an incentive for creditors to initiate early, involuntary bankruptcy ®lings. In the United Kingdom, France and Germany, both managers and the ®rm's bank face the possibility of civil or criminal penalties if the ®rm delays ®ling. In addition, creditors may initiate involuntary bankruptcy ®lings. In the United States, there are no

190 Corporate Bankruptcy

penalties for delay and it is dif®cult for creditors to initiate involuntary bankruptcy ®lings. However, managers are encouraged to ®le for bankruptcy voluntarily because Chapter 11 is favorable to their interests, as discussed above. Thus a justi®cation for violating the absolute priority rule and for treating managers favorably in bankruptcy reorganisation is that doing so reduces managers' incentives to waste the ®rm's assets in avoiding or delaying bankruptcy. Bargaining in Chapter 11 When ®rms ®le to reorganise under Chapter 11, creditors and equity must bargain over and eventually adopt a reorganisation plan; otherwise the ®rm is liquidated. The rules of Chapter 11 specify many aspects of the bargaining environment, such as that managers have the exclusive right to propose the initial reorganisation plan, that all classes of creditors and equity must consent to the plan, and that eventually the ®rm must be liquidated if no plan is agreed on. Bebchuk and Chang (1992) propose a model of the Chapter 11 bargaining process which shows how the rules of Chapter 11 affect the division of value of the ®rm between creditors and equity. To illustrate, suppose a ®rm that has just ®led under Chapter 11 has assets worth 10 and debt due immediately of 11. Managers ± representing equity ± have the right to offer the ®rst reorganisation plan and creditors must either accept or reject. If they reject, a second round of bargaining occurs in which either side may offer a reorganisation plan and the other side must accept or reject. If the second plan is not adopted, then suppose the bargaining ends, the ®rm's assets are liquidated, and the absolute priority rule is used to divide the proceeds. (More rounds of bargaining could be incorporated, but would complicate the example unnecessarily.) The ®rm's value is assumed to change over time. Because delay is costly, the value of the ®rm's assets declines by 1 during each round of bargaining. In addition, the value of the ®rm's assets either rises or falls by 3 during each round of bargaining. Thus at the end of round 1, the value of the ®rm's assets will either be 6 or 12 with equal probability. At the end of round 2, the value of the ®rm's assets will be 2 with .25 probability, 8 with .50 probability or 14 with .25 probability. In a common knowledge, sequential bargaining game, the party making the offer each period offers the other party an amount equal to what the recipient would get if agreement were reached one period later, which means that the offering party keeps the entire gain from reaching an agreement one period earlier. Given the offer, the recipient is indifferent between accepting the current offer or reaching agreement

Michelle J. White 191

one period later and, therefore, accepts the current offer. Using backwards induction, each period there is no incentive to delay reaching an agreement and, therefore, in equilibrium the parties reach agreement at the earliest opportunity. In the example, suppose creditors made the second round offer. Creditors would offer equity what equity expects to receive if no agreement were reached in the second round and the ®rm's assets were liquidated. In this case equity would receive 3 if the ®rm's assets were worth 14 and zero otherwise, so that equity's expected value in liquidation is .25(3) = .75. Now suppose equity rather than creditors made the second round offer. Equity would propose to keep what it would receive if the ®rm liquidated, or .75, plus equity would keep the savings from making an agreement during round 2, which is 1. Thus in round 2, equity would propose to keep 1.75 and to give creditors the rest. Since both sides have equal probability ex ante of proposing the second reorganisation plan, equity expects to receive 1.25 in round 2. Now consider round 1, where equity has the exclusive right to make the offer. Equity is assumed to offer creditors what they would receive if the parties agreed to a plan in round 2 rather than round 1. Since equity expects to receive 1.25 in round 2, it proposes to keep 1.25 plus the gain from an agreement being reached in round 1 rather than round 2, which is 1. Since the value of the ®rm in round 1 is 10, equity offers creditors 10 � 2.25 = 7.75 in round 1 and creditors have no better alternative than to accept. Thus the outcome is for the parties to agree in round 1 to a reorganisation plan which gives 2.25 to equity and 7.75 to creditors. Now consider how the features of Chapter 11 affect the division of value of the ®rm. The right to propose the plan each period is valuable because the party having the right keeps the value of making an agreement one period earlier. In the example, equity's share of the ®rm's value is 2.25, but it would be 2.75 if equity had the exclusive right to make the offer in both rounds rather than only the ®rst. The increase in equity's share results because equity keeps all ± rather than half ± of the gain from an agreement being made in round 2. A second aspect of Chapter 11 is that equity must agree to the reorganisation plan. Equity has an option on the ®rm and, because the value of the ®rm's assets is uncertain, equityholders have an interest in delaying agreeing to a reorganisation plan in order to play out the option. Creditors must therefore compensate equity for giving up the option by agreeing to a plan. It is well known that the value of options increases as the volatility of the ®rm's earnings increases, so that equity gets more in bargaining over a Chapter 11 plan as the ®rm's earnings become more risky. In addition, the value of the option rises at an

192 Corporate Bankruptcy

increasing rate as the initial value of the ®rm's assets increases. If the initial value of the ®rm's assets in the example is between 0 and 7 (but everything else remains the same), equity always receives 1.50. But if the initial value of the ®rm's assets is between 8 and 13, the value of equity increases by .25 for each unit increase in the initial value of the ®rm's assets. This pattern is consistent with the empirical pattern discussed above in which the payment to equity as a fraction of creditors' claims rose at an increasing rate as the payoff rate to creditors increased. Filtering failure in bankruptcy Some ®nancially distressed ®rms are economically inef®cient and should shut down, since the value of their assets is greater in some other use. An example is a Mexican restaurant in a city where the latest fad is Jamaican food, since the Mexican restaurant's furniture and equipment can easily be moved to a Jamaican restaurant where it would have higher value. There is an ef®ciency gain from liquidating the assets of these ®rms, thus allowing them to move to higher value uses. Other ®nancially distressed ®rms are economically ef®cient and should continue to operate ± at least temporarily ± since their assets have no greater value in any other use. Railroads are an example, since their embankments and rails are costly to move and have higher value if they remain in place and part of a rail network. There is an ef®ciency gain from saving these ®rms. Bankruptcy reorganisation in the United States in fact developed as a means of saving ®nancially distressed railroads whose secured creditors would otherwise have foreclosed on and liquidated particular sections of track (Baird, 1992). Thus there is an ef®ciency justi®cation for having two separate bankruptcy procedures: a liquidation procedure for ®rms that are economically inef®cient and ®nancially distressed and a reorganisation procedure for ®rms that are economically ef®cient but ®nancially distressed. However, while ®nancial distress is easily observable, economic ef®ciency depends on such unobservables as the earnings of the ®rm's assets in their best alternative use. Type I error occurs when ®rms that are ®nancially distressed but economically ef®cient ®rms are liquidated and type II error occurs when economically inef®cient, distressed ®rms are saved. The cost of type I error is loss of ®rm-speci®c human capital and extra transactions costs (since these ®rms are likely to be reopened), while the cost of type II error is that of retaining assets in inef®cient, outmoded uses. Filtering failure occurs when the bankruptcy procedure generates either type of error.

Michelle J. White 193 Figure 7.1 Filtering failure in bankruptcy

•

N

0.6

0.4

Efficient Firm Managers

Inefficient Firm Managers

•

•

Ch.7

Ch. 11 high Ch.11 low

Ch.11 low

•

•

(1, 1)

(3, 3) Creditors

• rej.

•

(2, 3)

• rej.

acc.

•

(4, 1.9)

acc.

•

•

(0.5, 1)

(2, 1.9)

Different countries use different means of assigning ®nancially distressed ®rms to bankruptcy liquidation or reorganisation procedures and, as a result, they have different levels of type I versus type II error. In the United Kingdom and Germany, bankruptcy reorganisations are rare, so that high levels of type I error probably occur. In France, bankruptcy court-appointed of®cials have the responsibility to decide whether or not to save failing ®rms. If these of®cials made unbiased decisions, then type I and type II errors would tend to occur with equal frequency. But since these of®cials' primary goal is to save distressed ®rms, high levels of type II error presumably occur. In the United States, managers have the right to choose between Chapters 7 and 11, which suggests that high levels of type II error are likely to occur.

194 Corporate Bankruptcy

White (1994) uses an asymmetric information game to model whether US bankruptcy procedures lead to ®ltering failure. Figure 7.1 gives an example. A chance event determines whether ®nancially distressed ®rms are economically ef®cient or inef®cient, where the proportion that are economically ef®cient is assumed to be .4. Managers are assumed to know whether their ®rms are ef®cient, but creditors do not. Managers of ef®cient ®rms always ®le under Chapter 11 and they propose a reorganisation plan in the form of a take-it-or-leave-it offer to creditors. There are two types of reorganisation plans: those involving high payments to creditors and those involving low payments to creditors. Managers of ef®cient ®rms choose which type of plan to offer. In contrast, managers of inef®cient ®rms choose between liquidating under Chapter 7 versus ®ling under Chapter 11 and offering low payment reorganisation plans which are identical to those offered by ef®cient ®rms. Creditors are all assumed to be identical. When managers ®le under Chapter 11, creditors must decide whether to accept or reject managers' proposed plans. Creditors always accept high payment plans, but they may either accept or reject low payment plans. Creditors are assumed to know the overall probability that failing ®rms are ef®cient, but not whether individual ®rms are ef®cient or inef®cient ± as indicated by the dotted line in Figure 7.1. If creditors accept low payment plans, then these plans go into effect and the game ends. If creditors reject low payment plans, then I assume that individual ®rms' types are revealed, either because the old managers are replaced by new managers who give creditors more information or because ®rms' assets are sold at auction. Creditors receive more when ®rms turn out to be ef®cient. Payoffs to managers and creditors, respectively, are shown in parentheses in Figure 7.1. Managers of ef®cient ®rms receive 3 with certainty if they offer high payment plans, but they gamble if they offer low payment plans, since they receive 4 if creditors accept, but only 2 if creditors reject. Managers are better off if creditors accept a low payment rather than a high payment plan, since the ®rm pays out less in the former case. But managers are worse off if they propose a low payment plan and creditors reject than if they had proposed a high payment plan. Managers of inef®cient ®rms receive 1 with certainty if they choose Chapter 7, but they gamble if they offer low payment plans, since they receive 2 if creditors accept, but only .5 if creditors reject. Managers of inef®cient ®rms bene®t from ®ling under Chapter 11 if creditors accept low payment plans, since the ®rm bene®ts from debt forgiveness and other subsidies to ®rms in reorganisation and since managers remain in control of the ®rm at least for a period. But they are hurt if they ®le under Chapter 11 and creditors

Michelle J. White 195

reject low payment plans, since managers are then replaced. Remaining outside of bankruptcy and ®ling eventually under Chapter 7 is managers' intermediate outcome, since they remain in control of the ®rm until it eventually shuts down. If managers offer low payment reorganisation plans, creditors receive a sure payoff of 1.9 if they accept. But they gamble if they reject, since they receive 3 or 1 if the ®rm turns out to be ef®cient or inef®cient, respectively. (Although the low payment amount is assumed to equal 1.9 in the example, it is endogenously determined.) The model incorporates many important features of US bankruptcy law. It captures the fact that managers of failing ®rms have the right to choose between ®ling under Chapter 7 or Chapter 11 and that they have the exclusive right to propose the ®rst reorganisation plan. It also captures managers' ability to control information ¯ows about the ®rm during at least the initial period of reorganisation, since creditors are unable to determine the ®rm's type at the time that managers propose the reorganisation plan. But if managers do not propose a plan that creditors are willing to accept, then bankruptcy courts eventually reduce managers' power and this is re¯ected in the model by the fact that managers learn ®rms' types if the reorganisation plan is rejected. The model ignores the fact that ®rms generally have multiple classes of creditors who are treated differently in bankruptcy. The solution to the model in Figure 7.1 is that managers of both ef®cient and inef®cient ®rms always offer low payment reorganisation plans under Chapter 11 and creditors always accept these plans. Creditors' expected return if they accept low payment plans is 1.9, which is higher than their expected return of .4(3) + .6(1) = 1.8 if they reject. Creditors therefore always accept low payment plans. And since creditors always accept these plans, managers always offer them. Managers of inef®cient ®rms receive a payoff of 2 if they offer low payment plans and creditors accept, compared to only 1 if they ®le under Chapter 7. Managers of ef®cient ®rms receive a payoff of 4 if they offer low payment plans and creditors accept, compared to a payoff of only 3 if they offer high payment plans. Therefore all failing ®rms ®le and successfully reorganise under Chapter 11. In this example, there is type II error, since economically inef®cient ®rms reorganise in bankruptcy even though the value of their assets would be higher if they liquidated. This equilibrium has the maximum amount of type II bankruptcy error, because all inef®cient distressed ®rms reorganise under Chapter 11. The inef®cient equilibrium occurs because managers of both types of failing ®rm bene®t when creditors cannot

196 Corporate Bankruptcy

distinguish between them: managers of ef®cient ®rms bene®t because creditors accept a lower payment than they would be willing to accept if they knew that the ®rm were ef®cient; while managers of inef®cient ®rms bene®t because they are able to reorganise under Chapter 11. Creditors are willing to accept low payment reorganisation plans because there are relatively few ef®cient ®rms in bankruptcy, so that their expected return when they reject these plans is low. Other types of equilibrium may also occur in the model, however. Suppose the probability that failing ®rms are ef®cient rises from .4 to .6. Now creditors' expected return if they reject low payment plans is .6(3) + .4(1) = 2.2. Since this is greater than their payoff of 1.9 if they accept, they always reject. As a result, all managers of inef®cient ®rms liquidate under Chapter 7 and all managers of ef®cient ®rms reorganise by offering high payment plans under Chapter 11. This equilibrium represents the best possible outcome in bankruptcy since there is no error of either type. Finally, partial pooling equilibria may also occur in which some but not all inef®cient ®rms succeed in reorganising under Chapter 11. Thus the model shows that the American system of having separate liquidation and reorganisation procedures in bankruptcy and allowing managers to choose between them could lead to either very high or very low levels of type II error in bankruptcy. It is therefore impossible to argue based on theory that the American bankruptcy system is more or less ef®cient than alternative bankruptcy systems. Non-bankruptcy workouts, costly bankruptcy, shirking and underinvestment Because bankruptcy procedures are costly, creditors and managers often attempt to avoid ®ling for bankruptcy by renegotiating the debt of ®nancially distressed ®rms outside of bankruptcy. Such renegotiations ± often called workouts ± usually lead creditors to forgive part of the ®rm's debt. Workouts are common in the United States (see Schwartz (1993), Gilson et al. (1990), Asquith et al. (1994), Gilson and Vetsuypens (1994), Franks and Torous (1994), and Tashjian et al. (1996) for discussion and data) and are probably even more common in countries such as the United Kingdom and Germany, where the incentive to avoid bankruptcy is even greater since reorganisations in bankruptcy are rare. However Hart and Moore (1989) suggest that this ex post incentive for creditors to forgive debt may give managers an incentive to cheat by declaring ®nancial distress even when the ®rm is solvent, since in an asymmetric information context creditors may not be able to determine the ®rm's true ®nancial

Michelle J. White 197 Figure 7.2 Managers' incentive to cheat in bargaining over workouts

•N (1–s)

s solvent

Solvent Firm Managers

insolvent

Firm • Insolvent Managers

• repay in full

d demand workout

demand workout

•

(3, 2.5)

• accept plan

•

(4, 2)

Creditors reject; firm repays in full

•

(2, 2.5)

• accept plan

•

(2, 2)

reject; bankruptcy

•

(1, 0) if bankruptcy

is costly

(1, 1.5) if bankruptcy is efficient

state. This in turn may lead to other inef®ciencies such as managers shirking and creditors underinvesting in the ®rm. Consider the example in Figure 7.2. Suppose in the past that creditors made loans to ®rms. A chance event determines whether each ®rm is solvent or insolvent at the time the loan comes due, where s denotes the probability that ®rms are solvent. Managers know whether their particular ®rms are solvent or insolvent and must decide whether to declare ®nancial distress and demand that creditors agree to a workout versus to repay their debt in full. Managers of insolvent ®rms are assumed always to declare ®nancial distress, but managers of solvent ®rms may either repay creditors in full or imitate managers of insolvent ®rms by declaring ®nancial distress. After managers declare ®nancial distress, they propose a workout in the form of a take-it-or-leave-it demand that creditors forgive part of the ®rm's debt. Creditors are assumed not to know whether individual ®rms are solvent or insolvent, although they know the overall

198 Corporate Bankruptcy

probabilities. They must decide whether to accept or reject the workout proposal. If creditors accept, then the workout goes into effect and the game ends. If creditors reject, then individual ®rms' types are revealed, because creditors declare the ®rm in default and may intervene in the ®rm's management and/or replace the existing managers. If the ®rm turns out to be solvent, then it repays creditors in full. If the ®rm turns out to be insolvent, then it ®les for bankruptcy (in the example it does not matter whether liquidation or reorganisation occurs in bankruptcy). Bankruptcy is assumed to be costly, so that all parties do badly. Figure 7.2 shows managers' and creditors' payoffs, respectively, in parentheses. An important assumption is that managers of solvent ®rms gamble when they declare ®nancial distress and offer a workout, since they receive 4 if creditors accept and 2 if creditors refuse, compared to 3 if they do not demand workouts and instead repay in full. Creditors similarly gamble if they reject managers' workout proposals. Creditors receive a payoff of 2 regardless of the ®rm's type if they accept workout proposals, while if they refuse they receive 2.5 from solvent ®rms and ± because bankruptcy is costly ± 0 from insolvent ®rms. Creditors also receive 2.5 if managers of insolvent ®rms choose to repay in full. Suppose initially that s = .3 and that bankruptcy is costly. In this case the equilibrium outcome is that all managers of solvent and insolvent ®rms declare ®nancial distress and offer workout proposals and creditors always accept. Creditors' certain return of 2 from accepting workout proposals when managers declare ®nancial distress exceeds their expected return of .3(2.5) + .7(0) = .75 when they refuse. Given that creditors always agree to workout proposals, managers of solvent ®rms always declare ®nancial distress, since their return of 4 if they do so exceeds their return of 3 if they do not. Thus when bankruptcy is costly, creditors and managers jointly gain when they avoid it by negotiating a nonbankruptcy workout. But this gives managers an incentive to cheat by declaring ®nancial distress even when their ®rms are solvent. Costly bankruptcy may also distort managers' incentive to expend effort in managing their ®rms and cause them to shirk. Suppose managers choose an effort level at the beginning of the game which affects the probability that their ®rms turn out to be solvent. In particular, suppose when managers use low effort, the probability that their ®rms turn out to be solvent is s = .3, as previously assumed. But when managers instead use high effort, s rises from .3 to .9. Working harder makes managers worse off, however, so that their Figure 7.2 payoffs all fall by .4 when they use high effort. (Creditors' payoffs remain the same.) When managers use high effort, a mixed strategy rather than a pure strategy equilibrium

Michelle J. White 199

prevails. Creditors accept workout proposals with .5 probability, so that managers of solvent ®rms are indifferent between repaying in full and proposing workouts, since their expected return of .5(4) + .5(2) = 3 when they propose workouts is the same as their certain return of 3 when they repay in full. Given that managers of solvent ®rms may either default or repay, creditors use Bayes' Law to determine an updated probability s0 that ®rms are solvent given that default has occurred, where s0 = ds/(ds + 1 � s) and d is the probability that managers of solvent ®rms default and propose workouts. Creditors are indifferent between accepting and rejecting workout proposals if their expected return when they reject, which is 2.5s0 + (1 � s0 )0 is the same as their certain return when they accept, which is 2. These conditions imply that the default rate d by managers of solvent ®rms must be .444. Thus when managers use high effort, a mixed strategy equilibrium occurs in which managers of solvent ®rms default and demand workouts with probability .444 and creditors accept workout proposals with probability .5. Managers of solvent ®rms are much less likely to cheat when they use high effort, since the probability of cheating is .444 under high effort compared to one under low effort. (Note that the model has no separating equilibrium in which managers of solvent ®rms always repay in full. This is because if managers of solvent ®rms always repaid in full, then the only ®rms that defaulted would be insolvent ®rms and creditors would then ®nd it in their interest always to accept workout proposals. But if creditors always accepted workout proposals, then managers of solvent ®rms would choose to default rather than repay.) Now consider whether managers make economically ef®cient decisions when they choose between high versus low effort. (See Table 7.2). Under low effort, managers' expected return is .3(4) + .7(2) = 2.6 and creditors' Table 7.2

Equilibrium outcomes in model of bankruptcy workouts

Bankruptcy is: Probability that creditors accept Probability that managers cheat Creditors' return Managers' return Cost of effort Total return

Low effort s = .3

High effort s = .9

Low effort s = .3

High effort s = .9

costly

costly

ef®cient

ef®cient

1 1 2.00 2.60 0 4.60

.5 .444 2.25 2.85 .40 4.70

1 1 2.00 2.60 0 4.60

.5 .111 2.40 2.85 .40 4.85

200 Corporate Bankruptcy

return is always 2. Together they receive 2.6 + 2 = 4.6. Under high effort, managers' expected return is .9[.444(.5(4) + .5(2)) + (1 � .444)3] + .1[.5(2) + .5(1)] = 2.85 and creditors' expected return is .9[.444(.5(2) + .5(2.5)) + (1 � .444)(2.5)] + .1[.5(2) + .5(0)] = 2.25. Together they receive 2.85 + 2.25 = 5.1, from which the cost of using high effort must be subtracted, so that the net return from using high effort is 5.1 � .4 = 4.7. Since 4.7 > 4.6, it is economically worthwhile for managers to use high effort. However, managers' expected return when they use low effort, 2.6, exceeds their net return of 2.85 � .4 = 2.45 when they use high effort. Thus managers use low effort even when high effort is economically ef®cient. Managers' incentive to use high effort is reduced because the change in the type of equilibrium allows creditors to capture some of the return from managers' extra effort. Because managers prefer the always cheat/use low effort equilibrium to the sometimes cheat/use high effort equilibrium, investment incentives are also distorted. Creditors' expected payoff in the low effort equilibrium is 2, while their expected payoff in the high effort equilibrium is 2.25. Assuming that the loan market is competitive, lenders who anticipate that the low effort equilibrium will prevail will be willing to make loans to the ®rm only if their opportunity cost of funds is less than 2. But if lenders anticipated that the high effort equilibrium prevailed, then they would be willing to make loans if their opportunity cost of funds were as high as 2.25. Assuming that lenders correctly anticipate managers' behaviour, they will refuse to lend if their opportunity cost of funds is between 2 and 2.25. Thus cheating by managers may cause underinvestment. Now suppose the bankruptcy procedure becomes more ef®cient, so that the assets of ®rms in bankruptcy are worth more. (See Roe (1983), Bebchuk (1988), Aghion et al. (1992), Baird (1993), Adler (1994), and Rasmussen (1994) for discussions of reform proposals intended to reduce bankruptcy costs.) Since a more ef®cient bankruptcy procedure mainly bene®ts creditors, suppose when insolvent ®rms ®le for bankruptcy, creditors' payoff in Figure 7.2 becomes 1.5 rather than 0, but managers' payoff remains the same. Does making the bankruptcy procedure more ef®cient improve the ef®ciency of managers' incentives? When managers use low effort and s = .3, the equilibrium is still a pure strategy equilibrium in which managers of solvent ®rms always default and creditors always accept workout proposals. Because bankruptcy never occurs, managers' and creditors' expected payoffs remain the same as when the bankruptcy procedure was costly. When managers use high effort and s = .9, the equilibrium is again a mixed strategy equilibrium. Creditors still accept workout proposals with probability .5, but solvent ®rm managers'

Michelle J. White 201

probability of defaulting and demanding workouts falls from .444 to .111. Managers' expected return under high effort remains the same as when the bankruptcy procedure was costly, 2.85, but creditors' expected return under high effort rises from 2.25 to 2.40. Because managers' expected return does not change when the bankruptcy procedure becomes more ef®cient, managers still prefer to use low effort even though high effort is economically ef®cient. But the probability that managers of solvent ®rms default falls from .444 to .111, so that less cheating occurs. And because creditors' expected return increases from 2.25 to 2.40, creditors are more willing to lend to the ®rm and the distortion to investment incentives is smaller. Thus when bankruptcy procedures are costly, managers and creditors have an incentive to renegotiate loans ex post when the ®rm is in ®nancial distress and this gives managers an incentive to use low effort even when high effort is economically ef®cient. Costly bankruptcy also gives managers of solvent ®rms a stronger incentive to cheat by declaring ®nancial distress even when their ®rms are solvent and it gives creditors an incentive to reduce the amount that they lend to the ®rm. Note that the same points would hold if negotiations over a reorganisation plan in Chapter 11 were substituted for the workout negotiation discussed here and if costly bankruptcy liquidation under Chapter 7 were substituted for the generic costly bankruptcy procedure discussed here. See Bebchuk (1991), Aghion and Bolton (1992), Schwartz (1993), Rasmussen (1994), È f and von Thadden (1994), and Bolton and Bester (1994), Berglo Scharfstein (1996) for further discussion of these issues, including how managers' incentives change when the ®rm has multiple creditors. See Schwartz (1996) for discussion of the possibility that `bankruptcy contracts' made when creditors lend to the ®rm could solve these problems and give managers incentives to behave ef®ciently. Note: This chapter is more or less a reproduction of the same author's article in The New Palgrave Dictionary of Economics and Law.

References Adler, B. (1994) `Finance's Theoretical Divide and the Proper Role of Insolvency Rules', Southern California Law Review, 67, pp. 1107±150. Aghion, P., O. Hart and J. Moore (1992) `The Economics of Bankruptcy Reform', Journal of Law, Economics, and Organization, 8, pp. 523±46. Aghion, P. and P. Bolton (1992) `An ``Incomplete Contracts'' Approach to Bankruptcy and the Financial Structure of the Firm', Review of Economic Studies, L I X , pp. 473±94.

202 Corporate Bankruptcy Asquith, P., R. Gertner and D. Scharfstein (1994) `Anatomy of Financial Distress: An Examination of Junk Bond Issuers', Quarterly Journal of Economics, 109, 625±58. Baird, D. (1993) `Revisiting Auctions in Chapter 11', Journal of Law and Economics, 36, pp. 633±53. Baird, D. (1992) The Elements of Bankruptcy, Westbury, NY, Foundation Press. Bebchuk, L.A. and H.F. Chang (1992) `Bargaining and the Division of Value in Corporate Reorganization', Journal of Law, Economics, and Organization, 8, pp. 523±46. Bebchuk, L.A. (1991) `The Effects of Chapter 11 and Debt Renegotiation on Ex Ante Corporate Decisions', Harvard Law School working paper. Bebchuk, L.A. (1988) `A New Method for Corporate Reorganization', Harvard Law Review, 101, pp. 775±804. Berglo È f, E. and E.-L. von Thadden (1994) `Short-term versus Long-term Interests: Capital Structure with Multiple Investors', Quarterly Journal of Economics, 109, pp. 1055±84. Bester, H. (1994) `The Role of Collateral in a Model of Debt Renegotiation', Journal of Money, Credit and Banking, 26, pp. 72±86. Bolton, P. and D.S. Scharfstein (1996) `Optimal Debt Structure and the Number of Creditors', Journal of Political Economy, 104, pp. 1±25. Bulow, J. and J. Shoven (1978) `The Bankruptcy Decision', Bell Journal of Economics, 9, pp. 437±56. Eberhart, A.C., W. Moore and R. Roenfeldt (1990) `Security Pricing and Deviations from the Absolute Priority Rule in Bankruptcy Proceedings', Journal of Finance, 44, pp. 747±69. Franks, J. and W. Torous (1989) `An Empirical Investigation of U.S. Firms in Reorganization', Journal of Finance, X L I V , pp. 747±69. Franks, J. and W. Torous (1994) `A Comparison of Financial Recontracting in Distressed Exchanges and Chapter 11 Reorganizations', Journal of Financial Economics, 35, pp. 347±70. Gertner, R. and D. Scharfstein (1991) `A Theory of Workouts and the Effects of Reorganization Law', Journal of Finance, X L I V , pp. 1189±222. Gilson, S.C. and M.R. Vetsuypens (1994) `Creditor Control in Financially Distressed Firms: Empirical Evidence', Washington University Law Quarterly, 72, pp. 1005±25. Gilson, S.C., K. John and L. Lang (1990) `Troubled Debt Restructurings: An Empirical Study of Private Reorganization of Firms in Default', Journal of Financial Economics, 27, pp. 315±55. Hart, O. and J. Moore (1989) `Default and Renegotiation: A Dynamic Model of Debt', working paper 520, Department of Economics, Massachusetts Institute of Technology, Cambridge, MA. Hotchkiss, E. (1995) `Postbankruptcy Performance and Management Turnover', Journal of Finance, 50, pp. 3±21. Jackson, T.H. (1986) The Logic and Limits of Bankruptcy Law, Cambridge, MA: Harvard University Press. LoPucki, L. (1983) `The Debtor in Full Control: Systems Failure under Chapter 11 of the Bankruptcy Code?', American Bankruptcy Law Journal, 57, pp. 99±126 (part I) and 57, pp. 247±73 (part II). LoPucki, L. and W. Whitford (1990) `Bargaining over Equity's Share in the Bankruptcy Reorganization of Large, Publicly Held Companies', University of Pennsylvania Law Review, 139, pp. 125±96.

Michelle J. White 203 Rasmussen, R. (1994) `The Ex Ante Effects of Bankruptcy Reform on Investment Incentives', Washington University Law Quarterly, 72, pp. 1159±211. Roe, M.J. (1983) `Bankruptcy and Debt: A New Model for Corporate Reorganization', Columbia Law Review, 83, pp. 527±602. Schwartz, A. (1996) `Contracting About Bankruptcy', working Paper, Law School, Yale University, New Haven, CT. Schwartz, A. (1993) `Bankruptcy Workouts and Debt Contracts', Journal of Law and Economics, 36, pp. 595±632. Tashjian, E., R. Lease and J. McConnell (1996) `Prepacks: An Empirical Analysis of Prepackaged Bankruptcies', Journal of Financial Economics, 40, pp. 135±62. Webb, D. (1991) `An economic evaluation of insolvency procedures in the United Kingdom: Does the 1986 Insolvency Act satisfy the creditors' bargain?', Oxford Economic Papers, 43, pp. 139±57. Weiss, L.A. (1990) `Bankruptcy Resolution: Direct Costs and Violation of Priority of Claims', Journal of Financial Economics, 27, pp. 285±14. White, M.J. (1996) `The Costs of Corporate Bankruptcy: A U.S. ± European Comparison', in J. Bhandari and L. Weiss (eds.), Corporate Bankruptcy: Economic and Legal Perspectives, Cambridge: Cambridge University Press. White, M.J. (1994) `Corporate Bankruptcy as a Filtering Device: Chapter 11 Reorganizations and Out-of-Court Debt Restructurings', Journal of Law, Economics, and Organization, 10, pp. 268±95. White, M.J. (1989) `The Corporate Bankruptcy Decision', Journal of Economic Perspectives, 3, pp. 129±51. White, M.J. (1980) `Public Policy Toward Bankruptcy: Me-First and Other Priority Rules', Bell Journal of Economics, 11, pp. 550±64.

8

Executive Option Plans and Incentives to Take Risk in Levered Firms: Equity Value or Firm Value Maximisation?* Gerald T. Garvey and Amin Mawani

Financial leverage does not distort investment decisions if executives are paid to maximise total ®rm value rather than equity value. Existing models of this idea imply that stock-based incentives should be negatively related to ®rm leverage, a prediction that has little empirical support. We show that the risk distortions induced by ®nancial leverage can be overcome without diluting effort incentives by adjusting the exercise price of executive stock options. We also show that the necessary adjustments are similar to the common practice of granting options at-the-money. We then empirically examine the risk incentives of executive stock option plans in a large sample of Canadian ®rms. The evidence consistently supports the hypothesis that executive stock options mitigate the risk-taking incentives of shareholders in levered ®rms.

1

Introduction

The notion that risky debt distorts investment decisions, also known as the `asset substitution problem', is a central feature of modern capital structure research (see Harris and Raviv, 1991). As originally pointed out by Jensen and Meckling (1976), investing in risky projects bene®ts shareholders at the expense of bondholders. But investment decisions are made by managers, not shareholders. Brander and Poitevin (1992) and John and John (1993) show that levered ®rms can nullify the assetsubstitution problem by appropriately de-coupling the manager's and the shareholders' incentives.1 We present a direct test by examining the relationship between capital structure and the risk incentives in CEO stock option contracts. 204

Gerald T. Garvey and Amin Mawani 205

At present, there is little evidence to support the appealing idea that incentive contracts are chosen to induce ®rm value rather than stockprice maximisation.2 Yermack (1995) and Begley and Feltham (1998) test John and John's (1993) hypothesis that ®rms with more outstanding debt or weaker bond covenants should exhibit a weaker linkage between executive pay and stock price performance. Yermack (1995) ®nds no relationship between debt/equity ratios and the extent to which options tie the executive's pay to the stock price. Begley and Feltham (1998) consider restrictive covenants as well as overall leverage ratios, with ambiguous results; restrictive covenants are more frequent when the CEO has a larger share of his ®rm's equity, but there no relationship between covenants and the dollar value of such stockholdings. We show that the above papers overlook an important component of actual option schemes, in a way which understates the role that such contracts can play in overcoming leverage-induced distortions. John and John's (1993) central result is that asset substitution problems in highly levered ®rms are nulli®ed by weakening the linkage between the manager's and the equityholders wealth. We show that executive stock option plans can control the manger's incentive to take risk without diluting her incentives to exert effort or otherwise increase ®rm value, by adjusting the exercise price of the option. The idea becomes immediately apparent if one views levered equity as a call option on the ®rm's underlying assets with a strike price equal to the face value of its debt. The executive's option now has an effective strike price equal to the face value of the debt plus the exercise price of her option. In principle, two ®rms with drastically different leverage ratios could present their executives with identical payoff pro®les by adjusting the strike price of the options. While this scenario is consistent with John and John's (1993) general point that management incentive contracts can mitigate con¯icts of interest between shareholders and bondholders, it does not lead to their prediction that more levered ®rms will provide weaker incentives to their managers. Indeed, it is consistent with Yermack's (1995) ®nding that leverage has no discernible effect on the strength of option-based incentives. To illustrate and preview our results, consider two ®rms in our sample with vastly different ®nancial structures at the end of 1993; Cambridge Shopping Centers which had a debt to total value (book debt plus market equity) ratio of 67% and Moore Business Forms with a ratio of only 3.9%. The standard asset-substitution model would suggest that an executive at Cambridge Shopping Centers would have a far greater incentive to take

206 Executive Option Plans and Risk Incentives in Levered Firms

risk than would an executive at Moore Business Forms. The actual stock options granted to their CEO suggest that if anything the opposite is true. The key is to recognise that shareholders owe only the face value of the debt, while the manager must pay the exercise price of the option as well as his share of the ®rm's debt. Converting debt and equity to a per-share basis, we arrive at a ratio of ®xed claims (debt per share plus exercise price of option) to total value (debt plus equity per share) of over 120% for Moore versus less than 90% for the highly leveraged Cambridge Shopping. The reason is simple; Moore's stock options had a high exercise price and were out-of-the money at the end of 1993 while Cambridge's options were granted in-the-money. This pattern holds for our entire sample: the average executives' claim is far more leveraged than that of his shareholders, and there is a negative cross-sectional relationship between the two. Based on leverage ratios, it appears that executives of highly levered ®rms will be less rather than more disposed to take on excessive risks. When we account for the costs the executive may bear in the event of ®nancial distress (due to an assumed loss of future salary and bonus), we ®nd that the average executive on net is somewhat risk-averse and that this attitude is unrelated to his employer's capital structure. This supports the idea that executives are motivated to maximise total ®rm value rather than just equity value. Our theoretical model begins by showing that ®rst-best investment can be costlessly implemented either by adjusting the exercise price, or by adjusting the number of options granted in the original John and John (1993) setting. In the appendix, we con®rm that the ®rm strictly prefers to adjust the exercise price to overcome leverage effects when we add an optimal capital structure, management wealth constraints, or effort incentive problems to the John and John (1993) model. We then model the standard practice of setting the exercise price equal to the value of equity at the date of the grant. We con®rm that by so doing, a more levered ®rm automatically chooses a lower dollar strike price than a ®rm that uses more equity ®nance. This practice alone substantially attenuates John and John's (1993) predicted relationship between ®rm leverage and CEO pay-for-performance, and helps explain why Yermack (1995) failed to ®nd supporting evidence. Our ®rst set of empirical tests show that shareholder and executive leverage is negatively related over the entire sample, even in two-stage least squares estimates which allow for endogenous capital structure choices. While leverage ratios are useful for calibrating risk-taking incentives, they are extremely crude. We ®rst use our version of the John and John (1993) model to produce an empirical speci®cation of the

Gerald T. Garvey and Amin Mawani 207

relationship between the manager's compensation and equity risk. We use basic option-pricing techniques with amendments developed in Guay (1998) to characterise the manager's dollar gains from taking on riskier projects. These gains depend on the exercise price of the option, but also on the level of risk and the number of shares granted. Our key ®nding is that the manager's incentive to increase equity risk is negatively related to ®rm leverage. This is consistent with the hypothesis that incentive contracts mitigate risk-taking incentives in levered ®rms, because a given change in equity risk implies a larger increase in underlying asset risk as ®rm leverage increases. The result continues to hold in a simultaneous equation setting which allows for endogenous capital structure choices. Finally, we restrict attention to options granted to the CEO and calibrate their net risk-attitudes by weighing the gain in option value from an increase in equity variance against the cost the manager bears from the increased chance of ®nancial distress. With this more complete measure of the manager's risk attitudes, we ®nd even stronger evidence that leverage does not increase managers' inclination to take on risky projects. While all of our empirical results point in the same direction, several caveats remain. We have characterised incentives to take risk in two ways. First, we have included the exercise price of stock options as a part of the ®xed claim faced by the manager. Second, we have used BlackScholes to compute the payoff the manager receives from increased volatility. These procedures overlook some important features of actual executive stock options. Most obviously, we have valued the options as if they were traded and therefore neglect issues of vesting and maturity. While these considerations are important on average, they do not very greatly across our sample and should not bias our cross-sectional tests. A less obvious but equally important caveat is that we only examine options granted in the years 1993±95. We do not capture later option grants, the anticipation of which could certainly affect investment incentives. The practice of granting options with exercise prices in the neighborhood of the prevailing stock price, however, may further support our conclusions. The reasoning is similar to that for the crosssection; a ®rm whose asset values fall will have an automatic increase in ®nancial leverage, but since its stock price will also fall there will be a countervailing reduction in the exercise price of future stock options. Spreading out options over time effectively reduces the sensitivity of the executive's risk incentives to asset value shocks. A more systematic study of the timing and bunching of option grants would, however, clearly be of value.

208 Executive Option Plans and Risk Incentives in Levered Firms

The paper is organised as follows. We ®rst present a modi®ed version of the John and John (1993) model to derive our hypotheses in Section 2. Section 3 presents the empirical evidence on the managers' and shareholders' incentives to take risk. In the appendix we generalise the John and John (1993) model and show that it is optimal for ®rms to use the exercise price of stock options to overcome asset-substitution problems.

2

Option contracts and optimal risk incentives

Option design: ®rm value versus shareholder value maximisation As mentioned in the introduction, this section adopts exactly the assumptions of John and John (1993) except where indicated. A riskneutral manager chooses between a safe project which returns I with probability one and a risky project which returns H > I with probability q and L < I with probability 1 ± q. The manager privately observes q before deciding to invest and the ®rm has an outstanding debt claim of face value F. While the manager is neutral with respect to income risk, we follow John and John and the evidence in Gilson (1989) by assuming that the manager loses her job which she values an amount J if the debt is not paid in full. For reasonable parameter values enumerated below, this assumption makes the manager effectively risk-averse so that she requires some incentive payment to induce her to ever choose the risky project. John and John restrict incentive contracts to be a linear share  of equity, that is, of terminal cash-¯ows above the face value of debt F. Consistent with actual options, we maintain the assumption of linear sharing rules but allow the ®rm to choose the (non-negative) price X which the manager must pay in order to exercise her option. To simplify the notation we follow John and John's assumption that the manager is paid out of gross cash ¯ows. To summarise, the payoffs to a representative shareholder (denoted S0) as a function of project choice can be written:3 8 8 > > < qH ‡ …1 � q†L � F; F  L < I � F; F  I S…saf e; F† ˆ ; S…risky; F† ˆ q…H � F†; H  F > L …1† > > : : 0; F > I 0; F > H As we con®rm below, shareholders are inclined towards excessively risky projects when the face value of debt is in the range (L, H). Since the manager loses J in the event of distress, she will also favor the risky project even if  = 0 if the ®rm has extremely high leverage (F > I). In this case the ®rm will default for sure unless she takes the risky project and no incentive

Gerald T. Garvey and Amin Mawani 209

contract within the allowable class will induce her to do otherwise. We therefore restrict attention to the relevant range I  F < H, in which case we can write the payoffs to the manager as: ( …I � F � X†; F ‡ X  I M…Saf e; F; X† ˆ ; 0; F ‡ X > I 8 > q…H � F � X† � J…1 � q†; F ‡ X  H > M…risky† ˆ > …2† : � J…1 � q†; F ‡ X > H The primary theoretical insight of this paper is can be seen by comparing (1) and (2) when the face value of debt falls in the relevant range I  F < H. Shareholders are tempted to take on the risky project even if the probability of success is low because they secure an expected wealth transfer from bondholders of (1 ± q)(F ± L). The manager's incentives are to some degree aligned with those of bondholders since she also bears a cost in the event the risky project does not pay off. John and John (1993) restrict the exercise price of the executive's option X to zero, but since there is only one choice (whether or not to take the risky project) the shareholders can implement a value-maximising policy by the appropriate choice of . It is apparent from (2) that adjusting the exercise price is an even more direct solution to the incentive problems of leverage. While the ®rm faces a ®xed claim of F, the executive faces a per-share ®xed claim of F + X so that changes in the exercise price can directly undo any incentive effect of leverage. That executive stock options have positive exercise prices which differ across ®rms is not in dispute. The important economic question is whether such options induce behaviour which is closer to shareholder wealth-maximisation taking the face value of debt as given (as assumed in the asset substitution literature), or whether they provide incentives to maximise total ®rm value (which is also consistent with shareholder wealth-maximisation at the time the debt is issued). Proposition 1 characterises option plans which implement these alternative investment policies for all face values of debt in the range I  F < H. Proposition 1: The value-maximising investment policy is to choose the risky project only if q  q*  (I ± L)/(H ± L). The manager will implement this policy if her option plan (X, a) satis®es: X ˆ J =   ‡ L � F

…3†

The asset substitution investment policy, which maximises shareholder wealth taking the face value of debt as given, is to take the risky project only if q  qs } 

210 Executive Option Plans and Risk Incentives in Levered Firms

(I ± F)/(H ± F) < q*. The manager will implement this policy if her option plan (X, ) satis®es: X ˆ J = 

…4†

Proof: At the time the manager chooses the investment project, she knows that total expected returns to the risky project equal qH + (1 ± q)L, which only exceeds the certain return I if q > q*. From (2), the manager's payoff from choosing the risky project is q(H ± F ± X) ±J(1 ± q ) and her payoff from choosing the riskless project is (I ± F). The exercise price in (3) equates these two payoffs when q = q*. For the asset substitution policy, the shareholders' expected return from the risky project from (1) equals q(H-F) which exceeds the certain return from the riskless project I ± F only if q > qs . The exercise price in (4) equates the manager's payoffs to choosing the risky and riskless projects when q = qs . QED By satisfying (3) or (4), shareholders can induce desired investment choices investment for any combination of leverage (F) and managerial incentives () by an appropriate choice of X. If shareholders wish to implement the asset substitution policy, the exercise price X is not adjusted as leverage changes since shareholders favor riskier and riskier investment choices.4 In order to implement the value-maximising investment policy, by contrast, greater leverage requires either a reduction in  as stressed by John and John, or a reduction in S to directly undo the leverage effect. Strictly speaking, the ability to choose S merely adds a redundant degree of freedom since as John and John show, shareholders can also induce value-maximising investment by ®xing S at zero and allowing  to vary.5 In the appendix, we show that the ®rm will want to use both instruments when we introduce managerial wealth constraints, target capital structures, or effort incentive problems. We present two sets of empirical tests of Proposition 1. First, we estimate the relationship between exercise prices and ®nancial leverage as a direct test of the value-maximising contract in (3) versus the asset substitution contract in (4). We then present a more general test of the key implication of (3); that the incentive to take risky investments is unrelated to ®nancial leverage. Before presenting our empirical results, however, we ®rst show that even a stylised model of actual option granting practice to a large degree implements the adjustments envisioned in (3). Investment incentives when options are granted at the money We now assume that the ®rm follows the common practice of setting the exercise price S equal to the value of the equity at the date of the grant. As

Gerald T. Garvey and Amin Mawani 211

before, we assume that the grant is made under conditions of symmetric information, before the manager has observed the realisation of q. We denote the shareholders and managers' common ex ante beliefs about q by the cumulative density function G(q). We can now characterise the exercise price by: ^ ˆ E0 ˆ X

… q^ 0

…I � F†dG…q† ‡

…1 q^

q…H � F†dG…q†;

…5†

where q^ is the critical level of q above which the manager w\ill choose the risky rather than the riskless investment project. A ®rm that sets its exercise price according to (5) automatically makes adjustments in the direction of undoing leverage incentives. Formally, we have: …1 …1 i h i h ^ @E0 @X ˆ ˆ � G…q^ † ‡ qdG…q† ˆ � 1 � G…q†dq @F @F q^ q^

…6†

The ®nal expression in (6) shows that the adjustment is not complete. The reason is that the equity value and thus the exercise price do not fall dollar-for-dollar when the face value of debt is increased because equity holders have the option of abandoning the ®rm when the manager takes the risky project and it provides low returns. In order to ensure optimal investment, the ®rm must now satisfy (3) ^ Granting options at-the-money reduces the with the exercise price X. need to dilute the executive's claim on the ®rm to overcome risk-shifting problems. This can be seen directly by differentiating * with respect to ^ the face value of debt F when the strike price is set at X: 8 9 ^ �:1 ‡ @@FX;

„1 � q G…q†dq @ ˆ ˆ @F …F ‡ X � L†2 …F ‡ X � L†2 

…7†

In (7), the prediction of John and John (1993) that  must be reduced when the ®rm is more highly levered, continues to hold because the exercise price is not fully adjusted. Since the ®rm defaults on its debt only when the risky project is selected and returns only L rather than H, the absolute value of the numerator in (7) is the probability that the debt is not paid in full. For the ®rms that are in Yermack's (1995) or any other sample of large US corporations, this number is less than one by at least an order of magnitude, and varies across ®rms. This fact may explain why little relationship is detected between option incentives and leverage.

212 Executive Option Plans and Risk Incentives in Levered Firms

3

Empirical evidence

The most fundamental implication of the theory is that managers' incentives to take on risky projects should be unrelated to the ®rm's ®nancial structure. Our primary tests rely on an option pricing model to compute the manager's net gain to increasing risk. There are, however, two interrelated problems with this approach. First, as pointed out by Huddart (1994) and Cuny and Jorion (1995), strong assumptions are required to apply option pricing methods to the restricted options granted to executives. Second, our estimates are a non-linear and opaque function of the strike price, the current stock price, the number of shares granted, the variance of the ®rm's past returns, and assumptions about the manager's exercise policy (see Lambert, Larcker and Verrechia (1991) for simulation results). To make our analysis more transparent, we begin with a direct test of (3) versus (4), documenting how the exercise price differs for ®rms of differing leverage. We then turn to the executive's gains to increasing risk, and then attempt to calibrate the executive's loss in the event of ®nancial distress in order to provide a fuller picture of actual risktaking incentives. Under Ontario Securities Law, incentive option awards to senior managers are deemed to be insider trades, details of which must be reported on a timely basis to the Ontario Securities Commission (OSC).6 Information about option awards such as the number of options granted, the exercise price, and the date of grant are published in weekly editions of a trade publication known as the OSC Bulletin. Our sample begins with all 1,662 option awards by 232 ®rms over the period 1993±95. Other ®rmspeci®c variables were retrieved from Compustat. 48 observations of option awards were deleted due to missing data on Compustat, and 11 observations were deleted due to obviously erroneous data in the electronic version of the Insider Reports. The resulting dataset has 1,603 option awards granted by 226 Canadian ®rms over the period 1993±95. Unlike the Standard and Poor's ExecuComp dataset for US companies that is based on SEC disclosure requirements, our database does not exclude out-of-money options. An option award to a manager is considered an insider trade regardless of whether it is at-, in- or out-ofthe-money money. Our data reveals considerable cross-sectional variation: out of the 1,603 option awards, 314 (19.6%) are at-the-money, 753 (47%) are in-the-money, and 536 (33.4%) are out-of-the-money. While most options held for several years will likely be in-the-money, the nondisclosure of out-of-the-money options may introduce a bias in the pay sensitivities in studies such as Guay (1998) (who like us studies the

Gerald T. Garvey and Amin Mawani 213 Table 8.1

Summary statistics for the full sample

Variable

N

Mean

Std Dev

Median

Min

Max

Assets MVE D/(D+MVE) P X E

1,603 1,603 1,603 1,603 1,603 1,603

13,501 16,41.9 0.321 15.68 15.26 0.63

39,584 24,01.2 0.225 11.59 11.68 1.44

759.8 561.5 0.293 13.375 12.69 0.74

0.33 2.838 0.00003 0.10 0.10 0.000013

218,000 137,600 0.875 108 89.325 18.2

Notes: This table presents summary statistics for the 1,603 option awards granted by 226 ®rms over the period 1993=95. MVE is the market value of equity. D/(D + MVE) is the debt to total value ratio. P is the stock price on option grant date, and X is the exercise price of the incentive options granted. Out of the 1,603 option awards, 314 (19.6%) are at-the-money, 753 (47%) are in-the-money, and 536 (33.4%) are out-of-the-money.

sensitivity of option value to risk) and Yermack (1995) (who studies sensitivity to share price) that rely on SEC-mandated proxy statement disclosures. Out-of-the-money options that remain out-of-the-money would not be disclosed, thereby potentially understating the pay sensitivities. On the other hand, small price changes that reclassify outof-the-money options into in-the-money (or vice versa) may overstate the pay sensitivities. The dataset used in this study is free from such potential biases. Table 8.1 presents summary statistics for our sample. Correlations are reported later for our more detailed sample of CEO's only, but are similar for the full sample. The mean (median) total assets of the ®rms is $13.5 (0.76) billion, while the mean (median) leverage measured as debt plus market value of equity over market value of equity is 1.72 (1.45). Option-granting ®rms are generally larger than ®rms that do not grant options to their senior managers, as con®rmed by Klassen and Mawani (1998). Almost all industry sectors are represented in this sample of option-granting ®rms. The mean (median) stock and exercise prices are $15.68 ($13.375) and $15.26 ($12.688), respectively. Out of the 1,603 option awards, 314 (19.6%) are at-the-money, 753 (47%) are in-themoney, and 536 (33.4%) are out-of-the-money. The variable E , explained in detail in the next section, measures the increase in the manager's option value from a unit increase in the variance of equity returns. Its average is approximately $630,000 and it varies widely across ®rms. In the analyses that follow, the 1,603 option awards are not aggregated by ®rm-years, even though the independent variables are largely ®rmspeci®c. In our ®rst set of tests, option grants are considered to be independent observations since within a given year, a ®rm could grant

214 Executive Option Plans and Risk Incentives in Levered Firms

options that are in-, at- or out-of-the-money, or a combination thereof. Since our other variables are measured at the ®rm level, we effectively characterise the average option grant for each ®rm-year. This approach is a valid characterisation insofar as investment decisions re¯ect the incentives of the `representative' executive rather than those of the CEO exclusively. This is defensible since managers other than the CEO could have formal or de facto authority over investment decisions because of special information, expertise, or delegation (Aghion and Tirole, 1997). Since we have different numbers of option grants per ®rm, however, we will inevitably have a more precise estimate for some ®rms than others leading to heteroscedasticity. We use heteroscedasticitity-consistent standard errors in all tests. In addition to heteroscedasticity, there is the additional problem that a single person could have multiple options so their incentive effects cumulate. This is not a problem for our simple tests which use only measures of the leverage of option grants, but we must deal with it in the next section where we attempt to estimate the dollar gain from increasing risk. How exercise prices undo cross-sectional differences in leverage This section tests directly whether exercise prices are determined according to (3) or to (4). Both equations are stated in raw dollar form, while it is clear from Table 8.1 that size varies widely across ®rms. The ®rst column of Table 8.2 reports the results when both X and D are normalised by total ®rm value de®ned as market value of equity plus book value of debt, and results are similar if we use book assets as the scaling factor. Both theories predict that X should be negatively related to the fraction or number of options granted, so we include as our measure of  the number of options in the grant as a fraction of total shares outstanding. Results are almost identical if we use the raw number of shares. We have no direct proxy for J until we restrict attention to the CEO-only sample. Since pay tends to increase with ®rm size we experimented with various size measures but none had any effect on our results. The estimates in the ®rst column of Table 8.2 strongly reject the asset substitution hypothesis in (4). Exercise prices de®nitely adjust to ®rm leverage. Indeed, they do so to an even greater degree than hypothesised in equation (3); the coef®cient on ®nancial leverage exceeds �1 in absolute value at the 99% con®dence level. The fraction of shares granted has the hypothesised negative effect but the coef®cient is not signi®cant at even the 10% level.

Table 8.2 Pooled OLS Estimates of the relationship between managers and shareholders' leverage (absolute t-statistics with heteroscedasticity-consistent standard errors in brackets) Independent Variable

Dependent Variable = (X* number of shares)/(D+E) (OLS)

Dependent Variable = X* number of Shares (OLS)

Dependent Variable = (X* number of shares)/(D+E) (two-stage-least squares, 1,478 observations)

Intercept

1.298 (27.4) �1.314* (12.2) �0.000774*** (1.73)

�70.37 (1.20)

1.205 (17.9) �1.020* (5.20) �0.00000648 (1.10)

Financial Leverage D/(D+E)  (fraction of shares in option grant) Face value of Debt Market `Value' of Equity Adjusted R2

10.9%

0.0702*** (1.70) 1.318* (17.0) 75.9%

10.3%

Note: *, ** and *** indicate two-tailed statistical signi®cance of coef®cients at the 1%, 5% and 10% level.

215

216 Executive Option Plans and Risk Incentives in Levered Firms Table 8.3 Identifying equation estimating leverage = D/(D + MVE) in the twostage-least squares estimates in the ®nal column of Table 8.2. Sales is average dollar sales for 1994±95, Market/Book is (D + MVE)/ASSET, EBIT is average earnings before interest and taxes, 1994±95 Intercept

Market/Book

Sales

EBIT



0.256 (40.2)*

�0.00437 (4.55)*

0.0000401 (14.17)*

�0.0000667 0.00000027 (2.72)* (2.05)**

Adj. R2 25.8%

The second column of Table 8.2 provides some more insight as to the cause of the large negative relationship between exercise prices and face value of debt. We now regress the dollar strike price against market equity and book debt per share. The coef®cient on market equity is greater than one, which implies that equation (5) understates the extent to which ®rms increase strike prices when equity ®nance increases. Interestingly, an increase in debt also increases the exercise price. But if we hold scale (debt plus market value of equity) constant, a substitution of debt for equity ®nance reduces the strike price more than dollar for dollar. This of course replicates the result in the ®rst column that the manager's leverage decreases rather than increases with the ®rm's ®nancial leverage. The ®nal column of Table 8.2 recognises that leverage is not an exogenous variable. We again normalise both X and F by total ®rm value, but now allow for leverage to be endogenously determined using a twostage least squares estimation with leverage determined by size, pro®tability, and market-to-book (see, e.g., Titman and Wessels (1988) on the cross-sectional determinants of ®nancial structure). As Table 8.3 shows, these three variables do a reasonable job of predicting leverage. We also include the fraction of shares in the option grant as a crude test of John and John's (1993) hypothesis that  should fall in leverage. Consistent with Yermack (1995) we ®nd no evidence of this effect. Indeed there is a tendency for more highly levered ®rms to provide a larger and not a smaller fraction of shares in their option grants. We investigate this issue more fully in the next section. Our treatment of executive leverage looks only at the sum D + X. This overlooks the possibility that the executive effectively holds a compound option, i.e., an option on equity which is itself an option. We circumvent these problems in the next section by focusing on the executive's incentive to increase equity risk, i.e., to increase the riskiness of the underlying security on which their option is written. It is also not clear that it is correct to value an executive's option just like the traded options

Gerald T. Garvey and Amin Mawani 217

on levered equity analysed in Toft and Prucyk (1997). Holders of derivatives such as traded options have no decision rights and must take the ®rm's default and other real decisions as given when valuing their securities. This is clearly not the case for executive stock options, who may well decide to keep their ®rm trading even when asset value falls below D in the hope of both avoiding the job loss J and in the hope that their option will pay off in the future. Whether stock option incentives affect the decision to wind up the ®rm is an open question for future research. Here we focus on the incentive to take risk in ®rms which are currently solvent. The above results strongly indicate that option exercise prices reduce the risk incentive effects of ®nancial leverage. The tests are incomplete, however, as we have not accounted for factors such as ®rm volatility which according to option pricing theory should affect the manager's risk incentives. We have also failed to account for the disincentive to take risk due to the manager's losses in the event of ®nancial distress. The rest of this section ®lls in these gaps. Firm leverage and managers' rewards for risk-taking 1

Hypotheses to be tested

While our theory is cast in terms of dollar returns, option pricing models use percentage returns, which has the advantage of removing direct scale effects from the model. To translate the model into returns, note that if the manager takes the riskless project then the ®rm's terminal cash-¯ow is 1 and the variance of returns is zero. If the manager takes on the risky project, expected terminal cash-¯ows are qH + (1 � q)L but could turn out to be either H or L. The implied increase in standard deviation of asset returns can be written as: =

dA  A …risky project† � A …riskless project† ˆ …H � L†‰q…1 � q†Š1 2 ‰qH ‡ …1 � q†LŠ

…8†

The key point of (8) is that the increase in risk re¯ects the relative values of the alternative projects and is not a function of the way the ®rm is ®nanced. If the ®rm wishes to induce the manager to take on the risky rather than the safe project, it must increase her wealth by at least J(1 ± q) to compensate for the reduction in her job security. For concreteness, focus on the contract at q*, the success probability where the manager is just indifferent between the two projects under an optimal incentive contract.7 The discrete version of the manager's marginal incentive to increase risk under the contract in (3) can now be written:

218 Executive Option Plans and Risk Incentives in Levered Firms

d…compensation†=dA  A ˆ J…1 � q †=dA …q †

…9†

If we could directly measure the sensitivity of the manager's option to changes in asset risk, (9) would imply that this increase is unrelated to leverage and positively related to the manager' loss in the event of ®nancial distress. In order to directly compute A , however, we would need a reliable measure of both total ®rm value and the risk of asset returns rather than equity returns. As shown in Giammarino et al. (1989) and Guay (1998), the adjustments required to convert (observed) equity values and risk into (unobserved) asset values and risk are not straightforward and could obscure the key forces driving the results. We instead use equity returns and adjust our hypotheses accordingly. First, we compute E , the sensitivity of the manager's call option value C to equity variance, as E ˆ n

8 9 p >1n…P=S† ‡ T…r ‡ E2 = 2†> @C > > > p  ˆ nP T N 0 > : ; @E TE

…10†

where n is the number of shares granted, P is the stock price, S is the exercise price, T is the time-to-maturity which we ®x at 10 years, r is the risk-free rate, and N' is the standardised normal density. For the risk-free rate we use the monthly average 91-day Treasury Bill rate retrieved from CANSIM We obtain the stock price from the TSE stock price ®le by looking up option grant date obtained from Insider Reports; if the grant date is not a TSE trading date, the previous trading date is used. For  E 2 we calculate the variance of daily returns over the previous 100 trading days, and the multiply by 250 trading days to convert to annual variance. We require at least 50 returns observations for the calculation of the variance of daily returns. The implication of our theory for E follows from substituting E (@ E / @ A ) for A in equation (13) and taking logs to arrive at: 9 9 8 8 > > J…1 � q †> @E > > > > > > > > � 1n: > 1n…E † ˆ 1n: …11† @A ; dA …q † ; The ®rst term on the right-hand side of (11) is the increased expected cost of job loss from taking the risky project. The second term converts observed equity variance to unobserved underlying asset variance. Modigliani and Miller's (1961) theorem 2 implies that the term is always positive, since equity risk is equal to asset risk for an unlevered ®rm and increases in leverage. This immediately leads to the implication that the manager's observed incentive to increase equity risk should decrease in

Gerald T. Garvey and Amin Mawani 219

®rm leverage. The reason is that the manager's incentive to increase asset risk should be unrelated to leverage, but for a more levered ®rm a given increase in asset risk implies a greater increase in equity risk. Thus, if managers' incentives to increase asset risk is unrelated to leverage, we should ®nd a negative relationship between ®rm leverage and the manager's incentive to increase equity risk.8 Another way to see the result is to recognise that shareholders by de®nition have an unlevered position in the ®rm's equity, so their incentive to increase equity risk is unrelated to the amount of debt in the capital structure. Of course, their incentive to increase asset risk increases with ®nancial leverage, and to undo this effect their managers' compensation must become less sensitive to equity risk as leverage increases. 2

Empirical evidence

2.1 Full-sample evidence. Table 8.3 presents our estimates of the relationship between E and leverage, using the functional form in (14) using the full sample. Consistent with our theory, the relationship is always negative and signi®cant. The intercept is signi®cantly different from zero, which is expected since leverage only captures the second term in (11). The leverage results become somewhat stronger, and the intercept falls, when we control for size. The positive and signi®cant effect of size on risk incentives is also consistent with (14) if J, managers' loss in the event of distress, increases in size. We construct a direct measure of the manager's loss in ®nancial distress in the next subsection. The ®nal equation in Table 8.3 shows that our results continue to hold when we allow for leverage to be endogenously determined using a two-stage least squares approach with leverage determined by size, pro®tability, and market-to-book (see, e.g., Titman and Wessels, 1988) on the crosssectional determinants of ®nancial structure). As Table 8.4 shows, these three variables do a reasonable job of predicting leverage, and none of them are signi®cantly related to E on their own. 2.2 Evidence for CEO's only. For our next set of tests, we assume that the largest option-grant in each ®rm-year is awarded to the CEO, resulting in a sample of 260 ®rm-years.9 We then compute the derivative of the option values with respect to its volatility, or E , as described in Guay (1998, pp. 28±29) or Hull (1989, pp. 200±201). The volatility, price of the underlying stock, time to maturity and risk-free rate are estimated as at the end of the 1995 ®scal year (rather than at the end of each ®scal year) for all 260 observations, since we wish to examine CEOs' risk incentives at the

220 Executive Option Plans and Risk Incentives in Levered Firms Table 8.4 Pooled Estimates of the relationship between the manager's incentive to increase equity return variance (E ) and leverage. The dependent variable in each model is log(E ). Models 1 and 2 are OLS estimates with 1,603 observations and Model 3 is a two-stage-least squares estimate with 1,478 observations and leverage determined by pro®tability, market/book, and total sales Variable Predicted Sign Model 1 Model 2 Model 3

Intercept

�1.58 (35.8)** �2.22 (�12.9)** 12.47 (116)**

log (Leverage)

log (ASSETS)

(�)

(+)

�0.166 (�4.94)** �0.310 (�6.16)** �0.551 (2.565)*

adj R2

0.0125 0.104 (3.79)**

0.0205 0.0117

Note: ** (*) indicates two-tailed statistical signi®cance of coef®cients at the 1% (5%) level.

end of ®scal 1995.10 Only the exercise price is ®rm-year speci®c. This allows us to aggregate the 260 ®rm-year E into 121 ®rm E measuring CEOs' risk incentives at the end of ®scal 1995.11 Vesting provisions prevent option grantees from exercising any options during the ®rst three years, thereby ensuring that options granted in 1993 and 1994 remain unexercised at the end of 1995. Thus we have the measure of risk incentives faced by 121 CEOs at the end of 1995 given the attributes of their outstanding options. This risk incentive re¯ects the increase in value of the CEOs' options from increasing equity risk, or E . In order to characterise the CEOs' risk-taking incentives more fully, we need to incorporate the value they attach to losing their job in the event of ®rm default. Annual cash compensation (or salary) data for the CEOs was collected from Micromedia's Disclosure database for the 260 ®rmyears, and average annual CEO cash compensation computed for the 121 ®rms for which aggregated E as at the end of 1995 were estimated. We also estimate the marginal increase in the probability of default resulting from an increase in equity variance. Since default occurs when D > A, the probability of default can be denoted by F(D ± A), where F is the cumulative normal distribution. The standard deviation of assets is expressed in dollar terms to compute a unique sensitivity for each ®rm. The resulting derivative is then multiplied by the market value of equity to make this Jobrisk measure consistent with our earlier E measure. Jobrisk, or (1 ± q)/d E , estimates the increase in the probability of default due to an increase in equity risk, and provides a complete characterisation of CEOs' risk-taking incentives. The CEO's average annual salary and Jobrisk

Gerald T. Garvey and Amin Mawani 221 Table 8.5

Summary statistics for the CEO-only sample

Variable

N

Mean

Std Dev

Median

Min

Max

Salary Jobrisk E Netrisk Mkt/Book Pro®t Sales

114 114 114 114 114 114 114

0.665 3.17 2.89 �9.05 1.35 0.501 1344

0.731 1.72 3.44 15.8 1.00 1.058 2563

0.540 3.04 1.57 �5.87 1.02 �0.053 553

0.060 �0.130 0.016 �139 0.489 �5.85 0

6.90 8.32 16.1 5.89 6.99 3.10 21400

Note: Salary and E are in millions of dollars. Jobrisk = @F (Debt ± Assets)/@ E where F is the cumulative normal. Jobloss = Salary* Jobrisk/0.2; Netrisk = E ± Jobloss. Average Netrisk = 0 at discount rate of 0.724.

variable are later multiplied to estimate a measure of the CEO's loss in the event of default, or Jobloss. Table 8.5 presents descriptive statistics for our sample of CEO's only, and Table 8.6 presents the simple correlations between our key variables. Canadian ®rms tend to be smaller than their US counterparts, and the average salary of just over $Cd 660,000 is approximately half that of Fortune 500 executives. Salary is positively correlated with size (measured as assets or sales), pro®tability, and leverage. The ®rst two relationships are standard in the literature (e.g., Murphy, 1985), and the correlation between salary and leverage is important if, as we assume, the manager loses the present value of continued salary in the event of ®nancial distress. Our Jobrisk variable takes on negative values in 5 cases, re¯ecting the fact that increases in variance do not always increase the chance of ®nancial distress (e.g., Castanias, 1983). It is, however, generally positive and strongly increases in leverage. Our E variable which captures the manager's dollar gain to increasing risk is far larger in the CEO-only than in the full-sample, re¯ecting the fact that CEO's receive both larger and more frequent option grants. The fact that both salary and Jobrisk are positively correlated with size implies that managers of larger ®rms would be more risk-averse if their compensation was not adjusted accordingly.12 The strong positive correlation between E and size implies just such an adjustment; managers of larger ®rms bear larger expected costs when they increase variance, but also receive larger rewards through their options. We con®rm in our regression tests that size has no discernible overall effect on risk incentives. Table 8.7 presents our primary regression results for the CEO-only sample. Our speci®cation is similar to that for the full-sample, but our

222

Table 8.6

Simple correlations for the CEO-only sample

D + MVE/MVE Assets/MVE Assets E Salary Job risk Sales Pro®ts Mkt/Book

D + MVE/MVE Assets/MVE

Assets

E

Salary

Job risk

Sales

Pro®ts

Mkt/Book

1 .931 .250 .0401 .341 .234 .150 .0027 �.381

1 .379 .194 .517 .926 .105 �.200

1 .0831 .186 .363 .0355 �.0472

1 .244 .164 .289 �.184

1 .453 .329 �.472

1 .118 �.211

1 �.0450

1

1 .255 .0311 .351 .268 .210 .0154 �.505

Gerald T. Garvey and Amin Mawani 223

theoretical expectations are different. The reason is that the variable Jobrisk computes the increased probability of ®nancial distress from an increase in equity variance, just as our E computes the increased value of the manager's option from increasing equity risk. Thus, to be consistent with our measurement approach, (11) needs to be rewritten as: 9 9 9 8 8 8 > > > J…1 � q † @E > @E > J…1 � q †> > > > > > > > � 1n> > > 1n…E † ˆ 1n: ; ˆ 1n> ; :@ > : d dE …q † @A ; A E

…12†

Our variable Jobrisk is a direct measure of the (1 � q))/d E , the extent to which increased equity variance increase the probability of ®nancial distress. Having stated all aspects of the manager's problems in terms of  E , we no longer expect leverage to have any effect. It is still true that for a highly levered ®rm an increase in asset variance implies a large increase in equity variance. But we have accounted for this effect since we also measure the manager's expected cost of risk-taking in terms of an increase in  E . If we ®nd that E is negatively correlated with leverage in our estimate of (12), we would conclude that the managers of levered ®rms are less and not more inclined to increase asset risk. By stating things in terms of equity variance, we should also ignore the risk incentives due to the executive's stockholdings since equity value should be unaffected by equity variance if the risk is diversi®able. We can also construct a proxy for J, the manager's loss in the event of ®nancial distress. Gilson (1989) and LoPucki and Whitford (1990) ®nd that distress both greatly increases the chance that executives and board members lose their jobs, and also reduces their chance of ®nding similar employment in listed ®rms, we assume that the CEO's dollar loss in the event of distress is related to their average annual salary over the years 1994±95. To convert the prospective salary loss to present value terms, we assume a constant discount factor  which summarises the effects of (1) the probability that the CEO would be dismissed in the event of distress, (2) the annual loss in pay that he would experience, (3) the number of years that the loss would occur, and (4) the CEO's inherent time preference. Since we have no additional information on these factors, we also assume that  is common across ®rms. With these admittedly strong assumptions, we arrive at our empirical speci®cation of (12): 8 9 > J…1 � q †> > > > > ˆ 1n…Salary 1n…E † ˆ 1n: dE ;



Jobrisk = † ‡ " ˆ �1n…† ‡

1n…Jobloss† ‡ "

…13†

224 Executive Option Plans and Risk Incentives in Levered Firms Table 8.7 Estimates of Equations (13) and (14) for the CEO-only sample. Dependent variable is E , sample size = 109 after deleting negative values of Jobrisk Intercept

0.164 (0.716) �0.434 (1.18) 0.228 (0.559)

log Jobloss

log Leverage

+ 0 0.371 �0.137 (2.84)** (0.306) 0.394 (2.69)**

0.0253 (0.054) �0.141 (0.316)

Market/ Book +

0.290 (2.02)*

Assets

log Jobrisk

+

+

log adj. R2 Salary/.2 +

0.000074 (2.02)* 0.333 (1.40)

0.066 0.094

0.402 (1.90)

0.057

Note: ** (*) indicates two-tailed statistical signi®cance of coef®cients at the 1% (5%) level.

The ®rst row of Table 8.7 presents our expected signs for each of the estimated coef®cients. Consistent with (13), we expect leverage to have no effect once the manager's job risk is measured in terms of equity risk. The ®rst set of estimates support all of our predictions save one. Most importantly, Jobloss has a positive and signi®cant effect on E while leverage ceases to have any effect. The prediction that fails to be supported by the data is that the coef®cient on Jobloss is not only positive but equal to one. The coef®cient is closer to one-third and we can reject the hypothesis that it equals one at the 1% level. There are two possible interpretations. First, it could simply re¯ect of measurement error, which is especially plausible since we have not accounted for any cross-sectional variation in the executive's vulnerability in ®nancial distress. If the true coef®cient is less than one, it implies that stock option plans do not entirely overcome differences in managerial risk-aversion due to the threat of ®nancial distress. This is the opposite of the standard asset substitution problem; managers of more highly levered ®rms are more and not less averse to risk. This follows from the strong positive correlation between both elements of Jobloss and ®rm leverage (all correlations above 30%), and the weak correlation between E and leverage. As mentioned earlier, however, the relatively small coef®cient could also re¯ect measurement error. A similar interpretation attaches to our estimate of the intercept term, � ln(), which implies a discount factor of nearly 70%. The original equation (11) which was the source of our empirical speci®cation required the manager to be effectively neutral with respect to risk; the increase in option value had to compensate for the increased prospect of job loss. If the average CEO's actual discount factor  is less than 70%, he will be averse to risk since he will weigh job loss more

Gerald T. Garvey and Amin Mawani 225

highly than an increase in option value from a mean-preserving increase in variance. Unfortunately, we have little a priori guidance for the size of  since it depends on the executive's time horizon, the probability of departure due to ®nancial distress, and outside job opportunities after such a departure. The next regression simply adds two additional controls to our speci®cation; size measured as assets and market-to-book. We specify size in linear form solely because it provides a better overall empirical ®t. Size could be expected to increase E for many reasons, the most obvious being that it captures elements of ®nancial distress costs J that we are not captured in the executive's annual salary. Market-to-book is also expected to increase E , because ®rms with more growth opportunities have more need to use stock-based compensation (e.g., Bizjak, Brickley and Coles, 1993). Also, Guay (1998) ®nds this effect in a sample of large US ®rms. The empirical results are consistent with the above expectations, and the coef®cient on Jobloss increases somewhat. Once again, leverage has no effect on risk incentives. The ®nal row of Table 8.7 recognises that equation (13) can be decomposed to isolate the effects of (1) the manager's loss in the event of ®nancial distress and (2) the effect of increased equity risk on the prospect of ®nancial distress: 1n…E † ˆ �1n…† ‡ 1n…Jobloss† ‡ " ˆ �1n…† ‡ 1n…Jobrisk†‡ 1n…Salary† ‡ "

…14†

The results are what we would expect given that the two components of Jobloss are positively correlated with one another. Each of them has a positive effect on E but the independent effects are weaker than that of the combined variable. In no case does leverage have any discernible effect.

4

Conclusion

Recently, researchers have come to recognise that many of the alleged distortions induced by ®nancial leverage rest on the factually incorrect premise that managers' interests are identical with those of shareholders. This paper studies the main device that is currently used to re-align these interests, namely executive stock options. Our primary theoretical point is that the leverage of the executive's claim on the ®rm depends on the exercise price of the option as well as the ®rm's ®nancial structure, and the common practice of writing options at-the-money undoes much of the relationship between the executive's and the shareholders' leverage. We

226 Executive Option Plans and Risk Incentives in Levered Firms

then showed that the incentives provided by stock option contracts in a large sample of Canadian ®rms do not suggest the presence of an assetsubstitution problem. Executives' incentives to take on risky projects varies widely across ®rms, but do not increase with the ®nancial leverage of their employer. A limitation of our research is that is almost exclusively cross-sectional. We have shown that the managers of highly leveraged ®rms do not in general have more incentives to take risk. The asset substitution problem could re-emerge in our setting, if ®rm value or leverage changes dramatically after the exercise price of the stock options are set. If the options are not appropriately re-priced then their risk characteristics would change dramatically. The practice of frequently granting new options with exercise prices in the neighbourhood of the prevailing stock price adds a degree of ¯exibility, but if grants are `lumpy' then they could be sensitive to major changes in ®rm value or leverage. We have captured some of these dynamic issues in our ®nal set of tests by cumulating stock options granted between 1993 and 1995 and evaluating their risksensitivity at the end of 1995. Future research into the dynamics of option granting, exercise, and repricing would clearly be of value. For researchers in contract theory and corporate governance, our results support the basic contention that managers interests differ from those of shareholders, and in so doing tend to reduce con¯icts of interest between various corporate stakeholders (see, e.g., Aoki, 1988; Garvey and Swan, 1994). For researchers in corporate ®nance, our results imply that ®nancial leverage is a poor measure for the ®rm's attitude towards risk. For capital budgeting and valuation applications, our measures support simple use of cash-¯ows to equity without the need to correct for the option attributes of stock. The reason is that real decisions need not be affected by leverage.

Appendix Option design with managerial wealth constraints and a target capital structure Here we study the simplest case where the ®rm wishes to support a speci®c optimal debt/equity ratio and still implement ®rst-best investment without overpaying the manager. We show that this provides a rationale for increasing the strike price well above zero. Denote the face value of debt at the ®rm's value-maximising capital structure by F*, and evaluate all relevant variables at this optimal face value. Given the tax bene®t of debt, the ®rm will never choose F < L so we restrict attention to the case of risky debt.

Gerald T. Garvey and Amin Mawani 227 While incentive contracts may induce optimal investment, it may require an unrealistically large up-front investment by the CEO or, alternatively, require the shareholders to give away some of the ®rm's value to the CEO. To formalise this idea while retaining the simplifying assumption of universal risk-neutrality, we now impose a wealth constraint in the form of a maximum level W which the CEO is able to commit to the purchase of options or shares in her ®rm. We assume that such purchases take place before q is revealed. If the investment decision to be implemented is to take the risky project whenever q > q*, and the exercise price S is ®xed at zero, we can write the value of the manager's claim on the ®rm as: "…  # … q

M…† ˆ  @M ˆ @

"…  q 0

0

…I � F†dG…q† ‡

…I � F†dG…q† ‡

…1 q

1

q

q…H � F†dG…q† #

q…H � F†dG…q† > 0

…A1†

The ®rm can implement ®rst-best investment, without giving away value to the executive, only so long as M(*)  W where * = J/(F* � L) as in (2). There is no guarantee that this condition will be satis®ed, and it is less likely to be satis®ed as the optimal debt level F* is reduced. However, if the strike price S is allowed to rise above zero, the condition M(*)  W can always be satis®ed. To see that this is the case, use the fact that * = J/(F* + X � L) to re-write the critical condition as: "… # … F  ‡ X < I; M… ; X† ˆ  …X†

q

0

…I � F  � X†dG…q† ‡

1

q

q…H � F  � X†dG…q†  W )

W…F  ‡ X � L† „1  IdG…q† ‡ q q…H � F  � X†dG…q† …  1  F  ‡ X  I; M… ; X† ˆ  …X† q…H � F  � X†dG…q†  W ) J  Y  „

q 0

q

J  Y  „1

q

W…F  ‡ X � L†

 q…H � F  � X†dG…q†

…A2†

Since the denominator of Y strictly decreases in X and the numerator increases without bound in X, there is always a value of X which satis®es (A2). Moreover, if we denote by X* the strike price which satis®es (A2) with equality for a given F*, it is straightforward to show that X* strictly decreases in the debt level. The intuition is simply that (A2) de®nes an optimal degree of leverage for the manager's contract. This leverage is a function of both the ®rm's ®nancial leverage and the exercise price on her option plan. If ®nancial leverage is reduced and everything else is held constant, the executive's incentive plan must provide additional leverage by increasing the exercise price.

Designing an option contract to motivate both effort and project selection Here we allow the manager to exert costly effort which increases the probability that the risky investment will pay off. For simplicity, we assume there are only two values of q, qH and qL and the condition

228 Executive Option Plans and Risk Incentives in Levered Firms qH H ‡ …1 � qH †L > I > qL H ‡ …1 � qL †L

…A3†

ensures that it is optimal to invest only when q = qH . If the manager takes no effort, qH occurs with probability pL and qL occurs with probability 1 � pL . If the manager takes high effort, she bears a cost c but the probability of the good investment state increases to pH > pL . Management compensation must now induce the manager to exert effort to increase the chance that the risky investment pays off, and to forgo the risky investment in the event he observes bad news about it (i.e., that it will only pay off with probability qL ). We proceed as follows. We ®rst ®x the exercise price of the option at zero, and show the conditions under which the ®rm is unable to induce both optimal effort and investment. We then show that, when the exercise price is set at zero, higher leverage always reduces the range of parameters for which optimal investment and effort can be achieved. It is in these circumstances that changing the exercise price to undo the leverage of the manager's incentive contract is of value. We begin with the problem of making optimal investments, since the manager's previous effort choice must be conditioned on what he expects to do when the time to invest arrives. The manager will overinvest, taking the risky project even when it will only succeed with probability qL , unless: …I � F†  qL …H � F† � J…I � qL † )   max ˆ

qL H

J…1 � qL † ‡ …1 � qL †F � I

…A4†

We refer to the solution to (A4) as max because the manager will overinvest in the risky project if she receives  above this level. Not surprisingly, her incentives to exert effort increase in  and given the private costs and productivity of effort there is a minimal level of  which will induce effort. If (A6) is satis®ed so that the manager takes the risky project only when q = qH , the value of equity if she does exert effort is pH qH (H � F) + (1 � pH )(I � F) and if she takes low effort the value is pL qH (H � F) + (1 � pL )(I � F). Since the manager also loses J with probability 1 � qH when she takes the risky project, she will exert effort only if: h   i …pH � pL †  qH …H � F† � …I � F† � J…1 � qH †  c )   min ˆ

…pH

c ‡ J…pH � pL †…1 � qH †   � pL † qH ‡ …1 � qH †F � I

…A5†

Combining conditions (A4) and (A6) into the single requirement max  min yields the range of parameter values for which the ®rm can induce both optimal effort and investment. In its most transparent form, the condition can be written:   J…qH � qL †…pH � pL †…H � I†  c qL H ‡ …1 � qL †F � I …A6† Condition (A6) is more likely to be satis®ed if the manager bears a large loss J in the event of ®nancial distress, because this rules out overinvestment even if the manager has a large share of equity. It is also easier to satisfy if effort is very costeffective so that c is small relative to (pH � pL ), because in this case the manager can be motivated to work hard with a small share of equity. Most importantly for our

Gerald T. Garvey and Amin Mawani 229 purposes, (A6) is less likely to be satis®ed as the ®rm's debt burden F grows. In the limiting case where F = I so that the ®rm is barely solvent if the manager takes the riskless project, (A6) reduces to J(qH � qL )(pH � pL )  cqL . There is no natural restriction on the parameters that ensures (A6) is satis®ed in this case.13 The severity of the two incentive problems has been overstated, however, because we have followed John and John in assuming that the ®rm's ®nancial leverage F also determines the leverage of the manager's incentive contract. If as before we allow the ®rm to freely choose the strike price X of the manager's option contract, (A6) is satis®ed with equality so long as X is set equal to X* which can be written: J

X ˆ c

…qH � qL †…pH � pL †…H � I† � …qL H ‡ …1 � qL †F � I† ˆ 1 � qL 8 9 …H � I†:cJ …qH � qL †…pH � pL † � qL ; I �F‡ 1 � qL

…A7†

Notes * 1. 2

3. 4. 5.

6.

Thanks to Patrick Bolton, Jim Brickley, Yukiko Hirao and Yasuhiko Tanigawa for helpful suggestions and comments. Prowse (1990) and Hoshi, Kashyap and Scharfstein (1991) argue that assetsubstitution problems are mitigated in Japanese keiretsu ®rms where major creditors can in¯uence investment policies. Persons (1994) and Garvey (1997) provide conditions under which the incentive contract `undoing' argument fails. Persons allows shareholders to costlessly renegotiate contracts after capital structure is chosen, and Garvey allows the manager to freely trade away her incentive contracts. The empirical relevance of these conditions is no better established than those of the original incentive arguments. While there is also little direct evidence that asset substitution problems are an important determinant of capital structure has proven elusive, there are at least two interpretations. It is certainly possible that such problems are offset by changes to incentive compensation, but it may also re¯ect the dif®culty of empirically capturing the severity of the asset substitution problem (see Titman and Wessels, 1988). We also follow John and John (1993) in de®ning payoffs gross of transfers to the manager which, as they show, is without loss of generality. See Persons (1994) for a rigorous justi®cation of this non-value-maximising policy based on the need to write a contract which shareholders will not renegotiate ex post. For examples of alternative schemes that induce optimal investment, see Haugen and Senbet (1981) on the use of put as well as call options and Harikumar (1996) on changes in the timing and structure of wage and stockbased pay. A majority of the Canadian option-granting large publicly-traded ®rms are listed on the Toronto Stock Exchange, and therefore fall within the jurisdiction of the Ontario Securities Commission. Firms traded on other smaller exchanges in other provinces are regulated by Alberta, British

230 Executive Option Plans and Risk Incentives in Levered Firms

7. 8.

9. 10. 11. 12.

13.

Columbia or Quebec Securities Commissions, and are not included in the scope of this study. In a more general model where the manager has a continuum of projects to choose from, such an indifference relationship will always exist between the optimal contract and the manager's next-most-favourite project. There is another advantage to using equity over asset risk. If we used asset risk, our hypothesis would that there is no relationship, a hypothesis we may fail to reject simply because of errors in variables. We instead look at relatively straightforward incentives to take equity risk, which leads to a more powerful and perhaps counterintuitive hypothesis; incentives to increase equity risk are negatively related to leverage. The 260 ®rm-years are comprised of 78 observations in 1993, 93 in 1994, and 89 in 1995. Vegas of ®rms that grant options in 1993 or 1994, but not in 1995 are also estimated based on the stock price, risk-free rate, volatility and time to maturity at the end of 1995. The 114 ®rms consist of 34 ®rms with data in one year, 35 ®rms with data in two years, and 52 ®rms with data in all three years. The analysis is correct for any model of optimal capital structure, since we consider only contracts which ®x the investment choice at the ®rst-best level. We restrict attention to the face value of debt in order to focus on ex post investment distortions. Since there is a one-to-one mapping between face value of debt and debt-equity ratio, this is without loss of generality. The requirement that effort be socially productive imposes the upper bound (pH � pL )(qH H + (1 � qH )L � I) on the cost of effort c. This does not ensure that (A4) is satis®ed.

References Aghion, P. and J. Tirole (1997) `Formal and Real Authority in Organizations', Journal of Political Economy, 105, pp. 1±29. Aoki, M. (1988) Information, Incentives, and Bargaining in the Japanese Economy, Cambridge: Cambridge University Press. Begley, J. and J. Feltham (1998) `An Empirical Examination of the Relation Between Debt Contracts and Management Incentives', Journal of Accounting and Economics, forthcoming. Bizjak, J., J. Brickley and J. Coles (1993) `Stock-Based Incentive Compensation and Investment Behavior', Journal of Accounting and Economics, 16, pp. 349±72. Brander, J. and M. Poitevin (1992) `Managerial Compensation and the Agency Costs of Debt and Equity', Managerial and Decision Economics, 13, pp. 55±64. Castanias, R. (1983) `Bankruptcy risk and optimal capital structure', Journal of Finance, 38, pp. 1617±35. Chance, D., R. Kumar and R. Todd (1997) `The ``Repricing'' of Executive Stock Options', working paper, Pamplin College of Business, Virginia Tech. Cuny, C. and P. Jorion (1995) `Valuing Executive Stock Options with a Departure Decision', Journal of Accounting and Economics, 20, pp. 193±205. Garvey, G.T. and P.L. Swan (1994) `The Economics of Corporate Governance: Beyond the Marshallian Firm', Journal of Corporate Finance, 1, pp. 139±74.

Gerald T. Garvey and Amin Mawani 231 Garvey G.T. (1997) `Marketable Incentive Contracts and Capital Structure Relevance', Journal of Finance, 52, pp. 353±78. Giammarino, R., E. Schwartz and J. Zechner (1989) `Market Valuation of Bank Assets and Deposit Insurance in Canada', Canadian Journal of Economics, 22,

pp. 109±27.

Gilson, S.C. (1989) `Management Turnover and Financial Fistress', Journal of Financial Economics, 25, pp. 241±62. Guay, W. (1998) `Compensation, Convexity and the Incentives to Manage Risk: An Empirical Analysis', working paper, The Wharton School. Harikumar, T. (1996) `Leverage, Risk-Shifting Incentives, and Stock Based Compensation', Journal of Financial Research, 19, pp. 417±28. Harris, M. and A. Raviv (1991) `The Theory of Capital Structure', Journal of Finance, 46, pp. 297±356. Haugen, R. and L. Senbet (1981) `Resolving the Agency Problems of External Capital Through Options', Journal of Finance, 36, pp. 629±47. Hoshi, T., A. Kashyap and D. Scharfstein (1991) `Corporate Structure, Liquidity, and Investment: Evidence from Japanese panel data', Quarterly Journal of Economics, 106, pp. 33±60. Huddart, S. (1994) `Employee Stock Options', Journal of Accounting and Economics, 18, pp 207±31. Hull, J. (1989) Options, Futures, and other Derivative Securities, Englewood [Cliffs] NJ: Prentice-Hall. Jensen, M.C. and W.H. Meckling (1976) `Theory of the Firm: Managerial Behavior, Agency Costs, and Capital Structure', Journal of Financial Economics, 3,

pp. 305±60.

John, K. and T. John (1993) `Top-Management Compensation and Capital Structure', Journal of Finance, 48, pp. 949±74. Klassen, K. and A. Mawani (1998) `The Impact of Financial and Tax Reporting Incentives on option grants to Canadian CEOs', working paper, University of British Columbia. Lambert, R., D. Larcker and R. Verrechia (1991) `Portfolio Considerations in Valuing Executive Compensation', Journal of Accounting Research, 29, pp. 129±47. LoPucki, L.M. and W. Whitford (1990) `Bargaining Over Equity's Share in the Bankruptcy Reorganisation of Large, Publicly Held Companies', University of Pennsylvania Law Review, 139, pp. 125±96. Modigliani, F. and M.H. Miller (1958) `The Cost of Capital, Corporation Finance, and the Theory of Investment', American Economic Review, 48, pp. 261±97. Murphy, K.J. (1985) `Incentives, Learning and Compensation: A Theoretical and Empirical Investigation of Managerial Labor Contracts', Rand Journal of Economics, 17, pp. 59±76. Persons, J.C. (1994) `Renegotiation and the Impossibility of Optimal Investment', Review of Financial Studies, 7, pp. 419±49. Prowse, S. (1990) `Institutional Investment Patterns and Corporate Behavior in the US and Japan', Journal of Financial Economics, 27, pp. 43±66. Saly, P. J. (1994) `Repricing Executive Stock Options in a Down Market', Journal of Accounting and Economics, 18, pp. 325±56. Titman, S. and R. Wessels (1988) `The Determinants of Capital Structure Choice', Journal of Finance, 43, pp. 1±19.

232 Executive Option Plans and Risk Incentives in Levered Firms Toft, K.B. and B. Prucyk (1997) `Options on Levered Equity, Theory and Tests', Journal of Finance, 52, 1151±80. Yermack, D. (1995) `Do Corporations Award CEO Stock Options Effectively?', Journal of Financial Economics, 39, pp. 237±69. Yermack, D. (1997) `Good Timing: CEO Stock Option Awards and Company News Announcements', Journal of Finance, 52, pp. 449±76.

9

Does the Decision to Retain Retiring Executives on the Board of Directors Help to Control Agency Problems in American and Japanese Firms?* James A. Brickley, Jeffrey L. Coles and James S. Linck

Prospects for promotion provide incentives to lower-level managers in America and Japan. Promotion incentives, however, do not exist for top managers who are at the apex of their ®rms' hierarchies. One little explored mechanism that might provide promotion-like incentives to top managers is the prospect of being retained on the board of directors after retirement (e.g., as chairman of the board). Incentives only exist if the retention decision is based on the top manager's performance during active employment. This paper summarises past and present research on the board retention/performance relation for retiring managers of Japanese and American ®rms. The evidence suggests a positive and statistically signi®cant relation between retention and performance for retiring managers in both countries. This study extends the existing research by providing new evidence on the importance of relativeperformance evaluation in the retention decision for US ®rms. We ®nd that absolute performance is more important than performance relative to the market or industry in explaining the retention decision.

1

Introduction

Most top managers of large corporations own a small fraction of their ®rm's outstanding stock. For example, Kaplan (1994) documents that chief executive of®cers (CEOs) from the 110 largest US companies typically own substantially less than 1 per cent of the outstanding shares (median = .05 per cent). His evidence indicates that the ownership positions are even smaller for the top executives of large Japanese ®rms. 233

234 Retiring Executives and the Control of Agency Problems

This dramatic separation of ownership and control gives rise to potential con¯icts of interest (agency problems) between top managers and residual claimholders. It is often asserted that top managers have strong incentives to use company resources to bene®t themselves at the expense of value maximisation. Previous researchers have examined a variety of external and internal corporate governance mechanisms that mitigate agency problems. Prominent in this research are studies of corporate takeovers, executive compensation and turnover, bank and block ownership, and the board of directors. The decision to retain retiring top executives on the board of directors is another potential, but little explored, mechanism to reduce agency problems. Board retention will provide incentives to managers if two conditions are met. First, managers must value board seats. Second, the retention decision must be based on past performance. This ®rst condition seems likely given the pecuniary and non-pecuniary bene®ts associated with board seats. Surveys indicate that US board members, who are not full-time employees of the ®rm, typically receive about $50,000 in cash compensation annually, while board chairmen are often paid hundreds of thousands of dollars.1 Board members also frequently receive pension plans and various other perquisites and fringe bene®ts. Perhaps as important are the nonpecuniary bene®ts (e.g., prestige and respect) that come from these high-status positions. This paper focuses on evidence related to the second condition: is there a relation between the likelihood of retention and past performance? We begin by summarising existing research on Japanese ®rms. This research documents that a `retiring' president in Japan (typically the most important position in the ®rm) is signi®cantly more likely to be retained as chairman of the board if his ®rm's performance (e.g., stock returns) was good during his last few years of employment. Next, we summarise our recent empirical work on the board retention decision in US ®rms and present new results on whether absolute performance or performance relative to the market/industry is more important in explaining the retention decision.2 Consistent with the evidence from Japanese ®rms, we ®nd a strong positive relation between performance and retention. Overall, the evidence is consistent with the hypothesis that board retention is an important control mechanism in both countries. In contrast to the predictions of standard agency theory, however, we ®nd that absolute performance is more important in explaining the retention decision than relative performance. The paper is organised as follows. Section 2 provides a description of US and Japanese boards of directors; discusses how concerns about post-

James A. Brickley, Jeffrey L. Coles and James S. Linck 235

retirement board service can mitigate agency problems among top managers, and summarises existing research on the retention/performance relation for Japanese ®rms. Section 3 describes our sample and summarises our past research on the retention/performance decision for US ®rms. Section 4 presents new evidence on the importance of relativeperformance evaluation. Section 5 presents the conclusions and discusses possible future research.

2 2.1

Post-retirement board service and agency problems Japanese and American board of directors

At the top of all corporations in Japan and the United States is a board of directors.3 These boards have the right to hire, ®re and compensate top managers. They also have decision rights for major issues, such as large asset sales and mergers. In the United States, board members are elected by shareholders to terms varying between one and three years, depending on the corporation. Typically the chief executive of®cer (CEO) is also chairman of the board (in about 80 per cent of the cases). In most of the remaining cases, the chairman is the former CEO who recently retired. Typically this chairman is not a full-time employee of the ®rm, but works part time helping to set the agenda for board meetings, running board meetings, and monitoring the new CEO. Normally, after a few years, the elderly chairman retires and the CEO assumes the title of Chairman and CEO. The typical US board has about 12 members. About half are outside directors who have no other direct af®liation with the ®rm other than being on the board. Outside directors consist of current or retired executives from other ®rms, academics, retired government of®cials, lawyers, and so on. About one-third of the typical board are inside directors consisting of current or former employees of the ®rm, while the remaining board members are classi®ed as `gray' in that while they are not employees of the company, they have substantial business dealings with the ®rm. In Japan, directors are elected at shareholder meetings to terms not to exceed two years. Typically, the president (shacho) is the highest-ranking, most powerful member. The majority of ®rms also have a chairman (kaicho) who is usually a former president. Typically, the chairman has less power than the president. However, this is not always the case (e.g., Sony and Toyota have had very powerful chairmen). Directors are divided into two classes. Representative directors have the legal right to represent the company, while nonrepresentative directors do not. A large Japanese

236 Retiring Executives and the Control of Agency Problems

®rm's board typically has about 21 directors (three or four are representative directors). They are usually comprised exclusively of employees and former employees, and seldom have `outside' board members (by US de®nitions). Many of the members are lifetime employees of the ®rm. Others come from af®liated companies, banks, and government ministries. 2.2

Board retention and incentives

It has long been recognised that the prospect of job promotion can provide powerful incentives (e.g., see Lazear and Rosen, 1981). Indeed, some scholars have argued that promotion opportunities are the most important incentive device for employees in both American (Baker, Jensen and Murphy, 1988) and Japanese ®rms (Itoh, 1994). A common view is that concerns about promotion are unimportant for a company's top executive because the manager is at the top of the organisation. Thus, signi®cant agency (horizon) problems exist among top managers unless they are offset by (a) contractual incentives, such as stock-based compensation or bonus plans, (b) threat of turnover (Warner, Watts and Wruck, 1988; and Weisbach, 1988), or (c) external control mechanisms, such as the market for corporate control.4 As we have discussed, however, many top managers in America and Japan continue to serve on the board of directors of the company after they step down from the top position. Given the associated compensation and prestige, the prospect of continued board service is likely to provide promotion-like incentives to top managers as long as the retention decision is based on performance during the manager's ®nal years of employment. Existing research has largely overlooked this potential incentive mechanism. 2.3

Past research for Japanese ®rms

Kaplan (1994) provides some evidence on the retention/performance relation for top managers of Japanese ®rms. He de®nes nonstandard turnover as cases where the retiring president either leaves the board completely or when the president is retained on the board but not given the title of chairman.5 His sample consists of 119 large Japanese ®rms over the period 1980±88. He ®nds that non-standard turnover is negatively and signi®cantly related to all performance measures used in his study except for sales growth. For example, he ®nds that presidents of ®rms with negative pretax income are at least 10 per cent more likely to experience non-standard turnover than presidents in other ®rms. He also ®nds that a two-standard-deviation decrease in stock returns (72 per cent) increases

James A. Brickley, Jeffrey L. Coles and James S. Linck 237

the likelihood of non-standard turnover by 8.2 per cent: 3.2 per cent in the same period and 5.0 per cent in the next 2-year period (p. 525).

3

Sample design and past results for US ®rms

In a recent study, we provide a detailed examination of the retention/ performance relation for retiring CEOs in US ®rms. In this section, we describe the sample used in this past research and summarise some of the key results. The subsequent section uses this sample to present new evidence on the importance of relative-performance evaluation in the retention decision. 3.1

Sample selection

We identify retired CEOs by reviewing the Forbes annual executive compensation surveys. Forbes compensation surveys, which cover approximately 800 CEOs annually, include CEOs whose ®rms are ranked among the 500 largest US companies on at least one of the following criteria: sales, pro®ts, assets, and market value of equity. For each survey year from 1991±93, we sort the Forbes data by CEO tenure, identifying those ®rms whose listed CEO has tenure of one year or less. For these ®rms, we examine previous Forbes surveys to identify each ®rm's previous CEO. If the ®rm is not listed in previous surveys, we examine the Dow Jones News Retrieval (DJNR) and/or company proxy statements to identify the previous CEO. Through these procedures, we identify 315 CEOs who potentially retired. For each CEO, we search DJNR to identify the actual departure date and the circumstances surrounding the departure. We eliminate CEO departures around reorganisations where the CEO remained as an executive of a related company, as well as CEO departures around bankruptcies and going-private transactions where public information is not available. Since we are concerned with post-retirement activities of the CEO, we eliminate CEOs whose departure was due to death. Using this methodology, we identify 277 CEOs, representing 257 companies, who left of®ce during 1989±93. As we examine the data, we ®nd that CEOs leave of®ce for at least six reasons: (1) ordinary retirement, (2) death, (3) illness, (4) take a job with another organisation, (5) dismissal, or (6) change in control. Previous research suggests that regardless of the motivation for the turnover, most departing CEOs `retire' from the active work force. Gibbons and Murphy (1992) ®nd that only 36 of 1,631 departing CEOs take jobs as CEOs with other ®rms (2.2 per cent). A small percentage accepts other corporate positions or go to work for law ®rms or universities. The vast majority,

238 Retiring Executives and the Control of Agency Problems

however, leave the active work force. Data from our sample portray a similar picture. We ®nd that only 9 of 277 departing CEOs (3.2 per cent) leave for other positions.6 Given this tendency, we refer to the period after departure as the post-retirement period. Similarly, we refer to CEOs who have left of®ce as retired. In our analysis, we report separate results for three samples of retired CEOs: (1) the full sample; (2) CEOs who depart at age 60 or older; and (3) CEOs who depart between the ages of 64 and 66. The latter two subsamples are analysed separately, because they are more likely to include CEOs taking `normal' retirement. This distinction is useful because we are particularly interested in whether the prospect of postretirement board service helps to control horizon problems in the ®nal years before anticipated retirement.7 Older CEOs, who are a part of the normal succession process, are more likely to know how much longer they will remain CEOs. 3.2

Pre-retirement data

For each CEO, we collect pre-retirement accounting information ®rst from Compustat and, in the few cases where the ®rm was not covered by Compustat, from the Disclosure SEC (D/SEC) database. We acquire preretirement stock returns from the CRSP (NYSE/AMEX/NASDAQ) tapes. For 219 (79.1 per cent) of our sample CEO-®rm observations, the ®rm trades on NYSE, and in the remaining 58 cases (20.9 per cent) the ®rm trades either on AMEX or over the counter. We focus on performance over the pre-retirement period, de®ned as the CEO's tenure as CEO or his last four years in of®ce, whichever is less. The primary reason to stress the most recent four-year performance period is to isolate retention concerns during the CEO's ®nal years when concerns about the external labour and takeover markets are likely to be least important. We measure performance using both accounting data and stock returns. Performance measures include: return on assets (ROA); industry-adjusted ROA; ®rm stock return, abnormal stock return (ABRET); and industry-adjusted stock return. Return on assets is the average annual return on average assets through the ®scal-year end closest to retirement. Industry-adjusted ROA is the yearly average of annual ROA net of the median ROA for all other ®rms (excluding the sample ®rm) on the Compustat tapes with the same two-digit SIC. All stock return measures include dividends and are calculated as the annual compound rate of return. Abnormal stock return is calculated as the ®rm return minus the CRSP value-weighted index. Industry-adjusted stock return is the ®rm

James A. Brickley, Jeffrey L. Coles and James S. Linck 239

stock return minus the median stock return for all other ®rms on the CRSP tapes with the same two-digit SIC (excluding the sample ®rm). Table 9.1 provides descriptive statistics for our sample. The average CEO is 61.2 years old, 71.5 per cent are at least as old as 60, and 31.8 per cent are 64, 65 or 66. The average CEO served as CEO for 9.6 years, and at retirement has worked for the ®rm for 27.5 years. The average CEO owns 1.6 per cent of the ®rm's stock. Mean (median) total asset book value is $13.4 ($4.2) billion, and mean (median) market value of equity is $3.7 ($1.4) billion. Table 9.1 also provides evidence on pre-retirement performance differences across our subsamples. Both accounting and stock performance are highest in the 64±66 subsample and next highest in the subsample of departing CEOs aged 60 and older. Difference-in-mean tests between each subsample and its complement suggest these differences are signi®cant for most of the accounting and market performance measures. That is, CEOs nearing ordinary retirement age appear to have better preretirement performance than other CEOs. This is consistent with the notion that the full sample is likely to contain at least some CEOs who left before normal retirement age because of poor performance.8 3.3

Past ®ndings

Table 9.2 presents mean performance measures for the CEO's in our sample classi®ed by whether the CEO continues to serve on his company's board two years after retirement.9 For the full sample, 137 CEOs continue to serve on their own boards, while 140 do not. All performance measures are signi®cantly higher for the subsample of CEOs who continue to serve on their boards (all p-values  0.01). The average annual return on assets and average annual stock return net of return on the CRSP value-weighted index are 4.3 per cent and �2.6 per cent for the CEOs with continued service, compared to 2.3 per cent and �13.5 per cent for the CEOs who do not continue to serve on their own boards. The differences in performance are fairly large: 2.0 per cent for ROA and 10.9 per cent for stock returns. Similar results hold for the CEOs who are over age 60 (at retirement), as well as for those who retire between the ages of 64 and 66. For both subsamples, the percentage of CEOs who continue to serve on their boards is higher than for the total sample. This result is expected since the full sample is more likely to contain CEOs who are ®red or leave the ®rm to take a job at another ®rm. Presumably, the likelihood of continued board service is lower for a ®red CEO than for a CEO who takes ordinary retirement. Outliers do not drive the results in Table 9.2. The Wilcoxon rank-sum test produces test statistics with similar p-values.

Table 9.1

Sample description: CEO and ®rm characteristics

Variable

Full Sample (N = 277) Mean Median

Age  60 (N = 198, 71.5%) Mean Median

64  Age  66 (N = 88, 31.8%) Mean Median

CEO Age

Tenure as CEO

Tenure with ®rm

Stock Held (% of ®rm)

Stock Held ($ millions)

61.2 9.6 27.5 1.59 30.8

63 8 32 0.18 2.6

64.1 10.7 31.2 1.13 32.5

64 9 36 0.17 2.8

64.4 10.6 31.2 0.77 14.7

64 8 37 0.18 2.7

CEOs' Firms Total Assets ($b)

MV of Equity ($b)

13.4 3.7

4.2 1.3

13.7 4.4

4.6 1.8

13.3 5.9

5.4 2.3

Pre-Retirement Performance (%) ROA pre-retirement perioda ROA year 0 ROA year �1

3.3 2.1 3.8

2.4 2.2 2.8

3.7b 2.4 3.9

3.6 2.8 3.6

4.5d 2.8 4.3

4.4 3.2 4.0

RTN pre-retirement perioda RTN year 0 RTN year �1

5.0 6.6 13.5

7.2 5.5 8.7

8.9d 12.7d 16.9c

11.1 8.7 13.5

11.4d 15.3d 13.5

13.0 9.7 12.2

240

We identify retired CEOs by using the Forbes annual executive compensation surveys. For each survey year from 1991±93, we sort the Forbes data by CEO tenure, and identify those ®rms whose listed CEO has tenure of one year or less. For these ®rms, we examine previous Forbes surveys to identify the ®rm's previous CEO. If the ®rm is not listed in previous surveys, we examine Dow Jones News Retrieval (DJNR) and/or company proxy statements to identify the previous CEO. We eliminate CEO departures around reorganisations where the CEO remained as an executive at a related company, CEO departures around bankruptcies and going-private transactions where public information is not available, and CEOs whose departure was due to death. Our ®nal sample consists of 277 CEOs who left of®ce during 1989±93.

Table 9.1

Sample description: CEO and ®rm characteristics (continued )

Variable ABRET pre-retirement perioda ABRET year 0 ABRET year �1

Full Sample (N = 277) Mean Median

Age  60 (N = 198, 71.5%) Mean Median

64  Age  66 (N = 88, 31.8%) Mean Median

�8.1 �6.7 �2.7

�4.0d �1.0d 1.1d

�1.2d 1.1d �1.1

�5.2 �7.6 �7.3

�1.8 �1.9 �3.6

�0.6 0.1 �4.2

Notes: a ± The pre-retirement period represents 4 years or the CEO's tenure, whichever is less. ROA represents the average annual return on assets; RTN represents average annual stock market performance and ABRET represents the abnormal return (RTN adjusted by contemporaneous return on the CRSP valueweighted market index). b; c; d ± Difference in mean tests (2-tailed; null: equal means; unequal variances) show signi®cant difference between the subsample and its complementary subsample (i.e., at least as old as 60 vs. 59 and under, and 64±66 vs. younger than 64 or older than 66) at the 0.10, 0.05, and 0.01 level, respectively.

241

Performance is measured over the tenure of the CEO or his last four years in of®ce, whichever is less. Return on assets is the average annual return on assets over the period (net income/average assets). Abnormal stock returns are average compound annual returns through the retirement date (return minus the CRSP value-weighted index). Industry-adjusted returns are net of the median industry return for the ®rm's industry (2-digit SIC code) excluding the sample ®rm. We use the standard ANOVA F-test to test for a difference in means and the Wilcoxon rank-sum test for a difference in medians (difference test columns report p-values from these tests). Sample consists of 277 CEOs who left of®ce from 1989±93. Mean

p-value for Difference Test

Median

On board

Off board

On board

Off board

Anova p(F)

Wilcoxon Rank-Sum p(2 )

Full Sample Number of Observations Return on Assets Industry Adj ROA Abnormal Stock Returns Industry Adj Stock Return

137 4.3 1.8 �2.6 6.3

140 2.3 0.2 �13.5 �3.2

137 3.8 0.5 �2.0 6.5

140 1.2 �0.2 �9.9 1.3

0.00 0.01 0.00 0.00

0.00 0.00 0.00 0.00

Over 60 Sample Number of Observations Return on Assets Industry Adj ROA Abnormal Stock Returns Industry Adj Stock Return

113 4.6 2.0 �0.2 8.0

85 2.5 0.4 �9.1 1.0

113 4.0 0.7 �1.1 7.5

85 2.4 0.2 �2.8 4.8

0.01 0.03 0.00 0.01

0.01 0.09 0.03 0.08

242

Table 9.2 Mean accounting and abnormal stock-return performance for CEOs classi®ed by whether they are on their own boards two years after leaving of®ce

Table 9.2 Mean accounting and abnormal stock-return performance for CEOs classi®ed by whether they are on their own boards two years after leaving of®ce (continued ) Age 64±66 Sample Number of Observations Return on Assets Industry Adj ROA Abnormal Stock Returns Industry Adj Stock Return

61 5.3 2.5 3.1 10.2

27 2.8 0.3

�11.0

0.7

61 4.6 0.9 0.7 9.7

27

3.0 �0.2 �6.9 1.3

0.02

0.03 0.00 0.01

0.04 0.09 0.00 0.03

243

244 Retiring Executives and the Control of Agency Problems

Furthermore, using industry-adjusted accounting and market returns leads to the same general conclusion that former CEOs who continue on their own board have signi®cantly better pre-retirement performance. The results reported in Table 9.2 are very robust to alternative speci®cations and controls. For example, we have estimated many logit models where the dependent variable is a dummy equal to one if the CEO stays on the board. Independent variables include stock and accounting performance measures and controls for age, tenure, shareholdings, founder status, ®rm size, industry, calendar year, and so on. Stock price performance, and to a lesser extent accounting performance, are positively and signi®cantly related to the retention decision. Interestingly, the stock returns for the ®nal two years in of®ce are more important in explaining the retention decision than stock returns in earlier years. Overall, the results strongly suggest that the retention decision is a potentially important incentive mechanism for top managers in the United States. Over 18 per cent of the CEOs in our sample are still chairman of the board (COB) two years after departure. Statistical analysis reveals that the performance of CEOs who remain as COB is not signi®cantly different from CEOs who simply remain as ordinary board members. It is plausible that being retained as COB is not so much a reward for good performance as it is part of the ordinary succession pattern used by some ®rms. Some ®rms use a `passing-of-the-baton' succession pattern, where the old CEO becomes chair for a short period before completely retiring and `handing off' the chair title to the new CEO. Some ®rms, however, appoint the new CEO to both titles at the same time (see Brickley, Coles and Jarrell, 1997; Vancil, 1987). These ®rms might not be expected to name a departing CEO as COB even if he performs well, but he might be asked to stay on the board.

4

New results on relative-performance evaluation

Holmstrom (1982) shows that it can be ef®cient to ®lter out common shocks from the evaluation of employee performance. Filtering common shocks provides a more precise measure of the agent's actions/ability and correspondingly lowers the costs of imposing risk on the agent. For example, basing the executive's compensation on performance relative to some benchmark, such as the performance of other companies in similar circumstances, can provide incentives while exposing executives to fewer risk-increasing factors beyond their control. This analysis suggests that market-wide and industry returns (which are positively correlated with

James A. Brickley, Jeffrey L. Coles and James S. Linck 245

the typical ®rm's returns) will be ®ltered in making compensation and retention decisions for top-executive of®cers. At least four papers examine whether CEO pay is based on absolute or relative performance (Antle and Smith, 1986; Barro and Barro, 1990; Gibbons and Murphy, 1990; and Janakiraman, Lambert and Larcker, 1992). The evidence is mixed. Gibbons and Murphy provide the strongest evidence that boards ®lter out market-wide movements in making CEO cash compensation decisions. Somewhat surprisingly, market-wide ®ltering appears more important than the ®ltering of industry performance. Blackwell, Brickley and Weisbach (1994) provide evidence that relative performance evaluation is important in the retention decisions of subsidiary managers in bank-holding companies. Table 9.3 provides evidence on whether the decision to retain CEOs on the board of directors is based on absolute or relative performance using the full sample of 277 observations. Each column contains an estimate of a logit model. In each case, the dependent variable is a dummy variable equal to 1 if the retiring CEO is on the board two years after retirement. Explanatory variables in the ®rst two models are based on stock returns. In the ®rst model, these variables include the company's stock return (unadjusted) and the contemporaneous return for the median ®rm in the same industry (2-digit SIC) over the tenure of the CEO or his last four years in of®ce, whichever is less. The second model replaces the industry return with the return on the CRSP Value-weighted index. The last two models in the table are based on accounting returns (ROA) rather than stock returns. If relative-performance evaluation is being used, we should ®nd a positive coef®cient on own-®rm performance measures and a negative coef®cient on the benchmark performance measure (holding own ®rm performance constant, the manager looks worse if the benchmark return is higher).10 The results do not support the relative-performance hypothesis. Although unadjusted ®rm performance is positive and signi®cant in all models, three of the four benchmark performance measures are positive and none of the four enter signi®cantly. When we estimate the same speci®cations from Table 9.3 for the over 60 and 64±66 subsamples, the results are similar. Table 9.3 presents results controlling for CEO tenure, percentage of stock held by the CEO, ®rm size and regulatory status. We have repeated the analysis eliminating these control variables. We have also included other control variables, such as the ®rms' market-to-book ratio. The results are very robust to these alternative speci®cations. It is curious that we ®nd no evidence of relative-performance evaluation. One possibility is that the performance benchmark provides

246 Retiring Executives and the Control of Agency Problems Table 9.3 Logit models predicting the probability of a CEO being on his own board two years after leaving of®ce as a function of his performance while on the job Performance is measured over the pre-retirement period de®ned as four years or the CEO's tenure, whichever is less. The dependent variable = 1 if the ex-CEO is on his own board 2 years after retiring, and 0 otherwise. Return on assets is net income/ average assets. Stock return is the compound average annual stock return. Industry median returns are returns for the median ®rm in the industry return (2-digit SIC) excluding the sample ®rm. Mean industry ROA is the mean return of all ®rms in the same industry (2-digit SIC) excluding the sample ®rm. The market return is the annualised return on the CRSP VW index. Sample consists of 277 CEOs who left of®ce from 1989±93. Control variables are the CEO's tenure (as CEO), the per cent of the ®rm's stock held by the CEO, the log of assets, and a regulation dummy which = 1 if the ®rm is a utility, bank or insurance company. (p-values reported in parentheses.)

Full Sample (N = 277) N Intercept Stock Returns Median Industry Stock Return Market Return

Dependent variable = 1 if ex-CEO on ®rm's board, 0 otherwise 1 2 3 4 260 1.0 (0.28)

260 1.5 (0.14)

1.9 (0.00) 0.7 (0.64)

2.2 (0.00)

Return on Assets

�3.4 (0.25)

Median Industry ROA Mean Industry ROA CEO Tenure Percentage Stock Held Log of Assets Regulation Dummy Model 2 p-value

0.0 (0.01) 0.0 (0.14) �0.2 (0.09) �0.3 (0.45) (0.00)

0.0 (0.01) �0.1 (0.11) �0.2 (0.09) �0.3 (0.43) (0.00)

260 0.4 (0.66)

260 0.6 (0.52)

4.6 (0.09) 9.8 (0.20)

5.5 (0.04)

0.0 (0.01) �0.1 (0.10) �0.1 (0.21) �0.3 (0.41) (0.00)

0.0 (0.01) �0.1 (0.09) �0.1 (0.20) �0.2 (0.52) (0.00)

2.7 (0.52)

James A. Brickley, Jeffrey L. Coles and James S. Linck 247

little additional information about managerial effort beyond what is already embedded in ®rm performance. Restated, if there is a common shock that is important (and thus should be ®ltered out), CEO/®rm performance and the performance benchmark should be positively and highly correlated. The Pearson correlation coef®cients between CEO/®rm performance and benchmark performance for the four speci®cations in Table 9.3 are as follows: Model 1, 0.239; Model 2, 0.088; Model 3, 0.206; and Model 4, �0.094. The largest correlation is 0.239 and the smallest is negative. Two out of the four are not signi®cant at the 10 per cent level. These results suggest that any common shock is relatively unimportant compared to CEO/®rm-speci®c effects, which helps to explain our results. Nevertheless, ®nding no evidence of relative-performance evaluation is somewhat surprising given the compelling nature of the theory and the fact that two out of four correlations are signi®cant.

5

Conclusions and future research

The evidence suggests a strong positive relation between the likelihood of continued board service for retiring top executives and prior ®rm performance in the United States. Similarly, in Japan there is a signi®cant positive relation between the likelihood of being retained as chairman of the board and prior ®rm performance for `retiring' presidents. Given the pecuniary and non-pecuniary rewards for board service, these ®ndings suggest that the prospect of continued board service provides promotionlike incentives to top managers in both countries. Our evidence suggests that, in contrast to the predictions of standard agency theory, this retention decision is based on absolute, not relative, performance. While existing research sheds light on the factors that affect the board retention decision for retiring executives, more research could be done. For example, it is likely that the bene®ts of keeping a former executive on the board vary depending on the environment of the ®rm. For example, if the top executive has valuable knowledge about investment opportunities, the costs of severing ties with the executive are potentially large. This condition appears more likely in high-growth ®rms than in lowgrowth ®rms. Similarly, the costs of not retaining a former CEO on the board might be large when the CEO has developed valuable relationships with either key customers or government regulators. Examining these issues in more detail represents a fruitful area for future research. Also, while we have provided evidence on the lack of relative-performance evaluation in the United States, there is no corresponding evidence for Japanese ®rms.

248 Retiring Executives and the Control of Agency Problems

Notes * We thank Mitch Rand, Director of Editorial Programming, Forbes, for providing data. Coles thanks the Dean's Council of 100 of the College of Business, Arizona State University, while Brickley thanks the Bradley Policy Research Centre at the University of Rochester for ®nancial support. Helpful comments were provided by the participants at the 36th Biwako Conference. 1. See Byrne (1996). 2. Our past research is reported in `What Happens to CEOs After They Retire? New Evidence on Career Concerns, Horizon Problems, and CEO Incentives', Journal of Financial Economics, 1999. 3. Our description of boards in the US and Japan is based on Brickley, Coles and Jarrell (1997), Kaplan (1994) and Yermack (1996). 4. For example, Bizjak, Brickley and Coles (1993), Gibbons and Murphy (1992) and Smith and Watts (1982) argue that explicit incentives, such as stock-based incentive plans and deferred compensation schemes can be important in offsetting horizon problems among aging top executives since they are less motivated by promotion and career concerns. 5. Retiring presidents typically stay on as chairman of the board. The chairman is a prestigious but typically less powerful position than president. 6. In rare cases, CEOs retire but later return to the workforce. In our sample, C.M. Harper left ConAgra (announced 12 April 1992) but eventually returned to be CEO of Nabisco (announced 27 May 1993), and Stanley C. Gault left Rubbermaid (announced 11 January 1991) and subsequently reversed his retirement to become CEO of Goodyear (announced 5 June 1991). 7. Past work on CEO departures indicates that CEO departures are concentrated among CEOs aged 64 to 66. This is true in our sample as well. In a further attempt to isolate ordinary retirements, we also examine the subsample of cases where the reported reason for departure is retirement. The results are qualitatively the same as those reported in the text. 8. The negative abnormal performance reported for the full sample is expected, since the sample is likely to contain some CEOs who were ®red for poor performance. This feature of our sample does not imply any obvious selection biases in our subsequent empirical analysis, since performance is an independent variable in our models. 9. Throughout, we repeat all empirical procedures based on board seats held three years after retirement. In general, the results are very similar. We also replicate our tests on the subsample of ®rms that are unregulated. We de®ne regulated ®rms as banks, utilities, and insurance companies (per Smith and Watts, 1992). 10. A similar speci®cation is used in Gibbons and Murphy (1990).

References Antle, R. and A. Smith (1986) `An Empirical Investigation of the Relative Performance Evaluation of Corporate Executives', Journal of Accounting Research, 24, pp. 1±39.

James A. Brickley, Jeffrey L. Coles and James S. Linck 249 Baker, G., M. Jensen and K. Murphy (1988) `Competition and Incentives: Practice vs. Theory', Journal of Finance, 43, pp. 593±616. Barro, J. and R. Barro (1990) `Pay, Performance, and Turnover of Bank CEOs', Journal of Labour Economics, 8, pp. 448±481. Bizjak, J., J. Brickley and J. Coles (1993) `Stock-based Incentive Compensation, Asymmetric Information, and Investment Behavior', Journal of Accounting and Economics, 16, pp. 349±72. Blackwell, D., J. Brickley and M. Weisbach (1994) `Accounting Information and Internal Performance Evaluation: Evidence from Texas Banks', Journal of Accounting and Economics, 7, pp. 331±58. Brickley, J., J. Coles and G. Jarrell (1997) `Leadership Structure: Separating the Positions of CEO and Chairman of the Board', Journal of Corporate Finance, 3,

pp. 189±220.

Brickley, J., C. Coles, and J. Linck (1999) `What Happens to CEOs After They Retire? New Evidence on Career Concerns, Horizon Problems and CEO Incentives', Journal of Financial Economics, 52, 341±77. Byrne, J. (1996) `And You Thought CEOs Were Overpaid', Business Week, August 26, p. 34. Gibbons, R. and K. Murphy (1990) `Relative Performance Evaluation for Chief Executive Of®cers', Industrial and Labour Relations Review, 43, pp. 30±51. Gibbons, R. and K. Murphy (1992) `Optimal Incentive Contracts in the Presence of Career Concerns: Theory and Evidence', Journal of Political Economy, 100,

pp. 468±505.

Holmstrom, B. (1982) `Moral Hazard in Teams', Bell Journal of Economics, 18,

pp. 234±40.

Itoh, H. (1994) `Japanese Human Resource Management from the View Point of Incentive Theory', in M. Aoki and R. Dore (eds.), The Japanese Firm, New York: Oxford University Press, pp. 233±64. Janakiraman, S., R. Lambert and D. Larcker (1992) `An Empirical Investigation of the Relative Performance Evaluation Hypothesis', Journal of Accounting Research, 30, pp. 53±69. Kaplan, S. (1994) `Top Executive Rewards and Firm Performance: A Comparison of Japan and the United States', Journal of Political Economy, 102, pp. 510±46. Lazear, E. and S. Rosen (1981) `Rank Order Tournaments as Optimal Labour Contracts', Journal of Political Economy, 89, pp. 841±64. Smith, C. and R. Watts (1982) `Incentive and Tax Effects of U.S. Executive Compensation Plans', Australian Journal of Management, 7, pp. 139±57. Smith, C. and R. Watts (1992) `The Investment Opportunity Set and Corporate Financing, Dividend and Compensation Policies', Journal of Financial Economics, 32, 263±292. Vancil, R. (1987) Passing the Baton: Managing the Process of CEO Succession, Boston, MA: Harvard Business School Press. Warner, J., R. Watts and K. Wruck (1988) `Stock Prices and Top Management Changes', Journal of Financial Economics, 20, pp. 461±492. Weisbach, M. (1988) `Outside Directors and CEO Turnover', Journal of Financial Economics, 20, pp. 431±60. Yermack, D. (1996) `Higher Market Valuation of Companies with a Small Board of Directors', Journal of Financial Economics, 40, pp. 185±211.

This page intentionally left blank

Part III

Capital Market

This page intentionally left blank

10

Viable Design of a Security with a Pre-existing Market*  Kazuhiko Ohashi

1

Introduction

Emergence of many new securities has been changing the world's ®nancial markets for the last three decades of the twentieth century. Now, innovation of new securities is one of the main services that the ®nancial institutions provide. Issuance of hybrid securities and securitisation of assets such as MBS (Mortgage Backed Securities), ABS (Asset Backed Securities) and CMO (Collateralised Mortgage Obligation) are examples of this innovation. One of the primary motives for this ®nancial innovation can be explained by the risk-based clientele (or the spanning) effect, where different investors have different demands for different risks, and hence will pay the securities premiums according to their own hedging needs. Such differences among the investors' valuation on different risks provide the security issuers with pro®table opportunities by carving the payoffs of the underlying assets into pieces to create new securities that will satisfy the different investors' hedging needs. This spanning effect is certainly one of the most important motives in the ®nancial innovation, but it does not seem to capture all parts of the picture. Especially, while the spanning-based explanation predicts the innovations of many securities with different payoffs, it seems that in reality, the issuers manage to create a huge number of securities with similar payoffs. For example, in the MBS and ABS markets, many securities are designed to reduce default and prepayment risks, to obtain higher credit-ratings, and consequently to have payoffs that are similar to those of T-bonds. It is true that this phenomenon can be partly explained for an institutional reason that some investors are regulated not to purchase low-

253

254 Viable Design of a Security with a Pre-existing Market

rated bonds. However, the persistence of the popularity for securities with similar payoffs suggests that there may be another non-institutional factor that plays an important role in determining the payoffs of the created securities. The purpose of this chapter is to propose an economic rationale for this somewhat puzzling phenomenon that is unexplained by the spanning theory. Throughout this chapter, I will consider the following situation. A security pre-exists, and another security is created. The payoff of the preexisting security is exogenously given, about which all investors have the same information. On the other hand, one investor (possibly the issuer) has private information about the payoff of an asset that constitutes a part of the payoff of the created security. This situation can be interpreted to describe the innovation of MBS and ABS. The T-bond markets existed before the creation of MBS and ABS. The value of T-bonds is determined by the interest rate about which no investor has private information. On the other hand, the value of MBS and ABS depends not only on the interest rate but also on the defaults and prepayments that are prone to asymmetric information, possibly between the issuer and the outside investors.1 I call the payoff of the created security viable if an equilibrium exists in the markets of these pre-existing and created securities, and investigate the condition under which the created security has a viable payoff. By focusing on the case where the degree of the asymmetric information between the informed investor and the uninformed investor is large, I show that for an equilibrium to exist, the payoff of the created security should not depend too much on private information. Such a payoff can be attained by reducing the part of the payoff that is subject to the asymmetric information.2 As a result, regardless of the issuer's objective, the viable payoff of the created security becomes similar to that of the pre-existing security about which all investors have the same information. Intuitively, this result can be explained as follows. When a new security is created, the informed investor will try to take advantage of his superior information. He can do it more effectively by trading the pre-existing security with the created security. This is because the former provides the informed investor the way to hedge the risk that is associated with the payoff of the latter about which he has no private information. However, if this hedging is too effective so that the informed investor is able to exploit the uninformed investor, the adverse selection in the markets will signi®cantly worsen, and scare the uninformed investor away from trading the created security. In the worst case, the markets will break

 Kazuhiko Ohashi 255

down. To avoid such a situation, the issuer has to make the payoff of the created security less dependent on the asymmetric information. Consequently, the payoff of the created security becomes similar to that of the pre-existing security about which the investors have symmetric information. Note that this result is also consistent with the observation that many MBS and ABS have the payoffs whose dependence on asymmetric information is small. In other words, this chapter provides an economic rationale of why in reality, asymmetric information does not seem to matter in trading MBS or ABS. Indeed, asymmetric information does matter in designing the payoffs of these securities. In order to be traded, their payoffs should not depend too much on the asymmetric information. Moreover, the existence of another security, such as T-bonds, that provide the informed trader the way to hedge the risk about which he has no private information, facilitates the required reduction of the dependence on the asymmetric information. As a result, the viable payoffs of the created securities cannot depend too much on asymmetric information. I emphasise the importance of the interaction between the created security and the pre-existing security through hedging, and analyse how the correlation between the payoff of the underlying asset and that of the pre-existing security affects the viable design of the payoff of the created security. I also investigate the difference of the viable designs between when there is a pre-existing security and when there is not. This chapter is closely related to DeMarzo and Duf®e (1993, 1995) who investigate securitisation of an asset under asymmetric information between an issuer and investors. Their argument is as follows. When the issuer has private information on the payoff, the security may not sell well because the investors, facing the adverse selection problem, require a discount for their informational disadvantage. To attain a higher expected revenue from the sales, the issuer designs the security to have the payoff less dependent on the private information. Consequently, the issuer with the private information on the asset creates the security backed by a part of the asset's payoff, which is relatively riskless and hence uninformative, and retains the rest of the asset's payoff.3 This chapter shares the idea with DeMarzo and Duf®e in that the issuer designs the payoff of the security to be less dependent on the asymmetric information in order to mitigate the adverse selection. However, it differs from them in analysing the interaction between the pre-existing security and the created security. For example, this chapter shows that the viable design of the created security depends on whether there pre-exists a

256 Viable Design of a Security with a Pre-existing Market

security or not, and on how much the payoffs of the securities are correlated. This point is important given the fact that many new securities are usually created in addition to some pre-existing securities. This chapter is also different from DeMarzo and Duf®e (1993, 1995) in its objective. The former seeks the condition that the viable security design should satisfy, while the latter looks for the optimal design for a particular objective (the expected revenue-maximisation) of the issuer. This difference enables us to obtain the restriction on the viable design of the created security, independently of the objectives of security issuers. That is, the payoff of the created security should be similar enough to that of the preexisting one, regardless of the objectives of the issuers. In this sense, this chapter provides the results complementary to DeMarzo and Duf®e. Demange and Laroque (1995) and Rahi (1996) also investigate the design of one security created by an issuer without any other pre-existing securities in a CARA-Gaussian setting. Again, this chapter is different from theirs in analysing the effect of the pre-existing security to the innovation of a new security, and in investigating the condition on the viable design of the created security rather than on the optimal security design for the issuer's particular objective. Regarding the methodology of the analysis, this chapter owes much to Bhattacharya et al. (1995) who thoroughly investigate the condition under which the market breakdown occurs for a set of securities with given payoffs. In this chapter, I take one step back and start from the payoffs of the underlying assets. Then, I allow the payoff of a security backed by (a part of) these assets to vary, and identify the condition for the created security to have an equilibrium. By doing so, I show that the similarity among the payoffs of many created securities in securitisation can to some extent be explained by the framework of Bhattacharya et al. In this sense, this chapter also reinforces the applicability of the methodology of Bhattacharya et al. to the analysis of ®nancial innovation. One important observation in this chapter is that by trading several securities with correlated payoffs, the informed investor can take advantage of his superior information more effectively than he could without any other securities. Thus, the adverse selection that the uninformed investor faces can be worsened when several securities are available for trading. Dow (1998) indicates this point in analysing the impact of adverse selection on the liquidity of securities and the welfare of  investors in a setting without market-breakdown. Ohashi (1997, 1999) also investigates how the worsening of the adverse selection affects the determination of the number and design of the securities backed by the underlying assets, given the objective of the issuer.

 Kazuhiko Ohashi 257

Finally, although it does not necessarily explain the similarity of the payoffs of the created securities, Gale (1992) is the notable work which explains why created securities are standardised so that many new securities with similar payoff structure are created by the issuers. This chapter is organised as follows. Section 2 presents the model of the innovation of a security when another security pre-exists. Section 3 characterises the condition on the viable design of the payoff of the created security. Section 4 provides some discussions, and Section 5 concludes.

2 2.1

The model The economy

Consider an economy with three dates {0, 1, 2}, and single consumption goods at date 2. There are three underlying risky assets that have independent random payoffs {x1 , x2 , x3 }, respectively, in terms of the goods at date 2.4 Let x = (x1 , x2 , x3 )T . Two types of agents exist: a security issuer and investors. Two types of investors exist: an informed investor i, and an uninformed investor u. The issuer creates a security, and the investors trade the securities. The timing is as follows. At date 0, a security is exogenously given, and the issuer creates another security. At date 1, after investor i learns his private information, both investors trade the securities. At date 2 uncertainty is resolved, the securities pay off, and consumption occurs. No explicit assumption is made on the issuer's objective. However, I assume that the issuer creates a security only when there is an equilibrium in the markets for the pre-existing and the created securities. Informed investor i has the utility function Uu (W) = E[ � exp(� 1/ru W)] over consumption W at date 2, the random endowment of ei = (ei1 , ei2 , ei3 )T units of assets 1, 2 and 3, and a private signal s about the payoff x3 of asset 3. On the other hand, uninformed investor u has the utility function Uu (W) = E[� exp(� 1/ru W)] over consumption W at date 2, the random endowment of eu = (eu1 , eu2 , eu3 )T units of assets 1, 2 and 3, but has no private signal about the payoffs of the assets. Both investors may also be endowed with some securities, which will be speci®ed later. Two securities {a, b} are potentially available in this economy that have random payoffs fa and fb at date 2, respectively. Let F = (fa , fb )T . I focus on the situation where the investors have symmetric information on the payoff of security a, but may have asymmetric information on a part of the payoff of security b. To be more precise, I assume that fa = a1 x1 + a2 x2 and fb = b2 x2 + b3 x3 for some numbers a1 , a2 , b2 and b3 . That is, there is no

258 Viable Design of a Security with a Pre-existing Market

asymmetric information on the payoff fa of security a, but asymmetric information exists on the payoff fb of security b through x3 . Furthermore, the payoffs fa and fb may be correlated through x2 , about which the investors have symmetric information. I investigate the condition under which the markets of these securities have an equilibrium. As mentioned before, one example that this situation describes is the case of a treasury bond (T-bond) and a mortgage backed security (MBS).5 Imagine security a to be a T-bond, and security b to be an MBS.6 While asymmetric information plays a little role in the T-bond market, a part of the payoff (e.g., defaults and prepayments) of the underlying assets by which the MBS is backed may be subject to asymmetric information between the investors. Furthermore, the payoffs of the T-bond and the MBS are correlated through the movement of the interest rate, which presumably contains little asymmetric information. In this case, our question will be under what design (i.e., the combination of the assets to securitise) the MBS market will have an equilibrium together with the given T-bond. Denote by i = (ia , ib )T (resp. u = (ua , ub )T ) the units of securities a and b endowed to investor i (resp. u), and by  = (a , b )T the aggregate initial endowment of the securities. Then, i + u 4  where ia + ua < a (resp. ib + ub < b ) implies that some additional units of security a (resp. b) is supplied by the security issuer. In order to analyse the model in a simple rational expectations equilibrium (REE) framework, I assume the following. First, x, ei and s are joint normally distributed such that xj  N…E‰xj Š; 1†7 ; eij  N…E‰eij Š; V‰eij Š†; euj  N…E‰euj Š; V‰euj Š† …j ˆ 1; 2; 3†;and s N(0, V[s]), and that all these random variables are mutually uncorrelated except x3 and s; 6 0. (Note that for random variables y and z, E[y] denotes the i.e., C[x3 , s] ˆ expectation of y, V[y] denotes the variance of y, and C[y, z] denotes the covariance between y and z.) Second, informed investor i is large and trades the securities strategically by taking account of the effect of his trade on the security prices. Finally, there is a riskless asset, one unit of which pays one unit of the goods at date 2, and the riskless interest rate between date 1 and 2 is zero.8 2.2

Equilibrium and market breakdown

Denote by i = (ia , ib )T investor i's security demand at date 1, and by p(i ) = (pa (i ), pb (i ))T the REE price vector of securities a and b at date 1, respectively. At date 1, informed investor i demands i of the set of securities F so that

 Kazuhiko Ohashi 259

�1 i 2 Arg max E‰� exp… Wi †jei ; sŠ  ri T

s:t: Wi ˆ eTi x ‡ i P…† ‡ ‰F � P…†ŠT :

…1†

On the other hand, uninformed investor u observes P(i ), infers the information that investor i has on the securities' payoffs, and demands u = (ua , ub )T of the securities F so that �1 Wu †jeu ; P…i †Š u 2 Arg max E‰� exp…  ru T

s:t: Wu ˆ eTu x ‡ u P…i † ‡ ‰F � P…i †ŠT :

…2†

The market-clearing condition requires that i ‡ u ˆ :

…3†

For a given set of securities, I say that the security markets break down if there is no equilibrium for that set of securities. In the following, I imagine the situation where security a pre-exists and where security b is created in addition to security a.9 I call the payoff fb viable if the security markets of the pre-existing security a and the created security b with the payoffs F = (fa , fb )T do not break down. I investigate the condition under which the payoff fb of security b is viable.10

3 3.1

Viable design of the created security Equilibrium with a linear pricing function

In the rest of this chapter, I assume that the REE prices P(i ) are linear in investor i's demand i ; i.e., P…i † ˆ  ‡ Ki

…4†

where  is a R2 -vector, and K is a non-singular R22 -matrix.11 Given this assumption, since the utility functions are negative exponential and the random variables are normally distributed, investor i's maximisation problem is re-expressed as max E‰Wi jei ; sŠ � 

T

1 V‰Wi jei ; sŠ 2ri

s; t; Wi ˆ eTi x ‡ i P…† ‡ ‰F � … ‡ K†ŠT :

260 Viable Design of a Security with a Pre-existing Market

The ®rst-order condition is 1 1 E‰FjsŠ �  ‡ KT i � C‰F; xjsŠei � …K ‡ K T ‡ V‰FjsŠ†i ˆ 0: ri ri

…5†

The second-order condition is 1 K ‡ K T ‡ V‰FjsŠ is positive definite: ri

…6†

If both the ®rst-order and second-order conditions are satis®ed, the optimal security demand of investor i is given by 1 1 i ˆ …K ‡ K T ‡ V‰FjsŠ†�1 …KT i �  ‡ E‰FjsŠ � C‰F; xjsŠei †: ri ri

…7†

Investor u infers the information that investor i has by observing the prices P(i ), and solves max E‰Wu jeu ; P…i †Š � 

1 V‰Wu jeu ; P…i †Š 2ru

T

s:t: Wu ˆ eTu x ‡ u P…i †‰F � P…i †ŠT : The ®rst-order condition is E‰FjP…i †Š � P…i † �

1 1 C‰F; xjP…i †Šeu � V‰FjP…i †Šu ˆ 0: ru ru

…8†

Since the second-order condition is always satis®ed, the optimal security demand of investor u is given by u ˆ ru V‰FjP…i †Š�1 fE‰FjP…i †Š � P…i †Š � P…i † �

1 C‰F; xjP…i †Šeu g: ru

…9†

De®ne Q ˆ E‰FjSŠ � 1i C‰F; jSŠei . Note that observing P(i ) is equivalent to observing Q for investor u; i.e., E[FjP(i )] = E[FjQ], V[FjP(i )], = V[FjQ], and C[F, xjP(i )] = C[F, xjQ]. From i's ®rst-order condition, (K + KT + r1i V [Fjs])i +  � KT i = E[Fjs] � r1i C[F, xjs]ei = Q With the market-clearing condition, 1  � i ˆ ru V ‰FjQŠ�1 fE‰FŠ ‡ C‰F; QŠT V ‰QŠ�1 ‰…K ‡ K T ‡ V‰FjsŠ†i ‡  � K T i Š ri 1 � C‰F; QŠT V‰QŠ�1 E‰QŠ � P…i † � C‰F; xjQŠeu g:

ru

 Kazuhiko Ohashi 261

Recall that P(i ) =  + Ki . Matching the coef®cients, in equilibrium one obtains  ˆ …I � C‰F; QŠT V‰QŠ�1 †�1 ‰E‰FŠ � C‰F; QŠT V ‰QŠ�1 …KT i ‡ E‰QŠ†� 1 1 C‰F; xjQŠeu � V‰FjQŠŠ ru ru

…10†

and K is the solution of K�

1 1 V ‰FjQŠ ˆ C‰F; QŠT V‰QŠ�1 …K ‡ K T ‡ V‰FjsŠ†: ri ru

…11†

Note that investor i's second-order condition should be satis®ed in equilibrium. Hence, an equilibrium exists in the security markets with F if and only if equation (11) has a solution K such that K is non-singular, and that K ‡ KT ‡ r1i V ‰FjsŠ is positively de®nite. Finally, the equilibrium price function P(i ) =  + Ki is obtained by substituting  and K obtained from (10) and (11).

3.2

Viable design of security b

Recall that this article considers the situation where security a exists before security b. Since the investors have symmetric information on the payoff of security a, it is easy to prove that if a were the only one security available to the investors, there would exist an equilibrium in the market of security a. Now, imagine that security b is created in addition to security a. Then, what is the condition under which an equilibrium exists in the markets of both securities {a, b}? In the following, a property is said to hold for almost every parameter value of (ri , ru , V[(xT , s, eT ) T ]), if that property holds except possibly for a closed set of parameter values of (ri , ru , V [(xT , s, eT )T ]) having Lebesque measure zero. A simple application of Bhattacharya et al. (1995) yields the ®rst result. Proposition 1: For almost every parameter value of (ri , ru , V[(xT , s, eT )T ]), there exists a linear equilibrium for securities fa and fb , if and only if C2 ‰fb ; sŠ c2 ‰fa ; x2 ŠV‰ei2 Š 1 † C2 ‰fb ; x2 ŠV‰ei2 Š ‡ < …1 � 2 V ‰sŠ C ‰fa ; x1 ŠV‰ei1 Š ‡ C2 ‰fa ; x2 ŠV‰ei2 Š ri2 1 2 C ‰fb ; x3 jsŠV‰ei3 Š: ri2

…12†

262 Viable Design of a Security with a Pre-existing Market

This proposition shows that, in order for an equilibrium to exist in the markets of securities {a, b}, the informed investor's trade of security b motivated by his information (represented by the left-hand side) should be small enough compared with that motivated by his hedging needs (represented by the right-hand side). Furthermore, the condition is affected by how closely the payoff of security a and that of the asset backing security b are correlated (represented by the term in ( ) of the right hand side). The more correlated the payoff of security a and that of the asset backing security b are, the less dependent should the payoff of security b be on the asymmetric information. The intuition behind this result is as explained in the introduction. When security b is created in addition to security a, the informed investor can take advantage of his information about security b more effectively by partially hedging the risk associated with a part of b's payoff, x2 , about which he has no information. How accurate i's information about security b is, and how effective the partial hedging is affect the degree of the adverse selection that the uninformed investor faces in the markets. If the information is accurate enough, and if the hedging is effective enough, it becomes bene®cial for uninformed investor u to mimic informed investor i so that u can avoid the adverse selection in trading with i. However, if u behaves so, then no trade occurs between i and u, and the markets break down. It is useful to re-express the restriction on the viable design of security b as follows. Proposition 2: For almost every parameter value of (ri , ru , V[(xT , s, eT )T ]), there exists a linear equilibrium for securities fa = a1 x1 + a2 x2 and fb = b2 x2 + b3 x3 , if and only if a21 V‰ei1 Š C2 ‰x3 ; sŠ 1 2 V‰ei2 Š 2 b23 … � 2 V ‰x3 jsŠV‰ei3 Š† < 2 b2 : : V‰sŠ ri a1 V ‰ei1 Š ‡ a22 V‰ei2 Š ri2

…13†

Based on this result, I investigate how the primitive conditions given to the economy affect the viable design of security b. 3.3

Characterisation of the viable design

Observe ®rst that, if b3 = 0, then a1 6ˆ 0 because securities a and b are assumed to be linearly independent, and b2 6ˆ 0 because fb ˆ 6 0. Hence, as long as V[ei1 ] > 0 and V[ei2 ] > 0, the inequality in Proposition 2 is satis®ed. Observe also that, if a1 = 0, then b3 6ˆ 0 since fa and fb are linearly

 Kazuhiko Ohashi 263

independent. However, then, the inequality in Proposition 2 will not be 2

satis®ed as long as C V‰x‰sŠ3 ;sŠ � r12 V 2 ‰x3 jsŠV ‰ei1 Š  0: These observations lead to i the following lemma. Lemma 1: For almost every parameter value of (ri , ru , V[(xT , s, eT )T ]) satisfying C2 ‰x3 ; sŠ V ‰sŠ

< r12 V 2 ‰x3 jsŠV‰ei3 Š; there exists a linear equilibrium for securities i

{a, b}. That is, if the adverse selection is weak, in the sense that the trade motivated by information is small enough compared with that motivated by hedging needs, then the security markets will not break down. Lemma 2: For almost every parameter value of (ri , ru , V[(xT , s, eT )T ]) satisfying  r12 V 2 ‰x3 jsŠV ‰ei3 Š;

C2 ‰x3 ; sŠ V ‰sŠ

i

(a) if a1 = 0, there exists no linear equilibrium for securities {a, b}, and (b) if b3 = 0, there exists a linear equilibrium for securities {a, b}. Thus, in trading security b, if a1 = 0 so that the informed investor can perfectly hedge the risk about which he has no information, and hence can trade only the risk about which he has superior information, then no equilibrium exists as long as the adverse selection on the underlying asset is strong enough. This is what (a) of this lemma implies. On the other hand, (b) states that if b3 = 0 so that no payoff with the asymmetric information is contained in security b, no market breakdown occurs. From now on, assume that a1 6ˆ 0 and b3 ˆ 6 0, and de®ne   j aa21 j and b2  j b3 j. Note that the squared correlation coef®cient between fa and fb is 2 2 a ;fb † 2 given by Corr 2 …fa ; fb †  VC‰fa…fŠV ‰fb Š ˆ 2 ‡1 2 ‡1 : That is, for a given payoff of security a, the larger is, the more correlated securities a and b will be, in the sense that the absolute value of the correlation coef®cient between a's payoff and b's payoff is higher. De®ne 2  ri2 …

2 1 C2 ‰x3 ; sŠ 1 2 ‡ †… � 2 V ‰x3 jsŠV ‰ei3 Š† V ‰sŠ V ‰ei1 Š V‰ei2 Š ri

…14†

and Corr 2 

2 2 : 2 ‡ 1 ‡ 1

2

…15†

264 Viable Design of a Security with a Pre-existing Market

Proposition 3: Suppose that a1 6ˆ 0 and b3 ˆ 6 0. Then, for almost every parameter value 2 ‰x3 ; sŠ  r12 V 2 ‰x3 jsŠV‰ei3 Š there exists a of (ri , ru , V[(xT , s, eT )T ]) satisfying C V‰sŠ i linear equilibrium for securities {a, b}, if and only if jCorr(fa , fb )j > jCorrj, or equivalently 2 > 2 . Here, jCorrj represents the lower bound of the absolute value of the correlation coef®cient between the payoffs of securities a and b for the successful innovation of security b. That is, for an equilibrium to exist, the payoffs of securities a and b should be correlated to the degree that the absolute value of their correlation coef®cient is higher than jCorrj. In particular, as it is expected in securitisation, if the signs of a2 and b2 are the same, i.e., a2 b2  0, this result implies that the payoffs of securities a and b should be similar enough to each other. Therefore, this proposition provides a non-institutional rationale for the phenomenon that the security issuers manage to create many new securities that have similar payoffs. Now, I proceed to investigate how the lower bound jCorrj is affected by the change of the primitive conditions. Proposition 4: Under the condition of Proposition 2, jCorrj (resp. ) is 2

‰x3 ; sŠ (a) an increasing function of ri ,  and C V‰sŠ , and (b) a decreasing function of V[ei1 ], V[ei2 ] and V[ei3 ].

The interpretation of this result is straightforward. The larger ri is, the more risk-tolerant informed investor i is, and the more aggressively i trades the securities based on his information. This makes i's trade of security b more motivated by information, and worsens the adverse selection in the markets for a given payoff of security a. To mitigate this problem so that an equilibrium is sustained, the part of b's payoff suffering from asymmetric information should be reduced, which results in higher and hence higher jCorrj. Similarly, the higher  is, the more effective the hedging with security a is. This deteriorates the adverse selection problem in the markets, requiring a smaller portion of the asset subject to the asymmetric information in security b, which results in the higher or the 2 ‰x3 ; sŠ higher jCorrj for an equilibrium to exist. Finally, higher C V‰sŠ implies that investor i has more accurate information. To balance off this stronger asymmetric information so as to sustain an equilibrium, security b should be less dependent on i's information, and hence and jCorrj should be higher.

 Kazuhiko Ohashi 265

On the other hand, larger V[ei1 ], V[ei2 ] and V[ei3 ] implies that investor i's trade is more motivated by his hedging needs. This reduces the adverse selection problem so that security b can contain a higher ratio of the payoffs sensitive to the private information, which results in a smaller and therefore a smaller jCorrj.

4

Discussions

4.1

The case without a pre-existing security

What would happen if there were no pre-existing security a? The answer is given by the following proposition. Proposition 5: Suppose that b3 6ˆ 0 and that there is no pre-existing security. Then, for almost every parameter value of (ri , ru , V[(xT , s, eT )T ] ), there exists a linear equilibrium of security b with the payoff fb = b2 x2 + b3 x3 , if and only if 2 > 2no where 2no 

ri2 C2 ‰x3 ; sŠ 1 2 � 2 V ‰x3 jsŠV‰ei3 Š †: … V‰ei2 Š V‰sŠ ri

…16†

Observe that 2  2no where 2 is de®ned by equation (14). In this situation, informed investor i cannot hedge the part of the risks in security b's payoff about which he has no superior information, and hence trades security b less aggressively based on his private information than when security a is available. This reduces the adverse selection that uninformed u faces in trading security b. Consequently, the restriction on the possible design of the security is loosened. This results in 2  2no . Note also that since the model here has the mean-variance structure, the investors' demands for securities with uncorrelated payoffs are independent. Thus, 2no = 2 when fa and fb are uncorrelated, or when a2 = 0. 4.2

The case of non-linear price function

The results above also hold for the case where the price function is nonlinear. In fact, as Bhattacharya et al. (1995) show, the existence of an equilibrium involving trade in the securities with a linear price function is necessary and suf®cient for the existence of any equilibrium involving trade in the securities. Thus, one obtains the following proposition corresponding to Proposition 1.

266 Viable Design of a Security with a Pre-existing Market

Proposition 6: For almost every parameter value of (ri , ru , V [(xT , s, eT )T ]), there exists an equilibrium for securities fa and fb , if and only if C2 ‰fb ; sŠ C2 ‰fa ; x2 ŠV‰ei2 Š 1 † C2 ‰fb ; x2 ŠV‰ei2 Š‡ < …1 � 2 V‰sŠ C ‰fa ; x1 ŠV ‰ei1 Š ‡ C2 ‰fa ; x2 ŠV ‰ei2 Š ri2 1 2 C ‰fb ; x3 jsŠV‰ei3 Š: ri2

…17†

Given this proposition, it is clear that the similar argument applies to the other claims. Hence, the results above also hold for the case with a non-linear price function.

5

Concluding remarks

The results above imply that the design of a new security is crucially affected by the availability of the pre-existing securities and the correlation of their payoffs with those of the underlying assets backing the created security. This point has an important implication on the ®nancial innovation. For example, in the context of securitisation, the design of the new securities may be more restricted in MBS and ABS than in CAT-bonds (bonds that securitise the insurance risks on natural disasters), because the underlying assets of MBS and ABS have the payoffs that are highly correlated with the pre-existing T-bonds, while there are few securities whose payoffs are correlated with the underlying assets of CAT-bonds, i.e., the damages caused by natural disasters. Of course, this argument applies to many other cases, and also suggests that some sort of path-dependence may appear in the evolution of ®nancial innovation. Second, in this chapter, the informed investor could trade the preexisting security freely to hedge the part of the risks of the created security about which he has no private information. This would enable the informed investor to take advantage of his superior information effectively, which would worsen the adverse selection that the uninformed investor faces, and cause the market breakdown. Short-sales constraint makes it dif®cult for the informed investor to trade the preexisting security, and hence mitigates the adverse selection by preventing him from aggressively trading the securities based on his superior information. As a result, innovation of the new securities may be facilitated. Although a formal investigation is required, this could be

 Kazuhiko Ohashi 267

another role of short-sales constraint in ®nancial innovation in addition to that analysed by Allen and Gale (1991). Third, for illustration, I interpreted the pre-existing security to be a Tbond and the created security to be an MBS, and assumed that the informed investor traded both securities strategically. However, if such interpretation is taken literally, it seems more natural to assume that the pre-existing security is traded competitively, while the created security is traded strategically. Although it is possible to show that qualitatively similar results still hold under this dichotomy, a more careful modeling on the behaviour of the investors is needed to scrutinise the described situation. This is a subject of further research. Finally, the result of this chapter has an important implication on the relation between ®nancial innovation and investment decision. As discussed in the notes, imagine that the model describes the situation of IPO where the pre-existing security represents a market index futures (e.g., S&P 500 or Nikkei 225), and where the created security represents an IPO stock. In this case, the choice of the payoff of the created security corresponds to the choice of the underlying investment project. Then, the results of this article imply that the IPO stock cannot be sold if its payoff, which is backed by the underlying project, is not viable. Consequently, the entrepreneur can invest only in the projects that generate viable payoffs to their stocks. This way, viability of a security affects ex ante investment decision, and depending on the objective of the entrepreneur, one can examine their relation. Demange and Laroque (1998) takes a step forward in this direction of research. Investigating the relation between ®nancial innovation and ex ante investment decision seems to be a promising and fruitful avenue for future research.

Appendix: proofs Recall that Q = (qa , qb )T where qa = E[fa js] � r1i C[fa , x]ei and qb = E [fb js] �r1i , C[fb , x]ei . Now,

de®ne

T a ; qb Š a  qa ; b  qb � C‰q V‰qa Š qa ; and   …a ; b † .

a  fa ; b  fb �

C‰qa ;qb Š V‰qa Š fa

Also,

de®ne

T

and   …a ; b † . Consider an economy with the

securities   (a , b )T . Since (a , b )T are just the payoffs of the portfolios of (fa , fb )T , it is easy to check that the markets of F do not break down, if and only if the markets of  do not break down. In Proposition 1, I utilise the following version of the theorem by Bhattacharya et al. (1995).

268 Viable Design of a Security with a Pre-existing Market Theorem BRS (Bhattacharya, Reny and Spiegel (1995), Theorem 4.2): Except possibly for a closed set of parameter values of (ri , ru , V[(xT , s, eT )T ]) having Lebesque measure zero, there is a linear equilibrium in securities   (a , b )T , if and only if every eigenvalue of the matrix C[, ]T V[]�1 is less than 12. Proof of Proposition 1:

0 One can ®rst show that C‰; ŠT V ‰Š�1 ˆ ‰ 0 C[, ]T V[]�1 are 0 and

C‰b ; b Š V‰b Š .

0

C‰b ; b Š Š. V‰b Š

Hence, the eigenvalues of

Theorem BRS now implies that the markets

T

of securities   (a , b ) (or equivalently the markets of securities F  (fa , fb )T ) do not break down, if and only if C‰fa ; qa Š ˆ 0;

C‰b ; b Š V‰b Š

Substituting 1 ri2

< 21

the

if

V‰qb Š ˆ

C ‰fb ; sŠ V‰sŠ

V‰qb Š �

C2 ‰qa ; qb Š V‰qa Š

‡

1 ri 2

and

1 ri2

2

< 12. Now, since C‰fa ; qb Š ˆ C‰fb ; qa Š ˆ

only

equalities

C‰fa ; x2 ŠC‰fb ; x2 ŠV‰ei2 Š; V‰qa Š ˆ 2

C‰b ; b Š V‰b Š

that

if

2

2C‰fb ; qb Š < V‰qb Š � C V‰q‰aq;a qŠ b Š : 2

‰fb ; sŠ C‰fb ; qb Š ˆ C V‰sŠ ; C‰qa ; qb Š ˆ

…C2 ‰fa ; x1 ŠV ‰ei1 Š ‡ C2 ‰fa ; x2 Š

V ‰ei2 Š†,

and

2

…C ‰fb ; x2 ŠV‰ei2 Š ‡ C ‰fb ; x3 jsŠV ‰ei3 Š†, one can show that 2C[fb , qb ] <

b ; b Š (and hence C‰V‰ < bŠ

1

2 is

equivalent to inequality (12).

Proof of Proposition 2: Substitute the following equalities into inequality (12): C2 [fa , x1 ] = a22 V2 [x2 ] = a22 , C2 [fb , s] b23 C2 [x3 , s], C2 [fb , x2 ], = b22 V2 [x2 ] = b22 , and C2 [fb , x3 js] = b32 V2 [x3 js]. Then, one obtains the desired result.

Proof of Lemma 1:

Under the condition of Lemma 1, the left-hand side of the inequality (13) in

Proposition 2 is strictly negative if b3 6 ˆ 0, while the right hand side is positive. If b3 = 0, then the left hand side is zero, and the right hand side is strictly positive because b2 6ˆ 0. In any case, inequality (13) is satis®ed, and Proposition 2 leads to the desired result. Proof of Lemma 2: Under the condition of Lemma 2, the left hand side of the inequality (13) in Proposition 2 is non-negative. Now, if a1 = 0, then the right hand side of (13) is zero so that the inequality cannot be met. On the other hand, if b3 = 0, then from the 6 0. Hence, the left-hand side is zero, and argument prior to Lemma 1, a1 6ˆ 0 and b2 ˆ the right-hand side is strictly positive in the inequality (13) as long as V[ei1 ] > 0 and V[ei2 ] > 0. Proof of Proposition 3: Divide by b23 both sides of inequality (13), and divide by a1 the numerator and the b denominator. Then, using the de®nition   aa21 and  b2 ; (13) yields the result 3 that there exists an equilibrium for the securities {a, b}, if and only if

 Kazuhiko Ohashi 269 2

‰x3 ; sŠ  2 > ri2 …V‰e ‡ V‰e1i2 Š†…C V‰sŠ � r12 V 2 ‰x3 jsŠV‰ei3 Š†  2 . This proves the former part of the i1 Š i proposition. The latter part is just the consequence of the facts that 2 2 2 a ;fb † Corr 2 …fa ; fb †  VC‰fa…fŠV‰f ˆ 2‡1 2 ‡1 ; and that Corr2 (fa , fb ) is a monotone bŠ 2

increasing function of 2 . Proof of Proposition 4: Note that Corr 2 

2 2 2 ‡1 2 ‡1

and

that

Corr2

is

a

monotone

increasing function of 2 . Hence, we only have to prove that the claims in Proposition 4 hold for 2 . However, this is immediate from the de®nition: 2 ‰x3 ;sŠ 2 2  ri2 …V‰e ‡ V‰e1i2 Š†…C V‰sŠ � r12 V 2 ‰x3 jsŠV‰ei3 Š†. i1 Š i

Proof of Proposition 5:

Set F  fb , and repeat the argument to prove Propositions 1 and 2. Then, we obtain

the desired result.

Proof of Proposition 6:

For the case with a non-linear price function, apply the general theorem (Theorem

5.1) in Bhattacharya et al. (1995), and repeat the same argument as Proposition 1.

Notes * This chapter was previously titled `Possible Design of a Security with a Preexisting Market'. I would like to thank Jerome Detemple, Jiang Wang and Dimitri Vayanos for their valuable comments and suggestions. Conversation with Franklin Allen at the 5th annual meeting of the NFA (Tokyo, 1997) was helpful in developing the related ideas. I would also like to thank Patrick Bolton, Guy Laroque, Masahiro Okuno-Fujiwara, Mark Latham, Hiroshi Osano, Abraham Ravid, Makoto Saito, and all the other participants at the 36th Biwako Conference (Shiga, 1998) and the NFA/APFA First Joint International Conference (Tokyo, 1998) for their helpful comments. Financial support by grant-inaids for scienti®c research from the Ministry of Education of Japan, the Tokyo Commodity Exchange, the Tokyo Grain Exchange, the Commodity Futures Association of Japan and the Nomura Foundation for Social Science is cordially acknowledged. Of course, I am solely responsible for any remaining errors. 1. Although it is beyond the focus of this paper, one can also interpret this situation to describe the IPO of a ®rm: an entrepreneur invests in a project and starts a ®rm. In the IPO stage, or when the entrepreneur issues and trades shares of the ®rm with the outside investors, he will have some private information about the value of the ®rm. The value of the ®rm depends also on the movement of the whole market, which can be traded through the market index futures (e.g., S&P 500 or Nikkei 225) about whose payoff all investors have symmetric information. The choice of the payoff of the created security is determined by the choice of the underlying investment project. See also the concluding remarks of this chapter.

270 Viable Design of a Security with a Pre-existing Market 2. In the case of IPO, such a payoff can be attained by choosing a project about whose payoff the entrepreneur has less accurate private information. 3. The result also suggests that securities created in securitisation look similar in the sense that their payoffs are close to that of the riskless asset. 4. One may also think of {x1 , x2 , x3 } as the underlying risk-factors in this economy. 5. As discussed before, another example is the case of a stock market index futures, such as S&P 500 or Nikkei 225, and an IPO stock. Of course, there are many other cases of securities and their derivatives that ®t for this situation. 6. Precisely speaking, one has to assume that date 2 is strictly before the maturity of the T-bond for this interpretation to make sense. 7. That is, the variance of the payoff of each asset is normalised to satisfy V[x j ] = 1 for j = 1, 2, 3. Since a1 , a2 , b2 and b3 can be arbitrarily taken, this normalisation is without loss of generality. 8. This assumption is unnecessary if both a and b are futures (or forward) contracts. 9. This is the case where ia + ua = a and b = 0. 10. In the terminology of Bhattacharya et al. (1995), this chapter investigates the condition under which introducing security b has no destructive interference to security a. 11. Bhattacharya and Spiegel (1991) and Bhattacharya et al. (1995) show that, with CARA (negative exponential) utility functions and normally distributed shocks, the existence of an equilibrium with a linear pricing function involving trade in the securities is necessary and suf®cient for the existence of any equilibrium involving trade in the securities. See also Rahi (1996).

References Allen, F. and D. Gale (1991) `Arbitrage, Short Sales, and Financial Innovation', Econometrica, 59, pp. 1041±68. Allen, F. and D. Gale (1994) Financial Innovation and Risk Sharing Cambridge, MA: MIT Press. Bhattacharya, U., P.J. Reny, and M. Spiegel (1995) `Destructive Interference in an Imperfectly Competitive Multi-Security Market', Journal of Economic Theory, 65, pp. 136±70. Bhattacharya, U. and M. Spiegel (1991) `Insiders, Outsiders, and Market Breakdowns', Review of Financial Studies, 4, pp. 255±82. Demange, G. and G. Laroque (1995) `Private Information and the Design of Securities', Journal of Economic Theory, 65, pp. 218±32. Demange, G. and G. Laroque (1998) `Investment, Security Design and Information', Chapter 11 in this volume. DeMarzo, P. and D. Duf®e (1993) `A Liquidity-Based Model of Asset-Backed Security Design', Kellogg School of Management, Northwestern University. DeMarzo, P. and D. Duf®e (1999) `A Liquidity-Based Model of Security Design', Econometrica, 67, 65±99. Dow, J. (1998) `Arbitrage, Hedging, and Financial Innovation', Review of Financial Studies, 11, pp. 739±55.

 Kazuhiko Ohashi 271 Duf®e, D. and R. Rahi (1995) `Financial Innovation and Security Design', Journal of Economic Theory, 65, pp. 1±42. Gale, D. (1992), `Standard Securities', Review of Economic Studies, 59, pp. 731±55.  Ohashi, K. (1997) `Expected-Revenue-Maximizing Pass-Through Securitization under Sever Adverse Selection', working paper series No. 30, Faculty of Commerce, Hitotsubashi University.  Ohashi, K. (1999) `Security Innovation on Several Assets under Asymmetric Information', Japanese Economic Review, 50, pp. 76±96. Rahi, R. (1996), `Adverse Selection and Security Design', Review of Economic Studies, 63, pp. 287±300.

11

Investment, Security Design and Information* Gabrielle Demange and Guy Laroque

We consider the decisions of a venture capitalist who starts a new project. The prospects for the resale values of the newly created ®rm(s) are important features of the choice of the project. We assume that they crucially depend on the state of information when the ®rms are offered on the stock market. Such information is likely to be asymmetric, which typically discourages the general public to invest in the new ®rms. This market imperfection in¯uences the initial choices of the capitalist: it favours the projects where the asymmetry of information is the least pronounced (or, equivalently in our model, the risks born by the investor are smaller), and projects whose risk characteristics may be useful as a hedge to the market at large. The paper illustrates this argument in a two stage CARA gaussian model with the following structure. In a ®rst stage, the venture capitalist chooses her/his investment, as well as the asset structure of her/his ®rm(s). In the second stage, after some information has been revealed on the investment pro®tability, the ®rms are ¯oated on the market. The stock market is competitive, with some noise that prevents prices to reveal all the available information. The ®rst stage decisions are analysed under rational expectations of the outcome of the stock market: we study security design (should the venture capitalist put all her/his projects into a single or several ®rms?), and the investment choice.

272

Gabrielle Demange and Guy Laroque 273

1

Introduction

We consider a venture capitalist who is willing to undertake risky projects. To ®nance these projects he simultaneously chooses their incorporation into one or several companies. These decisions are made at an ex ante stage and are irreversible. It is likely that the capitalist, and possibly some other traders, will bene®t from some new privileged information at the time when stocks are traded. Without restrictions on insider trading, this may deter other uninformed investors from trading in the stocks. Our aim is to study, in a simple model, how the expected distortions of the functioning of the stock markets and the nature of risks to be hedged, in¯uence the ex ante investment choice and security design of the venture capitalist. To analyse these issues we consider a two-stage model. In the ®rst stage, the ®rm picks a selection of projects with various risk characteristics. At the same time, it chooses how to ¯oat its investment on the stock market: while we impose that the total investment is listed on the market, the venture capitalist has the possibility to create several subsidiaries, and therefore to design several securities. We work under the simplifying assumptions often used in the ®nance literature: the utility functions exhibit constant absolute risk aversion, and the distributions of the project returns are assumed to have a jointly normal distribution. The securities are constrained to be a linear combination of the underlying project returns and their return are normal as well. At an interim stage, i.e., after the investment choice and the design of the securities, but before trade, the venture capitalist and possibly a section of the public receives some exogenous privileged information on the likely outcome of the risky activities. In the second stage, the securities are exchanged against the sure (numeraire) good on a competitive stock market. The description of the functioning of the market follows the principles of a rational expectations equilibrium under asymmetric information aÁ la Grossman and Stiglitz (1980). All participants in the market are assumed to behave competitively. There are three groups of competitive agents who trade on the market. First, the risk averse venture capitalist, who has privileged information on the security returns. Second risk averse traders, called the public, with a random endowment of fundamental risks, who trade both for hedging and speculative purposes. Finally, the market makers, who are uninformed risk neutral agents, and extract all relevant information from the behaviour of aggregate demand. Since the size of the endowment is unknown, they cannot separate the two sources of variability, new

274 Investment, Security Design and Information

information or endowments shocks, from the observation of prices, and the equilibrium is not fully revealing.1 We consider two speci®cations of the information received by the risk averse public: either they receive the same signal as the venture capitalist, or they behave according to their prior knowledge of the model, ignoring the information revealed by the prices. Because of the risk neutral agents, the market always works, providing insurance to the risk averse traders, and prices are equal to the expected value of the securities payoffs conditional on the publicly available information. This is in contrast with the works of Battacharya and Spiegel (1991), Bhattacharya, Reny and Spiegel (1995), where there are only risk averse agents. Then the market may collapse if the information asymmetries are too large, the potential insurers refusing to trade with the better informed ones due to a `lemon' effect. We assume that, in the ®rst stage, the ®rm has rational expectations as to the outcome of the second stage. It chooses its investment and designs the securities so as to maximise its expected ex ante utility. As is usual in CARA Gaussian models, this ex ante expected utility is the product of two terms. The ®rst term corresponds to speculative gains. It is directly related to the insider's informational advantage over the general public and does not depend on the investment choice of the ®rm, but only on the design of the securities. The second term re¯ects the gains from the insurance opportunities provided by the securities. As in Hirshleifer (1971), public information is harmful since it prevents the market from providing insurance. The corresponding insurance gains vary with the risk born by the venture capitalist, and are the larger, the lesser the information revealed in the market. We consider the ®rst stage decisions of security design and of investment choice separately. Given an investment choice, we ®rst ask whether it is better for the venture capitalist to have a single asset, incorporating the whole project, or to separate the ®rm into several subsidiaries, each to be ¯oated on the market. It turns out, not surprisingly, that the answer depends on the precise shape of the demand of securities of the public, which determines the information revealed by prices. In the ®rst case that we study, when the risk averse public receives the same signal as the investor and behaves rationally, the best choice to maximise the insurance gains is to have a single ®rm, whatever the risk characteristics of the investment. A lot of information then is anyhow revealed by the price, and introducing more securities could only worsen the situation from the capitalist's viewpoint. In the second case, when the risk averse public is uninformed and does not extract information from the price, the venture capitalist's best choice of asset(s) is less easy to

Gabrielle Demange and Guy Laroque 275

characterise. When the shape of the covariance matrix of the fundamental risks is much changed by the signal, the venture capitalist's informational advantage allows in some circumstances, depending on the precise shape of the hedging demand of the risk averse public, to do best by diversifying her/his activity into several subsidiaries, which then reduces the information contained in the prices. The above results concern the design of securities to maximise the insurance gains. When one also takes into account the speculative gains, diversifying becomes more advantageous and may be pro®table, even when the risk averse public has the same information as the capitalist. Speculative gains depend on the informational advantage of the ®rm, which typically may spread along other directions than the return on the aggregate single ®rm.2 Finally, for given risk characteristics, the investor chooses a level of investment that increases with the insurance provided by the market, i.e., decreases with the information revealed by prices. We take a ®rst look at the risk composition of investment that maximises the insurance gains. It seems to depend, in a rather complicated way, both on the informational content of the signal available at the time of trade and on the hedging needs of the risk averse public. The paper touches a number of themes that have received a lot of attention in the recent literature. First, security design has been considered from a variety of viewpoints. Allen and Gale (1988) assume that the designer of the securities differentiates his products to best ®t the insurance needs of the buyers, subject to a transaction cost that depends on the number and shape of the securities he puts on the market. For Duf®e and Jackson (1989), market organisations, whose revenues are based on a fee proportional to trade volume, facing some operating costs, create markets in a non-cooperative setup. Both of these works assume symmetric information, and the insurance needs of the potential buyers create the incentives for designing securities. At the opposite, in Boot and Thakor (1993), the motivation for designing securities is not to improve risk sharing ± all agents are risk neutral ± but to provide a better dissemination of information. Here, as in Demange and Laroque (1994, 1995b), we neglect the transaction costs studied in these previous works and focus on the trade-off between insurance and information: as pointed out in a celebrated paper of Hirshleifer (1971), advanced information on a risky project precludes risk sharing. Second, the paper is related to the literature on initial public offerings, initiated with the seminal work of Leland and Pyle (1977), and with the pros and cons of insider trading (see Leland, 1995). Here the information

276 Investment, Security Design and Information

is revealed after the investment decision, and insider trading is a nuisance. It reduces the insurance provided by the market, and because it arrives too late, it does not help to select the projects. Finally, the design of securities is linked with the incompleteness of markets (see Demange and Laroque, 1995a; Rahi, 1995). While the introduction of differentiated subsidiaries would be in the general interest of the public, we ®nd here that it is not always in the interest of the venture capitalist, due to the adverse effects of the information revealed on the new markets. This points to another motive, aside from transaction costs or lack of liquidity (Allen and Gale, 1994; Cuny, 1993), for the incompleteness of markets. The paper is organised as follows. The model is laid down in the second section. Section 3 then describes the functioning of the stock market under asymmetric information, building upon the Grossman±Stiglitz and Battacharya±Spiegel models, when the agents have risky initial endowments. Section 4 computes the ex ante utility expected by the venture capitalist from her/his trades in the securities and derives some properties of the optimal security design associated to the capitalist's investment choice. Section 5 studies the optimal investment choice. All proofs are gathered at the end of the paper in Section 6.

2

The model

The economy has a single good. At the outset, the venture capitalist can invest in a number of projects whose returns are random. A level of ~ where ~ is a zero mean, unit investment K will yield (K) + (K)0 , variance normal vector of dimension k, and (K) and (K) are well behaved functions of IR‡ into, respectively, IR‡ and IRk‡ .

We consider a sequence of three dates t = 1, 2, 3.

At date 1, the capitalist decides to go public, by incorporating the whole project into a single ®rm, or by setting up several ®rms. His decision is described by the stocks returns "~ of the different ®rms he sets up. We assume that any return is a linear combination of the underlying securities. Therefore, when the number of ®rms is n, the relationship between the vector of stock returns and the underlying project returns is described by a matrix A of dimension (k  n): "~ ˆ A0 ~ It will not be useful to have more than k ®rms and, without loss of generality, the matrix A is normalised, so that A0 A = In . We impose the constraint that the whole ®rm is incorporated on the market so that:

Gabrielle Demange and Guy Laroque 277

…K†0 ~ ˆ S0 "~ for some (n  1) matrix S. At date 2, the venture capitalist privately observes an advanced signal ~ on the likely outcome of his undertakings. ~ has a jointly normal distribution with the fundamental risks, has zero mean and unit variance. At date 3, stock markets are open. Apart from the capitalist, risk neutral rational traders operate, as well as traders whose random wealth may be ~ correlated with . The main decisions are taken at date 1, where securities are designed and at date 3 where trade takes place. We ®rst investigate the outcome of the stock market, since the design of the securities depend on the gains the investor expects to realise on the market.

3

The stock market

The stock market allows the venture capitalist to insure some of the risk s/he bears, and possibly to bene®t from his/her private information. The investment decisions have been made and to simplify notation we drop the argument K in  and . 3.1

The venture capitalist

~ The portfolio The venture capitalist's initial wealth is equal to  + S0 . choice of the capitalist is conditional on the signal ~. The capitalist's wealth, when he buys a portfolio X, can be written as: ~ ˆ  ‡ S0 "~ ‡ X0 …"~ � p† W Assuming CARA preferences, with a coef®cient of risk aversion equal to a, the investor maximises:

i:e:;

~ ~ � a var Wj ~ ~ EWj   2 0 0 X …E "~j � p† � a…X ‡ S† var…"~j†…X ‡ S†

Therefore his security demand is equal to: 1 X…; p† ˆ �S ‡ var…"~j†�1 …E"~j~ � p† a

…1†

It is composed of two parts: the hedging demand, �S, and the speculative demand which is proportional to the expected return, i.e., to the spread between the expected value and the price. In a risk neutral market without

278 Investment, Security Design and Information

asymmetric information, the speculative demand is null: the capitalist sells his whole risky project without paying any risk premium. 3.2

Traders with wealth correlated with the investment project

There are a large number of small traders, with CARA preferences. Apart from their initial wealth they all share the same information IT . At the time of trade, a trader knows the fraction of his wealth that is correlated with ~ . Therefore the aggregate security demand3 of these traders at date 3 is given by the standard formula:   ~ IT �P E"j ~ ~ †jIT Š y x…p; y† ˆ var…"~jIT †�1 � cov‰…"; …2† b where b denotes the aggregate risk aversion coef®cient of these traders, ~ Traders y 0 ~ the fraction of their endowment that is correlated with . know their own endowment only and perceive the aggregate value y~ as a random variable, which is drawn from a k dimensional gaussian distribution of zero mean and variance var (y~ ), and uncorrelated with the investment returns. As for the information structure IT we shall consider two polar cases: 1. (Informed public) The small traders share the same information as the ~ venture capitalist, IT = . 2. (The public is uninformed and does not extract information from the price) Small traders do not extract information from the observation of the price: IT is reduced to the prior information of the model, E"~jIT = 0, var ("~jIT ) = In . 3.3

Price formation

The price is ®xed on the market through competitive risk neutral traders. They are rational and use all public information IP , which includes the price, their knowledge of the correlation of the underlying risks with the securities, the signal structure, and more importantly the shape of the security demands (1) and (2). Risk neutrality implies: p ˆ E"~jIP Denoting their demands as z(p) the equilibrium condition may be written as: X…; p† ‡ x…p; y~ † ‡ z…p† ˆ 0: Inspection of (1) and (2) shows that the observation of the price is informationally equivalent to that of:

Gabrielle Demange and Guy Laroque 279

~ var…"~jIT †�1

~ ˆ E"~j~ ‡ avar…"~j†



E"~jIT ~ ~†jIT Š y~ � cov‰…"; b



If the risk averse public is informed (case 1), we have: ab ~ ~†jŠ y~ cov‰…";

~ ˆ E"~j~ � a‡b while in case 2: ~ ~† y~

~ ˆ E"~j~ � a var…"~j†cov…"; In both cases the price is given by: p ˆ E‰"~j Š

4

Optimal security design

The capitalist's incentives to design one or several securities depend on the utility he expects from the trades. 4.1

The ex ante utility of the venture capitalist

The CARA Gaussian structure of the market allows to compute the ex ante utility u of the capitalist at the end of date 1, when he has decided on the security structure, before he has received the signal. Let u0 be the capitalist's utility level if s/he is not allowed to transact (or equivalently if s/he does not go public). Since utility is negative, u0 /u increases with the utility level. Proposition 1: the ex ante utility level of the venture capitalist is given by:   2   u0 det var "~j ~ 1=2 a ˆ var‰ 0 ~j Š exp u 2 det var "~j

…3†

The two terms correspond respectively to speculative and insurance gains. The ratio of determinants is larger than 1, since the private information ~ is more precise than the public information ~. It corresponds to the speculative gains of the investor: it increases his ex ante utility level by a factor which is independent of the amount of capital invested. The term var[ 0 ~j ] corresponds to the gain of utility associated with the insurance. It is maximal when the market gets no information at all: the capitalist may get rid of all risks without paying any risk premium. The more information the market gets, the lower this gain: this is known as the information effect. Accordingly the information revealed to the market is the crucial determinant of these gains.

280 Investment, Security Design and Information

We now study in turn the consequences of this formula for security design, given an investment choice, and for the choice of investment. We are far from having a complete solution to the problem, and the results that we present are partial. We shall mainly focus on the circumstance where the investment is `large', so that the most important factor is the insurance gains, and we can neglect the speculative gains. For that purpose the following expression for the variance term in the insurance gains is useful: var‰ 0 ~ j Š ˆ  0  �  0 var…E~ j†A…var †�1 A0 var…E~ j† 4.2

…4†

Security design

Proposition 2: In order to maximise his insurance gains, the venture capitalist forms a single ®rm if the risk averse traders are informed. If they are uninformed, s/he takes into account their hedging needs. The crucial point when the risk averse public is as well informed as the venture capitalist is that, whatever securities are issued, the signal E‰ 0 ~ jŠ �

ab  0 var…~ j~† y~ a‡b

on the overall investment may be extracted from the prices.4 Therefore setting up only one security with the overall investment return "~ =  0 ~ always minimises the revelation of information, and therefore maximises the gains from insurance. The situation is different when the public is uninformed. In fact, from (4), all the differences between the two cases come from the term A(var )�1 A0 . The information that is revealed by the security collinear to the overall investment becomes: S0 ~ ˆ E‰ 0 ~jŠ � a 0 var…~ j†AA0 y~ It now depends (through AA0 ) on the other securities that can be exchanged on the market. One cannot exclude that issuing more securities affects the demand in such a way that less information is revealed. A simple example may help to understand why. Suppose that the fundamental risks are two-dimensional and that the project is given by  0 = (1, 0). The small traders' endowments are concentrated on the other branch of the fundamental risk, i.e., y1 = 0. If there is only one security "~ = ~ 1 , their hedging demand, without privileged information, is null, so that the private information  is entirely revealed to the market, and E[ 0 ~ j ] = E[ 0 ~ j]. If the two securities, with returns respectively equal to ~ 1 and ~ 2 , are marketed, small traders may now hedge their risks, by taking

Gabrielle Demange and Guy Laroque 281

a position on the second security. Then the public receives two signals, 1 and 2 , and in general cannot perfectly extract the values of E[~ 1 + ~ 2 j]. More precisely ~ y~

ˆ E~ j � a var…j† so that E‰ 0 ~ j Š ˆ E‰ 0 ~ jŠ � a 0 var…~ j†y~ When  0 var(~ j) y~ has a non-zero variance, i.e., conditioning with respect to  changes the shape of the covariance matrix of the fundamental risks,

the investor is better off issuing two securities rather than one. As in Allen and Gale (1994), the demand side of the market is important in the design of the security. Proposition 2 deals only with the insurance gains. It is of interest to know whether speculative gains could change the nature of the results. In particular, in case 1, when the risk averse public receives the same signal as the venture capitalist, can prospective speculative gains induce the capitalist to ¯oat several securities on the market, to bene®t from her/his privileged information? The answer to this question is yes. A simple intuition is that the investor has nothing to lose by trading a security that does not reduce her/his insurance gains. Consider as above an example where fundamental risks are two-dimensional and the investment project is given by  0 = (1, 0). Assume that the capitalist has two uncorrelated advanced signals, each of which brings information on ~ 1 or ~ 2 only. For example ~i = i + ~i , where each ~i is independent of all other variables. Assume furthermore that y~ 1 and y~ 2 are independent. If the two securities are ¯oated, the two markets are independent. The investor gains on the ®rst security are identical whether the second market exists or not while s/he gains from his superior information by trading on the second security.5

5

Investment choice

We now consider the optimal investment choice. It of course depends on the speci®cation of the functions (K) and (K). To make things as simple as possible and to concentrate on the choice of the risk pro®le, we suppose that the expected return by unit of investment is constant equal to some strictly positive . Therefore (K) = K for any K. At the opposite the venture capitalist can choose any risk pro®le for her/his investment: for any  such that  0  = 1, any K, there is a project with (K) = K. It is easy to determine a property on the optimal level of capital stock.

282 Investment, Security Design and Information

Proposition 3: Let the risk characteristics of the investment, , and the security design, A, be given. Then the optimal capital stock of the venture capitalist is given: Kˆ

 avar… 0 E~j †

The investment level increases with the expected return  and decreases with the investor risk aversion a. When no information passes through the prices, var 0 E~j = 0, and the optimal stock of capital is in®nite: the capitalist can fully hedge the risks without bearing any risk premium on the (risk neutral) market while earning a positive return  on each unit invested. In all other circumstances, the investment level is well de®ned and decreases with the information revealed on the market. There remains to understand the optimal risk characteristics of the venture capitalist's investment. We consider the case where the investment level is large, so that the speculative gains (the ®rst term in (3)) are negligible compared to the insurance gains. From Proposition 3, we see that the best direction for investment is the one that allows for the largest insurance possibilities on the market: it should combine the information brought by the signal , with the hedging needs of the public. Proposition 4: Consider a market where the risk averse traders are informed (case 1). The risk characteristics of the investment project that maximises the insurance gains from the market is a normalised vector ,  0  = 1, which satis®es: 2 2

a b ~j vary var~j† …varE~j ‡ …a‡b† 2 var 2…varE~j† ˆ � 2 b2 0 a 0  …var E~j†  …varE~j ‡ ~j vary var~j† 2 var …a‡b†

The choice of the investment risk characteristics is the solution of a non-convex programme, and the above proposition only gives the ®rst order condition associated with this programme. The ®rst order condition can be rewritten so that  appears to be an eigenvector of a linear combination of varE~j and of var~jvaryvar~j, with respective weights depending on the values of the two corresponding quadratic forms along the (unknown) direction . Since varE~j = Ik � var~j this condition ®nally depends on two matrices, the matrix var~j which describes the information contained in the signal, and the covariance matrix of the endowments, vary. The optimal direction may be a complicated function of the information motive and of the demand for hedging of the public. A simple situation occurs when the two above matrices have the same set of

Gabrielle Demange and Guy Laroque 283

eigenvectors. The ®rst order condition then says that the optimal direction is one of these eigenvectors. Going back to the expression for the insurance gains, if we let Vi , 0  Vi  1, and Yi denote respectively the eigenvalues of var ~ j and vary associated with the ith eigenvector, the optimal direction is the value of i that minimises: …1 � Vi †2 a2 b2 2 1 � Vi ‡ …a‡b† 2 Vi Yi Given Y, this expression is minimised for the largest possible value of V. Also, given V, one wants to have the largest possible Y. In words, when the hedging demand of the public is uniformly distributed (vary = YIk ), the venture capitalist chooses the direction where the signal reveals the less information (largest eigenvalue of varj). Conversely, when the information is uniformly revealing in all directions (varj = VIk ), the preferred investment is the one that best satis®es the hedging demand (largest eigenvalue of vary).

6

Proofs

Proof of Proposition 1: Given  and p, substituting the expression of the security demand (1) yields: ~ � a var Wj ~ ˆ EWj   2 …K† ‡ S0 E"~j � S0 …E"~j � p† ‡

1 …E"~j~ � p†0 var…"~j†�1 …E"~j~ � p† 2a

or ~ ~ ˆ …K† ‡ S0 p ‡ 1 …E"~j~ � p†0 var…"~j†�1 …E"~j~ � p† ~ ~ � a var Wj EWj  2 2a

…5†

~ j ~g ~ j � a varW We compute the mathematical expectation of exp �afEW   2 over the distributions of  and p, or equivalently over the distributions of  and y. Recall that p = E["~j ] and that ~ = E"~j~ + AVy~ for a suitable de®nition of the matrix V. Lemma 1 shows that the price p, and therefore the cost from hedging, S0 p, is independent of the speculative demand of the venture capitalist. Lemma 1: ~ E"~j ~ † ˆ 0 cov…E"~j~ � E"~j ;

284 Investment, Security Design and Information

and var…E"~j~ � E"~j ~ † ‡ var"~j~ ˆ var"~j ~ Proof of Lemma 1: This follows from the fact that E"~j~ � E"~j ~ is uncorrelated with ~ . Indeed, one can write: E"~j~ � E"~j ~ ˆ …"~ � E"~j ~ † � …"~ � E"~j~†

…6†

The ®rst term on the right hand side is uncorrelated with ~ by de®nition, while the second term is uncorrelated with ~, and uncorrelated with ~ by assumption, which implies that it is uncorrelated with ~ = E"~j~ + AVy~ . The same argument shows that (E"~j~ � E"~j ~ ) is uncorrelated with ("~ � E"~j~) so that the second equation follows. The utility level in the absence of trade is u0 = � E[exp � a {(K) + S0 ~ }]. From Lemma 1 we get u0 1 ˆ E‰exp �afS0 …~ � p†gŠ E‰expf …E"~j~ � p†0 var…"~j†�1 …E"~j~ � p†gŠ 2 u Lemma 2: let Z be a zero mean k-dimensional multivariate normal variable of variance covariance matrix , and let H be a symmetric matrix such that 2H + �1 is positive de®nite. Then: E exp…�Z0 HZ† ˆ

1 1

…det jIk ‡ 2Hj†2

Proof of Lemma 2: One just has to compute the integral, noting that: 1 1 �Z0 HZ � Z0 �1 Z ˆ � Z0 ‰2H ‡ �1 ŠZ 2 2 Using the shape of the distribution of the multivariate normal, the integral of the exponential of this expression is equal to: �1

…2†k=2 ‰det j2H ‡ �1 jŠ 2 The result follows.

Lemma 3: (speculative gains)

 1 1 det var…"~j ~ † 2 E exp �f …E"~j~ � E‰"~j ~ Š†0 var…"~j†�1 …E"~j~ � E‰"~j ~ Š†g ˆ 2 det var…"~j~†

Gabrielle Demange and Guy Laroque 285

Proof of Lemma 3: Let Z ˆ var…"~j†�1 …E"~j � E"~j †; and H ˆ

1 var"~j~ 2

The variance of Z is given by ~ �1 var…E"~j~ � E"~j ~†…var"~j~†�1  ˆ …var"~j† ~ �1 . Lemma 1 yields Moreover Ik + 2H = Ik + var(E"~j~ � E"~j ~ )(var"~j) ~ �1 Ik ‡ 2H ˆ var"~j ~…var"~j† and it suf®ces to apply Lemma 2. End of the Proof of Proposition 1: E["~j ] is normal with zero mean. Therefore Efexp �aS0 E‰"~j Šg ˆ exp �

a2 var…S0 E‰"~j Š† 2

The result follows from Lemma 3. Proof of Proposition 4: The capitalist seeks to maximise his/her insurance gains. Since there is a single security, proportional to the risk characteristics of the ®rm, using (4), this amounts to ®nd the value of , of norm 1, which minimises: … 0 varE~j†2 var where: var ˆ  0 varE~j  ‡

a2 b2  0 var~jvar…y~ †var~j  …a ‡ b†2

To simplify notations, let: B ˆ varE~j;

Cˆ

a2 b2 …a ‡ b†2

var~jvar…y~ †var~j

B and C are symmetric positive de®nite matrices, and we want to minimise

286 Investment, Security Design and Information

Mˆ

… 0 B†2 ‡ C†

 0 …B

subject to the constraint  0  = 1. Taking logarithms, the ®rst order condition for an extremum of M under the constraint can be written, where  is the multiplier associated with the constraint: 2B …B ‡ C† � ˆ   0 B  0 …B ‡ C† Multiplying from the left by  0 , and using the constraint shows that  = 1. 1 Therefore  is an eigenvector of the matrix 02B B � 0 …B‡C† …B ‡ C†, with eigenvalue equal to 1.

Notes * We have bene®ted from the comments of the participants in the Biwako conference, Kyoto University, and especially from Patrick Bolton. 1. An important feature of all the models with asymmetric information is the `noise' that prevents the uninformed traders from inferring the private information signals from the mere observation of the market prices. We use the modeling device introduced by Battacharya and Spiegel (1991) and Bhattacharya, Reny and Spiegel (1995). Here the noise is fully rational and comes from lack of information on some idiosyncratic risks. In our previous work (Demange and Laroque (1994, 1995b), as in a large part of the literature (Grossman and Stiglitz, 1980; Schleifer and Summers, 1990), we postulated an exogenous price inelastic random supply of securities. Since the price and information sensitivity of the security demand of the public is a priori an important element of the investment and security design decisions, this procedure is too crude for our current purpose. 2. This is in contrast with Rahi (1996) in which a ®rm that is only allowed to design a single security, does not ®nd it optimal to use its privileged information. However the result is somewhat trivial because the informational set up is such that the investor's information is always completely revealed to the market. Obviously then the capitalist has nothing to gain at trading in other directions than his own risks. 3. With obvious notation, each individual i's demand is: �1

x…p; yi † ˆ var…"~jIT †

  E"~jIT � p ~ ~†jIT Š yi � cov‰…"; bi

4. To see this, it suf®ces to multiply on the left by S0 the vector of signals ~:

ab ~ ~†jŠ y~ cov‰…";

~ ˆ E"~j~ � a‡b and to use S0 "~ =  0 ~.

Gabrielle Demange and Guy Laroque 287 5. Under the informational assumptions, E~ j~ = E~ i j~i and the matrix var(~ j) is diagonal. Consider the cases with two securities " = . We get ~ = E~ j~ � ab/a+b var(~ j) y~ , so that, as we already know, 1 is unaffected by the trading on the second security. As a consequence p1 and the capitalist's demand (see (1)) are also unaffected: the gains from the ®rst security are unchanged.

References Allen, F. and D. Gale (1988) `Optimal Security Design', Review of Financial Studies, 1, pp. 229±63. Allen, F. and D. Gale (1994) Financial Innovation and Risk Sharing, MIT Press. Bhattacharya, U., and M. Spiegel (1991) `Insiders, Outsiders and Market Breakdowns', Review of Financial Studies, IV (2), pp. 255±82. Bhattacharya, U., P.J. Reny and M. Spiegel (1995) `Destructive Interference in an Imperfectly Competitive Multi-Security Market', Journal of Economic Theory, 65 (1), pp. 136±70. Boot, A.W.A. and A.V. Thakor (1993) `Security Design', Journal of Finance, 48, pp. 1349±78. Borch, K. (1960) `The Safety Loading of Reinsurance Premiums', Skandinavrsk Aktuarieskrift, pp. 153±84. Cuny, C.J. (1993) `The Role of Liquidity in Futures Market Innovations', Review of Financial Studies, 6, pp. 57±78. De Long, J.B., A. Schleifer, L.H. Summers and R.J. Waldmann (1989) `The Size and Incidence of the Losses from Noise Trading', The Journal of Finance, 44 (3), pp. 681±96. Demange, G. and G. Laroque (1994) `Information asymeÂtrique et eÂmission d'actifs', Revue Economique, 45 (3), pp. 639±56. Demange, G. and G. Laroque (1995a) `Optimality of Incomplete Markets', Journal of Economic Theory, 65 (1), pp. 218±32. Demange, G. and G. Laroque (1995b) `Private Information and the Design of Securities', Journal of Economic Theory, 65 (1), pp. 233±57. Duf®e, D. and M. Jackson (1989) `Optimal Innovation of Futures Contracts', Review of Financial Studies, 2, pp. 275±96. Gantmacher, F.R. (1961) Theory of Matrices, New York: Chelsea Publishing Company. Grossman, S.J. and J. Stiglitz (1980) `On the Impossibility of Informationally Ef®cient Markets', American Economic Review, pp. 393±408. Hara, C. (1995) `Commission±Revenue Maximization in a General Equilibrium Model of Asset Creation', Journal of Economic Theory, 65 (1), pp. 258±98. Harris, M. and A. Raviv (1991) `The Theory of Capital Structure', Journal of Finance, pp. 297±355. Hirshleifer, J. (1971) `The Private and Social Value of Information and the Reward to Inventive Activity', American Economic Review, 61, pp. 561±74. Kyle, A.S. (1985) `Continuous Auctions and Insider Trading', Econometrica, 53, pp. 1315±35. Leland, H.E. (1992) `Insider Trading, Should It Be Prohibited?', Journal of Political Economy, 100 (4), pp. 859±97.

288 Investment, Security Design and Information Leland, H.E. and D.H. Pyle (1977) `Informational Asymmetries, Financial Structure, and Financial Intermediation', The Journal of Finance, X X X I I (2), pp. 371±87.  Ohashi, K. (1992) `Ef®cient Futures Innovation with Small Transaction Fee: Centralization vs. Decentralization', Institute for Social Economic Planning, University of Tsukuba.  Ohashi, K. (1995) `Endogenous Determination of the Degree of Market-Incompleteness in Futures Innovation', Journal of Economic Theory, 65, pp. 198±217. Rahi, R. (1995) `Optimal Incomplete Markets with Asymmetric Information', Journal of Economic Theory, 65(1), pp. 171±97. Rahi, R. (1996) `Adverse Selection and Security Design', The Review of Economic Studies, 63, pp. 287±300. Schleifer, A. and L.H. Summers (1990) `The Noise Trader Approach to Finance', Journal of Economic Perspectives, 4 (2), pp. 19±33.

Part IV

General Comments

This page intentionally left blank

12

General Comments Masahiro Okuno-Fujiwara

Altogether ten papers have been presented in this conference. Throughout the sessions, I enjoyed listening to new ideas and careful analyses presented by authors, thoughtful comments and helpful remarks made by discussants, as well as the lively ¯oor discussions that followed. The conference theme was `Banking, Capital Markets and Corporate Governance'. I believe that this theme was quite appropriate for this year's 36th Biwako Conference for the following three reasons. First, we have witnessed structural changes taking place over the last decade in capital markets worldwide. We should evaluate these changes and analyse their implications for resource allocations. Second, the Japanese economy, especially its ®nancial system, is also in a process of structural transition and, partly because of this, faces serious problems. How to solve these problems is, quite naturally, a hotly debated question at this time in this country. Corporate governance of banks and non-®nancial ®rms is an important related question. Third, not only the Japanese economy, but also many East Asian economies, are in a serious economic trouble. Many suspect that the way East Asian governments controlled their economies, in degrading the disciplinary effects of corporate governance systems, may be at the root of the problem. Let me discuss these problems in turn, the ®rst two problems rather brie¯y, and the last problem more extensively. Nevertheless, we should not forget that there is a fourth reason ± perhaps the most important one. Over the last two decades, we have seen the explosive development of analytical tools to help tackle these problems, ranging from the economics of information and incentives, to game theory and contract theory. 291

292 General Comments

First, regarding the structural changes in the capital markets, ®nancial innovations have changed the market. Concepts such as options and other ®nancial derivatives have become popular in economic terminology only in the last decade or so. Moreover, possibly re¯ecting the changes in the underlying ®nancial technology, bubbles in stock markets and price volatility in land prices have adversely affected the economies in many countries. Capital markets have also become tightly connected globally. Even in Japan, where foreign ®rms once had dif®culty in entering the market, the `Big Bang' in April 1998 removed legal barriers, and now foreign ®rms have become important players even in domestic retail markets. In sum, nations throughout the world seem much more closely connected to each other, but fragility of the global market may have increased. Thus, there is an increased need to understand how this new capital market works, including its informational asymmetries, and how recent developments have affected investment, management decisions, and the real sectors of the global economy.  Several papers, such as those by Guy Laroque, Kazu Ohashi and Gerald  Garvey, address these questions. Papers by Laroque and Ohashi provide elegant results, and prospects for further study in these areas seem extremely promising. In addition, as Garvey's paper shows, the effects of stock options on manager's incentive should also be worth studying further. Another paper by Nori Yanagawa asks why banks do not stop unproductive loans in times of economic boom. He suggests that a mechanism similar to the soft budget problem is behind this phenomenon. I was not fully convinced by his argument, but it seems worthwhile to examine more carefully the incentive structures behind the creation and growth of ®nancial bubbles. Second, Japan is in a process of system transition, and faces serious ®nancial problems. A clearer analysis of bank crises, and a better policy design for bank bailouts, such as those presented by Bolton and Osano, certainly help regulators to tackle these problems. Bolton provided a beautiful analysis for the design of bank bailouts. However, the real world is much more complex than this excellent paper assumes. For example, he emphasised that bank recapitalisation policies may create a soft budget problem for ®rm managers. However, what we saw in Japan was that anticipation of the strict BIS regulations imposed in April 1998 caused a serious credit crunch in December 1997. That is, a tough bank policy induced bank managers to adopt a harder attitude towards non-performing loans.

Masahiro Okuno-Fujiwara 293

It is clear why this discrepancy between the paper's predictions and the reality occurs. The real world is not a `one-shot' game, as Bolton assumes. Even if banks allow non-performing loans to live on, bank managers must expect that regulators will eventually discover the truth. Still, the paper is quite useful for understanding the nature of the bank crisis, and for designing an appropriate policy to solve it. However, at the heart of the Japanese problem, there seems to lie a lack of effective governance of bank managers, as the paper by Horiuchi indicates. This brings us to the question of corporate governance, the third part of the conference theme. I should emphasise, however, that it is not a problem of corporate governance narrowly de®ned, that is, it is not only a problem of how stockholders control managers. It should be considered as a problem of corporate governance more broadly de®ned, to take into account how various stakeholders, including depositors and, especially, regulators, control bank managers. If we start looking at the problem this way, we see immediately how corporate governance relates to the third problem I raised at the beginning. That is, many East Asian economies are in serious economic trouble. Many of these countries seem to be characterised by the authoritarian nature of their governments which, having a tight grip over the private sector, heavily intervened in inter-industry resource allocation, and controlled competition among private ®rms. This was once thought to be a factor behind the region's economic successes, because governments successfully coordinated the economy to direct resources towards growing export industries, often by creating rents in the ®nancial sector. This seems to suggest an important problem for economic sciences. As many conference papers, such as those by White and Tachibanaki, argue, institutional arrangements, such as bankruptcy laws and intercorporate stock holdings, can affect resource allocation in the economy. The ultimate form of resource allocation will depend not only on the country's resource endowments, level of technological development, and consumer preferences, but also on how the country's economic system is organised. Moreover, as we all know, and as Brickley explicitly states in his conference paper comparing incentive mechanisms for top corporate executives in the US and Japan, the economic systems of countries such as the US and Japan will differ in many ways, and such differences may account for the different performance of these economies. The recent problems of the East Asian economies may be explained by the uniqueness of their economic systems. Some observers suggest that

294 General Comments

the root of the problem lies in the authoritarian nature of these governments and the governance mechanisms they used. Capital ¯ights in Thailand and Korea may have been caused by the government's inability to control the domestic economy in the face of increasingly internationalised capital markets. Together with the experience of the Japanese government, this seems to imply that the East Asian strategy may not work well when the government must deal with increasingly open and technologically advanced markets. Lenders lost con®dence in Indonesia because its government provided assistance to a limited number of ®rms, based upon nepotism. Governments that are authoritarian and closed may tend to favour the interests of insiders in the ruling political regime. Certain governance structures may be effective if the economy is at a certain stage of development and/or the capital market is facing a particular type of external environment, while the effectiveness of these structures may be lost when the parameters change. Faced with these new and changing circumstances, the World Bank is now conducting a research project entitled `Rethinking the East Asian Miracle'. It is important to study not only how different institutions affect economic performance, but also to study how the effectiveness of an economic institution depends upon the nature of the wider economic environment. In view of recent experiences in Japan and other East Asian countries, one especially important point seems to be that emphasised by Horiuchi and Tachibanaki. That is, long-term relationships usually enhance the possibilities for cooperation. Many consider that Japan and other East Asian economies successfully promoted economic growth by adopting various mechanisms to encourage long-term relationship commitments. Examples of such mechanisms include such things as long-term employment and internal promotion policies, main bank relationships, discretionary and tight control over private ®rms by the government, as typi®ed by Ministry of Finance (MOF) banking polices and Amakudari. I should emphasise that these long-term relationships are institutionalised in Japan, in the sense that they have become prevalent nationwide. Because others are also committed to long-term relationships, no one can expect to obtain high enough payoffs from outside by terminating their own long-term relationship commitments. That is, because long-term relationships are prevalent across the economy, the outside option is low and each individual wants to retain current long-term relationships as much as possible. In a sense, people are captured by all long-term relationships, whether it is an employment, ®nancial, or government-business relationship. Recent experiences seem

Masahiro Okuno-Fujiwara 295

to suggest that long-term relationships can operate adversely precisely because of this. Sometimes, one partner wants to discontinue the relationship because it has outlived its usefulness. Nonetheless, being captured, he or she cannot discontinue the relationship. Much of the Japanese experience seems to have this feature in common with the East Asian experience. The results presented by Horiuchi and Tachibanaki provide examples of such experiences. In the discussion of the Horiuchi paper, Guy Laroque and Patrick Bolton remarked that deregulation policies, which were introduced to improve market competition and provide governance for bank managers, may have weakened the soundness of banks and created the bank crisis instead. They are probably correct, and deregulation policies may in fact be responsible for the current bank crisis. I personally, however, have found the following question to be more important. When faced with more global open competition, should we change the governance structures for our ®nancial markets? In particular, should we move from tight governmental protection of the industry, based upon the long-term relationship between the MOF and individual banks, to a harsher policy favouring more open market competition as a mechanism for providing governance? Alternatively, can long-term relationships be maintained, with an appropriate reform, so that a new economic system is created which enhances cooperation without causing the problems which long-term relationships can present? Questions such as this one, i.e., what systemic change is necessary when the economy's environment changes, seem to be important in view of recent events in this region. Finally, the participants of this 36th Biwako Conference enjoyed the intellectual discussion very much and bene®ted a great deal from lively international research exchanges. On behalf of all participants, I thank the conference organisers, Professors Hiroshi Osano and Toshiaki Tachibanaki for their great efforts, and congratulate them on the successful organisation and management of the conference.

This page intentionally left blank

Index

ABF model 32, 33±8 banks 34±6 conditional bank recapitalisations 41±4

®rms 33±4

regulator 36±8

absolute performance 244±7 absolute priority rule 183 accounting performance 238±44 adjustment speed of capital stock 159±65, 166±8 Adler, B. 200 administration orders 185 adverse selection 254±6, 261±5 agency approach 89 agency problems 152 board retention and 9, 233±49;

post-retirement service and

235±7

Aghion, P. 32, 33, 55, 86, 91, 200, 201, 214 Allen, F. 267, 275, 276, 281 Allen, L. 55, 78 Almon lag method 162±5, 168, 178 amakudari system 88, 142, 151±4 Ando, A. 110 Antle, R. 245 Aoki, M. 151±2, 154, 226 Asahi Bank 114, 125 Asquith, P. 196 asset backed securities (ABS) 253, 254, 255, 266 asset substitution investment policy 209±10 asset substitution problem 204±7, 225±6 see also executive option plans asset values 14, 15±16, 19±20, 29 ex post liquidity shock 22±5 ¯uctuation 25±8 total asset value 94, 96±102, 103, 107, 108±9, 130

asymmetric information between bank managers and government 47±8 security design 254±6, 257±8, 261±5 stock market functioning 277±9 see also information at-the-money options 210±11 Auerbach, A. 110 authoritarianism 294 bailouts, bank see bank bailouts Baird, D. 192, 200 Baker, G. 236 Baltagi, B. 94 bank bailouts 38, 40±1, 48±9 see also rescue programmes bank closures 58, 79 optimal threshold levels 61±2, 66±7, 71, 73 strict closure rules 38±40, 41 bank crises 5, 7, 31±50, 292±3 ABF model 33±8 banking policy in 1±2 closure v. bailouts 38±41 conditional bank recapitalisations 41±4

global 133

high-powered incentives 44±7

Japan see Japan

partial information asymmetry

47±8 quality of bank managers 48±9 see also injection of public funds; rescue programmes Bank for International Settlements (BIS) 49, 150, 155 Bank of Japan (BOJ) 85, 108, 141, 144, 149±50

amakudari system 152±3

emergency loans 142±3

bank managers 34±6 heterogeneous quality 48±9

297

298 Index bank managers (continued ) high-powered incentives 44±7 partial information asymmetry between government and 47±8 banking 4±7, 291, 292±3

fragility of system 1

liquidity demand of corporate

sector see liquidity demand of corporate sector

policy in banking crises 1±2

see also banks

bankruptcy 7±8, 183±203 economic issues and tradeoffs: 186±201; bargaining in Chapter 11 190±2; creditors' incentive to race to be ®rst 187±8; ®ltering failure 192±6; investment incentives, priority rules and bankruptcy decision 188±90; nonbankruptcy workouts, costly bankruptcy, shirking and underinvestment 196±201 legal procedures 183±6;

liquidation 183±4; payoff

patterns 186;

reorganisation 184±5

bankruptcy decisions 188±90 banks cost of capital at banks 110±29;

estimated results 122±9;

estimation methods 110±22;

motivation 110

governance structure see corporate governance injection of public funds see injection of public funds

bargaining, in Chapter 11 190±2

Barro, J. 245

Barro, R. 245

Bebchuk, L.A. 190, 200, 201

Begley, J. 205

È f, E. 201

Berglo

Berle, A.A. 89

Besanko, D. 55

Bester, H. 201

Bhattacharya, U. 256, 265, 267±8, 274

`big bang' 149, 292

Bizjak, J. 225

Blackwell, D. 245

board of directors Japan and US 235±6 retention of executives and incentives 236

see also retention of retiring

executives

Bolton, P. 32, 33, 55, 86, 91, 201

Boot, A.W.A. 155, 275

Brander, J. 204

Brickley, J. 225, 244, 245

bubble economy 94, 107, 129±30

Bulow, J. 188

Cambridge Shopping Centers 205±6

Campbell, T.S. 54

Canada: executive options 212±25

capital adequacy regulation 150±1

capital markets

and disciplining bank management 134; failure in Japan 171±3; lack of capital market discipline 140±1; safety net in Japan 140±7 securitisation of corporate

®nancing 1, 3±4

worldwide structural changes 291,

292, 295

see also security design

capital requirement, initial 77

capital stock, adjustment speed

of 159±65, 166±8

Castanias, R. 221

CAT-bonds 266

chairman of the board (COB) 235, 244

Chan, Y.-S. 54, 56

Chang, H.F. 190

Chapter 7 183, 184

Chapter 11 184±5

bargaining process 190±2

chief executive of®cers (CEOs) 233,

235

®rm leverage and rewards for risk-

taking 219±25

retention on the board 237±47; and performance in US 237±44; relative-performance evaluation 244±7 Chuo Trust Bank 119, 127

Index 299 city banks 93±129 cost of capital 110±29 productivity and structure of shareholders 93±109

closures, bank see bank closures

Coles, J. 225, 244

collateralised mortgage obligations

(CMOs) 253

common stocks 54, 58, 73, 74, 76,

80±1

competition 134

discipline of bank management in

Japan 147±9

competition-restricting

regulations 147±8

concentrated ownership 140

conditional bank recapitalisations

41±4 high-powered incentives 45±7 information asymmetry between bank managers and government 48

continuation decision 21, 22

convoy administration 148

corporate governance 1, 2±3, 7±9

bankruptcy see bankruptcy banks' governance structure and

business performance 6,

85±132; cost of capital 87,

110±29, 130; productivity and

structure of shareholders

91±109; role of debt-

holders 87±8, 91, 129

executive option plans see executive option plans East Asian economies 291, 293±5 Japanese banks see Japan retention of retiring executives see retention of retiring executives corporate sector: liquidity demand 4±5, 13±30 cost of capital 87, 110±29, 130

estimated results 122±9

estimation methods 110±22

motivation 110

costly bankruptcy 196±201

coverage caps 55, 78

cramdown 184

credit crunch 139±40

credit line 16, 28±9 creditors bankruptcy procedures and 183±5 incentive to race to be ®rst 187±8 Cuny, C. 212, 276

Daiichi-kangyo Bank 112, 123, 125

Daiwa Bank 115, 125, 171

debt 110

debt-holders 86

banks 87±8, 91, 129

Demange, G. 256, 267, 275, 276

DeMarzo, P. 255±6

deposit insurance 91

coverage caps 55, 78

injection of public funds into banks

under 5±6, 51±84 Japan 143±5 trustworthiness of the system 77±8 Deposit Insurance Corporation

(DIC) 51, 143±4, 145, 146, 147,

169

deposit insurance premium 53, 56

depositors 91

deregulation 88

delayed in Japan 148±9 policies 295

Dewatripont, M. 15, 20, 88

disclosure of non-performing loans 5,

31±50, 135±6, 137

dividends payout, propensity to 129,

130

Dollar/Yen agreement 149

Dow, J. 256

Dow Jones News Retrieval (DJNR) 237

Dreyfus, J.-F. 55, 78

Dryden, N. 147

Duf®e, D. 255±6, 275

East Asian economies 291, 293±5

Eberhart, A.C. 186

ef®ciency

®ltering failure in bankruptcy

192±6

management ef®ciency of

banks 91±109

effort 35

executive option contracts 227±9

shirking 196±201

300 Index emergency loans 142±3 employees, number of 94, 95±103 equity value maximisation 8±9, 204±32 v. ®rm value maximisation 208±10 ex ante utility 274, 279±80 executive option plans 8±9, 204±32 empirical evidence 212±25; exercise prices and cross-sectional differences in leverage 214±17; ®rm leverage and managers' rewards for risk-taking 217±25 option contracts and optional risk incentives 208±11; at-themoney options 210±11; option design 208±10 option design with managerial wealth constraints and target capital structure 226±7 option design to motivate both effort and project selection 227±9 exercise prices 209±10 at-the-money options 210±11 undoing cross-sectional differences in leverage 214±17 feasibility constraint 61

Feltham, J. 205

®ltering failure 192±6

®nancial innovation 3±4

and investment decision 267

security design with pre-existing

market 9±10, 253±71 Financial Supervision Agency 146±7 ®rms ABF model 33±4

governance structure and

management of 86, 89±90

liquidity demand of corporate

sector 4±5, 13±30

size 221, 224, 225

value maximisation 8±9, 204±32;

v. shareholder value maximisation 208±10

¯oating charge creditors 184

forbearance policy 55, 57, 59

danger of 143

Forbes compensation surveys 237

France 184, 185, 189, 193

Franks, J. 186, 196

Fries, S. 32, 33, 55

Fuji Bank 113, 123, 124, 171, 172

Gale, D. 257, 267, 275, 276, 281

Garcia, G. 155

Garvey, G.T. 226

general equilibrium 26±8

Germany 184, 185, 188, 189, 193, 196

Gertner, R. 188

Giammarino, R. 54, 218

Gibbons, R. 237, 245

Gilson, S.C. 196, 208, 223

global bank crisis 133

Gorton, G. 155

governance structure see corporate

governance

government

intervention and liquidity

demand 29

partial information asymmetry

between bank managers

and 47±8

role in bank managerial governance in Japan 149±54; bank capital and amakudari 151±4; capital adequacy regulations 150±1 see also Ministry of Finance

gradualism 148±9

Greenbaum, S.I. 56

gross income 158±9, 160, 161

Grossman, S.J. 273

Guay, W. 207, 212±13, 218, 219, 225

Harris, M. 204

Hart, O. 186, 196

Hashimoto, R. 149

hedging 280±1

pre-existing securities and 255, 262,

263±5, 266±7

Heiwa-Sogo Bank 141±2

Hellman, T. 148

high-powered incentives 44±7

Hirota, S. 91, 93

Hirshleifer, J. 274, 275

Hokkaido-Takushoku Bank 116, 125,

145

Holmstrom, B. 14, 19, 20, 27, 28, 244

Horiuchi, A. 88, 149, 152

hostile takeover 90

Index 301 Hotchkiss, E. 186 Huddart, S. 212 Hull, J. 219 Hyogo Bank 142 Ikeo, K. 87, 91 incentive compatibility constraint 60±2 incentives board retention 9, 233±49 effects of recapitalisation policy 5, 31±50 entrepreneurs and liquidity shocks 15, 20±5 executive stock options in levered ®rms 89, 204±32

high-powered 44±7

investment incentives 188±90

incomplete contract 86 incompleteness of markets 276 individual rationality constraint 60±1 Indonesia 294 Industrial Bank of Japan 120, 123, 128 information asymmetric see asymmetric information investment, security design and 10, 272±88 partial asymmetry 47±8 information effect 279 initial capital requirement 77 initial public offerings (IPOs) 267, 269, 275±6 injection of public funds 5±6, 51±84 extensions of basic model 74±9; caps on coverage of insured deposits 78; endogeneity of initial capital requirement 77; ex post moral hazard 74±6; interbank loan transactions 78±9; interest rate of subordinated bonds 76; more general class of forms of security 77; social cost of public funds 77; trustworthiness of deposit insurance system 77±8; variety of preferred stocks 76

Japan 51, 52, 172±3 model 55±62; basic environment 55±9; regulator's problem 60±2 optimal regulatory policy 62±74; common stocks 73; optimal security design 73±4; preferred stocks 67±73; subordinated bonds 62±7 inside directors 235 insider trading 275±6 insurance companies 92±109, 130, 140±1 insurance gains 274±5, 279±81 insurance premium 53, 56 interbank loan transactions 55, 78±9 intercorporate shareholding 6, 85±132 passim banks' productivity 91±109 cost of capital at banks 87, 110±29 governance structure and ®rms' management 89±90 interest rate of subordinated bonds 76, 81 internal governance capability 107 international comparisons 155±9 internationalisation of corporate ®nance 149 investment decisions ®nancial innovation and 267 liquidity demands of corporate sector and 4±5, 13±30; optimal investment level 22±5; sequence 17±18 security design, information and 10, 272±88; model 276±7; optimal investment choice 281±3; optimal security design 279±81; stock market 277±9 investment incentives 188±90 Itoh, H. 236 Iyo Bank 144, 145 Jackson, M. 275 Jackson, T.H. 187 Janakiraman, S. 245

302 Index Japan banking crisis 5, 7, 31±2, 133±4, 135±40, 295; danger of vicious circle 139±40; disclosure of non-performing loans 135±6; extended de®nition of nonperforming loans 136±9 `big bang' 149, 292 board retention of retired executives 9, 234, 235±7, 247 corporate governance of banks 2, 6±7, 85±132, 133±80, 293; adjustment speed of capital stock 159±65, 166±8; cost of capital 110±29; disciplinary in¯uence of market competition 147±9; international comparisons 155±9; productivity and structure of shareholders 91±109; revelation of vulnerability 165±73; role of government 149±54; safety net 140±7; vacuum of governance 155±65 injection of public funds into

banks 51, 52, 172±3

long-term relationships 294

system transition and serious

®nancial problems 291, 292±3 Yen/Dollar agreement 149 Japan Bond and Credit Bank 93, 121, 123, 128 Japan premium 139, 169±72 Jarrell, G. 244 Jensen, M. 15, 20, 204, 236 job loss, risk of 220±3, 224±5 John, K. 204, 205, 206±7, 208, 209, 211, 216 John, T. 204, 205, 206±7, 208, 209, 211, 216 Jorion, P. 212 Kahn, C.M. 79 Kanatas, G. 55 Kane, E.J. 143 Kaplan, S. 233, 236±7 Keeley, M.C. 153

Kim, K.A. 140 Klassen, K. 213 Kobayashi, T. 110 Korea 294 Kornai, J. 15 Kurasawa, T. 90, 110, 111 labour share 93, 94±5, 96±102, 103, 109, 130 Lambert, R. 212, 245 Larcker, D. 212, 245 Laroque, G. 256, 267, 275, 276 Lazear, E. 236 Leland, H.E. 275 leverage 8±9, 204±32 exercise prices and cross-sectional differences in 214±17 and managers' rewards for risktaking 217±25 Lewis, T.R. 54 life insurance companies 92±109, 130 Lindgren, C.-J. 133, 155 linear pricing function 259±61 liquidation 38±9, 58, 59 conditional recapitalisation 41±4 creditors' incentive to race to be ®rst 187±8

®ltering failure 192±6

procedures 183±4

see also bankruptcy

liquidity demand of corporate sector 4±5, 13±30 case with no ex post liquidity shock 18±20 ex post liquidity shock and soft budget constraint 20±5 ¯uctuation of liquidity asset value 25±6 general equilibrium analysis 26±8 model 16±18 provision of credit line 28±9 liquidity shocks 15±16, 17, 28±9, 56±7 ex ante 15, 17, 18, 19±20, 23, 27 ex post 15, 17±18, 20±5; with continuation 21; with termination 22 loans emergency 142±3

Index 303 interbank loan transactions 55,

78±9

non-performing see non-

performing loans

problematic 136±9

sales of 43

subordinated 51, 52, 140±1

London inter-bank offered rate

(LIBOR) 169, 171, 172

Long-term Credit Bank of Japan 93,

120, 123, 128

long-term credit banks 93±129 cost of capital 110±29 productivity and structure of shareholders 93±109

long-term relationships 86±7, 294±5

see also amakudari system;

intercorporate shareholding

LoPucki, L. 186, 188, 223

management ef®ciency 91±109 managers 89±90 banks see bank managers losses in the event of ®nancial distress 223, 225

wealth constraints 226±7

Marino, A.M. 55

market break down 258±9

Maskin, E. 15, 20

Mawani, A. 213

Means, G.C. 89

Meckling, W.H. 204

mergers and acquisitions 85

hostile takeovers 90

Miller, M.H. 218

Ministry of Finance (MOF) 91, 108,

148, 149

amakudari system 88, 152±4

capital adequacy regulation 150

inadequate monitoring 85

problematic loans 136±8

safety net 141±2, 143, 144; growing

dif®culty in bailing out bank failures 145

Mitchell, J. 33, 55

Mitsubishi Trust Bank 118, 126

Mitsui Trust Bank 118, 126

Miyazaki, M. 91±2, 93

Modigliani, F. 218

monitoring level 56, 60±2, 72±3

Moore, J. 196

Moore Business Forms 205±6

moral hazard 2, 14, 17, 91

in bank lending 35

injection of public funds into

banks 53±4, 66±7, 71±2, 73, 74;

ex post moral hazard 74±6

mortgage-backed securities (MBS) 253, 254, 255, 258, 266

Murdock, K. 148

Murphy, K. 221, 236, 237, 245

mutual insurance 79

Nagakubo, R. 90

Nagarajan, S. 55

Nickell, S. 147

Nicolitsas, D. 147

Nihon Trust Bank 119, 127

non-linear price function 265±6

non-linear transfer pricing

scheme 41±4

non-performing loans 152±3, 154

incentive effects of bank recapitalisation 5, 31±50; recapitalisations conditional on liquidation 41±4 Japan 135±40; danger of vicious circle 139±40; disclosure 135±6, 137; extended de®nition 136±9  Ohashi, K. 256

Okazaki, T. 110, 111

operating expenses 158±9, 160, 161

Osano, H. 77

outside directors 235

panics 79

Patrick, H. 151±2

payoff patterns 184, 186

per capita value-added 93±109

performance

banks' and their governance

structure 6, 85±132

and board retention of retiring

executives 9, 233±49

Poitevin, M. 204

Povel, P. 33

304 Index pre-existing securities 9±10, 253±71 preferred stocks 51, 52, 54, 58, 74, 76, 79 optimal regulatory policy 67±73 variety of 76 price-cash ¯ow ratio (PCFR) 111±22 price-earnings ratio (PER) 111±22 price formation 278±9 linear 259±61 non-linear 41±2, 265±6 priority rules 183, 188±90 privatisation 46 problematic loans 136±9 productivity, banks' 91±109 pro®tability 156±8 promotion prospects, as incentive 233, 236 prompt corrective action (PCA) 139, 146±7 Prowse, S.D. 140 Prucyk, B. 217 prudential regulation 150±1 public funds injection into banks 5±6, 51±84 social cost 77 Pyle, D.H. 275 quality of bank managers 48±9 Rahi, R. 256, 276 Rasmussen, R. 200, 201 rational expectations equilibrium (REE) 258±9 Raviv, A. 204 recapitalisation policy 5, 31±50 ABF model 32, 33±8 bank managers of heterogeneous quality 48±9 high-powered incentives 44±7 incentive effects 38±44; closure v. bailouts 38±41; conditional recapitalisation on liquidation of non-performing loans 41±4 partial information asymmetry 47±8 regional banks 92 amakudari system 152±4 regulation banks in Japan 85, 108±9; capital adequacy regulation 150±1;

disciplining bank management 134, 149±54 injection of public funds into

banks 5±6, 51±84; optimal

regulatory policy 62±74;

regulator's problem 60±2

regulator in Aghion et al. model 36±8 relative-performance evaluation 244±7 rents 148 Reny, P.J. 268, 274 reorganisation, bankruptcy 184±5 ®ltering failure 192±6 representative directors 235 rescue programmes bailouts 38, 40±1, 48±9

safety net in Japan 141±7;

limitation 145±7

see also injection of public funds

restructuring in banking international comparisons 155±6 transition economies 32±3 retention of retiring executives 9, 233±49 and agency problems 235±7 board retention and incentives 236 Japan 9, 234, 235±7, 247 relative performance evaluation 244±7 US see United States return on assets 238±44 Rhee, S.G. 140 risk executive option plans and incentives to take risk in levered ®rms 8±9, 204±32 investment, information and security design 282±3 Roberts, W. 79 Rochet, J.-C. 55, 78 Roe, M.J. 200 Rosen, R. 155 Rosen, S. 236 Saal, M.I. 155 safety net 140±7, 165±9 comprehensive 142±3 danger of forbearance policy 143 deposit insurance 143±5

Index 305 lack of capital market

discipline 140±1

limitation of traditional rescue

method 145±7

mechanisms 141±2

Saito, T. 92

Sakura Bank 112, 124

salary 221

see also staff costs

sale of loans 43

Sanwa Bank 114, 123, 124

Sappington, D.E.M. 55

Saunders, A. 55, 78

Scharfstein, D. 188, 201

Schwartz, A. 196, 201

Sealey, C.W. 55

secured creditors 183

securitisation of corporate

®nancing 3±4 security design 3±4 investment, information and 10,

272±88; investment

choice 281±3; model 276±7;

optimal security design 279±

81; stock market 277±9

optimal in injection of public

funds 54, 55, 73±4; state-

contingent securities 77

with pre-existing market 910, 253±71; case without preexisting security 265; characterisation of viable design 262±5; equilibrium with linear pricing function 259±61; equilibrium and market breakdown 258±9; model 257±9; non-linear price function 265±6; viable design of created security 259±65 shareholders 6, 86

intercorporate shareholding see

intercorporate shareholding role in ®rms' management 89±90 structure of and banks' productivity 91±109, 129±30 value maximisation 8±9, 204±32; v.

®rm value maximisation 208±

10

Sheard, P. 151±2

Shiller lag technique 162, 166±8, 178

Shimizu, K. 88, 152

shinkin banks 150

shirking 196±201

Shleifer, A. 90

short-sales constraint 266±7

Shoven, J. 188

size, ®rm 221, 224, 225

small traders 273, 278, 280±1

Smith, A. 245

social cost of public funds 77

soft budget constraint 15, 20±5, 27±8,

28±9

sogo banks 150

spanning effect 253

speculative gains 274±5, 279, 281

Spiegel, M. 268, 274

staff costs 156±8

state-contingent security design 77

Stiglitz, J. 148, 273

stock adjustment 159±65, 166±8

stock market 273, 277±9

stock options see executive option

plans

stock returns 238±44

strategic defaults 34, 35, 39

strict closure rules 38±40, 41

Stulz, R. 15

Suarez, J. 33

subordinated bonds 54, 57±8, 74,

75±6, 79

interest rate 76, 81

Japan 51, 52

optimal regulatory policy 62±7

subordinated loans 51, 52, 140±1

Sumitomo Bank 115, 124, 141±2

Sumitomo Trust Bank 119, 126, 171,

172

supervision by regulatory

authorities 134, 149±54

Swan, P.L. 226

T-bonds 254, 258, 266

Tachibanaki, T. 85, 88, 90, 107, 108

Takano, M. 111

Taki, A. 110, 111

Tashjian, E. 196

termination 22

Thadden, E.-L. von 201

306 Index Thailand 294

Thakor, A.V. 56, 275

Tirole, J. 14, 19, 20, 27, 28, 55, 78, 88,

214 Titman, S. 216, 219

Toft, K.B. 217

Toho Sogo Bank 144, 145

Tokai Bank 116, 125

Tokyo inter-bank offered rate

(TIBOR) 169

Tokyo Kyodo Bank 142

Tokyo-Mitsubishi Bank 113, 123, 124,

171, 172

Tokyo-Mitsubishi Trust Bank 127

top ten shares 94, 95, 96±102, 103,

104±6, 107

Torous, W. 186, 196

total asset value 94, 96±102, 103, 107,

108±9, 130

Toyo Trust Bank 118

transition economies 32±3

trust banks 93±129

cost of capital 110±29

productivity and structure of

shareholders 93±109

Ueda, K. 144

underinvestment 196±201

United Kingdom (UK) 184, 185, 188,

189, 193, 196

United States

bankruptcy 188; bargaining in

Chapter 11 190±2; ®ltering

failure 193±6; investment

incentives, priority rules and

investment decision 188±9;

procedures 183, 184±5, 186,

192

board retention of retired

executives 9, 234, 235±6,

237±44, 247; past ®ndings

239±44; pre-retirement

data 238±9; sample

selection 237±8

Yen/Dollar agreement 149

Vancil, R. 244

venture capitalist 10, 272±88 ex ante utility 279±80 investment choice 281±3 security design 280±1 and stock market 277±8 Verrechia, R. 212

Vetsuypens, M.R. 196

vicious circle 139±40

Vishny, W. 90

vulnerability 165±73

Warner, J. 236

Watts, R. 236

wealth constraints, managerial 226±7

Webb, D. 187

Weisbach, M. 236, 245

Weiss, L.A. 186

Wessels, R. 216, 219

White, M.J. 185, 188, 194

Whitford, W. 186, 223

workouts 196±201

World Bank: `Rethinking the East Asian

Miracle' research project 294

Wruck, K. 236

Yasuda Trust Bank 118, 126

Yen/Dollar agreement 149

Yermack, D. 205, 206, 211, 213, 216

Yonezawa, Y. 90, 91±2, 93, 110