Regional Financial Markets : Issues and Policies 9780313059278, 9781567205732

Citation preview

Regional Financial Markets

This page intentionally left blank

Regional Financial Markets Issues and Policies EDITED BY DILIP K. GHOSH AND MOHAMED ARIFF

+++++++ R

Westport, Connecticut London

Library of Congress Cataloging-in-Publication Data Regional financial markets : issues and policies / edited by Dilip K. Ghosh and Mohamed Ariff p. cm. Includes bibliographical references and index. ISBN 1-56720-573-9 (academic) 1. Capital market—Asia, Southeastern. 2. Finance—Asia, Southeastern. 3. Financial crises—Asia, Southeastern. I. Ghosh, Dilip K. (Dilip Kumar), 1942- II. Mohamed Ariff, 1940HG5740.8.A3R44 2004 332' .095—dc22 2003026529 British Library Cataloguing in Publication Data is available. Copyright © 2004 by Dilip K. Ghosh and Mohamed Ariff All rights reserved. No portion of this book may be reproduced, by any process or technique, without the express written consent of the publisher. Library of Congress Catalog Card Number: 2003026529 ISBN: 1-56720-573-9 First published in 2004 Praeger Publishers, 88 Post Road West, Westport, CT 06881 An imprint of Greenwood Publishing Group, Inc. www.praeger.com Printed in the United States of America

The paper used in this book complies with the Permanent Paper Standard issued by the National Information Standards Organization (Z39.48-1984). 10+9 8 7 6 5 4 3 2 1+++

Contents

Foreword by Ahmad Fazvzi b. Mohd Basri+++ii

+++

Acknowledgments

ix

Introduction by Dilip K. Ghosh

xi

1 From Turbulence to Tranquility: Reform and Restructuring in Malaysia Dilip K. Ghosh 2 Portfolio Investments in Australia: 1988-1997 W. Leong and ]. Wickramanayake

1 13

3 The Level of Managerial Ownership, Leverage, and Dividend Policies: Hong Kong Evidence Rohit Jain and Kam-wah Lai

33

4 Examining the Performance of the Malaysian Life Insurance Sector: Efficiency and Productivity Growth Alias Radam and Shazali Abu Mansor

51

5 Assessing Corporate Financial Distress in Malaysia M. S. Zulkarnain and M. Shamsher 6 Cross-Sectional Predictability of Stock Returns at the Colombo Stock Exchange Elyas Elyasiani, Priyal Perera, and Tribhuvan N. Puri 7 Linking Commercial Bank Lending into Determinants of Price Level in Malaysia Tuck Cheong Tang

73

95

117

vi

8 Optimal Financial Structure: A Modigliani and Miller Dynamic Model for Mexican Corporations Edgar Ortiz and Jonathan Torres 9 Thai Capital Market Integration Using Relevant Assets Pornanong Budsaratragoon and Sunti Tirapat 10 Intraday Systematic Patterns, Lead-Lag Relationships, and Pricing Efficiency: Evidence from the Kuala Lumpur Composite Index Futures Fauzias Mat Nor and Tea Lee Ghoo 11 The 1997-98 Asian Financial Crisis, Corporate Restructuring, and Earnings Per Share of Listed Companies Zainal A. Mohamed, Haim H. Abdullah, and Ahmad Yakob

Contents

137 157

181

223

12 The National Economic Recovery Plan: Perception of Financial Market 247 Nasruddin Zainudin, Nor Hayati Ahmad, Engku Ngah Engku Ghik, Nik Kamariah Nik Mat, and Che Ani Mad 13 Cost Efficiency and Economies of Scale in Asian Banking Mohd Zaini Abd Karim

265

14 Factors Influencing Capital Structure in Three Asian Countries: Japan, Malaysia, and Pakistan Mansor Isa and Mohamad Mahmud

283

Index

303

About the Editors and Contributors

315

Foreword

This book is a research monograph collaboratively worked on by two of our chair professors at the Universiti Utara Malaysia (UUM). It is with great pleasure that I write this foreword for the volume Regional Financial Markets: Issues and Strategies. About hundred scholars from UUM and from around the world pooled their talents to conduct research needed for writing the chapters in this book. The First International Conference on Banking and Finance, organized in late August 2000, provided the focus for these studies, and upon feedback and revisions on their original results, some of those studies finally got selected to be included in this volume. To these scholars, I would like to extend my sincere appreciation for the contributions. This final product would not have been possible but for the funding of the two chairs by the Kuala Lumpur Stock Exchange and Bumiputra Commerce Bank. I take this opportunity to thank these Malaysian corporations for their contributions to the scholarly work being done through their generous support of the UUM research efforts. Finally, it gives me great pleasure to compliment the two chair holders—Professor Dilip K. Ghosh and Professor Mohammed Ariff—for their dedication and interest to examine the issues of our time and to bring our university to this level of international visibility Kol. Prof. Dato Dr. Ahmad Fawzi b. Mohd Basri, Vice Chancello Universiti Utara Malaysia January 2002

This page intentionally left blank

Acknowledgments

Regional Financial Markets: Issues and Strategies is the twin volume of Global Financial Markets: Issues and Strategies. Both these projects started at the First International Conference on Banking and Finance under the tutelage of Universiti Utara Malaysia(UUM) and under the nursing care of Kuala Lumpur Stock Exchange endowment and Bumi Commerce endowment for the two chairs we respectively hold. At the end of the Asian financial distress of 1997-1998, numerous scholars began to post mortem the crisis, and there were lot of bickering all around. Instead of going with the glow of the moment, we decided to take a sober stand and gather scholars from all around the world who could analyze issues dispassionately and effectively. Our decision resulted in the Conference we have alluded to, and out of that Conference comes the genesis of this book. We chose to see the research from the global point of view and also from a regional point of view. Here we have the analysis of regional financial markets and the embedded issues and strategies. Several authors have looked at several issues that invited their attention, and we try to fit them in this book in selective way. We thank everyone who contributed to this project, and we sincerely hope our effort makes some difference. We cannot express out indebtedness to everyone by name, and yet we cannot avoid mentioning our dear helpers: our two dedicated secretaries, Mrs. Badariah Zahir and Mrs. Azura Othman, who always stood behind us with smiling face and tired fingers. They became the memory cells for instant information, and for that we remain immeasurably grateful. Mr. Hussein, whose dedication and honest work many may not recognize, was an unbelievable help. The university support staff, members of SWB

X

Acknowledgments+

faculty at UUM were superb in their acts of cooperation and commitment. We simply salute them. Finally, we must recognize our families, friends and well-wishers who kept us going on projects that involved many continents and many seasons of doubts and delight. Dilip K. Ghosh M. Ariff+

Introduction

Our world is a collection of islands, and each island is a nation. All nations maintain their independence and territorial integrity, and yet these nations are all dependent. Although outwardly paradoxical, the nations are all virtually and effectively interdependent. We all have something more than we need and we all have something less than we must have to maintain a healthy and wholesome well-being. International trade came into existence because of this realization, and later international investment— foreign direct investment and portfolio investment—followed suit in an effort to increase the potential economic gain. Of course, over the years, with the progress of time and technology, these so-called isolated islands have come closer and closer to few of them more than a few others because of similar resource endowments, geographical proximity, linguistic linkages, political alikeness, cultural affinity, and so on. Regional blocs have grown more and more, and thus the world appears to have been a number of oligopolists in their economic aspirations and financial strategies. The Treaty of Rome envisioned a United States of Europe like the United States of America as a power player, and it took many decades through different structures of snake, supersnake, and scrip currency like ECU to come to the current state o+ the European Monetary Union with its composite currency unit, the euro. The North American Free Trade Association (NAFTA) is already in existence. Many other trading blocs, such as LAFTA, AFTA, and custom unions, are the regional blocs in our days, operating, competing, and contesting in the market place. In the global village with global structure o+

XII

Introduction

financial markets, the regional markets are visible, potent, and rigorous with their functioning and trading. Against the backdrop of this reality many results of recent vintage need further scrutiny and analysis. Since the free fall of Thai baht, an Asian crisis took place, and the turbulence swept through the East Asian countries. It was more like a plague creating havoc to Thailand, Indonesia, Korea, Malaysia, and the Philippines mostly, but the tremor and strain of the virus afflicted other countries in the area beyond those regional economies. In this book, therefore, we initiate a study on the move from this turbulent period to tranquility through an attempted restructuring of the banking institutions in the Malaysian economy where the government defied the IMF prescription and rescue measures. Although no definitive answer can yet be given since the data are hardly available, a theoretical discussion is initiated to examine the issue of consolidation and mergers of many banks under several anchor institutions. In a separate study, we ask the question: do bankers make rational economic predictions? In another chapter, cross-sectional predictability of stock returns in the Colombo stock market is discussed. The characteristics of residuals of the market-adjusted return are important, and these characteristics are adumbrated and analyzed with reference to the Kuala Lumpur stock exchange. Prediction goes a step further in this regional structure, and we present forecasting corporate failure. The failure of 24 deposit-taking cooperatives and the determinants thereof are then brought to light in order to ascertain which cause(s) what. We then move from prediction to performance, and in this context, the issue of nonperforming loans is reviewed, and the extent of the severity of the problem is highlighted. But nonperforming loans by banks alone may not paint the picture of economic malaise, and so an attempt is made in another chapter to look into the performance, efficiency, and productivity of the life-insurance sector, and both studies are conducted in the Malaysian economy since its government and policy makers were single-handedly dealing with the crisis environment with no help from the outside world being solicited or accepted. In a different study we take a close look at portfolio investment in another region of the world: Australia. Money supply has a role on the propagation of price inflation, but how that relationship holds in a specific situation is analyzed empirically in a country-economy by examining the link between commercial bank lending and determinants of price level. In an apparently unrelated study we examine audit delay, its determinants, and ramification in the financial calculus in a regional context. An examination is made on the characteristics of residuals of marketadjusted return vis-a-vis single index market model based on Kuala Lumpur Stock Exchange data. Earnings per share have a lot of implications in the financial infrastructure, but often we see the relationship with dividend distribution, growth, leverage, and mergers under normal economic

Introduction

XIII

conditions. In a formal fashion the relationship amongst three flow variables—cash flows, earnings and dividends—are brought out. Here we further study earnings per share of listed companies under financial turmoil in a crisis-laced economy A chapter is devoted to the factors in capital structures in three separate Asian economies, one developed and two less developed, and observations are made on leverage on differing economic structures. Then optimal capital structure, discussed under the Modigliani-Miller value invariance paradigm in a static theoretical environment, is thrown into relief in a dynamic set-up in the financial environment of Mexico. Dividend policy and deregulation are then examined with reference to a specific industry, and it extends our theoretical background to a practical case. In a separate study we show intraday systematic patterns, intraday lead-lag relationship, and pricing efficiency with the evidence taken from a regional composite index futures. Many other issues are highly important, and yet we have limits to cover many of the issues and interests we have in search of analytical answers and empirical validity. So, with limited space and time we cover market timing and selection skill on retail superannuation funds. We also discuss and explore efficiency and scale economy with reference to the Asian banking industry. In that context we move to the issue of linking bank lending to price level in a regional economy in isolation. We then focus on a rather nascent development: the growth, the meaning, and the rationale for the Islamic financial system. Two important studies highlight the relevance and the efficacy of alternative regional, but potentially global, financial management. Finally, in a chapter we attempt to examine the perception and reality of the crisis-management policy of a country determined to go on its own way with a determined mind to restore equilibrium and order to its own economic house. The case in point is the national economic recovery plan of Malaysia. We never pretended that we would get all answers to every regional issue, but a start has been made through this book. Research is an ongoing process, and if others do forward our interest a bit further we will feel gratified. Dilip K. Ghosh

This page intentionally left blank

CHAPTER 1

From Turbulence to Tranquility: Reform and Restructuring in Malaysia Dilip K. Ghosh

Dramatic events in Asia, Russia, and Brazil have generated torrents of comments about exchange rates, hot money, exchange controls, currency boards, // dollarization, ,, and a new global financial infrastructure. Malaysia suffered as much as Thailand and other Asian countries in the Asian flu. It could not get inoculated against the financial contagion of the neighbors: Thailand, Indonesia, Korea, the Philippines, Hong Kong, and even Japan. It got sucked into this mess because of the international interdependence, trade linkages, and flows of net earnings and capital. It has been estimated that international capital flows were at the root of Asian financial crisis. They also remain critical to Asia's future. Note these inflows financed high investment rates and current account deficits, raising sustainable growth rates and living standards. However, large pre-crisis capital inflows also inflated Asian economies' real exchange rates, undermining export competitiveness and encouraging excessive resource allocation to non-traded sectors, and thus eventually destroying the foreign creditor confidence—all contributing to the crisis. Contraction in capital flows (to the tune of US$ 150 billion to five Asian countries: Hong Kong, Singapore, Thailand, Malaysia, and Indonesia) or capital flight caused this plague. Foreign direct investment (FDI) remained relatively stable, but portfolio investment and other flows, mainly bank lending, dropped significantly Just prior to this financial crisis, the stock market in Malaysia was capitalized at 323 percent of GDP, and by the end of 1997, the capitalization came down to 136 percent of GDP. The shares of net funds raised in Malaysian financial markets in 1996 were as follows:

2

Regional Financial Markets

Domestic Bank Market

50%

Domestic Debt Market

33%

Equity Market

13%

External Debt Market

4%

The ringgit lost its stability following the fall of the Thai baht and began to float freely It depreciated to RM 4.76 to the U.S. dollar in 1998, and some press reports indicated that in early 1998 RM 54 billion (20% of Malaysian GDP) got parked outside the jurisdiction of Malaysia. In August 1998, the prime minister announced the cabinet decision that the ringgit would be pegged at RM 3.80/U.S. dollar, effective September 1998, and other measures of control were put in effect. The ringgit was pegged at 3.80 per U.S. dollar, all ringgit trade outside was banned, and short-term deposits by foreigners were subjected to exit tax. In the pre-crisis period nonperforming loans were officially at 3.6 percent of the total bank loans, while the bank's average risk-weighted capital adequacy ratio was 11.8 percent—well above the BIS-mandated minimum of 8 percent. Malaysia's current account deficit was high at 5.8% of GDP, and yet it had a good match of foreign liabilities and assets, and foreign reserves exceeded short-term debt. Like Indonesia, Thailand, and Korea, Malaysia had high pre-crisis rates of private sector credit growth with increasing real estate and equity price inflation, real exchange-rate appreciation, and slowdown in exports. New credit expanded over 25% per year between 1995 through 1997; property and business services were the biggest destinations. Consequently, Malaysia's corporate sector became highly leveraged. By 1997, bank loans were around 197% of GDP. A good number of nonperforming loans (NPLs) came into being. Throughout 1998 and 1999, NPLs grew to levels where, without external assistance, some banks' capital adequacy ratios would have dropped below the BIS-mandated 8%. By the end of 1998 gross NPLs of the banking system reached 20% of outstanding loans. The growth rate of gross domestic product dropped ( — 7.5 percent in 1998 from its previous highs (7.8 percent in 1997, 8.6 percent in 1996, and 9.5 percent in 1995). Foreign exchange reserves to mobile capital reached from 171 in 1991 to 55.9 in 1997. The real exchange rate by the first half of 1997 was 70.6 compared to 103.1 (for Taiwan), 89 (Korea), 73 .2 (for Singapore), 68.9 (for Indonesia), and 72.3 (Philippines). It then became a period of financial turbulence for Malaysia. In the words of its prime minister, Dr. Mohamed Mahathir, "Then the unexpected happened. The Asian miracle was shattered and suddenly once fawning economists argued that all it really had been was a bubble, over-inflated by corruption, cronyism and bad loans." These steps already taken by the Malaysian government in response to this turbulence stemmed the upward mobility of both interest rates and

From Turbulence to Tranquility

3

inflation rate. In February of 1999, some relaxation on the rules of shortterm funds were announced, and business houses were assured that those controls would not be used to deny them of the access to foreign currencies for trading. However, unlike others, it did not seek help from International Monetary Fund. REFORM A N D R E S T R U C T U R I N G Prudential Regulation After the crisis surfaced, the government tightened many prudential regulations, although some were subsequently relaxed to ease the severity of the credit crunch. The government tightened loan classifications in September 1997, and a year later it relaxed the NPL classification to six months. For finance companies, the minimum capital requirement increased from RM 5 million to RM 300 million by the end of 1999 and to RM 600 million by the end of 2000. In addition, the government raised the risk-weighted capital adequacy ratio for these finance companies from 8% to 9% by December 1998 and to 10% by December 1999. For finance companies the minimum capital requirement increased from RM 5 million to RM 300 million by the end of 1999 and to RM 600 million by the end of 2000. In addition, the government raised the risk-weighted capital adequacy ratio for these finance companies from 8% to 9% by December 1998 and to 10%) by December 1999. These new policies are giving rise to mergers and acquisitions among finance companies and commercial banks. Bank Negara Malaysia (BNM) has been introducing consolidated supervision of all financial institutions in accordance with international best practice. BNM's new liquidity requirement policy replaces previous high liquidity ratios with liquidity requirements so that banks can meet the shortterm liquidity needs arising from their liabilities' maturity profiles. The government recognizes the need to reduce reliance on bank lending and upgrade capital market regulations. In 1999 new risk-based capital adequacy rules have forced risk-taking brokers to have adequate protection. The Government is also improving disclosure standards and moving stock-market regulation from a merit-based to a disclosure-based system. Other recent reforms include tightening regulations on related-party transactions, introducing new disclosure requirements for nominee accounts, tightening definitions and penalties for insider trading, and increasing Securities Commission inspection and enforcement powers. There is a lack of an over-the-counter market. Although the Central Depository System (CDS) is in place, still the impression exists that clearing system is relatively poor. SCORE and WINSCORE appear to be in good health. KLSELINK is a good information network in place. The approval process for corporate bonds is too slow and too long. BNM, the

Regional Financial Markets

4

Ratings Agency of Malaysia, and the Securities Commission take as many as six months at times to process an approval. In 1999, the SC established the Capital Market Strategic Committee to develop a capital market master plan by early 2000. In February of 1999 the Finance Committee on Corporate Governance proposed a code of corporate governance and recommended reforms to laws, rules, and regulations. Such reforms should improve confidence and market discipline in the securities markets. To eventually expand the corporate bond market and enable institutional investors to bring market discipline on corporations, the government has been relaxing restrictions on insurance companies holding corporate bonds. BNM has been introducing guidelines on securitization and revising its guidelines on issuing private-debt securities to streamline the corporate bond market. This development will diversify risk away from the banking system, increase the variety of fundraising instruments in the market, and widen the spectrum of papers available for investment. The capital market has been expected to grow, partly because of the economy's strong underlying fundamentals, and partly because of the large bond issues by Danamodal to recapitalize the banks and Danaharta's purchase of NPLs, issuance of government-guaranteed zero-coupon bond (MGS, raising RM 10.345 billion so far: RM 15 billion target amount), First Property Sale, and Restricted Tender of Foreign Loan Assets (= 9 Non-Ringgit Loans + 14 Marketable Securities totaling US$ 251.7 million: tender will close on February 22, 2000: Danaharta has a total of US$ 594.4 million). The capital market regulation and supervision reform, stamp duty exemption on corporate bonds, and relaxation on insurance companies' bond purchases should assist the capital market growth. One may find the excitement reflected through the recent KLSE indices in early January of 2000 and can easily ascertain that the growth of the capital market is no longer an anticipation—it has been a veritable reality. Bulls started roaring. The KLSE outperformed most other regional stock exchanges with spectacular gain for the past so many days in early January of 2000. Look at the eidolon of KLSE in those days (data from the New Straits Times): Selected KLSE Indices KLCI

Industrial

Finance

Jan 11, 2000

846.74 ( + 28.31)

1,414.95 ( + 30.23)

7,015.23 ( + 340.83)

Jan 12, 2000

869.62 ( + 22.88)

1,4.29.94 ( + 14.99)

7,380.52 ( + 365.29)

Jan 14, 2000

928.24 ( + 37.64)

1,509.06 ( + 74.68)

7,836.60 ( + 253.75)

Jan 17,2000

953.42(25.18)

1,541.50(32.44)

7,979.92 ( + 143.32)

From Turbulence to Tranquility

5

New Straits Times reports on January 17, 2000, "financial stocks led the splendid ride on the Kuala Lumpur Stock Exchange in the new year, gaining about 1,200 points." The KLSE Finance Index ended previous week at 7,836.6 points. Other indices were on the rise, too. BANKRUPTCY REFORMS The government has closed loopholes in the law to stop companies applying for restraining orders against creditors without their knowledge. However, companies can seek protection (like U.S. Chapter 11) against creditors while restructuring schemes are being worked out (Fitch IBCA, 1999). To protect creditor's rights, the restructuring companies are required to appoint independent directors to their boards, nominated by a majority of creditors, to oversee the restructuring process. The bankruptcy regime does not require independent judicial oversight of corporate workout schemes. This may be a good thing, but at times, it may create bottlenecks and delays in the resolution of conflicts and confusion. NPLs A N D RECAPITALIZATION In the middle of 1998 the government established three institutions to resolve serious banking sector NPL problems and to implement refinancing and restructuring: • Danaharta purchases NPLs from the banking system and maximizes the recovery value of acquired assets; • Danamodal infuses capital to strengthen banking institutions; • Corporate Debt Restructuring Committee facilitates out-of-court restructuring of corporate debt through voluntary agreements between creditors and debtors.

C O N S O L I D A T I N G F I N A N C I A L SECTOR The Malaysian government and Bank Negara Malaysia (BNM) came to the decision that to face competition from international banks and to stave off future financial crises, banking sectors must be further consolidated under a few anchor banks. In response to the strong reactions or objection from the banking sector, the timetable of merger/consolidation was moved, and the number of consolidations was increased to ten. All the 55 banks found partners by the end of this year. These financial institutions are on the move in this matter to generate synergy and gain strength. INSURANCE Malaysia had a high level of international participation in its insurance sector. The insurance law of 1996 forced foreign insurers to sell down their

6

Regional Financial Markets

shares in locally licensed companies to 51% of equity. It also introduced a 30% limit on new foreign investment in existing insurers. However, five foreign insurers were granted joint-venture licenses in February 1999, allowing them to convert their foreign branches into locally incorporated entities with Malaysian equity participation. Under the 1998 World Trade Organization financial services commitments, Malaysia agreed to maintain at last 51 percent foreign ownership of joint-venture companies, grant six new licenses for life reinsurance by June 30, 2005, and allow up to 30 percent of foreign shareholding of two government-owned reinsurance companies. Two observations are in order. It appeared that margin money was freely available and brokerage firms lent liberally. In this process they created sudden expansions and contractions in value-traded, causing market volatility. Secondly, it should be noted that the derivative market must be developed in line with the developed countries of the West. Forward and futures contracts, particularly with the ringgit, must be developed if Malaysia decides to revive the inflow of foreign capital and further growth of the economy. The role and effectiveness of Kuala Lumpur Options and Financial Futures Exchange (KLOFFE) must be streamlined to define the hedging and covered speculation in the interest of profitability and stability of trading environment. Note that not only foreign traders and investors gain by derivatives, but also private sectors and domestic investors can grow and earn foreign currency if they have good hedging instruments. It should be noted that the financial institutions consisted of the following: 36 commercial banks 40 merchant banks 30 foreign bank representative offices (correspondent banks) under 10 offshore banks 6 discount houses These institutions were under the direct supervision of BNM. However, there was no deposit insurance system as one finds in the United States (FDIC). The non-bank financial institutions were comprised of the institutions with large assets (Employees Provident Fund Board, National Savings Bank, Pilgrim Fund Board) and 13 leasing companies, 7 development banks, 60 proprietary and unit trusts, 4 housing finance units, and 59 insurance companies. Following the new policies that gave rise to mergers and acquisitions among finance companies and commercial banks, here came the approved consolidation scheme (from Bank Negara Malaysia, Press Release, February 14, 2000):

From Turbulence to Tranquility

Ten Anchor Banks, 2000 Banking Institutions in Group 1. Malayan Banking Berhad i. Malayan Banking Berhad ii. Mayban Finance Berhad iii. Aseambankers Malaysia Berhad iv. PhileoAllied Bank Berhad v. The Pacific Bank Berhad vi. Sime Finance Berhad vii.Kewangan Bersatu Berhad 2. Bumiputra-Commerce Bank Berhad i. Bumiputra-Commerce Bank Berhad ii. Bumiputra-Commerce Finance Berhad iii. Commerce International Merchant 3. RHB Bank Berhad i. RHB Bank Berhad ii. RHB Sakura Merchant Bankers Berhad iii. Delta Finance Berhad iv. Interfinance Berhad 4. Public Bank Berhad i. Public Bank Berhad ii. Public Finance Berhad iii. Hock Hua Bank Berhad iv. Advance Finance Berhad v. Sime Merchant Bankers Berhad 5. Arab-Malaysian Bank Berhad i. Arab-Malaysian Bank Berhad ii. Arab-Malaysian Finance Berhad iii. Arab-Malaysian Merchant Bank Berhad iv. Bank Utama Malaysia Berhad v. Utama Merchant Bankers Berhad 6. Hong Leong Bank Berhad i. Hong Leong Bank Berhad ii. Hong Leong Finance Berhad iii. Wah Tat Bank Berhad iv. Credit Corporation Malaysia Berhad 7. Perwira Affin Bank Berhad i. Perwira Affin Bank Berhad ii. Affin Finance Merchant

Regional Financial Markets

8 iii. Perwira Affin Merchant Bankers Bhd iv. BSN Commercial Bank Berhad v. BSN Finance Berhad vi. BSN Merchant Bank Berhad 8. Multi-Purpose Bank Berhad i. International Bank Malaysia Berhad ii. Sabah Bank Berhad iii. MBf Finance Berhad iv. Bolton Finance Berhad v. Sabah Finance Berhad vi. Bumiputra Merchant Bankers Berhad viiAmanah Merchant Bank Berhad 9. Multi-Purpose Bank Berhad i. Southern Bank Berhad ii. Ban Hin Lee Bank Berhad iii. Cempaka Finance Berhad iv. United Merchant Finance Berhad v. Perdana Finance Berhad vi. Perdana Merchant Bankers Berhad 10. EON Bank Berhad i. EON Bank Berhad ii. EON Bank Berhad iii. Oriental Bank Berhad iv. City Finance Berhad v. Perkasa Finance Berhad vi. Malaysian International Merchant vii.Bankers Berhad

To examine the effect of merger consider the following picture before the merger: acquiring firm (A) target firm (B)

Earnings

EB

Market value of a share

EA NA ++ VSCA,

NB + *S(B)

Earnings per share

EPSA. =+E A/NA

EPSB = EB/NB

P/E ratio

Vs^y NA/EA+

VS(B).NB/EB

Total equity

^S(A,.N A

V«B).NB

Numbers of shares

9

From Turbulence to Tranquility

If the merger takes place, then the cost of acquiring firm B is obviously VS{B).NB/+and hence addition to the acquiring firm's number of shares mu be U..VS(B)-NB/^S(B) = NN (the number of new shares issued). If ju = 1, obviously, it is a fair exchange. If ju is less than 1, the acquiring firm has discount price for merger, and for ju > 1, the merger is made with a premium price up front. In the case of merger in Malaysia// < 1, and apparently the takeover price is good for the anchor banks. What about the shareholders? Initially, of course, it is bad since in the post-merger situation the shareholders are not getting the fair market value. But a myopic view is not the right measuring rod for the success of any merger. The synergy may surface after a passage of time. Let us then consider the post-merger prospect, and to do so, consider the following:

combined earnings (after merger) = E total number of shares (after merger) = NA + Ns. The earnings per share (EPS) (after merger) = {EA + EB}/|NA + Ns] If the pre-merger P / E ratios are in these ranking situations: then

improves;

^s ( remains unchanged;

deteriorates.

A more analytical probe into the picture of this corporate combination yields the percentage change (growth or decay) in the EPS of the acquiring firm (after merger) by a simple logarithmic differentiation:

where

++++++++++++++++++++++++++++++++++++++++++++++++++++++++As A

10

Regional Financial Markets

O n a long-run d y n a m i c s of change, it is (1.1) that d e t e r m i n e s the efficacy of m e r g e r s in t e r m s of the percentage change in the earnings per share of the a n c h o r b a n k s after merger. O n the Malaysian scene, m e r g e r s are evolving, a n d h a r d d a t a are not available. Consolidation u n d e r anchor b a n k s m a y a p p e a r to give the failing b a n k s a n e w chance. But the "size mismatch/+ m a y create a h u g e burden. John Kitching notes that 84 percent of m i s m a t c h cases of m e r g e r s e n d s in failure, w h i c h reinforces the "critical m a s s " theory a d v a n c e d b y H. Igor Ansoff. O n e s h o u l d further note that the there are organizational m e t h o d s for dealing successfully w i t h mismatch. Malaysia has a m i n d s e t w i t h a VISION 2020. This is the goal of the c o u n t r y to join the club of the d e v e l o p e d nations. This consolidation m o v e m a y be the source of revitalization of the Malaysian financial structure. REFERENCE+ Boquist, J. A., G. A. Racette, and G. G. Schlearboum. (1975). "Duration and Risk Assessment of Bonds and Common Stocks: A Note," Journal of Finance, 3 (5), 1360-1365. Brealey, R. A., and S. C. Myers. (2000). Principles of Corporate Finance.+McGrawHill. Craig, B., and J. C. Dos Santos. (1997). "The Risk of Effects of Bank Acquisitions," Economic Review-Federal Reserve Bank of Cleveland, 33, 25-34. Edison, H., and C. Reinhart. (2000). "Capital Controls During Financial Crises: The Case of Malaysia and Thailand." International Finance Discussion Paper No. 662, Board of Governors of the Federal Reserve. Esty, B., B. Narasimhan and P. Tufano. (1999). "Interest-Rate Risk Exposure and Bank Mergers," Journal of Banking & Finance, 23, 221-249. Gomez, E. T, and K. S. Jomo. (1999). Malaysia's Political Economy. Cambridge, England: Cambridge University Press. Haggard, S., and L. Low. (2000). "The Political Economy of Malaysian Capital Controls." Unpublished manuscript, Harvard University. Hughes, J. P., W. W. Lang, L. Mester, and C. Moon. (1999). "The Dollars and Sense of Bank Consolidation," Journal of Banking & Finance, 23, 291-324. IMF. (1998). International Capital Markets (September). Jain, R. (1996). "Takeover Attempts, Poison Pills, and Corporate Pictures," The International Journal of Finance, 8, 2. James, G. S. (1951). "The Comparison of Several Groups of Observations When the Ratios of the Population Variances Are Unknown," Biometrica, 38 (December), 324-329. Joehnk, D. M., and J. F. Nielson. (1974). "The Effects of Conglomerate Merger Activity on Systematic Risk," Journal of Financial and Quantitative Analysis, 215-225. Milbourne, T. T., A. W. A. Boot, and A. V. Thakor. (1999). "Megamergers and

From Turbulence to Tranquility

11

Expanded Scope: Theories of Bank Size and Activity Diversity," Journal of Banking & Finance, 23, 195-214. Mohamad, M. (1999). A New Deal for Asia. Kuala Lumpur, Malaysia: Pelanduk. Prakash, A. J., R. M. Bear, K. Dandapani, G. L. Ghai, T. Patwa, and A.M. Parhizgiri. (1999).The Return Generating Models in Global Finance. Elsevier Science. Rhoades, S. A. (1996). "Bank Mergers and Industrywide Structures 1980-1994." Staff Study 169, Board of Governors of the Federal Reserve System. Washington D.C. Thompson, R. S. (1983). "Diversifying Mergers and Risk: Some Empirical Test," Journal of Economic Studies, 10, 12-21. World Bank. (1998). East Asia: The Road to Recovery. Washington, D.C: World Bank. . (2000). East Asia: Recovery and Beyond. Washington, D.C: World Bank.

This page intentionally left blank

CHAPTER 2

Portfolio Investments in Australia: 1988-1997 W. Leong and J.

Wickramanayake

INTRODUCTION The globalization of the world economy has facilitated greater levels of portfolio investments. Such capital inflows have increased the allocative efficiency of investment projects and supplemented the level of domestic savings in the recipient. However, portfolio capital flows can also destabilize an economy because of their volatile nature. The surge of capital inflows to Latin America and Southeast Asia and their subsequent outflows in the 1990s highlight the potential volatility of portfolio investments. Thus, an understanding of the behavior of portfolio investments may ensure that their benefits are maximized while minimizing their adverse consequences. The purpose of this chapter is to examine the effects of interest rates, exchanges rates, and scale variables on the level of portfolio investments in its various forms in Australia. A cointegration approach based on an unrestricted error-correction model (UECM) is used to analyze this relationship. The chapter is structured as follows. The section on evidence gives the theoretical background and existing empirical evidence on the determinants of portfolio investments. In the section on model formulation and data, the empirical model is formulated, and a discussion of the

data used in this study is provided. The results section presents a di sion of the empirical results obtained in this study Finally, the concluding section summarizes the major findings of the chapter. EVIDENC++ Changes in the level of gross portfolio capital flows into a country are influenced by several factors: economic fundamentals (i.e., growth in

14

Regional Financial Markets

wealth), official policies (i.e., capital controls) and market distortions (i.e., taxation) (Goldstein, Mathieson, & Lane, 1991). This chapter will focus specifically on the macroeconomic determinants of portfolio investment inflows, which may indirectly represent the latter two factors: official policies and market distortions. A further justification for modeling only the macroeconomic variables is that a study by Ghosh and Ostry (1993) found that portfolio capital flows are generally driven by economic fundamentals. A number of studies have analyzed the importance of macroeconomic variables in explaining portfolio investments. Some commonly used economic variables include interest rates, exchange rates, and scale variables (Spitaller, 1971). Portfolio capital movements have traditionally been explained by interest rates. Branson (1968) stated that the purchase of and demand for securities was a function of (among other things) domestic and foreign interest rates, as postulated by the yield theory. Since wealth maximization is an objective, foreign assets will only be held if foreign interest rates are higher than domestic interest rates; conversely, domestic assets will only be purchased if domestic rates are higher than foreign rates. Using monthly and quarterly U.S. inward portfolio data, Branson (1968) found that changes in U.S. long-term (domestic) interest rates had a powerful impact on the foreign purchase of US securities. However, this result was inconsistent with the "asymmetric effect of interest rates" as proposed by Miller and Whitman (1970). They state that the foreign purchase of domestic securities (i.e., inward portfolio investments) should be affected more by foreign interest rates than by domestic rates. That is, a change in the foreign interest rate has a greater impact on the purchase of domestic securities than an equivalent change in the domestic rate. Kreicher (1981), extending the stock-adjustment model used by Miller et al. (1970), also investigated the relationship between long-term portfolio capital inflows and outflows and real interest rates for three European countries (United Kingdom, West Germany, and Italy) and the United States. Using quarterly portfolio investment data, Kreicher (1981) found convincing evidence supporting the hypothesis that international portfolio flows (inflows and outflows) are responsive to real interest rates. However, not all domestic interest rates were statistically significant in explaining portfolio capital inflows to their respective countries, which accords well with the "asymmetric effect of interest rates." In addition, Kreicher (1981) showed that changes in U.S, interest rates were significant in influencing portfolio investments into the U.S., which is inconsistent with Miller et al. (1970), but supports Branson's finding of the quantitative importance o+ U.S. interest rates. Studies by Fausten (1990, 1995) on Australia and Germany also used interest rates (scaled by wealth) to explain portfolio investments abroad and found that interest rates were statistically significant

Portfolio Investments in Australia

15

and accorded with theory. A more recent review of these issues can be found in Feldstein (1999). Branson (1968) also suggested that exchange-rate expectations influenced portfolio capital flows. There is conflicting evidence of the relationship between expected exchange rates and portfolio flows. On the one hand, studies by Lee (1969), Miller et al. (1970) and Ueda (1990) concluded that exchange-rate expectations had a statistically significant association with portfolio capital movements. On the other hand, Spitaller (1971) found no evidence that exchange-rate expectations influenced the behavior of portfolio capital movements. Exchange-rate expectations can influence capital flows because a possible devaluation for a certain currency would increase the risk associated with holding assets denominated in that currency, while an actual devaluation would lower the return on the asset (Spitaller, 1971). Spitaller (1971, p. 208) argues that studies that only use interest rates to explain portfolio capital movements assume there is an "infinite elastic supply of, and demand for loanable funds." The inclusion of a scale variable (income or wealth) overcomes this problem because the coefficient o+ the scale variable will affect the interest rate coefficients. That is, the volume of capital inflows to a country in response to changes in interest rates will depend on the size of the portfolio (given by the scale variable) (Spitaller, 1971). This has been acknowledged as many writers incorporate interest rates with a scale variable when modeling portfolio capital flows (Branson, 1968; Fausten, 1990,1995; Kreicher, 1981). On one hand, Branson (1968) suggests that increases in domestic economic activity may attract (pull) portfolio capital inflows, but only by raising domestic rates of returns. On the other hand, an increase in domestic activity may reduce the level of inward portfolio investments if the "spillover effect" results in the repurchase of domestic securities from foreigners (Branson, 1968). An increase in foreign economic activity may either increase (spillover effect) or decrease (Tobin effect) inward portfolio investments. A foreign activity "spillover effect" occurs when the overall foreign demand for domestic securities increases, while the "Tobin effect" takes place when the total foreign demand for domestic securities decreases (Branson, 1968; Kreicher, 1981). Ultimately, any change in the scale variable (as a proxy for portfolio size) would stimulate portfolio adjustment as investors rebalance their portfolios to achieve the preferred equilibrium position, thus influencing portfolio capital flows. This process of rebalancing the portfolio reflects

the desire of investors to diversify their assets in an attempt to balance the risk-return tradeoff,1+which has been disturbed (Lee, 1977; Tesar & Werne 1995). Kreicher (1981) found that portfolio inflows to the United Kingdom, Germany, Italy, and the United States were influenced by at least one activity variable, with the majority being affected by foreign activity variables.

16

Regional Financial Markets

Pain (1993), Ueda (1990), and Fausten (1990,1995) also included a scale variable when modeling portfolio investments. Although different scale variables were used in these studies they were found to be statistically significant in explaining portfolio capital flows. Pain (1993) used the growth of wealth as a scale variable and found that it was an important

determinant of the level of portfolio investment inflows into the United

Kingdom. Ueda (1990) also concluded that wealth was important in explaining Japanese capital outflows. Further, Fausten (1990,1995) concluded

that domestic and foreign income (scale variable) strongly influenced Australian and German (nonofficial) portfolio investment outflows through the

asset demand function. Fausten (1990,1995) argues that increased income (domestic economic activity) affects the asset demand function by altering risk perceptions—that is, domestic expansion reduces the perceived riskiness of domestic returns (by reducing default risk and bankruptcies) while also enhancing the perceived creditworthiness of domestic borrowers. Therefore, domestic investors will find the risk associated with holding domestic assets is lower during domestic economic expansion than in times of economic slowdown (Spitaller, 1971). Thus, increases in domestic income raise demand for domestic assets and facilitate substitution of foreign for domestic assets. Portfolio investments are the sum of various financial flows, which are influenced differently by similar factors (Spitaller, 1971). In an attempt to capture these differences, numerous measures of portfolio investments (dependent variable) have been used in empirical research. For example, some studies used an aggregate figure of portfolio capital investments, while some used a disaggregated measure. For instance, Pain (1993), Fausten (1990, 1995), Kreicher (1981), and Branson (1968) used an aggregate measure of portfolio investments. However, more recent studies by Chuhan, Claessens, and Mamingi (1998), Taylor and Sarno (1997), Brennan and Cao (1997), and Fernandez-Arias (1996) have employed a measure of portfolio capital flows disaggregated into debt and equity components. For this study, "portfolio investment inflows: nonofficial" into Australia are to be disaggregated into debt and equity, and an aggregate measure of inward portfolio investments into Australia will also be used as the dependent variable. Moreover, unlike previous studies, the results obtained in this study will be subjected to rigorous diagnostic tests. The above discussion indicates that macroeconomic variables, specifically interest rates, exchange rates, and scale variables, have a fundamental role in determining the level of portfolio investments. However, the influence of these variables and their relative importance varies across countries. The lack of empirical studies on Australia's portfolio inflows

justifies an empirical investigation to determine the relative impact o

changes in economic variables on inward portfolio investments.

Portfolio Investments in Australia

17

MODEL FORMULATION A N D DATA The following long-run relationship w a s p o s t u l a t e d for analysis:

empirical

Where PI t

= Level of portfolio i n v e s t m e n t s (in real terms)

AURI t

= A u s t r a l i a n (real) interest rates*

USRI t

= U.S. (real) interest rates (proxy for foreign interest rates)*

EXR, AURGDP t

= Exchange rate (USD = A u s t r a l i a n real gross d o m e s t i c product**

USRGDP t

= US real gross d o m e s t i c p r o d u c t (proxy for foreign gross d o m e s t i c product)**

t DU+

= D u m m y variable***

£t

= R a n d o m error term.

*A n u m b e r of real interest rates a n d real interest rate differentials will be u s e d in s e p a r a t e regressions. **A n u m b e r of scale variables will also b e e m p l o y e d in different regressions. ***A n u m b e r of d u m m y variables will b e e m p l o y e d i n t e r c h a n g e a b l y for

quarterly data. Following the discussion in the p r e v i o u s section the expected signs o+ the m o d e l p a r a m e t e r s are as follows:

The predicted sign of Australian interest rates (AURIt)+is positive as sug-+ gested b y the yield theory (Lee, 1977; Spitaller, 1971). That is, a n increase in domestic (Australian) interest rates, ceteris paribus, s h o u l d increase portfolio inflows as investors seek to benefit from higher interest rates to m a x imize their w e a l t h (Kreicher, 1981). Conversely, a negative relationship exists b e t w e e n foreign (U.S.) interest rates a n d i n w a r d portfolio investm e n t s . W h e n foreign interest rates (USRIt)+increase, foreign investors will shift i n v e s t m e n t s from other countries (i.e., Australia) to take a d v a n t a g e of higher yields abroad. The exchange rate (EXRf)+a n d the scale variable+ (AURGDP++ a n d USRGDPt)+h a v e i n d e t e r m i n a t e signs as empirical studies+ t

have provided contrasting results. Portfolio investment theory suggests that a depreciation of the domestic (host country) currency (AUD) lowers

the rate of return on domestic assets for foreigners (holding these assets) (Spitaller, 1971). In contrast, marcoeconomic theory predicts a positive

18

Regional Financial Markets

relationship between a depreciation in the domestic (host country) currency and foreign investments, the converse is true for an appreciation. Since a depreciation reduces the cost of domestic assets (Australian assets) to foreigners (the rest of the world) they are more likely to purchase domestic assets (Australian securities). Branson (1968) suggests an increase in gross domestic product (GDP) of the host country (AURGDPt)+may increase portfolio capital inflows by increasing expected returns, or decrease inflows as domestic investors repurchase domestic securities from foreigners. Increases in the source country's GDP (USRGDPt++may also increase or decrease portfolio+ capital inflows (into Australia) if foreign investors rebalance their portfolios or switch from foreign to domestic assets (Branson, 1968; Kreicher, 1981). After preliminary experimentation with a number of dependent variables, two measures of inward portfolio investments in Australia were used in this study, namely, "portfolio investment equity: nonofficial" (PIENO) and "portfolio and other investment total: nonofficial" (POITNO). Only the nonofficial component of portfolio investments was used because the inclusion of official figures may have distorted the results. This is because portfolio inflows to the official sector are less affected by changes in market conditions, since international institutions and foreign governments tend to hold government securities (bonds and notes) for risk diversification purposes or as international reserves (Lee, 1977). Thus, these inflows are not motivated by wealth maximization and are determined independent of interest-rate differentials (Lee, 1977). Moreover, the use of a nonofficial portfolio investment measure (POITNO or PIENO) is consistent with the approach adopted by Fausten (1990, 1995). Three U.S. interest rates will be used in this study. The interest rates consist of one long-term U.S. interest rate (USRIET) and two short-term U.S. interest rates for three and six months, respectively (USRICPRAV3 and USRICPRAV6). The long-term U.S. interest rate to be used is the 10-year government bond rate (USRIET) for which data was collected from the IMF International Financial Statistics (IFS) Yearbook. The short-term US interest rate was represented by a three- and six-month commercial-paper rate (USRICPRAV3 and USRICPRAV6), which was collected from the Australian Bureau of Statistics (ABS) Time Series database. Real interest rates were calculated by subtracting the change in the corresponding Consumer Price Index (CPI) lagged one period from the nominal interest rate in the current period. Exchange rate figures are given as "market rates" quoted as U.S. dollars per Australian dollars (USD/AUD) at "period averages" (EXRIFSAV), obtained from the IMF International Financial Statistics (IFS) Yearbook. Both domestic and foreign scale variables were used in this study. Domestic scale variables consisted of Australian gross domestic product (AURGDP) and the Australian share price index (AUSPI), while

Portfolio Investments in Australia

19

foreign scale variables included U.S. gross domestic product (USRGDP), U.S. share price index (USSPI) and U.S. industrial production (USIP). All scale variable data were extracted from the IMF International Financial Statistics (IFS) Yearbook. The use of income (i.e., GDP) as a scale variable is consistent with studies by Branson (1968), Kreicher (1981), and Fausten (1990,1995). Brennan et al. (1997) have used share price indices as a proxy for portfolio size and expected returns. Further, the use of industrial production as a scale variable is congruent with studies by Chuhan et al. (1998) and Taylor et al. (1997). All variables with financial values were adjusted to real terms to present a more comparable analysis undistorted by inflation. The corresponding consumer price index (CPI) (i.e., Australian and U.S.) (1990 = 100) was used to transform nominal values into real values. Moreover, as a common practice when working with time-series data, logarithmic transformations were applied to provide a smoother series. Preliminary experimentation with the sample data series indicated that dummy variables to account for the economic recession in Australia in 1990:2 to 1991:2 and seasonal dummies were not good variables for the sample period (1988:1 to 1997:4) used in this study. Furthermore, the use of Australian interest rates yielded poor results as they entered with the incorrect sign. In addition, when quarterly "portfolio investment bond: nonofficial" (PIBNO) (in Australia) figures were used in initial experimentation they did not yield any satisfactory results. The methodology adopted for this study is similar to that of Taylor et al. (1997), who used cointegration techniques to reveal that both domestic and global factors explained bond and equity flows from the United States to developing countries. This approach is followed because of the similarity of the phenomenon under investigation, namely, the determinants of portfolio capital flows. To establish the time series properties of the data, Augmented DickeyFuller (ADF) and the Phillips-Perron (PP) unit root tests were used. The results of both unit root tests as presented in Appendix 1 show that all variables have unit roots, that is, they are first difference stationary: I (1). Variable definitions used in the regressions are shown underneath Appendix 1. Thus a cointegration and error-correction modeling approach is justified. The principle behind error-correction models is that there exists a longrun steady state relationship between two economic variables. However, in the short term there may be disequilibrium, which must be accounted for (Maddala, 1992). The error-correction model achieves this by allowing a proportion of the disequilibrium in one period to be corrected for in the following period. Hence, the error-correction process attempts to integrate short-run dynamics with long-run equilibrium. The short-run and longrun effects can be reconciled by using either the Engle and Granger (EG)

20

Regional Financial Markets

(1987) two-step procedure or by modeling both effects simultaneously using the UECM. The UECM estimates the long-run parameters and short-run effects simultaneously in a single equation using ordinary least squares (OLS) (Banerjee, Dolado, & Mestre, 1998). It has been shown that the UECM (simultaneous modeling) approach is superior to the EG procedure in terms of statistical properties (Inder, 1993). Further, the UECM procedure is the most suitable approach for small sample sizes (as used in this study) because there could be a strong bias when using the EG test procedure to estimate the long-run relationships (for small samples) (Banerjee, Dolado, Hendry, & Smith, 1986). Therefore, the UECM specification will be adopted in this study. Since the modeling objective of this study was to use the recent cointegration test proposed by Banerjee et al. (1998), it was decided not to use the Johansen procedure (Johansen, 1988; Johansen & Juselius, 1990). In empirical work a number of error-correction formulations can be adopted (Phillips & Loretan, 1991). This study closely follows Arize (1994) and Morling and Subbaraman (1995) and formulates the following UECM representation of equation (1):

where S and +t are the constant and random error term, respectively. A denotes the first difference of the variables defined in model (1). This UECM (2) does not include the dummy variable for simplicity. The above representation captures the long-run equilibrium and short-run dynamics. The differenced independent variables account for short-run behavior, while the long-run parameters are given by the lagged independent variables. An important advantage of the UECM formulation is that the lag length is not restricted before the commencement of the modeling process (dependent on the sample size) (Hendry, 1995). This is beneficial because it gives greater freedom in determining whether seasonal effects are present. In addition, since error-correction models are generalized versions of the partial adjustment model they implicitly account for expectations (Maddala, 1992). Thus, the need to explicitly include variables of expectations (i.e., expected interest rates and exchange rates) is diminished. Once modeling starts, the conventional "general-to-specific" procedure for narrowing down independent variables will be employed. This process involves systematically removing the least significant (independent) variables from the "general" (broad) model until they all be-

Portfolio Investments in Australia

21

come significant, ultimately arriving at the "specific" model (Hendry, 1983, 1995). Davidson, Hendry, Srba, and Yeo (1978) and Hendry (1983) suggest that this approach facilitates the estimation of a parsimonious error-correction model. The long-run elasticities can be calculated from the coefficients of the lagged-level variables in the equation because they have a comparable base (all variables are in natural logarithms). As a result, the UECM approach incorporates the cointegration test in the tstatistic of the error correction term, given by a k in model (2) (Hendry, 1995). The t-statistic is then compared with critical values provided by Banerjee et al. (1998) to determine whether a long-run relationship (cointegration) exists. If the t-statistic (in absolute terms) of the error correction term is greater than the critical values, the null hypothesis of no cointegration is rejected. RESULTS Quarterly data from 1988:1 to 1997:2 were used in the estimation of all regression equations, except equation 1, which used sample period data from 1988:3 to 1997:4 due to a lack of data starting from 1988:1. The results of the eight equations generated by OLS regressions using the UECM in model (2) are shown in Tables 2.1a and 2.1b. Although a number of dummy variables, as discussed previously, were included in all the regressions they were not statistically significant once the "general-tospecific" methodology was applied. Thus, they were omitted from the eight estimated equations shown in Tables la and lb. The t-statistics (in absolute terms is given in parentheses) for all the variables are statistically significant (except for the constant term in equation 1 and 2). The F-statistic for all the estimated models was significant; thus, the null hypothesis that all the coefficients in the regression are zero is rejected—meaning that the estimated regression equation is useful in explaining inward portfolio investments. The adjusted R2 is satisfactory and is relatively consistent with the results obtained by Taylor et al. (1997) for portfolio inflows to Latin American and Asian countries, and for Kreicher (1981) for inward portfolio flows to the United Kingdom, Germany, and the United States. The DW (Durbin & Watson, 1950) statistic and the LM (LM1 and LM4) (Godfrey, 1978) test provide no evidence of first-order and fourth-order serial correlation in the estimated equations. The results of the LM1 test also verify the DW test and the null hypothesis cannot be rejected, while the LM4 test suggests that no fourth-order residual correlation exists. The HET test (White, 1980) and ARCH tests (ARCH1 and ARCH4) (Engle, 1982) indicate that all equations do not exhibit heteroskedasticity. All estimated equations also satisfy the assumption of normality as given by the JBN test (Jarque & Bera, 1980). The JBN statistic is compared with the

critical values from the+%2+table, and there is insufficient evidence to reject

Regional Financial Markets

22

Table 2.1a Estimates of the Unrestricted Error-Correction ModeL Equation Sample

Number Period

Dependent

Variable

CONS 1 ANT AAl'SPl',++ AFXRIFS-1 I •,

I

2

3

4

1988i3-+ 1997:4+ PIENO, +

1988:1++ 1997:+ POIT\Ot+++

1988:1-+ 1997:2+ POITNO, +

1988:1++ 1997:2+ POITNO++

-0.30n.5~n" 0._83 (5729)***"

0.06 ( O . n j

4.66 (6.35)***

~"3.63_5.66)*** "

-0 4 4 ( 2 78)***

irssr[,+ Al SRILJ, ~A4l RGDP , APIF\+++M ~APOIT\0 .,.,

APorr\o^++ +lA AFXRlhS4\'+ Z AF\RIFS4\+,_

All

l2 SP/

AUSP/,,

0.28 (2.04)* -0.32(4.41)*** -0.94 0 . 7 3 ) * -0.37 (3.05)*** -0.34(3.02)*** 0.28(2.57)** 0.34(1.97)* 1.66(4 95)*** -0.39(2.59)** -0.42(2.91)***

Ail+RGDP,

AA I RGDP+M Al SSPl',.z Al SRI I 7 ,_, M'SRI 1.1,., ~Al 'SRILT^ 'p/F\0,+ ++

POlT\O

~ 0.r8(2.16)** 0 19(2.56)** 0.16(2.34)** [ " 0.27~(3.25)*** -0.36(4 45j***»

h

TlSPl,_

EAR IFS 41 ,_+ 'MRGDP,.++ 1++SSPfn I SRILT,'+" Adjusted R : F-Statistic ~D\V ~JB\ A I / O ) . (4) 4RG//(\).++(4)+ "//AT" RESET{\l+{A)'++ I'DT

a icnv' hORG

1.25(2.30)** - 3 6 0 (3 72)***

-1 5 6 ( 2 92)***

0.80(4.35)*** '_ -1 05 (5.80)***

-0.52 (6 89)***«

-0 49(6.06)***«

-0 42(4.33)*** 1.00(603)***

-0 37(3.80)*** ~ 0.59(3.83)*** [

0.72 1145 2 71 0 78 1.43.1.75"" 0.15.0.34 1 85 0.47."(")47 " 0.68 "l 03 ' 1.07

0.58 11 18 1.75 0.45 0.29. 0.20 1.16.0.72 0 61" 1 65._0.67 0 80 0.90 1.33

"-0.54(6 40)***•

0.22(4.58)**"* 0.63 '. 9-92

1.70 1.01 0.67. 0.34 1.32.0.52

0.31 0.66. 0 47 0.72 1.01 1.24

-0.39(6.29)*"** 0.71 * " 9.32

[ f.65 1.07 " 0.21.0.53 * 1.20.0.92 1.09 ~ 0 91. 1 11 _" * 0.82 0 97 1.88

+

Seasonal dummies included in all the regressions deriving the above results. ***, ** and * denote significance at 17c, 5%, and 10% respectively. •Passes the cointegration test.

the null hypothesis that the residuals of the estimated equations are normally distributed. The RESET test (Ramsay 1969) indicates that all equations (models) display no problems of functional form misspecification. The VDT indicates that the omitted variables could not have made a significant contribution in explaining portfolio investments; thus, the // general-to-specific,/ approach is appropriate. Finally, all the estimated

Portfolio Investments in Australia

23

Table 2.1b Estimates of the Unrestricted Error-Correction Model + 6

7

1988:1-1997:2+++

1988:1-1997:2++

1988:1-1997:2++

1988:1-1997:2+++

POIT\0,+++

POIT\Ot+++

POIT\0,+++

POIT\0, +++

1 9 6 ( 5 09)***

0 74 (2 ( ) V

5

Equation++\umber++ Sample

Period

Dependent

Variable

CONSI ANT AISR1GPR43+++, Ai\R/G~PRU6r+++++ APO/l\On~ +++ APO/T\() ,.; Al S RIG PR 113,^ ' Al SRIGPRU6 ,++ ACSR1CI;R416,, ++ Ail+++++RGDP h2 All RGDP ,, r Al'S RGF)P,[ ++ AGSRGDP,., " lZ +++ AlSIP++

POIl\0,++++ I SRIGPR 41 3, I++SRIGPRU6,++ i I RGDP', I+++SRGDP r ~l S//\ : 1 \djustcd R~ i--Statistic DW ~JB\ ~~ L\1(\) (4) iR( fl(\) (4)

lfU.'.".

RESLT(\).(4j++ \DT ~C110\\ 'lORC " f

()"92 (2 47)** -0 13 a++65)**

-5 9 9 ( 5 11.)***.. -0 11(2 16)** ""

8

-0 14(2 94)*** 0 2 6 ( 2 13)** -0 3 9 ( 3 71)*** "() 14(2 67)** " " 0 18(3 56)-** 0 11(1 90)* 0 14(2 81)*** -1 3 3 U 3 4 ) * * -1 61 (2 82)***

-1 0 7 ( 2 04)** -1 4 4 ( 2 5 6 ) * * -2 6 0 ( 2 86)*** -0 5 8 ( 6 76.)***« -0 10(5 24)***

-0 6 8 ( 7 80)***« -012(635)***

-1 0 6 ( 2 0 5 ) * * -057(650)***» -0 11 (488)***

-0 61 (7 3 ! ) * * * • -0 11 (5 81)*** 1 13 (6 52.)***

1 04 (5 87)*** 1 65 (6 74)*** 0 65 12 48 I 56 1 22 1 03 0 61 " 0 71 0 24 0 60 " 0 38 0 37 0 61 1 26 " 1 69

0 69 13 03 1 83 0 80 0 21 0 60 0 28 0 97 1 58 1 92 1 30 0 29 1 37 0 54

1 0 8 ( 5 48)*** 0 61* 12 62 1 76 2 35 0 30. 0 77 0 12.0 57 0 90 1 12. "l 83 0 57 0 70 0 49

0 69 10 26 1 68 2 68 0 65. 0 39 1 J 3 . 0 33 " 0 52 0 79 1 90 0 64 1 28 0 89

Seasonal d u m m i e s were included in all the regressions deriving the above results.

***, ** and * denotes significance at 1%, 5%, and 10% respectively. •Passes the cointegration test.

equations pass the two Chow tests (CHOW and FORE) (Chow, 1960) of parameter stability. The null hypothesis of the CHOW test of parameter constancy over different sub-periods could not be rejected. Further, the FORE test provides satisfactory results for forecasts for the last year (four quarter forecast observations) based on the rest of the sample observations. Since the null hypothesis cannot be rejected, it is concluded that the variance between the actual and predicted values for the forecast period is small (Maddala, 1992). Thus all the satisfactory diagnostic tests shown in Tables 2.1a and 2.1b indicate that the estimated equations are robust. The null hypothesis of no cointegration is rejected when the t-statistics of the two error-correction terms of PIENOt_, (portfolio investment equity: non-official) and POITNO^ (portfolio and other investment total: non-

Regional Financial Markets

24

official) in Tables la and l b are compared with the 10, 5, and 1 percent level critical values for three level variables for 25 and 50 sample observations in Table 2.2 below as provided by Banerjee et al. (1998). The estimated error-correction terms indicate that the speed of adjustment back to long-run equilibrium after shocks to the system are reasonable. For example, the coefficient of the error correction term POITNO^ (portfolio and other investment total: nonofficial) in equations 2 to 7 indicates that approximately half (a low of 49% to a high of 68%) of the short-run disequilibrium caused by disturbances in the system is eliminated in the following quarter. The adjustment speed for the other errorcorrection term PIENOt_2 (portfolio investment equity: nonofficial) is slower, with slightly over a third (36%) of the disequilibrium displaced after one quarter. On average these results are superior to those provided by Kreicher (1981). Kreicher's study found that the speed of adjustment ranged from 19 percent to 63 percent (after one quarter) for portfolio investment inflows for different countries. A likely explanation for the superior results of this study is that financial deregulation, especially among developed countries, has reduced the barriers and costs of international portfolio flows. As a result, international portfolio flows can adjust faster to (disequilibrium) shocks to the system. Thus, the adjustment speeds recorded in this study should intuitively be higher, which was confirmed

by the larger coefficients of the error-correction terms.

The long-run elasticity coefficients for the explanatory variables included in the eight equations are shown in Table 2.3 below. The signs of all the coefficients are consistent with the theory. An analysis of the elasticity coefficients indicates that changes in economic conditions overseas have an offsetting impact on the level of portfolio investments received by Australia. It is shown in Table 3 that all domestic scale variables—Australian share price index (AUSPI) and Australian gross domestic product (AURGDP)— and foreign scale variables—U.S. share price index (USSPI), U.S. gross domestic product ( USRGDP), and U.S. industrial production (USIP)—are positively correlated with portfolio investments into Australia. Thus, AusTable 2.2 Critical Values for the Test of Cointegration (k = 3) Level of Significance*

Sample Observations

(]%)+(5%)++10%++++++++++++++++++++++

25

-4.92

-3.91

-3.46

50

-4.59

-3.82

-3.45

*The critical values provided are for estimated equations with three regressors (k = 3) and a constant term.

25

Portfolio Investments in Australia

Table 2.3 Long-Run Elasticity Coefficients* Level Variables

AUSPI,

"

AIRGDP+++

EXRJFSAV++

USRJCPRA V3,

USRICPFL4V6,+

Sample Period

Table Number (Equation Number)

Elasticity Coefficient

1988:3-1997:4

l a ( l )+

2.23

1988:1 - 1 9 9 7 : 2

l a ( 2 )+

1.94

1988:1 - 1 9 9 7 : 2

la (4)

1.10

1988:1 - 1 9 9 7 : 2

lb(5)

1.78

1988:1 - 1997:2

lb (8)

1.85

1 9 8 8 : 3 - 1997:4

la(l+

-2.93

1988:1 - 1997:2

la (2)

-0.81 -0.76

1988:1 - 1997:2

la (3)

1988:1 - 1 9 9 7 : 2

lb(5)

-0.17

1988:1 - 1997:2

lb (6)

-0.18

1988:1 - 1997:2

lb(7)

-0.20

1988:1 - 1 9 9 7 : 2

"lb (8)

-0.18

USRILT,"

1988:1 - 1997:2

la(4)

-0.72

USSPI,+

1988:1 - 1997:2

1a(3)

0.45

USRGDP,+

1988:1 - 1 9 9 7 : 2

lb (6)

2.44

USIP+

1988:1 - 1997:2

lb (7)

1.91

*The long-run elasticity coefficients were calculated by dividing the coefficient of the (level) variable by the coefficient of the error-correction term (at the level) and reversing the sign.

tralian scale variables acted as a "pull" factor, while U.S. scale variables acted as a "push" factor. In this study, an increase in Australian (domestic) economic activity (measured by AUSPI+and AURGDP)+and an increase in U.S. (foreign) economic activity (measured by USSPI, USRGDP+and+USIP) attracted portfolio inflows into Australia. According to Branson's explanation, an increase in Australian activity attracts portfolio inflows because it raises domestic rates of return. An alternate explanation is provided by Fausten (1990,1995), who suggests that increases in domestic (Australian) income reduces the risk associated with holding domestic assets (Australian) as influenced through the asset demand function; thus, this raises the demand for domestic assets—causing portfolio capital inflows. Kreicher (1981) found similar results for the United Kingdom, where increases in U.K. domestic activity led to a boost in portfolio investments into the United Kingdom. In the long run, a 1 percent increase in the Australian share price index (AUSPI) causes an increase of 2.23 percent in "portfolio investment equity: nonofficial." An equivalent 1 percent increase in Australian gross domestic product (AURGDP)6+produces an in

26

Regional Financial Markets

crease in "portfolio and other investment total: nonofficial" (POITNO) ranging from 1.10 percent to 1.94 percent. The results of this study could not be compared (directly) to that of Fausten (1990) because their study focused on Australian non-official portfolio investment outflows, whereas this study examined Australian portfolio investment inflows. An increase in the foreign scale variables of U.S. share price index (USSPI), U.S. gross domestic product (USRGDP), and U.S. industrial production (USIP) also increased portfolio investments into Australia. Branson (1968), Spitaller (1971), and Kreicher (1981) suggest this is the result of portfolio adjustment. Since the scale variable is a proxy for the size of the foreign portfolio any subsequent increases would force the investor to rebalance the portfolio (i.e., purchase new assets-flow adjustment), until a new equilibrium position is achieved. For this study the results suggest that the rebalancing process has the effect of increasing the demand for Australian securities (the spillover effect). Table 3 indicates that there is a positive relationship between all foreign scale variables (USSPI, USRGDP, and USIP) and the level of "portfolio and other investment total: nonofficial" (POITNO). U.S. gross domestic product (USRGDP) has the largest impact, with a 1 percent increase causing a 2.44 percent increase in portfolio capital inflows into Australia. These results are similar to those of Kreicher (1981), who found that all countries were significantly influenced by at least one activity variable. Furthermore, it accords well with Kreicher's finding that the majority of countries were influenced by a foreign activity variable. However, the results are inconsistent with studies on developing countries. Chuhan et al. (1998) and Taylor et al. (1997) found that increases in foreign economic activity (measured by U.S. industrial production) led to decreases in capital inflows to developing countries, whereas this study suggests increases in foreign activity result in an increase in capital inflows. The exchange rate (EXRIFSAV) variable had the greatest impact of all variables on "portfolio investment equity: nonofficial" (PIENO) in Australia. However, the impact of the exchange rate (EXRIFSAV) variable on "portfolio and other investment total: nonofficial" (POITNO) was more moderate. The exchange rate had a significant negative effect on portfolio investment inflows, which accords well with the foreign direct investment (FDI) theory. On average, an increase of 1 percent (represents a depreciation of the USD) in the exchange rate produces a 2.93 percent decrease in capital inflows to Australia. This effect is consistent with the theory that a depreciation of the currency reduces the purchasing power of investors by increasing the price of Australian assets (in this case). Therefore, investors are less likely to purchase Australian securities and the level of inward portfolio investments falls. Thus in Australia, a change in the exchange rate also influences the level of portfolio capital inflows by changing the relative wealth of investors or institutions, which is con-

sistent with the findings of Klein and Rosengren (1994) for inwar

Portfolio Investments in Australia

27

eign direct investment for the United States. Finally, the interest rate variables represented by three- and six-month U.S. commercial-paper rates (USRICPRAV3,USRIGPRAV6) and a 10-year U.S. government bond yield rate (USRIET) entered all estimated equations with a negative elasticity coefficient, as suggested by the yield theory and the "asymmetric effect of interest rates." Thus, an increase in foreign (U.S.) interest rates reduces "portfolio and other investment total: nonofficial" (POITNO) into Australia. This finding conforms with those obtained by Chuhan et al. (1998) and Taylor et al. (1997), who found that increases in foreign interest rates (proxied by U.S. interest rates) resulted in lower levels of capital inflows to developing countries. Kreicher (1981) also concluded that increases in foreign interest rates reduced the level of portfolio capital inflows for all the countries in the study (i.e., the United Kingdom, Germany, Italy, and the United States). On average, the short-term interest rates (USRICPRAV3 zmdUSRICPRAV6) provided similar results; however, the long-term interest rate (USRIET) provided the greatest influence on portfolio investment inflows. In the long run, a 1 percent increase in the longterm U.S. interest rate (USRIET) produces a 0.72 percent decrease in the level of portfolio investments into Australia. The results provided by the interest rate variable in this study suggest that U.S. interest rates have a strong "push" effect on the level of portfolio investments into Australia. In other words, a decrease in U.S. interest rates tends to "push" capital abroad (some of which is received by Australia), which increases inward portfolio investments to Australia. These results are similar to those of Chuhan et al. (1998), Taylor et al. (1997) and Fernandez-Arias (1996), who

found that U.S. interest rates were the most important variable in+

mining the level of capital inflows to developing co CONCLUSION++ In this study, an attempt was made to investigate empirically the relationship between changes in interest rates, exchange rates, and scale variables on Australia's portfolio capital inflows, using quarterly data for the period 1988:1 to 1997:4. After testing for time-series properties of the variables, robust empirical results that pass numerous diagnostic tests were derived from the UECM. This study indicates that there are long-run equilibrium relationships between interest rates, exchange rates, and scale variables and portfolio investment inflows into Australia. Specifically, U.S. interest rates and exchange rates (USD/AUD) had a statistically significant negative effect on portfolio capital flows. Further, both domestic (Australia) and foreign (U.S.) income (scale variables) had a complementary effect on inward portfolio capital flows. These results suggest that various U.S. interest rates had a "push" affect—that is, decreases in U.S. interest rates forced

28

Regional Financial Markets

investors to look abroad for more attractive interest rates (returns). The implication of this finding is that Australia may see a reversal of portfolio capital inflows when U.S. interest rates increase. However, this reversal is unlikely to be in the same magnitude of those experienced in developing countries, because this will be mitigated by sound economic fundamentals (Goldstein et al., 1991) in the Australian economy. The results of modeling with Australian interest rates and an interest rate differential indicated that an increase in the Australian interest rates or an increase in the Australian interest rate differential led to a decrease in portfolio investments in Australia. A possible explanation for this result is proposed by Miller et al. (1970), who suggest foreign (U.S.) interest rates are quantitatively more important than domestic (Australian) interest rates, maybe to the extent that domestic interest rates have an insignificant impact. An implication of this finding is that raising Australian interest rates, ceteris paribus, may not directly provide a desired increase in portfolio investment inflows as postulated by the yield theory. Thus, this action is unlikely to alleviate the problem of a low level of domestic resource mobilization to supplement investment or moderate the current account problems. The results also indicate that changes in exchange rates may also affect portfolio investment inflows by altering the relative wealth of foreign investors. The positive relationship between domestic (Australian) income (scale variable) and inward portfolio investments indicates that domestic income may "pull" portfolio capital inflows into Australia by raising returns. Moreover, the importance of foreign (U.S.) income suggests that portfolio investments are "pushed" abroad as the foreign investors portfolio expands. Thus, for this study there appears to be a "spillover effect" when foreign income rises. That is, an increase in foreign scale variables (U.S.) leads to an overall increase in the demand for domestic (Australia) securities, thus increasing portfolio investment inflows to Australia. Therefore, Australia may experience portfolio capital inflows regardless of whether there are improved economic conditions within the country. The responsiveness of portfolio capital movements to income has significance for economic policy. If an increase in domestic economic activity attracts portfolio investments as found in this study, it lessens the adverse affects of balance of payments and exchange rate consequences of domestic expansion, while also reducing the need to use interest rate policies to offset the external account (Fausten, 1990, p. 98). In conclusion, the level of portfolio investment inflows received by Australia is influenced by two foreign factors (proxied by U.S. variables), U.S. interest rates and U.S. economic activity, and one domestic factor, namely, domestic economic activity. Thus, this study suggests, on balance, that external factors are relatively more important than internal factors in determining the level of portfolio investment inflows received by Australia. Since changes in foreign factors are beyond the control of a small open

Portfolio Investments in Australia

29

e c o n o m y like Australia, the policy m a k e r s can i m p l e m e n t policies that stimulate domestic economic activity (while k e e p i n g an eye on the exogen o u s factors) to attract portfolio investments. ACKNOWLEDGMEN++ The a u t h o r s w o u l d like to t h a n k Michael Skully, Vincent Dropsy, other participants of the First International Conference on Banking Finance for their valuable c o m m e n t s on a n earlier draft of this report. views expressed in the chapter are those of the a u t h o r s , w h o take

and and The sole

responsibility for any remaining errors or omissions. NOT++ 1. Studies by Grubel (1968), Solnik (1974), and Grauer and Hakansson (1987) recognize the potential gains from diversification of investment portfolios across nations, suggesting that the risks of a portfolio can be reduced by incorporating foreign securities. R E F E R E N C E S+ Arize, A. C. (1994). "Modelling the Demand for Broad Money in the United States," Atlantic Economic Journal, 22 (3): 37-51. Banerjee, A., J. J. Dolado, D. H. Hendry, and G. W. Smith. (1986). "Exploring Equilibrium Relationships in Econometrics through State Models: Some Monte Carlo Evidence," Oxford Bulletin of Economics and Statistics, 48(3): 253-277. Banerjee, A., J. J. Dolado, J. W. Galbraith, and D. H. Hendry. (1993). Co-Integration, Error Correction, and the Econometric Analysis of Non-Stationary Data. New York: Oxford University Press. Banerjee, A., J. J. Dolado, and R. Mestre. (1998). "Error-Correction Mechanism Tests for Cointegration in a Single-Equation Framework," Journal of Time Series Analysis, 19(3): 267-283. Branson, W. H. (1968). Financial Capital Flows in the US Balance of Payments. Amsterdam: North-Holland Publishing Company. Brennan, M. J., and H. H. Cao. (1997). "International Portfolio Investment Flows," Journal of Finance, 50 (5): 1851-1880. Chow, G. C. (1960). "Tests of Equality Between Sets of Coefficients in Two Linear Regressions," Econometrica, 28: 591-605. Chuhan, P., S. Claessens, and N. Mamingi. (1998). "Equity and Bond Flows to Latin America and Asia: The Role of Global and Country Factors," Journal of Development Economics, 55: 439-463. Davidson, ]., D. F. Hendry, F. Srba, and S. Yeo. (1978). "Econometric Modelling of the Aggregate Time Series Relationships between Consumers' Expenditure and Income in the United Kingdom," Economic Journal, 88 (December): 661692. Dickey, D. A., and W. A. Fuller. (1981). "Likelihood Ratio Statistics For Autoregressive Time Series With a Unit Root," Econometrica, 49 (4): 1057-1072.

30

Regional Financial Markets

Durbin, J., and G S. Watson. (1950). "Testing for Serial Correlation in Least Squares Regression," Biometrica, 37: 409-428. Engle, R. F. (1982). "Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of United Kingdom Inflation," Econometrica, 50: 987-1007. Engle, R. E, and C. W. J. Granger. (1987). "Cointegration and Error Correction: Representation, Estimation and Testing," Econometrica, 55 (2): 251-276. Fausten, D. K. (1990). "The Influence of Income on International Capital Movement: Some Preliminary Australian Evidence," Economics Letters, 33 (1): 95-100. Fausten, D. K. (1995). "Exploratory Estimates of the Influence of Economic Activity on German Portfolio Investment Abroad," Economia Intemazionale, 48 (4): 497-514. Feldstein, M. (Ed.) (1999). International Capital Flows. Chicago: The University of Chicago Press. Fernandez-Arias, E. (1996). "The New Wave of Private Capital Inflows: Push or Pull?," Journal of Development Economics, 48 (2): 389-418. Froot, K., and J. Stein. (1991). "Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach," Quarterly Journal of Economics, 106 (4): 1191-1217. Ghosh, A. R., and J. Ostry. (1993). "Do Capital Flows Reflect Economic Fundamentals in Developing Countries?," IMF Working Papers 93/94. (Washington, D.C: International Monetary Fund). Godfrey, L. G. (1978). "Testing for Higher Order Serial Correlation in Regression Equations When the Regressors Include Lagged Dependent Variables," Econometrica, 46: 1293-1302. Goldstein, M., D. J. Mathieson, and T. Lane. (1991). "Determinants and Systemic Consequences of International Capital Flows," Occasional Paper No. 77. Washington D.C: International Monetary Fund. Grauer, R. R. and N. H. Hakansson. "Gains From International Diversification: 1968-1985 Returns on Portfolios of Stocks and Bonds," Journal of Finance, 42: 721-741. Grubel, H. G. (1968). "Internationally Diversified Portfolios," American Economic Review, 58: 1299-1314. He, Y, and S. C Sharma. (1997). "Currency Substitution and Exchange Rate Determination," Applied Financial Economics, 7: 327-336. Hendry, D. F. (1983). "Econometric Modelling: the Consumption Function in Retrospect," Scottish Journal of Political Economy, 30 (3): 193-220. Hendry, D. F. (1995). Dynamic Econometrics. London: Oxford University Press. Hinton, T. (1997). "Australia's Foreign Investment Policy and the Role of the Foreign Investment Review Board," Economic RoundUp, (Summer): 57-61. Inder, B. (1993). "Estimating Long-Run Relationships in Economics: A Comparison of Different Approaches," Journal of Econometrics, 57: 53-68. International Monetary Fund (IMF). (1998). International Financial Statistics Yearbook. (Washington, D.C: International Monetary Fund). Jarque, C M., and A. K. Bera. (1980). "Efficient Tests for Normality, Homoskedasticity and Serial Independence for Regression Residuals," Economics Letters, 6: 255-259. lohansen, S. (1988). "Statistical Analysis of Cointegration Vectors," Journal of Economic Dynamics and Control, 12: 231-254.

Portfolio Investments in Australia

31

Johansen, S., and K. Juselius. (1990). "Maximum Likelihood Estimation and Inference in Cointegration with Application to the Demand for Money," Oxford Bulletin of Economics and Statistics, 52 (2): 169-210. Klein, M. W., and E. Rosengren. (1994). "The Real Exchange Rate and Foreign Direct Investment in the United States," Journal of International Economics, 36 (3-4): 373-389. Kreicher, L. L. (1981). "International Portfolio Capital Flows and Real Rates of Interest," Review of Economics and Statistics,+63 (1): 20-28. Lee, C. H. (1969). "A Stock-Adjustment Analysis of Capital Movements: The United States-Canadian Case," Journal of Political Economy, 77: 512-523. Lee, C. H. (1977). "A Survey of the Literature on the Determinants of Foreign Portfolio Investments in the United States," Weltwirtschaftliches Archiv—Review of World Economics, 113 (3): 552-569. MacKinnon, J. G. (1991). "Critical Values for Cointegration Tests." In R. F. Engle and C W. J. Granger (Eds.), Modelling Long-Run Economic Relationships (pp. 267-276). London: Oxford University Press. Maddala, G S. (1992). Introduction to Econometrics (2nd ed.). New York: Macmillan Publishing Company. McTaggart, D., C , Findlay, and M. Parkin. (1992). Economics. Sydney: AddisonWesley Publishing Company. Miller, N. C , and M. V. N. Whitman. (1970). "A Mean-Variance Analysis of United States Long-Term Portfolio Investment," Quarterly Journal of Economics,++ 175-196. Morling, S., and R. Subbaraman. (1995). Superannuation and Saving. Reserve Bank of Australia Research Discussion Paper 9511. Sydney: Reserve Bank of Australia. Pain, N. (1993). "Financial Liberalization and Foreign Portfolio Investment in the United Kingdom," Oxford Economic Papers, 45 (1): 83-102. Phillips, P. C B., and M. Loretan. (1991). "Estimating Long-Run Economic Equilibria," Review of Economic Studies, 53: 407-436. Phillips, P., and P. Perron. (1988). "Testing for a Unit Root in Time Series Regression," Biometrica, 75: 335-346. Ramsey, J.B. (1969). "Tests for Specification Errors in Classical Linear Least Squares Regression Analysis," Journal of Royal Statistical Society, 31: 350-371. Solnik, B.H. (1974). "Why Not Diversify Internationally Rather Than Domestically?," Financial Analysts Journal, 30: 91-135. Spitaller, E. (1971). "A Survey of Recent Quantitative Studies of Long-Term Capital Movements," IMF Staff Papers, 18: 189-217. Taylor, M. P., and L. Sarno. (1997). "Capital Flows to Developing Countries: Long and Short-Term Determinants," World Bank Economic Review, 11 (3): 451470. Tesar, L. L., and I. M. Werner (1995). "Home Bias and High Turnover," Journal of International Money and Finance, 14 (4): 467-492. Ueda, K. (1990). "Japanese Capital Outflows," Journal of Banking and Finance, 14 (5): 1079-1101. White, H. (1980). "A Heteroskedasticity-Consistent Covariance Estimator and a Direct Test for Heteroskedasticity," Econometrica, 48: 817-838.

Regional Financial Markets

32

APPENDIX 2.1 Time-Series Properties of the Level Variables (1988:1-1997:4)+ Level Variables* PIENO

Sample Observations 37

ADF(lags)**++P (lags)***+++Property*++ -2.196(1) -2.295(3) 1(1) -2.909(3)

POITNO

37

-3.001 (3) -3.001(3)

EXRIFSAV

39

-2.864(3)

-2.643(3) -2.643 (3)

1(1)

USRICPRAV3*

39

-2.490(2) -2.490(2)

-1.326(3) -1.326(3)

1(1)

USRICPRAV6*

38

-2.156(2)

-1.378(3)

1(1)

-2.548(3)

1(1) 1(1)

1(1)

USRILT

39

-2.960(1)

AURGDP*

37

-2.415(3)

-1.220(3)

AUSPI

39

-2.908(1)

-2.477(3)

1(1)

USRGDP

39

-0.508(1)

-0.341 (3) -0.341(3)

1(1)

USSPI

39

-0.803(1)

-0.694(3)

1(1)

-0.269(1) USIP 39 0.129(3) 1(1) + Seasonal effects were investigated in all unit root tests using seasonal dummies. In addition, a dummy variable for the recession in the early 1990s was included. * All the variables (defined below) are in natural logarithms (i.e., were logarithmically transformed), and all financial variables are in real terms. A time trend variable was included in each of the above ADF and PP regressions since for most economic time series, the main alternative to the presence of a unit root is a deterministic linear trend. The latter alternative is called trend stationary or 1(0) process while the former is called difference stationary or 1(1) process. **The critical values for ADF and PP tests are given by MacKinnon (1991). The approximate critical values for the ADF test for a sample of 37 is -4.232, -3.539 and -3.201 at the 1%, 5% and 10% level of significance, respectively. ***The t-statistics reported in the PP (Phillips-Perron) column are Dickey-Fuller statistics, with the modifications suggested by Phillips and Perron (1988). The approximate critical values for the PP test for a sample of 37 are the same as for the ADF test provided above. • All these variables become stationary after first differencing. •These variables are first difference stationary as indicated by the PP unit root test. However, the ADF test suggests that these variables are second difference stationary.

Variable Definitions: PIENO

= Portfolio investments equity: nonofficial (in Australia)

POITNO

= Portfolio and other investments total: nonofficial (in Australia)

EXRIFSAV

= Exchange rate average (USD/AUD)

USRICPRAV3

= U.S. (real) interest rate (3-month commercial-paper rate average)

USRICPRAV6

= U.S. (real) interest rate (6-month commercial-paper rate average)

USRILT

= U.S. (real) interest rate (10-year [long-term] government bond yield)

AURGDP

= Australian (real) gross domestic product

AUSPI

= Australian share price index

USRGDP

= U.S. (real) gross domestic product

USSPI

= U.S. share price index

USIP

= U.S. industrial production

CHAPTER 3

The Level of Managerial Ownership, Leverage, and Dividend Policies: Hong Kong Evidence Rohit Jain and Kam-wah Lai

INTRODUCTIO++ Since Jensen and Meckling (1976), managerial ownership has been recognized as an important determinant of capital structure (e.g., Kim & Sorensen, 1986, and Friend & Lang, 1988) and as a proxy for outside investors' monitoring on dividend policy (e.g., Rozeff, 1982). Recent research has investigated the relationship of managerial ownership with leverage and dividend policies. Most studies do not treat mutual ownership as an endogenous variable; that is, the level of mutual ownership determines leverage and dividend policies, and in turn, leverage and dividend policy affects the level of mutual shareholding. Crutchley and Hansen, in studying ownership, leverage and dividends (1989), concluded that managerial ownership and leverage and dividend policies were determined within a policy mix to minimize agency cost. However, they did not disentangle the direct effect among the policies from the indirect effect through the operating characteristics of the firm. Jensen, Solberg, and Zorn (1992) disentangled the direct effect but found that managerial ownership determined the two policies and not vice versa. Their result essentially shows that the nature of managerial ownership is exogenous in relation to the formulation of leverage and dividend policies. This chapter studies managerial ownership, leverage and dividend policies and reexamines the exogenous/endogenous issue using Hong Kong company data, where managerial ownership is higher than for U.S. firms. We find that managerial ownership is endogenous. Unlike Crutchley et al. (1989), who hypothesized a negative relationship

34

Regional Financial Markets

between every pair of the variables—managerial ownership, leverage, and dividend policies—we argue that it is sufficient for only one pair of variables to have a negative relationship. For Hong Kong firms we find that managerial ownership, leverage, and dividend policies are mutually determined. In the next section, the literature review is presented; hypotheses, research methods, and sample selection are stated in the section on research and design. The results section presents and discusses the results, while the last section concludes the chapter. LITERATURE REVIEW In a publicly held firm, managers run their firms on behalf of investors. Whereas ex-post settling up of owner-managers with themselves is inevitable (Fama, 1980), managers can pursue their own interests without bearing the full cost of doing so. Divergence of interest from investors is greatest when managers do not own any share of their firms, giving rise to higher equity agency conflict (Jensen et al., 1976). Limited liability of investors is another feature of modern corporations. Essentially, investors hold an option on the value of the firm and can transfer wealth from lenders by increasing the riskiness of the firm (Black and Scholes, 1973). This conflict between investors and lenders is termed debt agency conflict. When managers also own shares in their firms, they will be involved in the debt agency conflict as well. Lenders impose restrictions on the firm via covenants (Smith and Warner, 1979). Monitoring from investors also occurs. The agency cost arising from conflict is borne in part by investors (Jensen et al., 1976; Roberts and Viscione, 1984). Managers who own shares in their firms have an inherent interest to minimize agency cost. The hypothesis that managerial shareholding, leverage, and dividend policies are interdependent rests on the notion that they are complementary measures at the disposal of managers in reducing agency cost in a policy mix. Equity agency cost can be reduced when the managers own a larger proportion of their firms (Jensen et al., 1976). Debt agency cost is directly proportional to the amount of debt (Green and Talmor, 1986). An increase in managerial ownership may reduce the reliance on external funds and make managers bear more of the cost of their wealth-maximizing behavior. Agency cost, of both equity and debt, can be reduced by managers when they increase their shareholding. Both dividends and leverage reduce equity agency cost. By paying dividends, managers raise the debt-to-equity ratio of their firms and expose their firms to the scrutiny of the capital market when they raise funds externally later on. Shareholders can thus cut down on routine monitoring

The Level of Managerial Ownership, Leverage, and Dividend Policies

35

(Easterbrook, 1984). The raising of debt reduces the reliance on external equity and hence its attendant agency cost. Managers can use either dividend policy or leverage policy, or both, to minimize agency cost. Despite its rapid growth in the literature, the extant agency theory provides no indication on the exact relationship among managerial shareholding, leverage, and dividend policies in a policy mix to reduce agency cost. Crutchley et al. (1989), in the "trade-off hypothesis/' predicted that the relationship between every pair of the variables—managerial shareholding, leverage, and dividend policies—was negative. This may not necessarily be the case. Given that leverage and dividend policies reduce equity agency cost, it is reasonable to expect a negative relationship between them. If debt agency conflict is intense, due to the use of either dividend policy or leverage policy, managers may decrease their shareholding to mitigate the intensive debt conflict. Managerial ownership then bears a positive relationship with either one of dividend or leverage policies. Hypothesis Given the high managerial ownership in Hong Kong firms, we hypothesize that a negative relationship exists between any two of the variables, managerial shareholding, leverage, and dividend policies. We call this the "modified trade-off hypothesis." RESEARCH D E S I G N Analysis We apply three stage least square (TSLS) to the following system of simultaneous equations:

where a0-a4, b0-b8/ and c0-c6 are the parameters to be estimated and e0, ed, and ex denote the random term pertaining to ownership, dividend, and leverage, respectively. TSLS is a system-of-equations technique, which gives unbiased esti-

36

Regional Financial Markets

mates of parameters when there is interdependence among variables. In this, managerial shareholding, leverage, and dividend policies are hypothesized to be mutually dependent. In the above equations, they appear on both the right-hand side and left-hand side of the equations—as independent variables but also as dependent variables. In TSLS, each equation is expressed in its structural form. Each dependent variable appears on the right hand side of other equation(s) and has its own set of independent or regressor variables. We do not use the technique of Crutchley et al. (1989). They used reduced form equations in which all dependent variables are regressed on the same set of regressor variables and do not appear on the right hand side of other equations. The significance and sign of the parameter estimate of the same regressor variable establish mutual dependence across the equations. There is no rule of thumb to decide how many of the coefficients have to be significant before we can conclude or reject interdependence. In TSLS, interdependence can be concluded by direct examination of the parameter estimates of the managerial shareholding, leverage, and dividend variables on the right-hand side of the equations. Consistent with our discussion in the previous section, we expect the parameters a3, a4, b 7 , b 8 , c5, and c6 to be significant; and of the pairs b 7 and a3/ c5 and a4, and c6 and b 8 , at least one pair is expected to be negatively correlated. We define managerial ownership (OWNSHIP above) as the average of the ratio of the total number of ordinary shares held by managers (DSHRS) to the total number of outstanding ordinary shares (TOTSHRS). DSHRS includes shares held through affiliates and family members. Dividend (DIVIDEND) is the average of the ratio of the amount of ordinary share dividends to the total market capitalization of the firm. The latter is the product of TOTSHRS and the year-end closing market price of ordinary shares (MPRICES). Leverage (EEVERAGE) is the average of the ratio of total debt (DEBT) to the sum of total debt (DEBT) and market value of ordinary shares held by outside investors. The latter is the product of total number of ordinary shares held by outside investors and MPRICES. The definition of DEBT includes short-term bank loans. In Hong Kong, there are no debentures. Lending from banks represents the major source of capital. Frequently, bank loans are short-term but rolled over at maturity. Thus, bank loans are treated as long-term financing of the firms. Other variables are added to the above system of equations to avoid spurious correlation among managerial shareholding, leverage, and dividend. In the managerial ownership equation (Equation 1), we add diversification loss (DIVERSE) and firm size (FIRMSIZE). When managers change their ownership to control agency cost, their portfolio will not be well diversified. Managers will suffer a diversification loss. DIVERSE is inversely proportional to such loss. Following Crutchley et al. (1989),

The Level of Managerial Ownership, Leverage, and Dividend Policies

37

we define DIVERSE as the ratio of the excess return of a firm's ordinary shares over its standard deviation (STDRET). The excess return is the difference between the expected return on ordinary shares, measured as the monthly average return over the period January 1, 1987-June 30, 1991, and the risk-free interest rate measured as the interest rate on Hong Kong dollar savings account over the same period. The reason for the choice is that the managers, by holding shares in their firms, hold a less well-diversified portfolio, which consists of their ownership in their firms and other investments. Such a portfolio has an opportunity set that is different from a well-diversified portfolio, which consists of the market portfolio and the risk-free investment. The steeper the slope of the manager's opportunity set, the closer it is to the welldiversified portfolio and the lesser the diversification loss. DIVERSE is used to measure such slope. We expect that coefficient aT will be positive. FIRMSIZE is a proxy for the liquidity cost of maintaining shareholding. The larger the firm, the greater the amount of cash required to hold a given proportion of the firm. FIRMSIZE is measured as the average of the total assets (ASSETS) of the firm. We expect coefficien++a2 to be negative. In the dividend equation (Equation 2), we have included variability in equity return (STDRET), FIRMSIZE, GROWTH, liquidity (LIQUIDS), profitability (PROFIT) and retained reserves (RESERVES). STDRET and FIRMSIZE proxy for flotation costs. The greater the variability in equity return (STDRET), the higher the fee charged by issuing houses. The larger the firm (FIRMSIZE), the cheaper the per dollar flotation cost as a result of economies of scale. We expect coefficient+b 2 to be negative and b^ to be positive. STDRET is the standard deviation of returns on ordinary shares. FIRMSIZE is defined as above. GROWTH is assumed to proxy for the requirement of capital. The greater the need for funds, the less the dividends will be paid. GROWTH is measured as the average of the ratio of the increase in sales of a year over the same in the preceding year to sales in that year. We expect coefficient+b3 to be negative. LIQUIDS, PROFIT, and RESERVES stand for liquidity, profitability, and retained reserves, respectively. With more of these attributes, the firm can pay more dividends. We expect coefficient+ +4 , b 5 , and b 6 to be positive. LIQUIDS is defined as the average of the ratio of cash and cash-equivalent assets to total assets (ASSETS). Cash-equivalent assets are cash and bank balances and marketable securities. RESERVES is measured as the average of the ratio of distributable reserves to total equity Distributable reserves are the sum of retained earnings and revenue reserves. PROFIT is the average of the ratio of net operating income (NOI) to ASSETS. The period covered is the same as EARNVOL below. For the leverage equation (equation 3), we add earnings volatility (EARNVOL), debt capacity (DEBTCAP), growth (GROWTH), and debt

38

Regional Financial Markets

agency cost (INVEST). When the firms' expected bankruptcy cost (proxied by EARNVOL) or debt agency cost (proxied by INVEST) rises, the firms will decrease leverage. If the firms need funds (proxied by GROWTH) or have a greater debt capacity (DEBTCAP), they might increase leverage. We expect coefficient++c2 and c4 to be negative and ++A+and+++c3+to be po We measure EARNVOL as the standard deviation of the ratio of net operating income (NOI) to total assets (ASSETS). The period covered for this variable varies according to the individual firm. The sample is drawn such that each firm has at least eight years' data in the test period 19871990. Debt capacity (DEBTCAP) is the average of the ratio of net fixed assets to total assets (ASSETS). The more the fixed assets, the larger the amount of debt that can be secured and hence the higher the debt capacity. Debt agency cost (INVEST) is the average of the ratio of the sum of marketable securities, investment and development properties, intangibles, and other investments to total assets (ASSETS). The numerator comprises assets, the value of which is fluctuating. We use this variable in place of research and development expenditure to proxy for debt agency cost. Only a few firms in Hong Kong undertook research and development

activities during the sample perio Sample The sample period is 1987 to 1990. Variables are measured for this period unless mentioned already in the previous section. The sample is drawn from listed companies in Hong Kong. There were 196 nonfinancial firms listed on the Hong Kong Stock Exchange through the period 1987 to 1990. From this, firms with all the requisite data are selected. The sample size is 109. The data is taken from the Pacific-Basin Capital Markets Databases (PACAP), except for the managerial ownership information, which comes from the annual reports of listed companies. PACAP is a database that keeps track of capital markets data for eleven Pacific-Basin countries including Hong Kong on a continuous and systematic basis. Table 3.1 shows the descriptive statistics for the sample firms. Managerial ownership ranges from zero to 78% with a mean of 42%. This mean is considerably higher than 6.4% in Crutchley et al. (1989) and 15% in Jensen et al. (1992).] The managerial ownership data indicates the difference between the Hong Kong environment and the United States environment. On average, the dividend payout ratio of the firms in this sample is low, and the range is narrow (from 0 to 0.13). The mean ratio is comparable to that of Crutchley et al. (1989) but not to that of Graham, Bromson, Ma, and Pak (1992b). This is due to different formulae used. 2 Leverage of the sample firms ranges from 0 to 0.89. The mean ratio is 0.33. It is a

little higher than Crutchley and Hansen's sample of 0.27. It will be also

The Level of Managerial Ownership, Leverage, and Dividend Policies

39

Table 3.1 Descriptive Statistics for the Sample Firms

Minimum

Maximum

Mean

Standard deviation

OWNSHIP+

0

0.7813

0.4185

0.2468

DIVIDEND+

0

0.1306

0.0384

0.0244

LEVERAGE++

0

0.8925

0.3288

0.2170

FIRMSIZ++

37,746

41,438,739

3,533,577

7,482,772

STDRE++

0.0619

0.7467

0.1759

0.0875

GROWT++

-0.2651

60.3511

1.3052

6.031

PROFIT

-0.0636

0.5285

0.0832

0.0778

LIQUIDS

0.002533

0.7697

0.1552

0.1383

RESERVES+

-3.6669

1.2295

0.1573

0.5292

DIVERSE+

-0.2036

0.4395

0.1546

0.0908

EARNVO+

0.0105

0.2604

0.0691

0.0459

0.0000951

0.9265

0.269

0.2178

0

0.9525

0.3099

0.2355

Variable

DEBTCAP+ INVES+

FIRMSIZE is in thousands of Hong Kong dollars.

higher than the sample of Graham et al. (1992a).3 In our sample, there is one firm with zero managerial shareholding, five firms with zero dividend payout ratio, and one firm with zero leverage ratio. The zero managerial ownership firm and the zero leverage ratio firm are among the five firms with zero dividend payout ratio. As our objective is not to compare the "zero ratio" firms with others, they are retained in the sample. The firm size of the sample covers a wide range. The largest firm is Hong Kong Land Holdings Limited. The average firm size is 3,533 million Hong Kong dollars ($HK). The distribution may not be normal. It is biased towards the lower end, that is, there are more smaller firms than larger firms. There are 13 firms that experienced negative average growth in the sample (1987-1990) period. Nineteen firms in the sample experienced more than 100 percent growth. The average growth rate is 1.31. Likewise, nine firms in the sample exhibited average negative retained reserves over the sample period. Out of the nine, two firms also had negative growth. Only one firm had a retained reserves ratio that is larger than 1.0.

40

Regional Financial Markets

The average ratio of the variable DIVERSE is 0.15. For the vast majority of the sample firms, DIVERSE is positive. Only four firms have a negative DIVERSE. Out of the four firms, three have negative monthly average returns while the remaining one has a small positive return. This indicates that they have performed badly. Table 3.2 shows the correlation matrix of the variables of the sample firms. All variables are included in the table, be they independent variables or endogenous variables. None of the Pearson correlation coefficients have an absolute value which is greater than 0.5. Only 10 variables have correlation coefficients of 0.3 or greater. Of the 10 variables, only 6 relate to pairs of independent variables. The result indicates that the problem of multicollinearity is not serious for the sample data. We also notice from Table 3.2 that none of the correlation coefficients for pairs of managerial shareholding, leverage, and dividend go beyond - 0.3 or + 0.3. This indicates that the interdependence among the three policies does not arise from spurious correlation. RESULT++ The results of the TSLS analysis are shown in Tables 3.3 to 3.5 for managerial shareholding, dividend, and leverage equations respectively The coefficients+a3, a4, b7, bs, c5,+and+c6+are statistically significant at 1 pe cent. This supports the notion that managerial ownership is endogenous. Jensen et al. (1992) also used TSLS (with different independent variables to control for spurious result) for U.S.-listed firms, but could only conclude that managerial ownership determined leverage and dividend policies but not vice versa. Their result implied that managerial ownership is exogenous. Of the pairs of coefficients+b7 and a3, c5 and a4, and c6 and bs, the pair c6 and b8 is negative while the other two pairs are positive. This supports our modified trade-off hypothesis. Leverage and dividend policies are traded off by managers in reducing agency cost. Managerial ownership supplements the role of leverage and dividend policies The results pertaining to the independent variables are not significant. We reran the analysis by two stage least squares (2SLS), another systemof-equations technique. The result obtained, however, is not substantially different from that with TSLS reported in Tables 3.3 to 3.5. Hence, our result is not sensitive to the research technique used. We also reran the analysis by dropping the managerial shareholding, leverage, and dividend variables from the right-hand side of the equations. This essentially reduces the system of simultaneous equations to three separate independent equations. We aim to discover what the situation will be without interdependence among managerial shareholding, dividend, and leverage. The results are shown in Tables 3.6, 3.7, and 3.8.

Table 3.2 Correlation Matrix—Sample Firms OWNSHIP OWNSHIP

LEVERAGE

DIVIDEND

FIRMSIZE

STDRET

RESERVES

GROWTH

PROFIT

LIQUIDS

DIVERSE

EARNVOL

DEBTCAP

1

LEVERAGE

0.265

1

DIVIDEND

0.110

-0.169

1

FIRMSIZE

-0.279

-0.024

-0.146

1

STDRET

0.086

0.398

-0.241

-0.129

RESERVES

0.110

-0.192

0.358

0.046

-0.329

GROWTH

0.135

0.259

-0.005

-0.042

0.137

-0.003

PROFIT

0.002

-0.462

0.370

-0.048

-0.405

0.386

-0.052

1

LIQUIDS

0.049

-0.371

0.397

-0.144

-0.187

0.108

-0.038

0.291

DIVERSE

0.029

-0.154

0.248

-0.087

-0.057

0.037

-0.112

-0.008

0.062

1

EARNVOL

0.216

-0.021

0.033

-0.258

-0.026

-0.301

-0.014

0.030

0.167

0.103

1

DEBTCAP

-0.207

-0.141

-0.114

-0.090

-0.026

-0.171

-0.175

0.056

-0.245

0.122

-0.042

1

INVEST

0.092

0.339

-0.129

0.165

0.163

-0.002

0.167

-0.180

-0.116

-0.005

-0.150

-0.491

The Pearson correlation coefficient is shown in the cells.

1 1 1

1

INVEST

42

Regional Financial Markets

Table 3.3 Results for the Managerial Ownership Equation (dependent variable OWNSHIP) Parameter estimate

0

+

-0.258

t-statistic (Prob>jt|) -1.864* (0.0652)

a

l

^-ve

-0.0262

-0.198 (0.844)

FIRMSIZE

a

2

-ve

-9.25X10' 10

-0.457 (0.648)

DIVIDEND

a

3

~ve/-ve

+9.593

4.418*** (0.0001)

LEVERAGE

a4

-ve-ve

-0.958

4.318*** (0.0001)

Variable

Coefficient

Intercept

a

DIVERSE

Expected sign

•

System weighted R-square: 0.4691 The f-statistic is for the null hypothesis that the parameter estimate is zero. Prob> I M is the probability of observing a f-value that is greater than the absolute value of t under the null hypothesis. Significance: 1% ***, 10% *

The R-square of the managerial shareholding, dividend, and leverage equations is 6.1 percent, 25.5 percent, and 12.9 percent respectively. All of+ them are significantly lower than the system weighted R-square of TSLS, which is 46.9 percent. This shows that mutual dependence among managerial shareholding, leverage, and dividend policies is an important explanatory factor in their determination. In Table 3.6, the liquidity cost of maintaining ownership is the prime factor for considering the fraction of total equity of the firm to be held by managers. FIRMSIZE is significant at 1 percent. For the dividend equation, Table 3.7 shows results that are similar to Table 3.4. Variables LIQUIDS and RESERVES become significant determinants of dividend policy in the absence of managerial ownership and leverage. PROFIT is only marginally insignificant at 10 percent. These three variables relate to the ability to pay dividends and seem to be the prime factors in dividend policy. This appears to be consistent with the findings of Graham et al. (1992b). The authors conducted a survey of listed companies in Hong Kong on the views of chief financial officers as to the determinants of dividend policy. It was found that current profitability ranked at the top and liquidity, which was more important to smaller firms, ranked second. For the leverage equation in Table 3.8, variables GROWTH and INVEST are significant determinants of leverage for our sample. INVEST does not

The Level of Managerial Ownership, Leverage, and Dividend Policies

43

Table 3.4 Results for the Dividend Equation (dependent variable DIVIDEND) Parameter estimate

t-statistic (Prob>|t|)

+0.031

2.404** (0.018)

+++

+ 1.78xl0" 12

0.007 (0.995)

b

2

+++ ++

-0.00034

-0.022 (0.982)

GROWTH

b3

+++

+0.0000837

0.259 (0.796)

LIQUIDS

b4

+++

-0.00228

0.203 (0.839)

RESERVES

h

5

+++

-0.00021

0.062 (0.95)

PROFIT

*6

+++

-0.00066

0.028 (0.978)

OWNSHIP

h

+++++

+0.0943

3.722*** (0.0003)

LEVERAGE

*8

+-ve/-ve

-0.0979

-2.866*** (0.0051)

Expected sign

Variable

Coefficient

Intercept

++

+

FIRMSIZE

"1

STDRET

The f-statistic is for the null hypothesis that the parameter estimate is zero. P r o b > I H is the probability of observing a f-value that is greater than the absolute value of t u n d e r the null hypothesis. Significance: 1% ***; 5% *

have the expected sign. When the regression is rerun by using+INVEST+ (INVEST+++excludes investment and development properties, the valu+ which does not decline in the sample period and hence might attract less debt agency conflict), INVEST++becomes insignificant+++t =+0.792 an Prob> I f I = 0.43). The only significant variable is+GROWTH (t =+2.433 and Prob> 111 = 0.0167). (This result is not reported in a table.) It appears that variables INVEST+and INVEST!+do not capture debt agency costs well, at least for the current sample. CONCLUSION This contributes to an understanding of the determination of leverage and dividend policies, especially in an environment of high managerial ownership of firms. This proposes and tests a modified "trade-off hypothesis" and sheds new light on the possible relationships among man-

44

Regional Financial Markets

Table 3.5 Results for the Leverage Equation (dependent variable LEVERAGE) Parameter estimate

t-statistic (Prob>|t|)

+

-0.307

2.759*** (0.0069)

++

+++

-0.00662

-0.136 (0.892)

++

++

-0.0393

0.119 (0.906)

c

3

+++

-0.00057

0.27 (0.788)

++ ++

++

-0.00393

0.097 (0.923)

Variable

Coefficient

Intercept

++

DEBTCAP EARNVOL+ GROWT++ INVEST+

Expected sign

OWNSHI++

c

5

+++++

-0.9585

3.384*** (0.001)

DIVIDEN++

++

+++++

-9.95

-5.895*** (0.0001)

The f-statistic is for the null hypothesis that the parameter estimate is zero. P r o b > I f I is the probability of observing a f-value that is greater than the absolute value of f u n d e r the null hypothesis. Significance: 1% ***

Table 3.6 Result of Independent Equation for Managerial Shareholding (dependent variable—OWNSHIP)++

Variable

Expected sign of coefficient

Parameter estimate

t-statistics

Prob>|t

Intercept

+

0.4488

9.426**'

0.0001

DIVERSE

++ +++

0.0140

0.055

0.9561

FIRMSIZE

+++

-9.2xl0 - 9

-2.975***

0.0036

The f-statistic is for the null hypothesis that the parameter estimate is zero. P r o b > I f I is the probability of observing a f-value that is greater than the absolute value of f u n d e r the null hypothesis. F-value for model: 4.476; Prob>F: 0.0136; Adjusted R-square: 0.0605. Significance: 1% ***

The Level of Managerial Ownership, Leverage, and Dividend Policies

45

Table 3.7 Result of Independent Regression of Dividend Equation (dependent variable—DIVIDEND)

Variable

Expected sign of coeffi cient

Parameter estimate

t-statistics

Prob>|t|

Intercept

-

0.0283

4.502***

0.0001

FIRMSIZE

^ve

-3.7xl0" 1 0

-1.331

0.1861

STDRET

-ve

-0.0156

-0.587

0.5586

GROWTH

-ve

0.000074

0.218

0.8277

PROFIT

+ve

0.0500

1.629

0.1064

LIQUIDS

+ve

0.0524

3.373***

0.0011

RESERVES

+ve

0.0116

2.722***

0.0076

The f-statistic is for the null hypothesis that the parameter estimate is zero. P r o b > I f I is the probability of observing a f-value that is greater than the absolute value of t under the null hypothesis. F-value for model: 7.167; Prob>F: 0.0001; Adjusted R-square: 0.2552. Significance: 1%***

Table 3.8 Results of Independent Regression of Leverage Equation (dependent variable—LEVERAGE+

Variable

Expected sign of coefficient+

Intercept

+

0.1920

2.896*++

EARNVOL

++

0.1691

0.39

0.6974

DEBTCAP

++

0.0656

0.629

0.5307

GROWTH

++

0.0072

2.34**

0.0212

INVEST

++

0.3141

3.228***

0.0017

Parameter

estimate+++++++++++++++++++++++++ ++++++

The f-statistic is for the null hypothesis that the parameter estimate is zero. P r o b > I f I is the probability of observing a t-value that is greater than the absolute value of t u n d e r the null hypothesis. F-value for model: 4.992; Prob>F: 0.001; Adjusted R-square: 0.1288. Significance: 1%***; 5%**

46

Regional Financial Markets

agerial shareholding, leverage, a n d d i v i d e n d policies. We find that leverage a n d d i v i d e n d policies are jointly d e t e r m i n e d b y m a n a g e r i a l shareholding. We also establish that m a n a g e r i a l o w n e r s h i p is e n d o g e n o u s ; that is, it is also d e t e r m i n e d b y leverage a n d d i v i d e n d policies. We p r o v i d e evidence that m a n a g e r s m a k e trade-offs a m o n g leverage a n d d i v i d e n d policies a n d their o w n e r s h i p to m i n i m i z e agency cost, as envisaged b y agency theory. NOTE++ 1. Crutchley et al. (1989) used 1981 to 1985 as the sample period, while Jensen et al. (1992) performed cross-sectional analysis at two points in time, that is, 1982 and 1987. The 15% in the main text refers to 1987 data. The mean managerial ownership for 1982 is 16%. This shows that although sample period and the sample may differ, the average managerial ownership of firms in the United States is not likely to rise above that in Hong Kong. 2. Most of the ratios reported by Graham et al. (1992b) were over 0.4. Although the formula used was not disclosed, the note to Table 2 discerned that negative ratio was permitted in the formula. In contrast, the formula used in this paper allows only positive figures. 3. Graham et al. (1992a) did not include debt in the denominator of the ratio as we do. Therefore, their ratio will be adjusted from around 0.31 downwards in order to make a comparison. REFERENCES Black, S., and M. Scholes. (1973). "The Pricing of Options and Corporate Liabilities/' Journal of Political Economy, (May/June), 637-654. Crutchley, C. E., and R. S. Hansen. (1989). "A Test of the Agency Theory of Managerial Ownership, Corporate Leverage and Corporate Dividends," Financial Management, 18, 4 (Winter), 35-46. Easterbrook, F. H. (1984). "Two Agency-Cost Explanations of Dividends/' American Economic Review, (September), 650-659. Fama, E. F. (1980). "Agency Problems and the Theory of the Firm," journal of Political Economy, (April), 288-307. Friend, I., and L. H. P. Lang. (1988). "An Empirical Test of the Impact of Managerial Self-interest on Corporate Capital Structure," Journal of Finance, (June), 271282. Graham, B. R., G. Bromson, A. Ma, and S. L. Pak. (1992a). "The Capital Structure Puzzle," Securities Journal, (June), 20-22. . (1992b). "The Dividend Decision," Securities Journal, (July), 20-22. Green, R. C , and E. Talmor. (1986). "Asset Substitution and the Agency Costs of Debt Financing," Journal of Banking and Finance (Netherlands), 10,3 (October), 391-399. Jensen, G. R., D. P. Solberg, and T. S. Zorn. (1992). "Simultaneous Determination of Insider Ownership, Debt, and Dividend Policies," Journal of Financial and Quantitative Analysis, (June), 247-263.

The Level of Managerial Ownership, Leverage, and Dividend Policies

47

Jensen, M. C , and W. H. Meckling. (1976). "Theory of the Firm: Managerial Behavior, Agency Costs and Capital Structure," Journal of Financial Economics, (October), 305-360. Kim, W. S., and E. H. Sorensen. (1986). "Evidence on the Impact of the Agency Costs of Debt on Corporate Debt Policy," Journal of Financial and Quantitative Analysis, (June), 131-144. Roberts, G. S., and J. A. Viscione. (1984). "Note on Who Pays the Agency Costs of Debt," Financial Review, (May), 232-239. Rozeff, M. S. (1982). "Growth, Beta and Agency Costs as Determinants of Dividend Payout Ratios," Journal of Financial Research, (Fall), 249-259. Smith, C. W , Jr., and J. B. Warner. (1979). "On Financial Contracting: An Analysis of Bond Covenants," Journal of Financial Economics, (June), 117-161.

APPENDIX 3.1 Results of Integration Tests Series

Level:

First difference:

ADF

PP

ADF

PP

Actual gross revenue

-3.31

-2.45

-3.01**

-5.09**

Anticipated gross revenue

-2.81

-2.61

-3.27**

-7.09**

Actual capital expenditure

-2.34

-3.19

-2.76

-8.91**

Anticipated capital

-2.29

-2.73

-3.69**

-7.27**

Actual employment -3.08+++++++++

-3.08

-3.48

-4.05

-6.47

Anticipated+ + + + +2.3 + + -3.39+ +-3.40* + +6.23*

-2.33

-3.39

expenditure

employment

-3.40

-6.23

Notes: Lag length chosen for ADF was 3, based on Schwert (1987) formula: int{ +++++++++++++++++++++++++++++++++++++++ T is the total n u m b e r of observations. For PP, the lag length chosen was also 3, based on the Bartlett are - 3 . 5 2 (5%). ADF critical values for series in first differences are - 2 . 9 3 (5%). PP critical values for series in levels are - 3 . 5 1 (5%). PP critical values for series in first differences are - 2 . 9 3 (5%). See MacKinnon, 1991. Asterisk (**) denotes statistically significant at 5 percent level.

APPENDIX 3.2 Results of U n b i a s e d n e s s Tests Dependent variable is actual values of: Gross revenue Capital expenditure Employment Constant (a)

0.2283 (1.2794)

0.6612 (2.1345)**

0.0259 (0.4270)

Slope (ft

0.9731 (44.660)**

0.8059 (13.016)**

0.9917 (57.104)**

R-squared

0.9798

0.8051

0.9875

CRDW ADF{\)+ PPO)

2.26** -4.63** -7.41**

1.64** -3.32** -5.43**

1.96** -4.48** -6.29**

F-statistics (or=0,/H)

0.879 [0.4221

18.084 [0.000]**

0.285 [0.753]

LM/(4)

2.496 [0.645]

4.948 [0.292]

2.133 [0.711]

Notes: Critical value for CRDW at 5% level is 0.78 (see Engle and Yoo, 1987). Critical values for ADF and PP is -1.94 (5%) (see MacKinnon, 1991). The LM chi-square statistic for serial correlation with four lags is 9.48, with four degrees of freedom at 5 percent level. Figures in square brackets are p-values. Asterisk (**) denotes statistically significant at 5 percent level.

APPENDIX 3.3 Results of Efficiency Test Lag length

Test with respect to lagged forecast error: Gross Capital Employment revenue expenditure

+-statistics with

+^ngth:

Test with respect to lagged actual values: Gross Capital Employment revenue expenditure

respect to lag

1

0.668 [0.517]

11.954 [0.000]**

0.177 [0.838]

0.165 [0.847]

11.313 [0.000]**

0.186 [0.830]

2

0.408 [0.747]

8.660 [0.000]**

0.121 [0.946]

0.119 [0.947]

7.912 [0.000]**

1.650 [0.194]

3

0.330 [0.855]

7.253 [0.000]**

0.141 [0.965]

0.143 [0.964]

5.927 [0.001]**

1.199 [0.328]

Notes: Figures in square brackets are ^-values. Asterisk (**) denotes statistically significant at 5 percent level.

CHAPTER 4

Examining the Performance of the Malaysian Life Insurance Sector: Efficiency and Productivity Growth Alias Radam and Shazali Abu Mansor

INTRODUCTION The interest in the study of efficiency has continued to draw considerable passion among researchers. To an economist efficiency is inherently important, positing a direct relationship with productivity. Calls for liberalization of trade by the developed nations have been pervasively pushed at the World Trade Organization, where the North insists that developing countries should open up their markets for goods and services. Thus, under the era of globalization, improving efficiency and productivity have been seen as the important factors contributing to the survival of firms in the developing countries. While various techniques have been developed over the period of time to study the efficiency growth of firms and industry, most have concentrated on goods. Limited attempts have been made to study the efficiency in the service sector, and the bulk of these studies have been confined to income and profitability. Insurance has been an important instrument to transfer risk by providing financial assistance to those who suffer the effects of risks, whereby the individual or the business enterprise can shift some of the uncertainty to the shoulders of insurance companies with a small amount of premium. Without insurance, there would be a great deal of uncertainty experienced by an individual or enterprise, not only as to whether a loss would occur, but also as to the degree of a loss if one did occur. While losses incurred by firms and individuals as a result of accidents and death could not be avoided resulting in economic loss, with an efficient insurance sector, resource could be utilized at the utmost possibility, minimizing the loss in an economy.

52

Regional Financial Markets

This chapter attempts to examine efficiency and productivity of the insurance industry, namely the life insurance sector. The next section presents the historical backdrop of the insurance industry, followed by a literature review on productivity and subsequently results and implications for policy. HISTORICAL BACKDROP The life insurance concept was conceived as early as 2800 B.C., when it was being applied by merchants to protect themselves against the loss of their property. In Malaysia, the first rumblings of the insurance industry took place in the eighteenth century, where it was mainly based on the British system because of the advent of colonialism. The insurance industry experienced a spiraling growth in 1950s, and a substantial portion of the market was controlled by British and American firms. Local participation was negligible because of ignorance and lack of expertise. After independence in 1957, encouraged by nationalistic sentiments, domestic companies were poised to make inroads into the insurance market. However, the boom was marred by many companies, that did not have sound underwriting practices. A number of companies went out of business, leaving their policyholders uninsured. In 1963, the government stepped in to remedy the situation by introducing the Insurance Act. The Act required that all insurance companies in Malaysia be registered. The Insurance Division was created in the Finance Ministry to monitor and regulate the insurance industry in Malaysia. Apart from the compulsory registration, the Act also required that insurance companies lodge a deposit, which could not be used as an asset to cover liabilities of the insurance fund, but would be used to protect the policyholders in the event of a winding up of the company. After the Insurance Act was enforced in 1963, the life insurance industry witnessed a steady growth. Between 1965 and 1969, the average growth of business in force, indicated by the total sums insured in force and the annual premium in force, was more than 10 percent. 1 The growth rate was very encouraging, considering the industry was still in the infancy stage with a loose institutional framework. The insurance industry continued to experience a high growth over the years. As shown in Table 4.1, the amount of sums in force has increased from RM 60,390.70 million in 1988 (69.2 percent of GNP) to RM 336,795.70 (125.1 percent of GNP) in 1998. This represents an annual growth of about 18.5 percent within this period. The Insurance Division has played a major role to gear the insurance industry in tandem with economic development of Malaysia. In line with the New Economic Policy implemented in 1969, the Insurance Division had taken measures to encourage the participation of the local companies

Examining the Performance of the Malaysian Life Insurance Sector

53

Table 4.1 Sum Insured in Force RM Million

% of GNP

1988

60,390.7

69.2

1989

73.033.6

73.5

1990

86,678.0

76.0

1991

105,666.5

82.3

1992

129.568.9

90.8

1993

161.410.7

98.5

1994

202.162.3

108.7

1995

246,228.5

116.1

1996

282.605.2

116.8

1997

321.852.7

120.6

1998

336.795.7

125.1

Source: Adapted from Bank Negara Malaysia, 1999

in the insurance industry. The government urged the foreign companies to restructure their insurance branch offices in Malaysia with a view to seeking domestic incorporation of their business in Malaysia with equity participation by Malaysians. As a result, the number of domestic companies increased; and in 1979, domestic insurance companies outnumbered the foreign insurance companies for the first time in the history of insurance in Malaysia. The number of local companies expanded from 6 in 1963 to 51 in 1988, compared to nine branches of foreign incorporated companies. Following the requirement of the Insurance Act of 1996,2 there is only left with two foreign-incorporated insurers in the direct insurance market in Malaysia in 1999. However, the foreigners still hold 45.8 percent of the total equity of insurance companies and control 74.3 percent of life insurance premiums (Bank Negara Malaysia, 1999). The growth of the insurance industry has contributed to employment opportunities in Malaysia. Total employment in the insurance industry increased from 10,911 in 1988 to 19,280 in 1998. Measures were taken to upgrade the skill of the insurance workforce. Various training programs in insurance have been made available by the public and private higher learning institutions. The life insurance industry can be viewed as a stra-

54

Regional Financial Markets

tegic industry in generating technical skills for the nation, as a high degree of technical know-how is involved in this industry, namely actuarial science, underwriting, risk management, and information technology. In light of this, measures were also taken to expand the insurance training facilities, such as the University of ITM and the Insurance Institute of Malaysia. The purpose of the expansion was to supply the industry with qualified professionals to meet the demand of the growing insurance industry. THE M E A S U R E M E N T OF P R O D U C T I V I T Y CHANGE Since Solow's (1956) paper on U.S. aggregate growth, productivity measurement has had an important role in applied economics. Theorists have improved their understanding of the relationship between productivity and other economic variables while applied economists have improved their understanding of the components of productivity growth. This improved understanding has coincided with data-processing capabilities. Therefore, numerous methodologies for measuring productivity have developed over the last three decades. The commonly accepted indices of productivity change are the Tornqvist index (Tornqvist, 1936), the Fisher Ideal index (Fisher, 1922), and the Malmquist index (Malmquist, 1953). The popularity of the Tornqvist and Fisher Ideal indices result from two desirable features they share (Grifell-Tatje and Lovell, 1995). First, both can be calculated directly from price and quantity data, and it is not necessary to recover the structure of the underlying best practice production frontier and how it shifts over time whether by using econometrics techniques to estimate the parameters of functions characterizing the frontier or by using mathematical programming techniques to construct the frontier. Second, both are consistent with flexible representations of the frontier—that is, both are superlative indices (Caves, Christensen, & Diewert, 1982; Diewert, 1992). The popularity of the Malmquist index stems from three quite different sources. First, it is calculated from quantity data only, a distinct advantage if price information is unavailable or if prices are distorted. Second, it rests on much weaker behavioral assumptions than the other two indices, since it does not assume cost-minimizing or revenue-maximizing behavior. Third, provided panel data are available, it provides a decomposition of productivity change into two components. One is labeled technical change, and it reflects improvement or deterioration in the performance of best practice firms. The other is labeled technical efficiency change, and it reflects the convergence toward or the divergence from best practice on the part of the remaining firms. The value of the decomposition is that it provides information on the source of overall productivity change in the firms. We implement the Malmquist index by solving a series of linear

Examining the Performance of the Malaysian Life Insurance Sector

55

programming problems to construct the distance function that make up the Malmquist index. These distance functions characterize the best practice production frontier at any point in time, and they also characterize shifts in the frontier over time as well as movements of the producers toward or away from the frontier. The nonparametric approach, introduced by Farrell (1957), is used largely because it does not require prices and leads directly to simple efficiency comparisons and the Malmquist index. The Farrell technical efficiency measure is defined so that the isoquant, which is the locus of the efficient points that form the boundary of input requirements set, designated the minimal set of inputs, xt, resulting in the unit level of output of yt. The efficiency of the other firms is measured radially relative to this isoquant. A useful feature of the total Malmquist productivity index, first noted by Fare, Grosskopf, Lindgren, and Roos (1995), is that it decomposes into the product of an index of technical efficiency change and an index of technical change, as follows: (1) 0

1

1

1

1

where M{(y , y , x°, x ) is Malmquist productivity index, E{(y°, y , x°, x ) is an index of relative technical efficiency change, T^y0, y1, x°, x1) is technical change of component of productivity, y° is output at time period 0, y1 is output at time period 1, x° is input at time period 0, and x1 is input at time period 1. Productivity changes arising from changes in technical efficiency can be measured as the ratio of two distance functions at different points in time, or as follows: (2)

An index of relative technical efficiency measures the ratio of technical efficiency at time period 0 and time period 1. This is a measure of a firm i catching up to a frontier representing best-practice technology. This index is greater than, equal, or less than unity depending on whether the relative performance of producer i is improving, unchanging, or declining. The second component of total Malmquist productivity index is an index of technical change. Fare et al. (1995) calculate the technical change component of productivity as the geometric means of two ratios of output distinct function as follows. (3)

The four distance functions defined the shift of the technical progress frontier. The ratios compare year t observations with the t + 1 reference tech-

56

Regional Financial Markets

nology, or vice versa. For example, in the first ratio, the numerator measures the technical efficiency in time period 1 relative to technology in time period 0. This is the mixed-distance function. The denominators measure technical efficiency in time period 1 relative to the technology in period 1. The technology index measures the shift in the frontier. This index shows whether the best practice relative to which firm i is compared is improving, stagnant, or deteriorating. This component to greater than, equal to, or less than unity depending on technical change is positive, zero, or negative, on average, at the two observations (y°, x°) and (y1, x1). The Malmquist productivity index and its two components are local indices, in the sense that their values can vary across firms and between different time periods. One firm may exhibit an increase in technical efficiency, another may exhibit a decrease, and either can change over time. Similarly, one firm may exhibit technical progress, another may exhibit technical regress, and either can change over time. E S T I M A T I O N OF M A L M Q U I S T P R O D U C T I V I T Y INDEXES We develop the Malmquist productivity estimates from mathematical programming models of the frontier production function. For a recent survey of this approach see Fare, Grosskopf, and Lovell (1994) and Seiford and Thrall (1990). Calculation and decompositions of the Malmquist productivity index require the calculation of four output distance functions, for each firm in each pair of time periods. We concentrated our attention on Malmquistbased productivity growth in the context of year-by-year improvements. The Malmquist index is computed for each firm in each year of the data using 1975 as the base year for comparison. We follow Arnade (1994) by using linear programming techniques to calculate these output distance functions observations. The reference technology must be defined, and the distance of the K observation from the reference technology must be measured. The programming problem used to calculate the Farrell measure of technical efficiency for a specific observation; K' in time period 0 is set up as subject to

(4)

Examining the Performance of the Malaysian Life Insurance Sector

57

Superscripts on the data represent the time period 0. Superscripts on functions represent the technology that defined by the data. Subscript K' refers to a specific cross-sectional observation. Subscript m and n refer to output and inputs. Mixed-distance functions are estimated by comparing observations in one time period with the best-practice frontier of another time period. The result is an estimate of the inverse of the mixed-distance function for observation K', which can be defined as [D%y\,x\)] subject to

' = min^

(5)

The technology is defined from data in time period 0, where the efficiency of the specific observation K' is defined using data from time period 1. In measuring the efficiency performance, we evaluate the Malmquist index of a sample of 12 Malaysian insurance companies over the 1987 to 1997 period. 3 We adopt the Malmquist index measures using three variables as the output (new policy issued, premium, policy in force) and three inputs, namely salaries, expenses, and other costs. The data are obtained from the Annual Report of the Director General of Insurance. RESULT A N D D I S C U S S I O N This study engages data envelopment analysis (DEA) to measure technical efficiency, technical changes, and factor productivity. A technical efficiency index measures the efficiency with which inputs are utilized in the production of outputs. DEA has been widely used to calculate and compare technical efficiency across individual firms. Among others are Arnade (1994), Fare, Grosskopf, Lindgren, and Roos (1992), Fare and Grosskopf (1994), Grifell-Tatje et al. (1995), Piesse, Thirtle, and van Zyl (1996), Chavas and Cox (1994), Caves et al. (1996), Cummins, Tennyson, and Weiss (1999), Sathye (2000), and others. We begin by looking at the whole production possibility set, consisting of observed inputs and related outputs produced by insurance firms over the period of 1975 to 1997. In Table 4.2, the constructed frontier is shown by the average Farrell efficiency index for each firm. The average technical efficiency for Malaysian insurance industries for the period of this study

is quite high, that is, 60.87 percent. Firms that experience hig

58

Regional Financial Markets

Table 4.2 Mean Technical Efficiency Index of Malaysian Insurance Company, 1975-1997 Company Malaysia National Insurance Company Bhd. (MNI) Malaysian Assurance Alliance (MAA) Malaysia Cooperative Insurance Society Limited (MCIS) Safety Life General Insurance Sdn. Bhd (SAFETY) Talasco Insurance Sdn. Bhd. (TALASCO) United Malaysian Insurance Company (UMI) American International Assurance Co. (AIA) Asia Life Assurance Society Ltd. (ASIA LIFE) Great Eastern Life Assurance Co. Ltd. (GREAT EASTERN) Oversea Assurance Corporation Ltd. (OAC) Prudential Assurance Co. Ltd. (PRUDENTAL) Malaysia Co operative Insurance Society7 (Housing) (MCIS (H)) AVERAGE

Index 0.9010 0.5074 0.5262 0.4113 0.5798 0.5065 0.5167 0.5802 0.7093 0.7173 0.3862 0.9627 0.6087

technical efficiency include Malaysia Co-operative Insurance Society (Housing Service) (96.27 percent) and Malaysian National Insurance (90.10 percent). The disparity between the highest (96.27 percent) and lowest (38.62 percent) average technical efficiency was large, accounting more than twofold between the two. Firms on the production frontier can be labeled as "best practice" and demonstrate optimum efficiency in resource utilization. An index measure of 1.0 indicates that a firm lies on the best-practice frontier, while an index measure of less than 1.0 indicates inefficient resource utilization compared to those on the best-practice frontier. An inefficiency index subtracted from one represents the largest proportional amount of input that can be reduced without reducing output (Chavas & Aliber, 1993). Annual technical efficiency results are summarized in Table 4.3. It shows that insurance firms under study provide, on average, a range from 33.11 percent (1978) to 82.93 percent (1997) of the output by the best-practice firm over the period 1975 to 1997. The increase in this range over time could be due to a gradual narrowing of the gap between the least efficient and most efficient firms. Table 4.4 shows the average estimate of the Malmquist index, technical efficiency index, and technical change index of the insurance industry from 1975 to 1997. Indices representing productivity growth due to technical change are calculated by estimating technical efficiency in one time period against the best-practice technology of another period. This study's

Examining the Performance of the Malaysian Life Insurance Sector

59

Table 4.3 Mean Technical Efficiency Index of Malaysian Insurance Company, 1975-1997 Year 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 rrowth (%)

+

Index 0.3010 0.6657 0.3564 0.3311 0.3347 0.6292 0.6159 0.5918 0.6417 0.7772 0.8034 0.6989 0.5660 0.5622 0.5100 0.6080 0.5334 0.6612 0.6970 0.6989 0.7037 0.7944 0.8293 2.50

estimates represent the inverse of the technology index defined by Equation (3), so a number greater than 1.0 represents an improvement in productivity due to technical change (Arnade, 1994). Index numbers are defined so that the 1975 observation equals 1.0. A Malmquist productivity index is calculated from a combination of technical efficiency change indices and technical change indices. The estimated indices represent the inverse of the Malmquist index described in Equation (1), so production improvements are greater than 1.0. The technical efficiency index for the insurance industry increased to 2.86 from the base year. It indicates that there was a 186 percent increase in technical efficiency during the 23-year time span, at an average annual growth of 2.5 percent. It also implies that with the same amount of resources used in 1975, it was able to increase output by 186 percent more in 1997. As shown in the table, within the same period of time, the insurance industry acquired a technical progress of 9 percent. The growth of technical progress in this industry is slightly lower than the efficiency

Regional Financial Markets

60

Table 4.4 Malmquist Index, Technical Efficiency Change Index, and Technical Change Index for Insurance Companies, 1975-1997 Year 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

Technical Efficiency Change Index 1.00 2.36 1.02 0.99 0.84 1.74 1.71 1.69 2.02 2.50 2.69 2.25 1.69 1.71 1.53 2.01 1.66 2.22 2.46 2.43 2.39 2.79 2.86

Technical Change Index 1.00 1.27 1.15 1.07 1.11 0.62 1.02 0.94 1.12 1.12 1.00 0.88 1.08 1.04 1.19 0.87 1.06 0.83 0.82 0.84 0.89 0.91 1.09

Malmquist Index 1.00 3.00 1.18 1.06 0.93 1.09 1.74 1.58 2.27 2.80 2.69 1.99 1.82 1.77 1.82 1.75 1.76 1.85 2.02 2.04 2.12 2.54 3.11

growth. The industry also experienced a 211 percent productivity growth from 1975 to 1997. This is equivalent to a 2.81 percent annual productivity growth rate. This suggests that in 1997, insurance industries yielded about 211 percent more output per unit of resource consumed as they produced 23 years ago. However, if compared to the average annual growth of GNP of above 8 percent, and sum insured in force, the productivity growth in this sector is considered to be relatively low. POLICY IMPLICATIONS A N D C O N C L U S I O N While it is difficult to measure efficiency in the service sector, this chapter has highlighted some aspects of productivity in the life insurance sector. It is not the objective of the chapter to argue about the techniques used

Examining the Performance of the Malaysian Life Insurance Sector

61

to measure productivity in the life insurance sector, however it has shown that despite a significant growth in business in this sector, the productivity growth is relatively low. The low productivity growth in the insurance industry was due to the low growth in the technical efficiency and technical progress. This result has policy implications. The development of the insurance industry in Malaysia over the past decades was generally achieved through the rapid opening of branches. Prior to the last decade of the twentieth century, less effort was given in widening of insurance products and acceleration of insurance-related activities. Along with the expansion of infrastructure outlet, the legal framework governing this industry has undergone changes. Despite the ability of these policies to garner growth in the industry, they have not made significant contributions in improving productivity and efficiency. Under the era of globalization the insurance firms need to improve their efficiency and productivity to be competitive. Being able to provide service in an efficient way would be an important source of comparative advantage under the era of globalization. The rapid opening of new branches over the last few decades led to the industry producing at suboptimum. Thus, mergers in the insurance industry are inevitable to reap the economies of scale associated to the service sector. Through mergers, firms in this industry could improve their operational efficiency. The study also shows that there is a large degree of disparity between firms7 efficiency. Hence efficiency performance should be incorporated in implementing the merger exercise as to allow inefficient firms to gain from the more efficient companies. Efficiency in this industry can also be enhanced through the upgrading of the distributive channel with gradual utilization of information technology. The conventional distributive agency problems can be overcome by employing information technology, thus increasing the operational efficiency of the industry. Transactional costs could be reduced by eliminating certain unnecessary middleman activities. The manpower development is another aspect that needs to be addressed. The skills of personnel working in the insurance industry should be in tandem with the era of globalization and information technology. Without a competent workforce, it is difficult to compete, particularly in this type of knowledge-based industry. Thus, learning institutions offering insurance-related courses should be prepared to incorporate a more dynamic curriculum in line with the rapid changing of the environment affecting the industry. Policy should also be addressed to educate the public on the importance of life insurance. Insurance should be viewed not only as a risk aversion instrument, but also as another investment portfolio. Thus, various fiscal incentives need to be considered in encouraging the public to invest in insurance.

62

Regional Financial Markets

As a conclusion, the insurance i n d u s t r y can capitalize on i m p r o v i n g technical efficiency a n d technical progress as a source of c o m p a r a t i v e adv a n t a g e in the era of globalization a n d information technology. The l o w productivity g r o w t h of the i n s u r a n c e sector p r o v i d e s a potential for future g r o w t h of the industry.

ACKNOWLEDGEMENT+++ We are grateful to Prof. Milind Sathye at the D e p a r t m e n t of Finance a n d Banking, University of Southern Q u e e n s l a n d , Australia, a n d seminar participants at the First International Conference on Banking a n d Finance for c o m m e n t s on an earlier version of the chapter.

NOTE++ 1. The are several ways of measuring the growth of insurance industry. One of the ways is to measure the increase of business in force. 2. This Act required all licensed foreign-incorporated insurers to transfer their Malaysian Insurance Businesses to public companies under the Companies Act of 1965. 3. The sample size is in line with the rule of thumb followed by Soteriou and Zenios (1988) and Dyson, Thanassoulis, and Boussofiane (1998), that is, that sample size should be larger than the product of the number of inputs and outputs.

REFERENCE++ Arnade, C. A. (1994). "Using Data Envelopment Analysis to Measure International Agricultural Efficiency and Productivity," Technical Bulletin No. 1831, U.S. Dept. Agr., Econ. Res. Serv., Feb. Avkiran, N. K. (1999). "An Application Reference for Data Envelopment Analysis in Branch Banking: Helping the Novice Researcher/7 International Journal of Bank Marketing+++17 (5): 206-220. Caves, D., L. Christensen, and E. Diewert. (1982). "The Economic Theory of Index Numbers and the Measurement of Input, Output, and Productivity/' Econometrica, 50: 1393-1414. Chavas, J., and T. L. Cox. (1990). "A Non-Parametric Analysis of Productivity: The Case of U.S.," The American Economic Review, 80 (3): 450-464. Chavas, J.P, and M. Aliber. (1993). "An Analysis of Economic Efficiency in Agriculture: A Nonparametric Approach," Journal of Agricultural and Resource Economics, 18. Cummins, J. D., S. Tennyson, and M. A. Weiss. (1999). "Consolidation and Efficiency in the US Life Insurance Industry/' Journal of Banking and Finance, 23: 325-357. Diewert, W. E. (1992). "Fisher Ideal Output, Input and Productivity Indexes Revisited," Journal of Productivity Analysis, 3: 213-248.

Examining the Performance of the Malaysian Life Insurance Sector

63

Dyson, R. G., E. Thanassoulis, and A. Boussofiane. (1998). "Data Envelopment Analysis. Warwick Business School," http://www.csv.warwick.ac.uk/ ~bsrlu/dea/deat/deatl.htm. Fare, R., and S. Grosskopf. (1994). "Estimation of Returns to Scale Using Data Envelopment Analysis: A Comment," European Journal of Operational Research, 79: 379-382. Fare, R., S. Grosskopf, B. Lindgren, and P. Roos. (1992). "Productivity Changes in Swedish Pharmacies, 1980-1989: A Nonparametric Malmquist Approach," Journal of Productivity Analysis, 3. Fare, R., S. Grosskopf, B. Lindgren, and P. Roos. (1995). "Productivity Developments in Swedish Hospitals: A Malmquist Output Index Approach." In A. Charnes, W. W. Cooper, A. Y. Lewin and L. S. Seiford (Eds.), Data Envelopment Analysis: Theory Methodology and Applications (pp. 253-272). Amsterdam: Kluwer Academic Publishers. Fare, R., S. Grosskopf, and C. A. K. Lovell. (1985). Measurement of Efficiency of Production. Boston: Kluwer Nijhoff. Fare, R., S. Grosskopf, and C. A. K. Lovell. (1994). Production Frontier. Cambridge, England: Cambridge University Press. Farrell, M. (1957). "The Measurement of Productive Efficiency," Journal of the Royal Statistical Society, Series A, 3: 253-290. Grifell-Tatje, E., and C A. K. Lovell. (1995). "A Note on the Malmquist Productivity Index," Economics Eetter, 47: 169-175. Malmquist, S. (1953). "Index Numbers and Indifference Surfaces," Tradajos de Estadistica, 4(1): 209-242. Nunamaker, T. R. (1985). "Using Data Envelopment Analysis to Measure the Efficiency of Non-profit Organizations: A Critical Evaluation," Managerial and Decision Economics, 6 (1): 50-58. Piesse, J., C. Thirtle, and J. van Zyl. (1996). "Effects of the 1992 Drought on Productivity in the South African Homelands: An Application of the Malmquist Index," Journal of Agricultural Economics, 47 (2): 247-254. Sathye, M. (2000). "X-efficiency in Australian Banking: An Empirical Investigation," Journal of Banking and Finance, 32. Seiford, L., and R. Thrall. (1990). "Recent Developments in DEA, The Mathematical Approach to Frontier Analysis," Journal of Econometrics, 46. Solow, R. (1957). "Technical Change and the Aggregate Production Function," Review of Economics and Statistics,++39 (3): 312-320. Soteriou, A., and S. Zenios. (1998). "Data Envelopment Analysis: An Introduction and an Application to Bank Branch Performance Assessment." In G. Marcoulides, Modern Methods for Business Research (p. 136). London: Lawrence Erlbaum Associates. Thornqvist, M.A. (1936). "The Bank of Finland's Consumption Price Index," Monthly Bulletin, Bank of Finland, 10: 27-34.

APPENDIX 4.1

Changing Definitions of NPL Recognition Date

Modification made

November. 1985

NPL period was set at 12 months. Accrued interest was required to be clawed back to day one of default. NPL period was tightened from 12 months to 6 months. The requirement on accrued interest was abolished. General provisions set at 1 percent of total outstanding loans, net of interest suspended and specific provision. Specific provision is set at 50 percent for doubtful debts and 100 percent for bad debts based on the value of the unsecured portion of the loan. NPL period was further tightened from 6 months to 3 months. The minimum requirements on general provision was increased from 1 percent to 1.5 percent of the total outstanding loans, net of interest suspended and specific provision for bad and doubtful debts. In addition, banking institutions were required to provide 20 percent specific provisions against uncollaterized portion of sub-standard loans. Banking institutions were also required to set aside provisions for off-balance sheet items. NPL classification was relaxed from 3 months to 6 months. However, financial institutions are required to report their NPL

1986

January 1998

September 1998

positions on a 3-month basis for monitoring purposes. Source: BNM Report—various issues

APPENDIX 4.2

NPL and Capital Adequacy Ratio (CAR) of Malaysian Banking System Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

NPL (RM million) 23,551 23,256 23,212 21,638 22,684 21,493 15,880 14,320 12,480 25,053 52,066 47,460 50,585

NPL/LOAN

CAF

(%) 30.1 24.8 20.0 15.4 14.5 12.3 7.8 5.5 3.7 4.1 11.8 6.6 6.5

Source: Bank Negara Malaysia Monthly Bulletin—various issues.

na na 9.8 9.9 10.9 11.4 10.9 10.9 10.7 10.5 11.8 12.5 12.8

APPENDIX 4.3 NPL Ratio of Malaysian Banking Institutions (Gross)

1998:05 1998:06 1998:07 1998:08 1998:09 1998:10 1998:11 1998:12

Banking system (%)

Commercial Bank

Finance Companies

Merchant Bank (%)

11.6 12.6 14.3 16.0 12.3 13.2 14.1 12.6

9.7 10.9 12.3 14.0 11.7 12.3 12.9 10.7

16.7 17.3 19.6 21.4 13.9 15.7 17.1 17.4

11.6 13.3 17.2 19.5 13.8 15.0 17.8 17.5

Source: BNM Statistical Monthly Bulletins—various issues

APPENDIX 4.4

Malaysian Regulatory Changes and Their Impacts on Malaysian Banks Period October 1997

January 1998

September 1998

Regulatory change Financing of hire-purchase loans reduced from 75% to 70% of purchase price of vehicle Repayment period shortened from 5.7 to not more than 5 vears Classification period for NPL tightened from 6 months to 3 months Minimum requirements on general provision increased from 1 percent to 1.5 percent of the total outstanding loans net (refer page 1) 20% requirement for specific provision against uncollaterized loan NPL classification relaxed from 3 months to 6 months Setting up of asset management company : Danaharta Setting up of capital injection vehicle: Danamodal

Source: BNM—A decade of change (2000): 202-204

Impact Reduced demand for hire purchase financing, curb loan growth of finance companies. Increased NPL Increased provisions for loan loss, lowered banks" profits.

Reduced NPL. eased liquidity and strengthened capital base of ailing banks.

APPENDIX 4.5

Non-Performing Loans by Sector of Banking Institutions Commercial Banks: Nonperforming 1997 1998 RM RM million %TL million % TL 7.220 12.8 Manufacturing 2.763 5.0 Broad property sector 5.052 5.0 13.216 12.2 2.595 22.1 Real estate 932 7.5 Construction 5.221 17.1 1,544 5.3 Purchase of residential 3.002 10.0 1.628 4.3 property Purchase of non4.4 948 residential property 2.398 10.4 4.130 121.3 Purchase of securities 2.005 8.2 Finance Companies: Nonperforming Loans by Sector 517 1.242 30.8 9.3 Manufacturing 3.347 10.8 Broad property7 sector 6.161 20.4 1.904 58.8 1.097 29.9 Real estate 934 Construction 1.722 22.6 11.3 Purchase of residential property 1.099 15.1 721 5.5 Purchase of nonresidential property 595 10.2 1.435 22.3 524 2.419 26.7 5.0 Purchase of securities Merchant Banks : Nonperforming Loans by Sector Manufacturing 161 5.8 555 121.0 3.4 1.454 20.4 Broad property sector 256 2.4 18.4 47 Real estate 297 Construction 208 4.2 1.110 22.6 Purchase of residential + property 1 0.5 1.1 Purchase of non+ residential property 9.4 46 + 5.8 974 Purchase of securities 251 24.6

Loans by Sector 1999 2000 RM RM million % TL million 7.171 13.0 7.562 13.463 12.3 13.699 2.107 18.2 2.496 5.210 19.4 5.347

%TL 13.4 12.9 21.0 21.1

3.694

3.343

6.9

2.451 11.5 2.039 i 13.3

2.513 2.329

12.1 15.4

1.032 4.979 550 1.989

33.5 18.1 32.2 28.5

768 5.911 587 2.060

26.6 28.1 38.4 33.6

1.212

15.5

1.712

21.1

1.228 1.730

22.5 24.8

1.551 1.345

29.4 22.8

476 1.372 418 852

20.3 24.1 25.5 23.6

403 1.057 471 518

17.8 22.0 33.3 16.8

26

24.8

0.4

0.5

76 1.059

22.4 36.5

68 907

30.2 27.9

9.5

APPENDIX 4.6

Asset Quality of Malaysian Banking Institutions (based on 6 months classification)

(RM '000)

Commercial banks Finance companies Merchant banks 3-month 6-month 3-month 6-month 3-month 6-month

Interest in suspense Specific provisions General provision Total provision NPL (npl) Net npl ratio (%) Total provisions/npl (%)

4201 13348 6555 24104 44896 9.7 53.7

Source: BNM 1998 Annual Report

13643 'l 1704 ;5693 21040 |32086 5.9 165.6

2640 3822 T824 8286 25122 21.8 33.0

2237 13551 Tfg24 7612 16320 12.2 46.6

624 1416 446 2486 7197 25.6 34.5

456 1189 1446 12091 3888 110.9 53.8

APPENDIX 4.7 Monthly NPL Ratio of Financial Institution.* Month May 1999 December 1999 December 2000

Commercial Bank (%) 6.5 5.7 5.4

* based on 6-month classification Source: BNM Bulletin, April 2001.

Finance Companies (%) 11.9 8.6 8.6

Merchant Bank (%) 12.8 12.3 12.3

Banking System (%) 7.9 6.6 6.3

APPENDIX 4.8

Profile of Malaysian Listed Institutions Based on 1999 Annual Reports Institution

Malayan Banking Berhad Hong Leong Bank Berhad Public Bank Berhad. Ban Hin Lee Bank Berhad Hock Hua Bank Berhad Pacific Bank Berhad Southern Bank Berhad RHB Sakura Merchant Bank Berhad Arab Malaysian Finance Berhad Public Finance Berhad

No. of ordinary shares issued (-000 units) 2.308,661 577.171 1.183.161 173.247 125,312 341.022 719.523 338.646

Nonperforming Loans (RM'OOO)

Shareholders' equity (RM'000)

Total Assets (RM'000)

7.896,764 1,564,801 3,307.823 621.350 744.981 849,518 1,898,311 601,953

87,591,952 15.094.683 30,898,341 7.369,522 5,431,293 9,632,270 10.210,943 4.275,516

5.368.299 990.452 678,374 391.633 467,544 1.011.174 724.831 191.596

452,646

611.046

15.149.034

2.614,507

330,000

956,752

11.524.025

399.067

APPENDIX 4.9 Summary of the Impact of NPLs on Big Banks versus Small Banks NPL/MKT.CAP RATIO Big Banks Small Banks NPL/EQUITY RATIO Big Banks Small Banks MKT.CAP/EQUITY RATIO Big Banks Small Banks

1997

1998

1999

0.14 0.16

0.44 1.03

0.28 0.64

0.20 0.53

0.42 0.76

0.51 1.06

1.84 2.79

1.00 0.86

1.90 1.36

CHAPTER 5

Assessing Corporate Financial Distress in Malaysia M. S. Zulkamain

and M. Shamsher

INTRODUCTION In a competitive business environment, market discipline ensures the survival of the fittest and corporate failures become a norm. Though corporate failures are perceived to be a problem of developing economic environments (Altman, Baidya, & Dias, 1979), firms operating in developed economies are no exception. It must be noted that despite the immense amount of research into this issue in developed economies, the problem of failure predictions is far from resolved. This is partially attributable to the nature of research findings from developed economies that cannot be generally applied to different economic environments such as emerging markets. The differences in market structure, provision, and implementation of law and accounting standards in these economies make it difficult to apply developed-economy prediction models in developing economies. The research issue of interest is not the incidence of corporate failures but the ability to predict impending failures using common identifiable attributes that can help formulate and implement preemptive measures to avoid such failures. This effort could avoid the financial distress to all stakeholders and reduce the costs of bankruptcy and contribute to the business and financial environment stability. The total cost of bankruptcy is substantial; Altman (1984b) showed that firms incur bankruptcy costs in the range of 11% to 17% of the firm value three years prior to bankruptcy in developed economies. It is expected that if failures can be prevented in developing economies, the cost savings could even be greater.

74

Regional Financial Markets

Predicting corporate failure is based on the premise that failure is a gradual process and a consequence of problems developed over many years, and also that the symptoms of the problems are identifiable and can be measured through some generally acceptable yardsticks. These common symptoms are observed and measured in the form of gradual decline over a period of time in profits, working capital, liquidity, asset quality, arrears interest, and loan repayment; delay in payment to suppliers, staffs, and all other creditors; and implementation of some form of austerity measures. The consistency of the symptoms enables researchers to formulate corporate prediction models in an effort to identify the potential failures and implement corrective measures to avoid a n d / o r minimize the cost of failures in cases where failure cannot be prevented. The numbers of corporate failures in Malaysia have increased over the last decade, though the failure rate is higher for small and unlisted firms. For example, the average number of firms liquidated in the 1990s was 1,000 firms a year. This comprised 97% of all failed firms, although only 3% of the firms that failed were listed firms, which supports the notion that the failure rate is higher for unlisted firms, which are usually smaller and less stable. Firms are considered failures or under financial distress when they apply to the court or relevant authorities for restructuring or a reorganization scheme based on a scheme of arrangement pursuant to section 176 of the Malaysian Companies Act of 1965. Firms under protection of this provision have to formulate survival options including proposals on strategies for corporate rescues and reconstruction. Highly leveraged corporations with severe financial problems might resort to outright liquidation. The objective of this study is to develop a model that discriminates between failed and non-failed firms based on accounting and market information attributes of firms. These attributes will provide important insights to policy makers and financial institutions to guide them in formulating effective preemptive measures to mitigate corporate failures. The chapter is divided into five sections. The chapter introduction explains the background to the corporate failures, the second section explains the documented findings on this issue, and the methodology section describes the research design. The fourth section discusses the discriminant model. The final section provides the implications of findings and conclusion. R E V I E W OF L I T E R A T U R E The earliest study on failure prediction by Beaver (1966) showed that corporate failure could be reliably predicted through the combined use of sophisticated quantitative techniques on selected financial ratios. Using a sample of 79 failed firms and non-failed firms and 30 financial ratios from

Assessing Corporate Financial Distress in Malaysia

75

five years prior to failure for the period 1954 to 1964, he found that the ratio of cash flow to total debt was significant in predicting failure. This ratio misclassified only 13 percent of the sample for one year before bankruptcy and 22 percent of the sample for five years before bankruptcy. These findings suggest that ratio analysis is useful for predicting failures at least five years before failure. Altman (1968) investigated a set of financial and economic ratios as possible determinants of corporate failures using multiple discriminant analysis. The study used 66 corporations from manufacturing industries comprised of bankrupt and non-bankrupt firms and 22 ratios from five categories—liquidity, profitability, leverage, solvency, and activity. Five ratios were finally selected for their performance in the prediction of corporate bankruptcy, and the derived model correctly classified 95 percent of the total sample (correctly classifying 94 percent as bankrupt firms and 97 percent as non-bankrupt firms) one year prior to bankruptcy. The percentage of accuracy declined with increasing number of years before bankruptcy. Altman, Marco, and Varetto (1994) reported the use of neural network in identification of distressed business by the Italian central bank. Using over 1,000 sampled firms with 10 financial ratios as independent variables, they found that the classification of neural networks was very close to that achieved by discriminant analysis. They concluded that the neural network is not a clearly dominant mathematical technique compared to traditional statistical techniques. Begley, Ming, and Watts (1995) incorporated the time "bias" factor into the classic business failure prediction model. Using Altman (1968) and Ohlson's (1980) models to a matched sample of failed and non-failed firms from the 1980s, they found that the predictive accuracy of Altman's model declined when applied against the 1980s data. The findings explained the importance of incorporating the time factor in the traditional failure prediction models. Mossman, Bell, Swartz, and Turtle (1998) conducted a study to compare four types of bankruptcy prediction models that are based on financial statement ratios, cash flows, stock returns, and returns standard deviations. They tested four bankruptcy models: Altaian's (1968) Z-score model based on financial ratios; Aziz, Emanuel, and Lawson's (1988) model comprised of cash flows; Clark and Weinstein's (1983) market return model; and Aharony, Jones, and Swary's (1980) market return variation model. They found that in the year prior to bankruptcy, the ratio model is the most effective in explaining the likelihood of bankruptcy. In the three years preceding bankruptcy, the cash flow model most consistently discriminates between bankrupt and non-bankrupt firms. The findings suggest different uses for the models, as stakeholders might be particularly interested in cash flow variables as "early warning" indicators of failure.

76

Regional Financial Markets

Alternatively, a large negative shift in accounting ratio variables could be a useful indicator of imminent financial collapse. METHODOLOGY The data used in this study is a set of financial ratios derived from financial statements of sampled failed listed companies for the period from 1980 to 1996. The paired sample design technique was employed, where each failed firm has a non-failed "match" in the sample for the same period. The use of a matched sample of failed and non-failed firms (one-to-one match) might introduce a potential firm failure bias (Palepu, 1986). It is claimed that the potential for failure is overstated using this technique. However, it is stressed that the bias may or may not be important depending on the usage of the model. If the model is used to rank the firms for the potential failure in order to perform a more detailed analysis, then the bias is not important. However, if the model is used to identify investment portfolio selection, then the bias is significant. Furthermore, Zmijewski (1984) reviewed 17 financial distressed studies that used this controversial method and found that although a choice-based sample bias was present, the results do not indicate significant changes in overall classification and classification rates. Finally, Platt & Platt (1990, 1991) urged that one-to-one sampling technique is still an acceptable method in failure prediction studies. Listed firms that qualify in any one or a combination of situations mentioned below are defined as failed firms. The definitions of "failure" are the firms protected under section 176 of the Companies Act of 1965 for the period from 1980 to 1996; the firms approved to undertake a restructuring scheme to revive their financial conditions by the Kuala Lumpur Stock Exchange (KLSE), Securities Commission (SC), or relevant authorities; and the firms that were put under receivership. Thirty-three failed firms were sampled and matched with 33 non-failed firms that were sampled based on the following criteria using the KLSE annual reports: 1. Same industry as the failed firm 2. Closest asset size 3. Similar age since incorporation The above criteria were set as control factors to ensure a minimum bias in selection of the control sample used in the development of the failure prediction model. Furthermore, the use of the one-to-one matched procedure is consistent with previous studies documented in Beaver (1966), Altman (1968), Blum (1974), and others.

Assessing Corporate Financial Distress in Malaysia

77

The failed firms were identified from information contained in the KLSE daily diary, and financial information of failed firms was sourced from annual reports published for five years prior to failure. The financial information of the non-failed firms was sourced from annual reports for the same fiscal years as those of their failed match sample. The sampled companies are from six different industries: 23 companies from the industrial sector, 6 companies from the property sector, and 1 company each from the consumer, finance, hotel, and mining sectors. Since the number of failed listed companies in different sectors is rather small, the analysis is done on the total sample of mixed industry sectors.1 The dependent variable is a dichotomous variable of failed or non-failed firms. The independent variables are the 65 selected financial ratios used by Beaver (1966), Altman (1968), and Ou and Penmen (1989). E S T I M A T I O N OF D I S C R I M I N A N T MODEL Using a sample of failed and non-failed firms as the dependent variable and the ratios as the independent variables, a forward stepwise multivariate discriminant analysis was used to determine the discriminating power of the variables. In stepwise estimation, independent variables were entered into the discriminant function one at a time on the basis of their discriminating power. The normal variables were entered into the discriminant analysis. 2 Three groups of potential variables were tested. 3 Finally, one group that gave the highest hit ratio was selected. The Mahalanobis D 2 measure was used to choose the variable that generated the greatest separation for the pair of groups, which are closest at particular step. The following prediction model was derived:

where Z = Overall Index XI = Log V12 (Total Liabilities Percent) X2 = Square Root V29 (Asset Turnover) X3 = V32 (Inventory Percent) X4 = Log V51 (Sales to Inventory) X5 = Log V65 (Market Value to Debts) X6 = Log V13 (Cash Percent) The failed group centroid (DV = 1) was found at 1.156421 and the nonfailed group (DV = 0) centroid at -1.156421. The overall mean or called cutting score is equal to zero.4

78

Regional Financial Markets

Assessing Overall Fit To determine the predictive accuracy of the discriminant function, classification matrices were developed for both the analysis sample and the holdout sample. In the previous section, the cutting score was calculated to be 0.00. The procedure for classifying firms is as follows: 1. Classify a firm as a failed firm if its discriminant score is positive. 2. Classify a firm as a non-failed firm if its discriminant score is negative. Using these criteria, the classification result of the analysis sample is shown in Table 5.1. The holdout sample accuracy rate at 91 percent (average of correct classification of DV = 0 at 94 percent and DV = 1 at 88 percent) is higher than 86.2 percent (average of correct classification of DV = 0 at 84.4 percent and DV = 1 at 88.1 percent) accuracy of the analysis sample. Further, the 91 percent classification accuracy is high compared to the priori chance of classifying individual firms correctly without the discriminant function.5 The 86.2 percent accuracy for the analysis sample is substantially higher than the proportional chance criterion of 50 percent. The final measure of classification accuracy was made using the Press Q statistic.6 The Press Q statistic is a statistical test for the discriminatory power of the classification matrix when compared with the chance model. The Press Q statistic of the analysis sample was 114.5 and 67.24 for the holdout sample. The critical value (the chi-square value for 1 degree of freedom at the 1% confidence level) is 6.63, implying that the predictions were significantly better than chance, as the Press Q statistic was higher than the critical value. Table 5.1 Classification Results

Original Sample CrossValidated Holdout Sample

% % %

DV 0 1 0 1 0 1

Predicted Group Membership 0 1 84.4 15.6 11.9 88.1 84.4 15.6 11.9 88.1 94 6 12 88

Total

100 100 100 100 100 100

Note: DV denotes dependent variable, DV = 0 is non-failed group, and DV = 1 is failed group

Assessing Corporate Financial Distress in Malaysia

79

Interpretation of the Results This section examines the relative importance of each independent variable in the discriminant function in discriminating between the failed and non-failed groups. Three methods were employed here: (1) standardized discriminant weights, (2) discriminant loadings (structure correlations), and (3) partial F-values. Table 5.2 shows the six significant variables, which were screened by the stepwise procedure, and provides the ranking of each variable in the function and among all variables tested. The discriminant loadings and univariate F-values showed the ranking of discriminating power of these variables. Among the six variables, V12 (Total Liabilities Percent) discriminate the most with the highest discriminant loading and F-statistics, and VI3 (Cash Percent) discriminate the least. Validation of the Discriminant Results This section examines the validity of the discriminant function using the holdout sample and U method (leave-one-out). The results for the holdout sample were tabulated in Table 1. The holdout sample used in this study came from the original sample. The model was verified when it correctly classified up to 91 percent (average of correct classification of DV = 0 at 94 percent and DV = 1 at 88 percent) in holdout validation. In the U validation method, the classification accuracy is 86.2 percent (average of correct classification of DV = 0 at 84.4 percent and DV = 1 at 88.1 percent). These results suggest that the model derived has significant internal validity and a potential for application in identifying potential failures. External Validation The predictive capability of the model developed was tested for external validity on a new sample of 33 failed firms in the year 1998. The model showed consistently accurate classification accuracy up to four years before failure with 59 percent accuracy rate (Table 5.3). However, if we take into consideration the overlapping area (gray area7), the model is considered accurate in all the years tested with the highest misclassification rate of only 17 percent for 5 years before a firm failed. Tests of Relationship Between the Final Variables in the Model, and Profit and Cash Flow Variables In the model development process, prime variables such as assets, liabilities, equity, sales, and expenses were considered for inclusion in the

80

Regional Financial Markets

Table 5.2 Summary of Interpretative Measures Variable

Standard Weights Value

VII VI2 VI3 V15 V29 V32 V5I V53 V65

NI 0.9501 0.243088 NI -0.642206 0.621284 0.478912 NI -0.314269

Discriminant Loading Value 0.208358 0.780804 -0.131900 0.047853 -0.196509 0.283862 -0.295212 -0.135180 -0.410612

Rank 5 1 8 9 6 4 3 7 2

Univariate F Ratio Value NI 177.7349 47.46414 NI 109.4794 78.45759 63.16183 NI 54.45438

Rank NI 1 6 NI 2 3 4 NI 5

Note: NI denotes not included.

Table 5.3 External Validation Procedures Year Before Failure 0 1 2 3 4 5 Avg

Correct Classification (°o) 100 97 79 67 59 48 75

Grey Area

(%) 0 3 18 21 25 35 17

Note: Avg denotes average.

model. Whereas variables that have potential negative values like profit, cash flow, and working capital were omitted due to the requirement of the transformation techniques (to normalize the variables) that use variables with positive values only. In order to provide fair treatment to all classes of financial ratios and to avoid misconceptions among users of accounting information who often use cash flow, profit, working capital, and net worth variables in their analysis, an additional analysis was performed to investigate whether the final variables selected in the prediction model have any relationship with the above mentioned variables (i.e., cash flows, profits, working capital, and net worth). A Pearson correlation test was employed to investigate this relationship, and the results are summarized in Table 5.4. The findings show a strong relationship among the final variables and cash flow, profit, working capital, and net worth variables. For example, total liabilities percent (V12) has a significant relationship with cash flow to sales (V01), cash flow to assets (V02), cash flow to total debt (V04), return on sales (1/05), return on assets (V07), working capital percent

Assessing Corporate Financial Distress in Malaysia

+

Table 5.4 Pearson Correlation Test

p r 0 f i t St C a s h F I 0 w y a r I a b 1 e s

vm

V02 V03 V04

vos vm

V07 V09 V16 V26 V27 V2S V37

V38 V47 V48 V49 VSO VS6 VS7 VS9 V60 V61 V62 V63 V64

Final Variable In Prediction Model V12 V29 V32 V13 0.097 0.122 -0.043 -&5S5 0J4? -0.031 ftj*7 + 4t7«« • -0.010 0.002 0.010 -0.056 §351 -0.034 •4.14* -0.038 S5( + 0.125 0.099 -0.079 0,056 -0.078 0.029 r -0.026 0.171 &2ii : "*m$ -0.132 -0.035 %sn . 0.124 -0.095 -4UOT • -0.043 0.116 -0.049 Q217 0.104 0.096 0.087 0.065 0.087 0.117 -0.058 SMM+ 0.361 0.17! -0.111 0.086 -0.032 0.032 0.010 -0.051 0.122 0.097 -0.043 -C$85 -0.088 -0.111 0.104 0.006 0.125 0.099 -0.038 ^565 -0.079 0.056 0.029 -0.078 0.191 0.130 0.041 0.03 J 0.039 -0.031 -0.051 0.134 0.085 -0.103 0.088 -0.002 -0.021 0.010 0.021 0.079 -0.056 0.123 0.090 Q.131 -0.029 0.028 0.060 0.004 0.004 0.001 -0.025 0.107 0.086 0.086 S£9&+

mm

mu+

*&m

mm

V5I 0.048 0.060 0.008 0.073 0.049 -0.001 0.059 0,062 0.053 -0.036 0.001 0.045 0.058 0.009 0.048 -0.042 0.049 -0.001 0.034 0.042 0.035 0.005 0.045 0.038 0.004 0.036

V6S 0.040 -0.012 0.006 0.151 0.040 0.007 -0.009 A.U2 0.039 0.045 -0.018 0.038 0.099 0.013 0.040 -0.007 0.040 0,007 -0.044 0.014 -0.004 0.015 0.078 -0.008 0.003 0.081

Note: The shaded cell showed the correlation is significant at the 1% level

(V16), net worth to sales (V28), profit before depreciation to sales (V47), pre-tax income to sales (V49), operating income to asset (V56), and earning power (V64). In other words, V12 is the image and considered representative of the variables mentioned above. The other variables in the final model also have a significant relationship with the cash flow, profit, working capital, and net worth variables. This significant correlation suggests that the variables selected in the final models are a good proxy for the commonly used variables by investors to evaluate the financial health of firms. CONCLUSION The failure model was tested for is predictive accuracy and subjected to both the internal and external validity tests. The model showed exceptional performance with high correct classification accuracy rate (more

Regional Financial Markets

82

than 80 percent) both in internal and external validity. The model showed great potential for application to predict firm failures. The finding suggests that the model is reliable and has good practical use for making decisions in real markets as the validation tests showed better than chance prediction. The percentage of correct classification of failed and non-failed firms was at 59% as far as four years before failure. In addition, 6 ratios were found significant out of 65 financial ratios used in this study to discriminate between failed and non-failed firms. The significant variables are ranked in descending order of their discriminating power: 1. 2. 3. 4. 5. 6.

Total Liabilities Percent (V12) Asset Turnover (V29) Inventory Percent (V32) Sales to Inventory (V51) Market Value to Debts (V65) Cash Percent (VI3)

Finally, a strong relationship was documented between the final variables in the prediction model and the most commonly used variables by investors to evaluate financial health of firms, namely, cash flow, profits, working capital, and net worth. The significant variables mentioned above could guide the policy makers or users of the model to develop an early warning system to either evade or mitigate impending failures. For example, bankers or creditors could use these models to assess the potential borrower's or credit applicant's credit risk and continuously assess the borrower's financial condition in making decisions to renew or extend the loan granted. Managers could use these models in their financial planning. If failure can be predicted three to four years earlier, management could take remedial action such as merger exercise or restructuring to avoid the potential bankruptcy costs. Regulating agencies such as the Securities Commission and Kuala Eumpur Stock Exchange might want to assess a firm's going-concern status, solvency, and compliance of certain important requirements. This is important to preempt any systematic financial failures and bailouts using public funds. Auditors could use these models to help them formulate a fair opinion of the overall financial condition and the status of the firm as a going concern. Unfortunately, in this study more than 60% of the failed firms had unqualified auditor reports even one year before the actual failure. Investors could use these models to give them early warning signals on financial conditions of firms in their portfolio to help them make better management decisions.

Assessing Corporate Financial Distress in Malaysia

83

NOTES 1. An analysis of the industrial sector firm's data showed similar results with that of the mixed industry sector. 2. Before the discriminant analysis, a normality test was carried out to all independent variables. Only one variable was normal, with nine variables lognormal and three variables square root normal. However, the variables that have the potential to be negative values were excluded from the analysis due to problems in transforming the data. 3. A correlation test was done, and it was found that a number of variables highly correlated each other. Instead of dropping/deleting the highly correlated variables, three potential groups were constructed that exclusively include the highly correlated variables. 4. The overall mean was calculated as follows: Z = (NOZO + N1Z1)/(N0 + NI), where Z is a critical cutting score, NO is the number of observations in the non-failed group, ZO is the centroid for the non-failed group, NI is the number of observations in the failed group, and Zl is the centroid for the failed group. 5. The proportional chance criterion is calculated as follows: C = p2 + (1 2 p) , where C is the proportional chance criterion, p is the proportion of firms in the failed group and 1 - p is the proportion of firms in the non-failed group. Substituting the appropriate values, C = 0.50. 6. The Press Q statistic is calculated by the following formula: Press Q = [N — (nK)]2/[N(K - 1)], where N is total sample size, n is number of observations correctly classified, and K is number of groups. 7. Gray area is the area between the two centroids of the failed and non-failed groups. This area is where the failed firm characteristics overlap with the nonfailed firm characteristics. For a market-based model, the gray area is between -1.156 and 1.156, and for a nonmarket-based model it is between-1.106 and 1.106. Practically, this area is where the successful firms deteriorate and finally succumb to failure, and for failed firms, this area is where they show their improvement and finally overcome their problem and become successful firms. Therefore, users of the derived model should note the existence of this area and qualify their findings of its existence. REFERENCE++ Altman, E. I. (1968). "Financial Ratios, Discriminant Analysis and The Prediction of Corporate Failure/' Journal of Finance, 22 (September), 589-609. . (1984a). "Introduction: Company and Country Risk Models," Journal of Banking and Finance, 8 (2): 151-152. . (1984b). "The Success of Business Failure Prediction Models: An International Survey." Journal of Banking and Finance, 8 (2): 171-198. Altman, E. I., T. K. N. Baidya, and L. M. R. Dias. (1979). "Assessing Potential Financial Problems for Firms in Brazil." Journal of International Business Studies, 10 (2), 9-24. Altman, E. I., G. Marco, and F Varetto. (1994). "Corporate Distress Diagnosis: Comparisons Using Linear Discriminant Analysis and Neural Networks (The Italian Experience)," Journal of Banking and Finance, 18 (May), 505-529.

84

Regional Financial Markets

Beaver, W. H. (1966). "Financial Ratios as Predictors of Failure," Journal of Accounting Research, 4 (Supplement), 71-111. Begley, J., J. Ming, and S. Watts. (1995). "Bankruptcy Classification Errors in the 1980s: An Empirical Analysis of Altman and Ohlson's Models." Unpublished manuscript. University of British Columbia. Blum, M. (1974). "Failing Company Discriminant Analysis." Journal of Accounting Research, 12 (Spring), 1-25. Mossman, C. E., G. G. Bell, M. Swartz, and H. Turtle. (1998). "An Empirical Comparison of Bankruptcy Models," Financial Review, 33 (2), 35-55. Ou, }. A., and S. H. Penman. (1989). "Financial Statement Analysis and the Prediction of Stock Returns." Journal of Accounting and Economics, 11, 295-329. Palepu, K. G. (1986). "Predicting Takeover Targets: A Methodological and Empirical Analysis." Journal of Accounting and Economics, 8, 3-35. Platt, H. D., and M. B. Platt. (1990). "Development of a Class of Stable Predictive Variables: The Case of Bankruptcy Prediction," Journal of Business Finance and Accounting, 17 (1), 31-51. . (1991). "A Note on the Use of Industry-Relative Ratios in Bankruptcy Prediction," Journal of Banking and Finance, 15 (December), 1183-1194. Zmijewski, M. E. (1984). "Methodological Issues Related to the Estimation of Financial Distress Prediction Models." Journal of Accounting Research, 22 (Supplement), 59-82.

APPENDIX 5.1

List of Ratios Examined Code VOl V02 V03 V04 V05 V06 V07 V08 V09

Ratio Name Cash flow to Sales Cash flow to Assets Cash flow to Net Worth Cash flow to Total Debt Return on Sales (ROS) Percentage Change in ROS Return on Assets Return on Equity Net Income to Total Debt

Code V34 V35 V36 V37 V38 V39 V40 V41 V42

vio

V43 V44

Vl2 V13

Current Liabilities to Total Assets Long Term Liabilities To Total Assets Total Liabilities To Total Assets Cash To Total Assets

V14 V15

Quick Assets To Total Assets Current Assets To Total Assets

V47 V48

V16 V17

Working Capital To Total Assets Cash to Current Liabilities

V49 V50

V18 V19

Quick Ratio Percentage Change in Quick Ratio

V51 V52

V20 V21 V22

Current Ratio Percentage Change in Current Ratio Cash Turnover

V53 V54 V55

V23 V24

Receivable Turnover Quick Asset Turnover

V56 V57

V25 V26 V27

Current Asset Turnover Working Capital Turnover Percentage Change in Sales to Working Capital Net Worth to Sales Asset Turnover Percentage Change in Sales to Total Assets Days Sales in Receivable Inventory To Total Assets Inventory Growth

V58 V59 V60

Ratio Name Sales Growth Depreciation Growth Dividend Growth Return on Opening Equity (ROOE) Percentage Change in ROOE Equity to Debt Percentage Change in Equity to Debt Equity to Long Term Debt Percentage Change in Equity to Long Term Debt Equity to Fixed Assets Percentage Change in Equity to Fixed Assets Times Interest Earned Percentage Change in Times Interest Earned Profit Before Depreciation to Sales Percentage Change in Profit Before Depreciation to Sales Pretax Income to Sales Percentage Change in Pretax Income to Sales Sales To Inventory Percentage Change in Sales to Inventory Sales to Fixed Assets Percentage Change in Total Assets Percentage Change in Working Capital to Total Assets Operating Income to Assets Percentage Change in Operating Income to Asset Percentage Change in Long Term Debt Dividends to Cash Flows Net Income to Cash Flow

V61 V62 V63

Operating Profit to Sales Return on Owners Equity Total Assets to Net Worth

V64 V65

Earning Power Market Value To Debts

Vll

V28 V29 V30 V31 V32 V33

V45 V46

APPENDIX 5.2

Descriptive Statistics for the Sample Firms Minimum

Variable

Maximum

Mean

Standard deviation

OWNSHIP

0

0.7813

0.4185

0.2468

DIVIDEND

0

0.1306

0.0384

0.0244

LEVERAGE

0

0.8925

0.3288

0.2170

FIRMSIZE

37.746

41.438.739

3.533.577

7,482,772

STDRET

0.0619

0.7467

0.1759

0.0875

GROWTH

-0.2651

60.3511

1.3052

6.031

PROFIT

-0.0636

0.5285

0.0832

0.0778

LIQUIDS

0.002533

0.7697

0.1552

0.1383

RESERVES

-3.6669

1.2295

0.1573

0.5292

DIVERSE

-0.2036

0.4395

0.1546

0.0908

EARNVOL

0.0105

0.2604

0.0691

0.0459

DEBTCAP

0.0000951

0.9265

0.269

0.2178

INVEST

0

0.9525

0.3099

0.2355

FIRMSIZE is in thousands of Hong Kong dollars.

APPENDIX 5.3

Correlation Matrix—Sample Firms OWNSHIP

LEVERAGE

DIVIDEND

FIRMSIZE

OWNSHIP

1

LEVERAGE

0.265

1

DIVIDEND

0.110

-0.169

1

FIRMSIZE

-0.279

-0.024

-0.146

1

STDRET

0.086

0.398

-0.241

-0.129

STDRET

RESERVES

GROWTH

PROFIT

LIQUIDS

DIVERSE;

EARNVOL

DEBK

+++

1

RESERVES

0.110

-0.192

0.358

0.046

-0.329

1

GROWTH

0.135

0.259

-0.005

-0.042

0.137

-0.003

1 1

PROFIT

0.002

-0.462

0.370

-0.048

-0.405

0.386

-0.052

LIQUIDS

0.049

-0.371

0.397

-0.144

-0.187

0.108

-0.038

0.291

1

DIVERSE

0.029

-0.154

0.248

-0.087

-0.057

0.037

-0.112

-0.008

0.062

1

EARNVOL

0.216

-0.021

0.033

-0.258

-0.026

-0.301

-0.014

0.030

0.167

0.103

1

DEBTCAP

-0.207

-0.141

-0.114

-0.090

-0.026

-0.171

-0.175

0.056

-0.245

0.122

-0.042

1

INVEST

0.092

0.339

-0.129

0.165

0.163

-0.002

0.167

-0.180

-0.116

-0.005

-0.150

+++++++ -0.49

The Pearson correlation coefficient is shown in the cells.

APPENDIX 5.4

Results for the Managerial Ownership Equation (dependent variable OWNSHIP) Expected Sign

Parameter estimate

-

-0.258

+ve

-0.0262

a

-ve

-9.25xl0" 10

DIVIDEND

a3

+ve/-ve

+9.593

LEVERAGE

a4

+ve/-ve

+0.958

Variable

Coefficient

Intercept

a

DIVERSE

a

FIRMSIZE

0 \ 2

t-statistic (Prob>|t|) -1.864* (0.0652) -0.198 (0.844) -0.457 (0.648) 4 4]g*** (0.0001) 4.318*** (0.0001)

System weighted R-square: 0.4691 The f-statistic is for the null hypothesis that the parameter estimate is zero. Prob> I f I is the probability of observing a f-value that is greater than the absolute value of t under the null hypothesis. Significance: 1% ***, 10% *

APPENDIX 5.5 Results for the Dividend Equation (dependent variable DIVIDEND) Parameter estimate

Variable

Coefficient

Expected sign

Intercept

b0

++

FIRMSIZE

b\

+ve

+1.78xl0" 12

STDRET

b2

++

-0.00034

GROWTH

h

++

+0.0000837

LIQUIDS

b4

+ve

+0.00228

RESERVES

b5

+ve

+0.00021

PROFIT

b6

+ve

+0.00066

OWNSHIP

bl

+ve/-ve

+0.0943

LEVERAGE

b*

+ve/-ve

-0.0979

+0.031

t-statistic (Prob>|t|) 2.404** (0.018) 0.007 (0.995) -0.022 (0.982) 0.259 (0.796) 0.203 (0.839) 0.062 (0.95) 0.028 (0.978) 3 722*** (0.0003) -2.866*** (0.0051)

The f-statistic is for the null hypothesis that the parameter estimate is zero. Prob> 111 is the probability of observing a f-value that is greater than the absolute value of f under the null hypothesis. Significance: 1% ***; 5% **

APPENDIX 5.6 Results for the Leverage Equation (dependent variable LEVERAGE) Variable

Coefficient

Expected sign

Parameter estimate

t-statistic (Prob>|t|) 2 759***

+

+0.307

\

+ve

-0.00662

(0.0069) -0.136 (0.892)

EARNVOL

c'2

++

+0.0393

0.119 (0.906)

GROWTH

c

+++

+0.00057

0.27 (0.788)

INVEST

c

+++

+0.00393

0.097 (0.923)

OWNSHIP

c

+ve/-ve

+0.9585

3.384*** (0.001)

DIVIDEND

c

++++

-9.95

-5 895*** (0.0001)

Intercept

c

DEBTCAP

c

0

3 4 5 6

The f-statistic is for the null hypothesis that the parameter estimate is zero. P r o b > I f I is the probability of observing a f-value that is greater than the absolute value of t u n d e r the null hypothesis. Significance: 1% ***

APPENDIX 5.7 Result of Independent Equation for Managerial Shareholding Dependent Variable—OWNSHIP Variable

Expected sign of coefficient

Parameter estimate

t-statistics

Prob>|t|

intercept

-

0.4488

9.426***

0.0001

DIVERSE

+ve

0.0140

0.055

0.9561

FIRMSIZE

-ve

-9.2x10"9

-2.975***

0.0036

The f-statistic is for the null hypothesis that the parameter estimate is zero. Prob> I f I is the probability of observing a f-value that is greater than the absolute value of f under the null hypothesis. F-value for model: 4.476; Prob>F: 0.0136; Adjusted R-square: 0.0605. Significance: 1% ***

APPENDIX 5.8

Result of Independent Regression of Dividend Equation Dependent Variable—DIVIDEND Variable

Expected sign of coefficient

Parameter estimate

t-statistics

Prob>|t|

Intercept

-

0.0283

4.502***

0.0001

FIRMSIZE

+ve

-3.7xl0- 1 0

-1.331

0.1861

STDRET

-ve

-0.0156

-0.587

0.5586

GROWTH

-ve

0.000074

0.218

0.8277

PROFIT

+ve

0.0500

1.629

0.1064

LIQUIDS

+ve

0.0524

3.373***

0.0011

RESERVES

+ve

0.0116

2 722***

0.0076

The f-statistic is for the null hypothesis that the parameter estimate is zero. Prob> U I is the probability of observing a f-value that is greater than the absolute value of t under the null hypothesis. F-value for model: 7.167; Prob>F: 0.0001; Adjusted R-square: 0.2552. Significance: 1% ***

APPENDIX 5.9 Results of Independent Regression of Leverage Equation Dependent Variable—LEVERAGE Variable

Expected sign of coefficient

Parameter estimate

t-statistics

Prob>|t|

Intercept

-

0.1920

2.896***

0.0046

EARNVOL

-ve

0.1691

0.39

0.6974

DEBTCAP

+ve

0.0656

0.629

0.5307

GROWTH

+ve

0.0072

2.34**

0.0212

INVEST

-ve

0.3141

3.228***

0.0017

The f-statistic is for the null hypothesis that the parameter estimate is zero. Prob> I f I is the probability of observing a f-value that is greater than the absolute value of f under the null hypothesis. F-value for model: 4.992; Prob>F: 0.001; Adjusted R-square: 0.1288. Significance: 1% ***; 5% **

This page intentionally left blank

CHAPTER 6

Cross-Sectional Predictability of Stock Returns at the Colombo Stock Exchange Ely as Elyasiani, Priyal Perera, and Tribhuvan N. Puri

INTRODUCTION Early equilibrium asset pricing models predict that portfolios held by investors are efficient in the mean-variance sense. This implies that (i) expected security returns are positive linear functions of the corresponding betas (/?, a measure of systematic risk) and (ii) betas are sufficient for describing the cross section of expected returns. However, recent empirical evidence from the United States and Japan, which constitute the two largest markets in the world, suggests that some fundamental variables such as size, price-earnings ratio (P/E), cash-flow yield, and book-to-market value of firms are better capable of explaining stock returns than the risk measure (3.1 Fama and French (1992) and Reinganaum (1988) report strong significance of size and book-to-market ratio in predicting returns of U.S. securities. Fama and French show that after adjusting for size and bookto-market factors, the systematic risk /? has no explanatory power. This finding stands in sharp contrast to the capital asset pricing model and constitutes an important anomaly to the notion of rational markets. 2 Size, P/E, and book-to-market effects are among the most puzzling phenomena in finance. Researchers have provided several interpretations of these effects. For instance, Fama (1991) argues that the anomalous relationship between these variables and stock returns can be explained within the framework of a rational expectation asset pricing model, in which the fundamental variables proxy some omitted risk factors. Alternatively, Fama (1991) and Lakonishok, Shleifer, and Vishney (1995) indicate that the anomalous relationship can be a reflection of inefficiency in

96

Regional Financial Markets

the market. Fama and French (1993) argue that the size and book-to-market ratio can be treated as proxies for systematic sources of risk. They use an APT model to show that stocks with higher factor loading (coefficients) on the size factor have higher average returns. They interpret these higher returns as reflecting the risk premium associated with the factor.3 Lakonishok, et al. (1995) alternatively argue that the observed anomalous behavior between fundamental variables and stock returns is an evidence of market inefficiency. According to them, stock analysts have a tendency to overprice the recent // winners ,/ and underprice the "losers." This tendency creates systematic errors in the forecasts of returns and renders the markets inefficient. A recent study by La Porta (1996) confirms this behavior. He finds that stocks for which analysts forecast low earning growth rates demonstrate a superior performance to those with forecasted high earnings growth rates. Daniel and Titman (1995) test whether the size and book-to-market ratio are the risk factors whose risk premia contribute to the anomalous returns. They find that the factor loadings for these factors do not provide any new information in predicting returns. This result is inconsistent with the position of Fama et al. (1993) mentioned earlier. Yet another interpretation put forward in the literature is that the anomalous behavior may arise from "data snooping." 4 According to this view, if empirical tests carried out to measure the impact of a set of variables on stock returns are repeated, the t-statistics for these variables will be overstated. The present study investigates whether the above anomalous behavior manifests in Sri Lanka, an emerging capital market. Specifically, this study examines whether it is possible to predict the cross-sectional returns of the Sri Lankan stocks with the help of some fundamental variables such as those used by Fama et al. (1992) to describe the U.S. stock market behavior. We use the P/E ratio and size (measured by market capitalization (MCAP)) as the fundamental variables and investigate the role of these variables in determining the cross sectional returns, once the market risk (P) is accounted for.5 Our approach employs three alternative procedures—univariate analysis, estimation of a capital asset pricing model extended to include size, and the P/E ratio as additional explanatory variables—and the Fama-MacBeth cross-sectional procedure. It is noteworthy that, although the Sri Lankan capital market is smaller than most other emerging markets of the region, it has generated significant interest among international and regional investors due to the high growth in the Sri Lankan GDP and international trade, as well as the passage of the 1989 legislation liberalizing foreign investment in this country 6 Moreover, Duff & Phelps Credit Rating Lanka (CDRL), a joint venture between the U.S.based Duff & Phelps, the World Bank's International Finance Corporation, and Sri Lanka, commenced operation in Sri Lanka in October 1999 with the purpose of rating the riskiness of the firms in this market. Market

Cross-Sectional Predictability of Stock Returns at the Colombo Stock Exchange

97

analysts suggest that this risk classification function is crucial for future development and fund-raising capability of the Sri Lankan corporations as it reduces information and transaction costs to investors and international organizations such as the World Bank and the International Monetary Fund. Among other significant developments is the upcoming liberalization of the Sri Lankan insurance sector and the mandatory requirement that foreign investment companies be listed on this country's stock exchange. 7 The finding on the predictability of stock returns in the emerging Sri Lankan market will provide indication of profit opportunities for investors and can also prompt the Sri Lankan regulators to take steps to improve the market design and efficiency These steps may include, for example, increased disclosure requirement and introduction of more advanced communication technologies to be used by the market participants. The plan of the chapter is as follows. The section on data description presents the sources and construction of the data, the section on measurement of the effect of fundamental variables describes the methodology and the econometric models, and the next section presents the empirical results and their implications. The final section concludes our study. DATA D E S C R I P T I O N The sample for the present study includes daily data on 40 firms listed on the Colombo Stock Exchange (CSE). Price data were acquired from the CSE. Income statements and balance sheet data of these companies were also obtained from various issues of annual reports filed by the companies with the CSE. Two criteria are used in selection of the firms to be included in the sample. First, only those firms whose most recent financial statements are available are considered. Second, out of these firms, only the ones that had more than 1,000 trades during the period January 1, 1990 to June 30, 1996 are selected. The purpose of the latter restriction is to select only firms whose stocks have been actively traded. This approach can be justified on the grounds that inclusion of stocks with infrequent trading may introduce two problems. First, it may result in serial correlation in the regression errors. Second, since actual trades occur over distant periods, prices may not impound all the relevant information between consecutive trades. It is noteworthy that selection of firms based on the availability of the financial statements and with the most trading frequency may introduce a bias of its own. This is because the most frequently traded firms are also likely to be the larger and healthier ones. However, given the constraints on data availability, the sample used is the best feasible sample. An alternative approach is to keep all the firms with complete data and

Regional Financial Markets

98

make the necessary adjustments for infrequent trading in order to correct for the bias. Several techniques have been suggested for this purpose in the literature. Unfortunately, however, it is doubtful whether these procedures do actually mitigate the problem. In particular, Bartholdy and Riding (1994), among others, have shown that the bias correction techniques do not provide incremental benefits over standard OLS estimation. In brief, based on their empirical results, these authors conclude that "from the perspectives of bias, efficiency, and consistency, none of the most commonly used correction procedures is superior to OLS estimation." The sample used here represents 20 percent of all the firms listed on the CSE. No attempt is made to focus on firms in some specific industrial sector because of the paucity of the data. 8 The monthly returns on the individual stocks and the All Share Price Index, which includes all of the companies listed on the CSE, are based on the end-of-the-month market prices. Continuously compounded returns are calculated by taking the difference of the log of prices for successive months. 9 Total dividends paid over a year were divided into 12 equal amounts and added to the end-of-the-month price in calculating the monthly returns. Returns are then adjusted for dividends, rights, and bonuses. Excess returns are obtained by subtracting the monthly returns on the three-month Sri Lankan Treasury bill from the monthly asset returns. The size variable (MCAP) is adjusted whenever shares are issued. A major point is that the question of survivorship bias is not a concern in this study because there are almost no firms that have been excluded from the list over the period under consideration. M E A S U R E M E N T OF THE EFFECT OF F U N D A M E N T A L VARIABLES The Effect of Beta, Size, and P/E We employ three techniques to analyze the effect of market risk (/?), size (MCAP), and the price-earning ratio (P/E) on the stock returns of the Sri Lankan market: univariate analysis, regression analysis of an extended capital asset pricing model, and the Fama-MacBeth procedure. In the univariate analysis the 40 stocks in the sample are ranked in ascending order of the market risk (/?) or of a specific fundamental variable (P/E, MCAP), and assigned membership into one of four portfolios according to their rankings. Each portfolio consists of 10 stocks. The mean return, /?, P/E ratio, and size of the largest-ranked and the lowest-ranked portfolios are then tested for equality. The second procedure involves estimation of an extended CAPM with market index, size (MCAP), and P/E as regressors. In this procedure, a system of 40 equations is estimated for the 40 firms in the sample, using

Cross-Sectional Predictability of Stock Returns at the Colombo Stock Exchange 99 the Seemingly Unrelated Regression (SUR) technique. This framework allows us to analyze the marginal impact of the P/E ratio and MCAP on stock returns, given that the market factor is already accounted for. The SUR technique takes into account the contemporaneous cross-sectional correlation in the residual returns across firms and allows statistical tests of parameter values across the 40 equations. 10 The SUR model is also used by Jaffe, Keim, and Westerfield (1989) to investigate the effect of earnings yields and market values of firms on the U.S. stock returns, and by Chan, Hamao, and Lakonishok (1991) to explain the variations in average crosssectional returns on the Japanese stocks by the earnings-price ratio, size, book-to-market value of stocks, and cash-flow yield. The SUR model employed here is described by the following system of equations: (1) where Rit: monthly return on firm / in month t, ++++++++++++++++++++++++++++++++++++++RF +++++++++++++++++++++++++++++++++++++++++++++++++++++ t: risk-free return in month t (monthly return on three-month T-Bill), ++++++++++++++++++++++++++++++++RM t: monthly return on the All Share Market index in month t, Sizelt: market capitalization of firm /' in the month t (price times outstanding shares), (P/E)lt: price-earnings ratio for firm / in month t, ++++++++++++++++ (), (I,, a,, a2: parameters,fi et: the error term. ++++++++++ In the third procedure, we estimate a cross-sectional regression following Fama and MacBeth (1973) in order to investigate the role of fundamental factors on the return predictability at the Colombo Stock Exchange and the pricing of the market risk. The Fama-MacBeth procedure involves two steps. First, /? values are estimated for each firm from equation (1) using the time-series data for that firm. The SUR technique is used here to carry out the estimation in this step. Second, the average excess returns for the firms are regressed on the corresponding (3 estimates, the average MCAP, and the P/E multiples. Specifically, the following model is estimated for this step:

(2) In this model, (R - RF)jr f}SURif and e, are, respectively, the average excess return, the beta coefficient estimated using the SUR technique, and the error term for firm /. The y coefficients are the model parameters, and P/E and MCAP are as defined before. If cross-sectional returns are deter-

Regional Financial Markets

100

mined by beta alone, as indicated by the CAPM, the coefficients of the P/E and the MCAP variables in this model must be zero. Rejection of the null hypothesis of zero coefficient for P/E a n d / o r MCAP would indicate that cross-sectional returns on the CSE are predictable, providing support for the Fama et al. (1992) results. The Price Effect Since stock price (P) is a common factor between market capitalization (MCAP) and price-earning ratio (P/E), it may be a good explanatory variable in predicting cross-sectional returns. Indeed, the relationship between price and returns, known as price effect, has been documented by Jaffe et al. (1989).n These authors examine the average share price of 30 E/P-sizesorted portfolios and observe that higher E / P is associated with lower share price. They imply that the observed effects of E / P and size may in fact be a proxy for a price effect. In order to determine the presence of a price effect at the CSE, we estimate the following equation, which is similar to the one used by Jaffe et al. (1989): +++++++++++++++++++++++++++++++++++++ (R - RF)t = }'o + yAuRi + VzWVi + yaOnP),- + e, (3) The above equation is similar to equation (2), except that the explanatory variable MCAP in the latter equation is replaced by the log of the price level (InP). If the cross-sectional returns do exhibit a price effect, the regressor InP in equation (3) will be statistically significant and its impact will be larger than the P/E. ANALYSIS OF EMPIRICAL RESULTS Univariate Analysis The mean values for the monthly returns, market risk (/?), price-earning ratio (P/E), and size (MCAP) for the four portfolios constructed are reported in the three panels of Table 6.1. The rows in these panels show the values for mean returns, /?, P/E, and MCAP for each of the four portfolios, respectively. The last column reports the t-test results for the null hypothesis that the mean values of the corresponding left-side variable are equal for the first (smallest) and the fourth (largest) portfolios in the sample. In Panel A, the stocks are ranked according to size, measured by market capitalization (MCAP). The results show that the mean value of risk (/?) for the portfolio of the 10 smallest-size firms is higher than the average (3 for the portfolio of the largest size firms at the 1 % significance level. This implies that the larger firms exhibit less systematic risk. The difference in

Cross-Sectional Predictability of Stock Returns at the Colombo Stock Exchange 101

Table 6.1 Mean Values for Portfolio Returns, /?, and Fundamental Variables (MCAP and P/E) Portfolio I (small)

Portfolio 2

Portfolio 3

Portfolio 4 (large)

T-statistic H0: Rl =R4

1.371

1.25 1.027 18.128 7.549

-0.851 4.478*** -1.117 -15.674***

-1.094

Panel A. Ranked by size (MCAP) Return Risk (0) P/E MCAP

0.519 1.546 -26.296 4.799

1.11 1.296 17.491 5.846

1.213 20.031 6.495

Panel B. Ranked by P/E Return

0.564

Risk (P)

1.299

P/E

-36.075

MCAP

5.615

1.616 1.324 12.39 5.981

0.454

6.588

1.611 1.368 37.12 6.505

1.511 1.196 17.339 6.537

2.071 1.369 12.304 5.999

0.366 1.708 11.641 5.365

1.09 15.92

-0.511 -1.845* -1.946*

Panel C: Ranked by Market Risk (p) Return Risk (P) P/E MCAP

0.297 0.809 12.678 6.788

-0.078 -12.134*** 0.183 3.872***

Portfolios are formed by sorting the stocks according to ascending values of fi or a fundamental variable (size (MCAP) or P/E), a n d dividing them into four groups. Each portfolio consists of 10 stocks. The last column of the table shows the results for a test of equality of means between the first a n d the last portfolios (Rl a n d R4). ^Significant at the 10% level ^Significant at the 5% level ***Significant at the 1 % level

the mean monthly returns of these two portfolios, however, is statistically insignificant. Three conclusions can be drawn. First, size cannot be used to predict returns, as larger firms seem to produce the same returns as the smaller ones. This is inconsistent with the Fama et al. (1992) results. Second, since riskier firms do not offer higher returns, either risk is not priced in the Sri Lankan market, or the CSE is inefficient. Any such test is always a joint test. Third, larger firms display a superior performance as they offer the same returns as the smaller firms, for a lower level of risk. In other words, large firms offer a better risk-return trade-off. In Panel B, stocks are ranked according to the P / E ratio to form the portfolios. The high P / E stocks are, on average, found to be bigger in size than the smaller P / E stocks. The differences in the mean monthly returns and market risks between the lowest and the highest P / E portfolios, how-

102

Regional Financial Markets

ever, are found to be insignificant. This indicates the failure of the commonly used P/E in predicting returns. In Panel C, the stocks are ranked according to market risk (/?), and portfolios are formed on that basis. The results in this case confirm the finding in panel A, namely, that the higher-risk stocks are on average smaller in size. The important point, however, is that the mean monthly returns for the highest and the lowest (3 portfolios do not differ at any usual significance level. This result implies that |3 is not a predictor of returns as required by the CAPM, and that the market risk is not priced at the CSE. This result is consistent with the findings of Fama et al. (1992). The calculations for the three panels are repeated by forming portfolios of the 20 highest-value and the 20 lowest-value stocks from the sample. The findings remain unchanged, indicating robustness of the results to the number of portfolios.12 Overall, the results of the univariate analysis suggest that (3 does not support the notion of the CAPM; either the Colombo stock market is inefficient or risk is not priced in this market. Note, however, that univariate test results are not dependable because this procedure does not take all of the factors into account at the same time. The Extended CAPM The asset pricing model used here generalizes the basic CAPM to include size (MCAP) and P/E as the explanatory variables, in addition to the market risk. The SUR methodology is employed to estimate the system of 40 equations described by (1). Table 6.2 reports the results for each of the 40 firms covered in this study, and Table 6.3 summarizes the statistics on the values and statistical significance of ft, P/E, and MCAP. The results on ft indicate that the equity returns of all of the firms in the sample are sensitive to market movements. However, the coefficients of the P/E and size variables are also found to be significant, indicating the presence of a P/E and a size effect at the CSE. Overall, 29 of the 40 firms show some return predictability in terms of the two fundamental variables, P/E and MCAP, presenting a contradiction to the basic CAPM. As Chan and Chen (1988) have pointed out, this anomaly suggests either the inability of the basic CAPM to correctly measure /? or its failure to account for all the relevant risk factors.13 The Market Risk The statistics in Table 2 show that in the extended CAPM the /? coefficient is positive and statistically significant for all of the firms included in the sample. It follows that stocks of all of the firms do respond to changes in the overall market. Panel A in Table 3 summarizes the statistics on the values and statistical significance of /?. According to the figures in this

Cross-Sectional Predictability of Stock Returns at the Colombo Stock Exchange 103

Table 6.2 Coefficient Estimates for Risk (/?) and Fundamental Variables P/E and Size (MCAP): The Sur Technique intercept RM,-REt P/E MCAP +++++++++++++++++++++++++++++++++++++++++++++++++Name of Company

AMW Korea C. F.

Haycarb

Richard Peries

A. Spence

LOLC

10.19

0.92

0.77

-7.55

(2.62)***

(2.60)***

(3.06)***

(-2.66)***

13 (0.85)

1.49 (6.78)** *

0.46 (1.23)

-2.84 (-0.92)

0.98 (7.27)** *

-0.06 (-0.45)

(2.53)**

-38.62 (-2.68)***

1.31 (5.48)***

-0.28 (-1.76)

6.54 (2.86)***

11.17

1.27

-0.11

-1.05

(0.72)

(6.71)***

(-0.75)

(-0.45)

7.79

1 (5.85)***

0.21 (0.41)

-1.32 (-0.41)

-39.26

1.58

0.05

6.69

(-3.97)***

(6.39)***

(1.71)

(3.26)***

(0.46) Keells Foods

5.42

-40.23 (-2.78)***

32.86

1.41

0.24

(2.23)**

(8.69)***

(1.35)

-5.82 (-2.21)**

-234.28 (-6.68)***

2.16

-0.66

50.19

(2.45)**

(-1.49)

(6.92)***

Lanka Tiles

-27.07 (-0.65)

1.55 (4.39)***

0.49 (0.57)

3.62 -0.42

ACL Cabels

27.19 (1.85)

1.27 (3.38)***

1.16 (3.06)***

-7.17 (-2.21)**

Ahungala Hotels

ASCO

table, of the 40 firms, 31 firms have /fe over 1.0, indicating an above average market risk. However, these results cannot tell us whether risk is fairly priced at the CSE—that is, whether higher returns are associated with higher risk, or whether the security market line is positively sloped. More can be learned on this issue by investigating cross-sectional return behavior of firms listed on the CSE. This is deferred to subsection C (Section IV), where results based on the Fama-MacBeth procedure are discussed. P/E Effect It is commonly believed that most of the investment advice given by the majority of the brokerage houses in Sri Lanka is based on P/E ratios.

104

Regional Financial Markets

Table 6.2 Coefficient Estimates for Risk (/?) and Fundamental Variables P/E and Size (MCAP): The Sur Technique (continued) -68.96

1.73

0.72

14.18

(-4.89)***

(5.20)***

(2.60)***

(4.23)***

Dipped Products

-91.78 (-2.99)**

0.62 (4.05)***

-0.69 (-2.24)**

15.47 (2.89)***

Shaw Wallace

-113.19 (-6.77)***

1.69 (7.84)***

-0.02 (-1.07)

18.2 (6.51)***

P r o p e r t y Dev.

-311.9 (-4.67)***

0.83 (5.08)***

-1.68 (-1.76)

45.42

-25.76 (-1.23)

1.02 (7.17)***

0.82 (5.14)***

1.53 (0.44)

Chemanex

Bata

C o m m e r c i a l Bank

MLL

(4.43)***

-28.57

0.69

0.02

(-1.21)

(3.51)***

(0.06)

3.68 (1.02)

-76.76 (-2.87)***

1.34 (5.44)***

-0.38 (-0.41)

15.56 (2.37)**

-7.45

1.08 (6.01)***

1.05 (3.21)***

-0.26

(-0.51) Ceylon Tobacco

-95.34 (-1.13)

0.83 (2.06)**

1.9 (3.94)***

7.7 (0.69)

C T C Eagle

-63.46 (-5.43)***

1.94 (10.64)***

-0.04 (-3.07)***

10.39 (5.45)***

G r a i n Elevators

40.66 (2.39)**

1.29 (6.71)***

0.19 (1.6)

-5.73 (-2.26)**

-60.44 (-2.35)**

1.21 (6.14)***

0.15 (0.64)

8.22 (2.13)**

C e n t r a l Finance

Nestle

(-0.10)

However, the extent to which P/E ratios are used in predicting stock returns varies considerably across the Sri Lankan brokerage houses, and the use of technical analysis based on other indicators is also prevalent. Usually, when a stock's P / E ratio is lower than the prevailing market and industry P / E ratios, the brokerage houses will consider it to be undervalued and will expect it to offer a higher return. Hence, a negative relationship is expected between the P/E ratio and the returns. As can be seen from Table 3 (Panel B), in the extended CAPM estimated here, the SUR-based coefficients on the P/E variable are statistically significant for 15 of the 40 firms. Of these 15 firms, 12 firms have significant

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ hh Table 6.2 Coefficient Estimates for Risk (/?) and Fundamental Variables P/E and Size (MCAP): The Sur Technique (continued) Singer SL

-4.29 (-0.22)

1.26 (7.69)***

1.74 (4.22)***

-2.81 (-0.80)

Regnis

-68.3 (-2.32)**

1.37 (6.18)***

-0.14 (-0.49)

15.69 (2.04)**

0.71 (4.39)***

-0.54 (-1.08)

10.91 (1.67)

0.8 (3.67)***

1.1 (3.22)***

18.43 (3.62)***

1.48 (8.00)***

0 (-0.22)

1.47 (0.63)

Commercial Devlp. -57.28 (-1.75) Rekitt & Colman

-131.99 (-4.24)***

Union Assurance L. -8.38 (-0.89) C Insurance

-16.19 (-1.30)

1.64 (10.07)***

0.34 (1.62)

2.16 (0.85)

HNB

-16.95 (-1.51)

0.71 (5.86)***

1.05 (4.90)***

0.08 (0.05)

Lanka Alum.

-28.74 (-2.08)**

1.67 (4.75)***

0.07 (1.09)

5.51 (2.00)**

E. B. Creasey

-42.83 (-2.79)***

1.31 (3.58)***

0.62 (3.78)***

7.06 (2.43)**

ACME Alum.

-39.02 (-2.39)***

1.55 (5.46)***

0.38 (2.21)**

5.04 (1.54)

JKH

34.2 (-1.45)

1.28 (6.16)***

0.31 (1.62)

-5.01 (-1.49)

positive coefficients while the remaining three firms have significant negative coefficients. The positive and significant values of P/E are contrary to the view that low P/E multiples predict high future performance and vice versa. Note, however, that the explanatory power of the P/E ratio in describing average monthly returns may be more appropriately determined if the forecasted earnings for a company, rather than historical values used here, are employed to calculate P/E. This is because it is the projected earnings that brokers use to make buy and sell recommendations. 14 Size Effect Panel C in Table 3 reports the frequency of statistical significance of the size effect. Based on these figures, over 50 percent of the firms analyzed

++++++++++++++++++++++++++

hTh

Table 6.2 Coefficient Estimates for Risk (/?) and Fundamental Variables P/E and Size (MCAP): The Sur Technique (continued) 1.26 (6.13)***

0.07

4.69

(0.82)

(1.2)

-104.74

1.39

(-4.81)***

(6.17)***

0 (-0.59)

(4.66)***

-45.69

1.18 (7.82)***

-0.38 (-2.55)**

6.59 (2.39)**

(-2.21)**

1.17 (5.54)***

-0.01 (-0.29)

6.74 (2.14)**

UML

-15.07 (-1.20)

1.23 (6.84)***

0.16 (0.5)

2.24 (0.99)

S a m p a t h Bank

-32.82 (-2.48)***

1.57 (8.18)***

0.05 (1.84)

4.42 (2.27)**

LMF

C T Land

Hay leys

-32.96 (-2.39)**

(-2.28)** P u r e Baverages

-44.32

19.3

^Significant at 10% **Significant at 5% ^Significant at 1% ++++++++++++++++++++++++++++++++++++++++++++++++++++ + where Rlt: monthly return on firm / in month t, RFt: risk-free return in month t (monthly return o +++++++++++++++++++++++++++++++++++++++++++++++++++++three-month T-Bill), RM capitalization of firm i in the month t (price times outstanding shares), (P/E),t: price-earnings ratio for firm / in month t. /?0, p,, al7 a2: parameters, et: the error term.

in this study show a significant size effect in predicting average returns in the market. About 18 firms exhibit significant positive coefficients for the MCAP variable, and the remaining four firms have a significant negative coefficient. The positive and significant size effect found here is contrary to the findings of the small-firm effect in many developed markets; the latter are characterized by an inverse relationship between size and rate of return (Fama et al., 1992). One explanation for the positive size effect may be the fact that large-cap firms are better managed, more likely to get preferential treatment from the government in terms of protection from foreign competition, and hence able to generate a superior risk-return trade-off (Errunza, 1994).15 These attributes frequently entice investors into larger holdings of large-cap stocks and create an excess demand for the stocks of such firms.16 Thus, for a fixed supply of large-cap stocks in the market, excess demand runs up the price, providing abnormal gains for the investors. 17 Chan et al. (1988) state that larger returns on small-cap firms manifesting a positive small-firm effect may reflect the premium for bearing risk rather than return predictability based on size. However, in

hhh Table 6.3 Summary Statistics for Values and Significance of p, P/E, and MCAP Panel A: Number of firms with greater/lower than market risk £>I Firms +++

Sig. 5%

Sig. 1%

31

30

p