Risk Management Post Financial Crisis : A Period of Monetary Easing 9781784410261, 9781784410278

Citation preview

RISK MANAGEMENT POST FINANCIAL CRISIS: A PERIOD OF MONETARY EASING

CONTEMPORARY STUDIES IN ECONOMIC AND FINANCIAL ANALYSIS Series Editors: Robert Thornton and J. Richard Aronson Recent Volumes: Volume 85:

The Irish Economy in Transition: Success, Problems and Prospects, Edited by J. R. Aronson, V. Munley and R. J. Thornton

Volume 86:

Asia Pacific Financial Markets in Comparative Perspective: Issues and Implications for the 21st Century, Edited by Jonathan A. Batten and Thomas A. Fetherston

Volume 87:

Developments in Litigation Economics, Edited by P. A. Gaughan and R. J. Thornton

Volume 88:

European Responses to Globalization: Resistance, Adaptation and Alternatives, Edited by Janet Laible and Henri J. Barkey

Volume 89:

Globalization: Perspectives from Central and Eastern Europe, Edited by Katalin E. Fabian

Volume 90:

Privatization in Transition Economies: The Ongoing Story, Edited by Ira W. Lieberman and Daniel J. Kopf

Volume 91:

Personal Injury and Wrongful Death Damages Calculations: Transatlantic Dialogue, Edited by John O. Ward and Robert J. Thornton

Volume 92:

Moving Beyond Storytelling: Emerging Research in Microfinance, Edited by Todd A. Watkins and Karen Hicks

Volume 93:

The Impact of the Global Financial Crisis on Emerging Financial Markets, Edited by Jonathan A. Batten and Peter G. Szilagyi

Volume 94:

Derivative Securities Pricing and Modelling, Edited by Jonathan A. Batten and Niklas Wagner

Volume 95:

The Developing Role of Islamic Banking and Finance: From Local to Global Perspectives, Edited by Fouad H. Beseiso

CONTEMPORARY STUDIES IN ECONOMIC AND FINANCIAL ANALYSIS VOLUME 96

RISK MANAGEMENT POST FINANCIAL CRISIS: A PERIOD OF MONETARY EASING EDITED BY

JONATHAN A. BATTEN Monash University, Australia

NIKLAS F. WAGNER University of Passau, Germany

United Kingdom
North America
Japan India
Malaysia
China

Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2014 Copyright r 2014 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. Any opinions expressed in the chapters are those of the authors. Whilst Emerald makes every effort to ensure the quality and accuracy of its content, Emerald makes no representation implied or otherwise, as to the chapters’ suitability and application and disclaims any warranties, express or implied, to their use. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-1-78441-027-8 ISSN: 1569-3759 (Series)

ISOQAR certified Management System, awarded to Emerald for adherence to Environmental standard ISO 14001:2004. Certificate Number 1985 ISO 14001

CONTENTS LIST OF CONTRIBUTORS

ix

PART I: A PERSPECTIVE ON THE FINANCIAL CRISIS AND GLOBAL RISK MANAGEMENT INTRODUCTION TO RISK MANAGEMENT POST FINANCIAL CRISIS: A PERIOD OF MONETARY EASING Jonathan A. Batten and Niklas F. Wagner

3

COMPLEXITY ANALYSIS AND RISK MANAGEMENT IN FINANCE Charilaos Mertzanis

15

THE EFFECTS OF MACROECONOMIC NEWS ANNOUNCEMENTS DURING THE GLOBAL FINANCIAL CRISIS Pilar Abad and Helena Chulia´

41

PART II: RISK AND INTEGRATION IN A POST-CRISIS SETTING THE PRO-CYCLICAL IMPACT OF BASEL III REGULATORY CAPITAL ON BANK CAPITAL RISK Guoxiang Song

59

NONPARAMETRIC EXPECTILE REGRESSION FOR CONDITIONAL AUTOREGRESSIVE EXPECTED SHORTFALL ESTIMATION Marcelo Brutti Righi, Yi Yang and Paulo Sergio Ceretta

83

v

vi

CONTENTS

THE IMPACT OF EXTERNAL SHOCKS ON STOCK PRICES IN THE EAST ASIAN DOMESTIC BANKING SECTOR Masahiro Inoguchi MEASURING FINANCIAL INTEGRATION: EVIDENCE FROM TEN INDUSTRIES IN A “US-EMERGING WORLD” Michael Donadelli

97

153

PART III: MONETARY POLICY AND CREDIT IN A POST-CRISIS SETTING WILL QUANTITATIVE EASING ENHANCE OR DRAIN THE AVAILABILITY OF FUNDS TO FINANCIAL MARKETS? Yasushi Suzuki

181

MONEY DEMAND CAUSALITY FOR TEN ASIAN COUNTRIES: EVIDENCE FROM LINEAR AND NONLINEAR CAUSALITY TESTS Ahdi Noomen Ajmi and Nicholas Apergis

193

PRODUCT MARKET COMPETITION AND INFLATION PERSISTENCE Amr Sadek Hosny

211

UNLOCKING CREDIT Ike Mathur and Isaac Marcelin

221

PART IV: COUNTRY STUDIES ON MONETARY POLICY, CREDIT, AND FINANCIAL INTERMEDIARIES MONETARY POLICY AND BANK LIQUIDITY IN CHINA Nan Shi, Xin Sun and Fan Zhang

255

vii

Contents

MONETARY POLICY AND BANK CREDIT RISK IN VIETNAM PRE AND POST GLOBAL FINANCIAL CRISIS Xuan Vinh Vo and Phuc Canh Nguyen

277

ANALYSIS OF FACTORS INFLUENCING AND CONTROLLING EXCESS CASH AND SHORT-TERM BANK LOANS IN TAIWAN Ma-Ju Wang and Yi-Ting Chang

291

BANK COMPETITION, MANAGERIAL EFFICIENCY AND THE INTEREST RATE PASS-THROUGH IN INDIA Jugnu Ansari and Ashima Goyal

317

FINANCIAL ARCHITECTURE AND MONETARY POLICY TRANSMISSION MECHANISM IN KENYA Roseline Nyakerario Misati, Alfred Shem Ouma and Kethi Ngoka-Kisinguh

341

PART V: COUNTRY STUDIES ON MARKETS AND RISK THE DIM SUM BOND MARKET IN HONG KONG Ike Mathur and Soumen De

367

SENTIMENT AND BETA HERDING IN THE BORSA ISTANBUL (BIST) Nazmi Demir, Syed F. Mahmud and M. Nihat Solakoglu

389

TESTING FOR RATIONAL SPECULATIVE BUBBLES IN THE BRAZILIAN RESIDENTIAL REAL-ESTATE MARKET Marcelo M. de Oliveira and Alexandre C. L. Almeida

401

CHALLENGES IN THE APPLICATION OF EXTREME VALUE THEORY IN EMERGING MARKETS: A CASE STUDY OF PAKISTAN Jamshed Y. Uppal and Syeda Rabab Mudakkar

417

INDEX

439

LIST OF CONTRIBUTORS Pilar Abad

Department of Economics, University Rey Juan Carlos and Riskcenter-IREA, Madrid, Spain

Ahdi Noomen Ajmi

College of Science and Humanities in Slayel, Salman bin Abdulaziz University, KSA; ESC de Tunis, Manouba University, Tunisia; IPAG Lab, IPAG Business School, Boulevard Saint-Germain, Paris, France

Alexandre C. L. Almeida

Departamento de Fı´ sica e Matema´tica, Universidade Federal de Sa˜o Joa˜o del-Rei, Ouro Branco, Brazil

Jugnu Ansari

Centre for Advanced Financial Research and Learning (CAFRAL), Reserve Bank of India, Mumbai, India

Nicholas Apergis

School of Economics and Finance, Curtin University, Perth, Australia

Jonathan A. Batten

Department of Banking and Finance, Monash University, Melbourne, Australia

Paulo Sergio Ceretta

Department of Business, Federal University of Santa Maria, Santa Maria, Brazil

Yi-Ting Chang

Taipei Fubon Bank, Taipei, Taiwan

Helena Chulia´

Department of Econometrics, University of Barcelona and Riskcenter-IREA, Barcelona, Spain

Soumen De

Department of Finance, Menlo College, Atherton, CA, USA ix

x

LIST OF CONTRIBUTORS

Marcelo M. de Oliveira

Departamento de Fı´ sica e Matema´tica, Universidade Federal de Sa˜o Joa˜o del-Rei, Ouro Branco, Brazil

Nazmi Demir

Department of Banking and Finance, Bilkent University, Ankara, Turkey

Michael Donadelli

Research Center SAFE, Department of Finance, Goethe University Frankfurt, Frankfurt, Germany

Ashima Goyal

Indira Gandhi Institute of Development Research, Mumbai, India

Amr Sadek Hosny

Regional Studies Division, Middle East and Central Asia Department, International Monetary Fund, Washington, DC, USA

Masahiro Inoguchi

Faculty of Business Administration, Kyoto Sangyo University, Kyoto, Japan

Syed F. Mahmud

Department of Economics, Bilkent University, Ankara, Turkey

Isaac Marcelin

School of Business, Management and Technology, University of Maryland Eastern Shore, Princess Anne, MD, USA

Ike Mathur

Department of Finance, Southern Illinois University Carbondale, Carbondale, IL, USA

Charilaos Mertzanis

Department of Management, The American University in Cairo, Cairo, Egypt

Roseline Nyakerario Misati

Central Bank of Kenya, Research and Policy Analysis Department, Nairobi, Kenya

Syeda Rabab Mudakkar

Lahore School of Economics, Centre for Mathematics & Statistical Sciences, Lahore, Pakistan

xi

List of Contributors

Kethi Ngoka-Kisinguh

Central Bank of Kenya, Research and Policy Analysis Department, Nairobi, Kenya

Phuc Canh Nguyen

School of Banking, University of Economics, Ho Chi Minh City, Vietnam

Marcelo Brutti Righi

Department of Business, Federal University of Santa Maria, Santa Maria, Brazil

Alfred Shem Ouma

Retirement Benefits Authority, Research Department, Nairobi, Kenya

Nan Shi

Durham University Business School, University of Durham, Durham, UK

M. Nihat Solakoglu

Department of Banking and Finance, Bilkent University, Ankara, Turkey

Guoxiang Song

Department of Accounting and Finance, Faculty of Business, University of Greenwich, London, UK

Xin Sun

Durham University Business School, University of Durham, Durham, UK

Yasushi Suzuki

Graduate School of Management, Ritsumeikan Asia Pacific University, Oita, Japan

Jamshed Y. Uppal

School of Business and Economics, Catholic University of America, Washington, DC, USA

Xuan Vinh Vo

School of Banking, University of Economics, Ho Chi Minh City, Vietnam

Niklas F. Wagner

Department of Business and Economics, University of Passau, Passau, Germany

Ma-Ju Wang

Department of Finance, College of Finance and Banking, National Kaohsiung First University of Science and Technology, Yanchao, Kaohsiung, Taiwan

xii

LIST OF CONTRIBUTORS

Yi Yang

The School of Statistics, University of Minnesota, Minneapolis, MN, USA

Fan Zhang

Liverpool John Moores University, Liverpool Business School, Liverpool, UK

PART I A PERSPECTIVE ON THE FINANCIAL CRISIS AND GLOBAL RISK MANAGEMENT

INTRODUCTION TO RISK MANAGEMENT POST FINANCIAL CRISIS: A PERIOD OF MONETARY EASING Jonathan A. Batten and Niklas F. Wagner ABSTRACT Financial markets have experienced considerable turbulence over the past two decades. The recent subprime and sovereign debt crises in the United States and Europe, respectively, have resulted in significant new regulatory responses. They also prompted the re-evaluation of how best to manage and measure financial risk. The 20 chapters in this volume provide a number of different perspectives on financial risk in the postcrisis period where monetary easing has become a predominant monetary policy. While asset price volatility has now returned to levels experienced in the mid-2000s many lessons remain. Among the most important is the need to accurately measure and manage the complex risks that exist in financial markets. Our hope is that the chapters presented here provide a

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 3
13 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096019

3

4

JONATHAN A. BATTEN AND NIKLAS F. WAGNER

better understanding of how best to do this, while also giving insights for next suitable steps and further developments. Keywords: Subprime crisis; global financial crisis; sovereign debt crisis; monetary easing; risk management; risk measurement JEL classifications: G01; G1; G2

In recent years the focus of much of the risk management literature has been on the measurement of financial and related risks, rather than necessarily the management of the underlying risky positions of financial sector intermediaries or corporations. This is not surprising given the impetus for increased regulation of the financial sector arising from revised capital adequacy guidelines and the events surrounding the 2007
2012 Global Financial Crisis (GFC). As is well known, these events had an unprecedented overall impact on the world economy (see, e.g. Stiglitz, 2009, 2010). As the vast impact of the GFC on the real economy was anticipated, policy makers worldwide decided to take determined action. The given economic policy responses included three major fields, namely (i) global bank and other financial institution rescues, (ii) immense economic stimuli from fiscal spending packages and finally (iii) monetary policy that provides ample liquidity and applies non-standard bond buying techniques often called ‘monetary easing’ or ‘quantitative easing’. Needless to say, these market interventions caused massive externalities. As many developing countries are depended on external capital flows and typically cannot afford large rescue packages, the impact of the GFC on emerging economies should not be underestimated, which became apparent with capital flows to emerging economies after the GFC and their recent abrupt reversal in 2013/2014. Although the European sovereign debt crisis of 2010
2012 has its own roots, it can well be seen as a follow-up crisis to the initial U.S. subprime crisis as subprime-related bank rescues in Europe led to massive increases in the amounts of government debt, which in turn, among other events, triggered several country crises more or less simultaneously.1 These coherences illustrate that the GFC and its evoked policy actions still impact, and for quite some time will continue to impact, the state of global economies and financial markets. The GFC arose in part due to the complexity of many products traded by financial market participants, and their lack of ability in managing and

Introduction to Risk Management Post Financial Crisis

5

measuring risks which subsequently brought attention from regulators.2 It was in this context and with the U.S. housing market collapse as a trigger, that many ad-hoc models were put under the regulatory microscope, especially with respect to the sufficiency of capital that supported underlying positions. Nonetheless the management of these positions via natural offset, or through the use of other financial products or derivatives as hedges, also remains a critical responsibility of both the corporate and financial sector. Many of these themes are discussed in this volume, where risk management is investigated very broadly in the post-GFC period. This period may be described as a period of monetary easing, where asset price volatility has diminished, financial sector balance sheets have been strengthened, or in many cases rebuilt, and where attention has shifted to the management of underlying risks either through the breakdown of these risks into their underlying components, or through consolidation.3 Consolidation, or risk aggregation, offers both the benefit of diversification, as well as the possibility of natural offset. Risk diversification offers particular benefits to credit risk management, whereas natural offset is clearly evident in the international positions of banks where offsetting asset-liability positions reduce, or in some cases, eliminate underlying risks. This special issue comprising 20 chapters is divided into four parts. These provide a perspective on global risk management, assess the effects of integration and risk measurement, the impact of monetary policy and two parts finally focus on country studies from a monetary policy, intermediaries and overall markets perspective.

PART I: A PERSPECTIVE ON THE FINANCIAL CRISIS AND GLOBAL RISK MANAGEMENT There are two chapters in this part, apart from this introduction. The chapter by Mertzanis assesses how complexity analysis of financial systems can be used in measuring and reporting risk despite its lack of a theoretical foundation. The author considers three dimensions of complexity: financial instrument complexity, financial process complexity and financial system complexity. These dimensions of complexity interact with each other in accordance with prevalent market structures, agent actions and the financial models used by financial market participants. The chapter argues that policy makers and practitioners need to take both a micro and macro view of financial risk, identify proper transparency requirements on complex

6

JONATHAN A. BATTEN AND NIKLAS F. WAGNER

instruments, develop dynamic models of information generation that best approximate observed financial outcomes and identify and address the causes and consequences of systemic risk. The chapter by Abad and Chulia´ considers the impact of news on financial asset prices. Their chapter focuses specifically on the impact of macroeconomic news announcements on bond markets in the Euro area, where they analyse the impact of changes in the interest rate, unemployment rate, consumer confidence index and industrial production index on the returns, volatility and correlations of European government bond markets. Their results suggest that while bond return volatility is strongly affected by these news announcements, the response is asymmetric, suggesting a complex interplay of price impacts from the macroeconomic news releases.

PART II: RISK AND INTEGRATION IN A POST-CRISIS SETTING The second part of the present volume contains four chapters that investigate the role of risk and financial market integration. The chapter by Song investigates the pro-cyclical impact of Basel III, including the effects on capital requirements and bank balance sheets. The chapter focuses on the regulatory capital ratios of six systemically important global U.S. banks and their development since 2007. The author designs various models to measure the direct impact of accounting rules on a bank’s regulatory capital ratios and investigates the impact when the Basel III regulatory capital definition is applied. The results indicate that Basel III regulatory capital will indeed enhance the pro-cyclicality of bank capital risk. The chapter by Righi, Yang and Ceretta investigates an alternative measure to the widely used Value at Risk (VaR), which is termed Expected Shortfall (ES) and today is also advocated within the Basel accord. As is well known, VaR represents the maximum loss given a confidence level during a pre-specified period, and it does not consider losses above the quantile of interest. Furthermore, it is not sub-additive, that is despite diversification, the VaR of a portfolio could be greater than the VaR of individual assets from the same portfolio. ES is defined as the average loss given that overcomes the VaR and thereby considers the magnitude of losses. The authors estimate the ES in conditional autoregressive expectile models by using a nonparametric multiple expectile regression via gradient tree boosting. This approach has the advantage of flexibility in data

Introduction to Risk Management Post Financial Crisis

7

assumptions and avoids the drawbacks and fragilities of a restrictive estimator such as historical simulation. The authors present various specifications with the results obtained from simulated and real market data indicating that this approach provides an important alternative for ESbased risk management. The next chapter by Inoguchi examines the price impact of foreign stock markets on the stock prices of domestic banks in Korea, Malaysia, Singapore and Thailand during the GFC. The chapter helps to clarify the debate on whether these banks were immune from events that were occurring in markets outside East Asia. The author employs a multinomial logit model to estimate how changes in the U.S. and Japanese stock markets affected the banking sectors. The results clearly show that foreign stock market indices exerted a larger impact on the prices of East Asian banking stocks during the 2000s than during the 1990s. Importantly, the authors conclude that increased foreign capital flows and foreign assets and liabilities greatly influenced domestic banking systems in East Asia during the 2000s. The final chapter in this part by Donadelli investigates the role of financial market integration in emerging markets. The chapter employs two newly introduced robust integration measures that rely on principal component analysis in order to investigate financial integration across 20 emerging equity markets and the United States. To capture the dynamics of the financial integration process, both integration measures are computed in four ad-hoc periods that span the period from January 1994 to July 2007. In addition, the national market (in each region or country) is divided into 10 different industries. The results show that the level of integration in the aftermath of the 2008 Lehman Brothers’ collapse was higher than the level of integration in the aftermath of equity market liberalizations (i.e. 1994
1998). This result holds across industries and suggests that de jure integration does not necessarily improve de facto integration. In most industries, financial integration slows between the first and second eras accelerating as financial markets and trade opens. The author concludes that his findings give rise to a ‘diversification benefits/insurance benefits trade-off’.

PART III: MONETARY POLICY AND CREDIT IN A POST-CRISIS SETTING The third part investigates the impact of monetary policy, liquidity and completion in the post-crisis environment. The part opens with Suzuki’s

8

JONATHAN A. BATTEN AND NIKLAS F. WAGNER

investigation of the impact of quantitative easing (QE) on financial markets. QE refers to changes in the composition and/or size of a central bank’s balance sheet which are intended to improve market liquidity and/ or credit conditions. The author’s analysis suggests that an appropriate level of market reference rate would encourage investors to absorb the relatively wider range of credit risk in the bond market. A higher market rate would discourage borrowers, while a lower market rate would drain ‘risk’ funds in the bond market. Overall the author argues that there is no clearcut mechanism based on economic theory for underpinning the commonly accepted view upon which the concept of QE is based. The next chapter in this part by Apergis and Ajmi estimates tests the causality properties between real money demand and a number of determinants, that is real output, the lending rate and the real exchange rate, across 10 Asian economies through linear and non-linear causality methodologies. Their results spanning the period 1990
2012 document both bidirectional and unidirectional causality between monetary aggregates (M1 and M2) and their determinants for different country groups. These empirical findings highlight the importance of money demand as a policy tool and offers useful policy recommendations on the formation of an Asian monetary union. This theme is continued with the next chapter by Hosny who investigates the role of various demand and supply factors in explaining inflation. Understanding the factors that drive the inflationary process is of critical importance to central banks given their typical objective of achieving price stability. The author employs a backward-looking Phillips curve framework in a dynamic panel framework for 105 countries. Over the years 2008
2011, he investigates the role of product market imperfections in explaining inflation and inflation persistence, especially among emerging market economies. The results suggest that product market competition does not have a significant impact on inflation persistence. On average, higher competition and efficiency in product markets reduces the inflation persistence effect especially in the MENA region and countries at lower stages of development. The final chapter in this part by Mathur and Marcelin investigates the association between collateral coverage, country-level governance and various institutional proxies. The authors note that overcollateralization may be economically and socially non-optimal since it may thwart efficient resource allocation. They investigate the role that credit insurance can play in increasing the level of private credit, investment and growth, where collateral spread is the main inhibitor of finance. Credit insurance enables a

Introduction to Risk Management Post Financial Crisis

9

lender to collect the debt on default and liquidate collateral assets at prices below outstanding loan values, since the lender’s loss is covered through the insurance.

PART IV: COUNTRY STUDIES ON MONETARY POLICY, CREDIT AND FINANCIAL INTERMEDIARIES Next, the volume provides five country studies that explore the relationship between monetary policy, credit and financial intermediaries. The chapter by Shi, Sun and Zhang’s is their investigation of recent monetary policy in China, with specific attention paid to the period around June 2013 when the Shanghai Interbank Offered Rate peaked at 13.4%. This liquidity crunch occurred despite the fact that the Chinese central bank (the Peoples Bank of China (PBoC)) frequently expressed concern to financial intermediaries about their need to reduce their levels of lending, and despite its statements that the previous expansionary monetary policy would not continue. As noted by the authors the PBoC refused to intervene, which was unexpected by the market, which subsequently had difficulty in adjusting. The role that monetary policy and credit plays in developing economies is considered in the chapter by Vo and Nguyen. The authors investigate the impact of monetary policy on bank risk in Vietnam pre and post the GFC. They employ a unique and disaggregated bank level data set from 2003 to 2012 and utilise a panel estimation technique. Their results support the importance of the bank lending channel and show that the transmission mechanism was affected by characteristics of commercial banks. The next chapter by Wang and Chang investigates the impact on operating performance, dividend policy, financial leverage and corporate governance of excess cash holdings and the use of short-term bank loans by firms in Taiwan during 2001
2010. One key novelty of the chapter is that the authors’ survey managers to support their use of variables employed in their regression analysis. Their findings show that excess cash and shortterm bank loans in all industries are negatively correlated. When the interest rate in the money market is low, firms tend to accumulate excess cash. The survey results highlight the importance of bank relationships in deciding about the allocation of short-term funds. The study by Ansari and Goyal examines interest rate pass through for Indian banks. Their sample period from 1996 to 2012 allows an investigation of the impact of financial reforms. The authors use a dynamic panel

10

JONATHAN A. BATTEN AND NIKLAS F. WAGNER

data methodology for a sample of 33 banks to estimate the determinants of commercial banks’ loan pricing decisions. Their results show that commercial banks consider several factors apart from the policy rate. More competition reduces policy pass through by decreasing the loan rate as well as spreads. Financial market reform has had mixed effects, while managerial inefficiency raises rates and spreads, while product diversification reduced both. The last chapter in this part by Misati, Shem and Ngoka-Kisinguh focuses on the importance of financial market architecture in emerging markets. With a focus on monetary policy transmission in Kenya, their study assesses monetary policy transmission channels in an evolving domestic financial structure during the period of the GFC. Their chapter employs trend analysis of financial variables to obtain an assessment of the effects of the financial crisis on financial variables and on monetary policy transmission. Their results imply that while a central bank can rely on the interest rate and exchange rate channels in its pursuit of price stability objective, this does not imply that the retail interest rate rigidities observed in practice, especially following a monetary policy loosening, have disappeared.

PART V: COUNTRY STUDIES ON MARKETS AND RISK The remaining four chapters in the present volume provide insights into specific risks and the importance of institutional structure in mitigating these risks. The first chapter in the part is the analysis of the emerging Dim Sum or the Renminbi (RMB) bond market in Hong Kong by De and Mathur. This market enables non-residents of China to invest in yuandenominated bonds. These securities are typically issued by a select group of pre-screened entities and financial institutions, outside of China. The authors discuss the important role that Hong Kong has played in promoting the Dim Sum bond market, although the recent development of offshore markets in Taiwan, Singapore and London provide alternatives for investors. With policy challenges going forward, the development of these markets is important for the long-term development of the Chinese RMB as an international currency. The next chapter by Demir, Mahmud and Solakoglu investigates another developing market, the stock market in Turkey, the Borsa Istanbul

Introduction to Risk Management Post Financial Crisis

11

(BIST). The study investigates sentimental herding in the BIST during the last decade using a state-space model, which employs cross-sectional deviations of systematic risk (Beta). The authors find evidence of time-varying herding behaviour that is both statistically significant and persistent over the period from 2000 to 2011. For example, there was evidence of herding during the financial crisis in 2000
2001, whereas in more recent times herding pattern was not so evident due to the difficulty in interpreting internal and external events. The real estate market of Brazil is investigated for speculative bubbles in the chapter by de Oliveira and Almeida. The authors show that the recent increase in Brazilian housing prices has justifiably fuelled regulatory concerns that the economy was experiencing a speculative housing bubble. The authors employ a recently proposed recursive unit root test in order to identify possible speculative bubbles in the residential real estate market. The empirical results show evidence for speculative price bubbles both in Rio de Janeiro and Sao Paulo, the two main Brazilian cities. As the test is able to identify the presence of asset bubbles at an early stage, it may therefore provide important insights to market participants in the construction of early warning systems. The final chapter of the volume is the study by Uppal and Mudakkar, which addresses the challenges in applying extreme loss risk estimates in the context of emerging markets. Their study of the Karachi Stock Exchange, the main equity market of Pakistan, offers important insights due to the high country and political risk. Returns in the market have shown significant periods of high volatility and the impact of extreme events. Nonetheless Pakistan has a well-established institutional and regulatory structure. The authors address the important issue of how best to optimally model the dynamics in such a turbulent environment. Their findings validate the use of Extreme Value Theory for a small frontier market like Pakistan and indicate its usefulness for other emerging markets. Risk management systems based on the Dynamic VaR with tail estimation by EVT may be more helpful than standard VaR models. The 20 chapters in this volume provide a number of different perspectives on financial market and institutional risk in the post-crisis period. Since the subprime crisis in the United States was surpassed by sovereign risk events in Europe, the global financial system and worldwide asset markets appear to have been subject to unprecedented levels of volatility and risk. Many financial market intermediaries and regulators were caught off-guard by these events and the turmoil in financial markets, especially those effects related to liquidity that unfolded. These consequences were

12

JONATHAN A. BATTEN AND NIKLAS F. WAGNER

exacerbated by the degree of financial market integration and the globalisation of financial market intermediaries. Thus asset volatility was quickly, and easily, transmitted to markets worldwide and along with it concerns by intermediaries over credit quality, liquidity and other forms of financial risk. While asset price volatility has now returned to levels experienced in the mid-2000s many lessons remain. Among the most important is the need to accurately measure and mange financial market risk. We hope that the chapters presented in this volume facilitate this process and will provide new perspectives, while also providing insights for next steps and further developments.

NOTES 1. For the impact of the GFC on emerging economies see for example Stiglitz (2010) and Batten and Szilagyi (2011). A discussion of the 2010
2012 European sovereign debt crisis can be found in Lane (2012). 2. See for example Pauget (2009) for a post-crisis outlook on financial regulation and supervision. For perspectives on the 2007
2009 subprime crisis see for example Stiglitz (2009) and Breitenfellner and Wagner (2010), for the 2007
2012 GFC see for example Elliott (2011) and Sikorski (2011). 3. See also the discussion on different approaches to the management of financial risks by Hull (2007, pp. 1
25).

REFERENCES Batten, J. A., & Szilagyi, P. G. (2011). The impact of the global financial crisis on emerging financial markets. In J. A. Batten & P. G. Szilagyi (Eds.), The impact of the global financial crisis on emerging financial markets, contemporary studies in economic and financial analysis (Vol. 93, pp. 3
16). Bingley, UK: Emerald Group Publishing Limited. Breitenfellner, B., & Wagner, N. (2010). Government intervention in response to the subprime financial crisis: The good into the pot, the bad into the crop. International Review of Financial Analysis, 19, 289
297. Elliott, L. (2011). Global financial crisis: Five key stages 2007
2011. The Guardian, August 7, 2011. Retrieved from http://www.theguardian.com/business/2011/aug/07/global-finan cial-crisis-key-stages Hull, J. (2007). Risk management and financial institutions. Princeton, NJ: Pearson Education. Lane, P. R. (2012). The European sovereign debt crisis. Journal of Economic Perspectives, 26, 49
67.

Introduction to Risk Management Post Financial Crisis

13

Pauget, G. (2009, September). Regulation-supervision, the post-crisis outlook. Financial Stability Review No. 13: The Future of Financial Regulation, Banque de France, Paris, France (pp. 117–122). Sikorski, D. (2011). The global financial crisis. In J. A. Batten & P. G. Szilagyi (Eds.), The impact of the global financial crisis on emerging financial markets, contemporary studies in economic and financial analysis (Vol. 93, pp. 17
90). Bingley, UK: Emerald Group Publishing Limited. Stiglitz, J. E. (2009). Freefall, free markets and the sinking of the global economy. London: Penguin Books. Stiglitz, J. E. (2010). The stiglitz report, reforming the international monetary and financial systems in the wake of the global crisis. London: The New Press.

COMPLEXITY ANALYSIS AND RISK MANAGEMENT IN FINANCE Charilaos Mertzanis ABSTRACT Standard financial risk management practices proved unable to provide an adequate understanding and a timely warning of the financial crisis. In particular, the theoretical foundations of risk management and the statistical calibration of risk models are called into question. Policy makers and practitioners respond by looking for new analytical approaches and tools to identify and address new sources of financial risk. Financial markets satisfy reasonable criteria of being considered complex adaptive systems, characterized by complex financial instruments and complex interactions among market actors. Policy makers and practitioners need to take both a micro and macro view of financial risk, identify proper transparency requirements on complex instruments, develop dynamic models of information generation that best approximate observed financial outcomes, and identify and address the causes and consequences of systemic risk. Complexity analysis can make a useful contribution. However, the methodological suitability of complexity theory for financial systems and by extension for risk management is still

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 15
40 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096001

15

16

CHARILAOS MERTZANIS

debatable. Alternative models drawn from the natural sciences and evolutionary theory are proposed. Keywords: Complexity; finance; risk management

INTRODUCTION The need for improvement of risk management systems in financial institutions has been decisively emphasized by the recent international financial crisis. The calls for improvement extended not only to financial institutions but also to the regulators and central banks in dealing with systemic risk. The risk management issues that were called into question include the theoretical foundations of risk management, the statistical calibration of risk models, and the micro-prudential character of regulation. Risk management is concerned with rational decision-making under uncertainty. Asset prices, financial return probability distributions and market agent preferences form the core theoretical foundations of rational decision-making in an uncertain world. Prices, probabilities, and preferences are inextricably linked within an Arrow
Debreu general equilibrium model of pricing, hedging and trading financial assets. This model determines that asset prices are determined by a discount factor structure. Discount factors are traded by arbitrageurs in frictionless and costless markets. Actual discrepancies in predictions are explained by the introduction of frictions and imperfections (e.g., transaction costs, liquidity, financial innovation) in the price adjustment mechanism. Asset prices are assumed to fully incorporate the expectations and information of all market participants, making price changes random and therefore unpredictable (random walk) (Lo and MacKinley, 1999). Randomness is the outcome of many active infinitesimal arbitrageurs attempting to profit from access to information. Motivated by self-interest, arbitrageurs aggressively seize any informational advantage by incorporating their information into market prices and quickly eliminate profit opportunities. In frictionless markets and costless trading, information-induced price change occurs instantaneously, and thus prices are assumed to always fully reflect all available information and therefore no more profits can be generated from information-based trading. Further, the statistical calibration of frictionless risk management models relies heavily on historical time series and cross-section data on asset returns used to construct statistical measures of risk. This data exhibits non-stationarities that are difficult to model. Thus, to arrive in credible

Complexity Analysis and Risk Management in Finance

17

predictions, calibration models require personal judgment regarding nonquantitative risks. While longer and more detailed data series have helped the calibration process, the persistent non-stationarity reduces the overall benefit of data addition. Finally, the micro-prudential approach to financial regulation has assumed away the complexity of the financial system as a whole and the important role of systemic risk. Risk management systems of financial institutions were incentivized to operate within a given market environment, and disregarded the impact of interactions among institutions and between institutions and the financial system as a whole. Today, risk management practice has to operate within a complex financial environment. As a result of financial innovation and changing demand patters financial instruments have evolved to include complex and opaque products traded in less transparent market environments. Actual financial data patterns are not adequately approximated by normal (Gaussian) distributions thus restricting predictability. The links and interactions among financial institutions and between them and the financial system as a whole are complex and instrumental in determining the system’s behavior that feeds back on the performance of institutions. The elucidation of complex instrument valuation, information generation dynamic processes, and system-wide interactions among market actors could allow financial institutions to respond more efficiently to events that not only impact their own positions but also cause feedbacks between market actions and asset valuation thus affecting performance. Complexity analysis can be used to address these issues and inform risk management practice. Risk management research should seek to identify the ways in which complex instruments can be effectively included in the optimization exercise, the dynamic processes that best generate realistic information, and the core channels through which system-wide risk is generated and amplified. The purpose of this note is to explore how complexity analysis might be used to inform the management of financial risks in modern complex environments. It is argued that complexity characterizes financial instruments, financial processes and financial systems. Each type of complexity is examined and the implications for risk management are explored. This note aims at highlighting the basic directions for integrating complexity analysis and risk management theory. While the merits and application to finance of complexity theory are still a matter of debate, complexity analysis could contribute to the formation of a coherent body of propositions that are capable of better approximating reality in financial systems, that is, explain the stylized facts in finance. In what follows, the section “Complexity in

18

CHARILAOS MERTZANIS

Finance” analyzes the nature of complexity in finance, the section “Complexity and Risk” analyzes the implications of complexity in finance for risk management, and the section “Conclusions” concludes.

COMPLEXITY IN FINANCE If complexity analysis is to be applied for understanding financial behavior, the question arises as to whether the financial system is a complex system and its behavior can be depicted by evolutionary science. Arthur (1995, 1999) argues that the financial system is a complex one in at least five respects: financial markets are open, dynamic systems, far from being in equilibrium; agents are made up of heterogeneous agents, lacking perfect foresight, yet able to learn and adapt over time; agents interact through various more or less robust networks; macro patterns emerge from micro behaviors and interactions; and evolutionary processes create novelty generating growing order and complexity over time. In order to explain financial behavior, complexity analysis draws on insights from various approaches that challenge conventional economic thinking. These approaches originate in institutional economics, evolutionary biology and natural sciences. These views are echoed by numerous scholars. Their exposition is beyond the scope of this note. Complexity is often loosely defined, but in all instances is meant to convey a difficulty to understand or apprehend and thus to predict financial actions and outcomes. The relevant literature has broadly identified three dimensions of complexity: financial instrument complexity, financial process complexity and financial system complexity. These dimensions of complexity interact with each other in accordance with prevalent market structures, agent actions and models used. Each dimension is analyzed below.

Complex Financial Instruments The first dimension of complexity refers to financial instruments and their pricing. Complex financial instruments, such as asset-backed securities, collateralized debt obligations and credit default swaps, have attracted much of the blame for the recent crisis. The complexity and the interconnections
the long chains of claims embodied in the financial

Complexity Analysis and Risk Management in Finance

19

engineering, re-packaging, and synthetic derivatives made it almost impossible for people to understand and price risks. The chains of borrowing and lending made it difficult to trace interdependencies among counterparties and with respect to the amount and character of the collateral securing liabilities. This was a problem with respect not only to the underlying assets, but also to the various instruments used for refinancing along the chain, such as repurchase agreements and securities lending. Under these circumstances it’s not surprising that risk is not well priced. While these instruments are not overtly complex themselves, the methods for pricing them almost certainly are (Christophers, 2009). The substantial difficulty in pricing complex financial instruments even when buyers know almost all relevant information and use simple models of asset pricing, goes beyond the asymmetric information problem in financial markets. Regardless the sophistication of financial agents, in such circumstances rationality cannot be assumed. Understanding complex financial instruments and their impact on financial behavior at both the micro- and macro-level requires comprehensible pricing mechanisms and investment strategies. The increasingly complex chain of tranching and distributing risks that characterizes the structure of complex instruments makes the derivation of fundamental values and risk profiles of underlying assets hard or impossible to reconstruct even by sophisticated investors. Complex instruments deepen intricate links among assets and counterparties thus concealing agency costs and obscuring the underlying capitalist processes. Since the value of complex financial instruments depends on the complex interaction of numerous attributes of the constituent elements, the issuer can easily tamper such instruments without an investor or risk manager being able to detect it within a reasonable amount of time (Arora, Barak, Brunnermeiery, & Ge, 2009). Each new complex instrument can be an opportunity for the issuer to extract more fees and trading profits leading to the proliferation of these deals. Designers of financial products can rely on computational intractability to disguise their information via suitable cherry picking. They can generate extra profits from this hidden information, far beyond what would be possible in a fully rational setting. Thus, higher complexity and lower transparency typically enhance profit margins and therefore provide the issuer the incentive to keep the innovation cycle running (Tett, 2009). Complex financial instruments have weakened the role of personal judgment in risk management. The recurring issuance of complex financial instruments over the past decade made the rating business both more necessary for valuation and therefore more profitable and powerful. Despite

20

CHARILAOS MERTZANIS

their dubious quality, credit ratings benefited from the demand caused by complex instrument deals by creating an illusion of risk predictability. However, increased reliance by asset and risk managers on credit ratings proved crucially destabilizing since unrealistically high ratings inadvertently contributed to the build-up of systemic risk (Bank of International Settlements (BIS), 2009). Outstanding exposures on complex financial instruments have potential system-wide risk implications. Such instruments contribute to systemic risk in two ways. First, the imprecise valuation of riskiness of complex instruments complicates both internal risk management and the evaluation of aggregate exposures. A concentrated position or a series of counterparty relationships pose the risk of joint failures if market participants and regulators fail to understand and accurately value these instruments. Second, the imprecise valuation of complex instruments can exacerbate procyclicality. Market booms are characterized by financial innovations which tend to create hidden, underpriced risks. Institutions feel confident to experiment, creating new, untested instruments that are difficult to understand and value. Investors tend to be highly optimistic about future economic conditions without seriously considering the possible risks when markets deteriorate; institutions have little incentive to convince them otherwise. However, as the boom begins to wane, the unseen risks materialize, deepening the retrenchment that is already underway. Although financial innovation is a source of progress, it can also become a source of procyclicality that exacerbates system-wide risk.

Complex Financial Processes The second dimension of complexity refers to financial processes generating information on market outcomes. Observed financial time series in many markets over several decades have exhibited some puzzling empirical regularities (“stylized facts”) which proved difficult to model. Stylized facts are universal regularities, independent of time, place and composition details, and they are taken as benchmarks for theory testing. The puzzling statistical properties (“anomalies”) of financial time series are well known but they remain a puzzle for standard financial theory (Keim, 2008; Shleifer, 2000; Simon, 1955). These puzzles imply reduced predictability of financial returns. The most important stylized facts in finance which are relevant to complexity analysis include (Bouchaud & Potters, 2003; Cont, 2001; Sornette, 2003): (a) Fat tails: the distribution of returns of financial assets, evaluated

Complexity Analysis and Risk Management in Finance

21

at high frequencies, exhibits fourth moments (kurtosis) that are anomalously large when superimposed over a normal (Gaussian) or lognormal probability distribution. The latter is bell-shaped and assigns greater probability to events at the center (higher peaks) than at the extreme parts (narrow tails). Observed time series of financial returns display a significantly larger number of extreme events than a Gaussian process would predict. The standard theory of finance cannot explain fat tails as it relies on the normal distribution. This implies that massive fluctuations (disruptions or financial crashes) are assigned a diminishing small probability and therefore cannot be adequately predicted. (b) Volatility clustering: periods of intense fluctuations and mild fluctuations of financial returns tend to cluster together: big price changes of either sign follow big price changes, and little ones of either sign follow little ones (conditional heteroskedasticity of returns). The standard theory of finance cannot explain volatility clustering for the underlying Gaussian process of time series generation that predicts a uniform time distribution of both large and small fluctuations in returns. (c) Volatility persistence (“long memory”): financial returns are interdependent over time following a non-linear pattern. This means that return volatility exhibits slowly decaying autocorrelation rather than a quick decay to zero as the efficient market hypothesis would predict and the Brownian motion model would explain. (d) Interaction of volatility clustering and persistence: volatility persistence is related to volatility clustering. The clustering itself generates excess volatility (fat tails). Explaining the clustering and long memory most likely constitutes an explanation of the fat tails. Overall, financial return volatility changes by too much, too often, and with too much “order” to fit the geometric Brownian motion model used by the standard finance theory. The latter cannot explain the quantity and frequency of large crashes that have been witnessed in the recent decades because it assigns lower probabilities to extreme events. On account of the clustering and interdependence, time series of financial returns exhibit too much predictability and therefore the relevant data cannot be assumed to be generated through a random walk process. (e) Leverage effect: financial return volatility tends to increase when financial price drops, exhibiting negative skewness. The leverage correlation is moderate and decays much slower for individual stocks, while it is stronger but decays much faster for stock indices. (f) Increasing downside correlations: cross correlations increase in high volatility market conditions, in particular when prices drop significantly, without assuming that financial returns are time-dependent. In order to explain the puzzling stylized facts, some financial economists and risk managers have used empirical models without adequate

22

CHARILAOS MERTZANIS

theoretical grounding, whose main purpose was to replicate the statistical properties of observed data patterns. In an important effort to move away from the random walk hypothesis, risk analysis turned from the conditional distribution to the marginal distribution of asset returns. The traditionally used normal (Gaussian) distribution and its variants has the desired property of being stable, but it does not adequately explain fat tails, especially in abnormal market conditions, which accord well with observed financial data patterns.

Complex Financial Systems The third dimension of complexity refers to the structure and behavior of the financial system as a whole. The financial crisis has highlighted the need to look at the links and connections between financial institutions and markets and the financial system as a whole. Failure of certain institutions and/ or major disruption in certain markets can rapidly spill over to other institutions or markets and eventually to the whole financial system. A remote triggering event, such as a financial institution failure or a market disruption, can cause a widespread disruption of the financial system as a whole, creating significant problems in otherwise viable institutions or markets (Brunnermeier et al., 2009; International Monetary Fund, 2009). The financial system may prove more or less resilient to contagion depending on the nature of major triggers and prevalent channels of contagion. The crisis has shown that an apparently robust financial system may in fact become fragile. This results from the large number of interconnections within the financial system that serve as shock amplifiers rather than shock absorbers. Financial institutions, linked through the interbank market, payment systems, monoline insurers and custodian banks, are financial networks with strong degree of interconnectivity and therefore systemically important. Understanding network structures is crucial for the identification of systemically important institutions and markets. The resilience of the financial system as a whole depends on proper maintaining of individual institutions’ liquidity buffers and capital reserves as well as on controlling large exposures and addressing interdependencies. A particular financial institution might not only be critical to the normal functioning of financial markets or infrastructures because other institutions are financially exposed to it, but also because other market agents rely on the continued provision of its services which will not be forthcoming. Thus, the impact of a failure of a given institution

Complexity Analysis and Risk Management in Finance

23

or market also hinges on the ability of the financial infrastructure to support its resolution and to facilitate the orderly unwinding of financial exposures. Modern financial systems are characterized by complexity and homogeneity (Haldane, 2009). Complexity means that the financial system is characterized by an increasingly knotted and uneven interconnectivity; more financial institutions do more business deals with more counterparties on a global scale (Fig. 1). Homogeneity means that the financial system becomes more adaptive since behavior is driven by optimizing agents who herd and blindly jump on the next big opportunity so long as their peers are profiting without regard to the negative impact of their move on the system as a whole. Being adaptive means converging: financial institution balance sheets grew all alike; their risk models were more and more assimilated and the associated risk management strategies grew alike; strategic behavior grew alike; financial regulation grew alike through unified rulebooks regarding market operations characterized by free mobility of financial capital. Financial institutions looked alike and responded alike. Most market participants have instant electronic access to risk-return data for financial assets worldwide out of Bloomberg as everyone else. Diversification strategies by individual firms generated a lack of diversity across the system as a whole. The common pursuit of return and the uniform risk management practices explain the reduction in diversity in the financial sector. Financial institutions were racing for return on equity, which led them to pursue high-yield activities. The result was that business strategies were replicated across the financial sector. Simultaneously, risk management models became homogenous, in part because credit ratings were hardwired into regulation and Basel II provided the same rules for everyone. The consequence was a highly homogenous financial system that was less resistant to aggregate shocks, the same as ecosystems where diversity is lower. Markets segmentation is disappearing, investment behavior becomes more and more alike decided by ever larger pools of institutional players worldwide. The combination of complexity and homogeneity of a financial system causes fragility and instability. In complex financial systems, scaling up risks may result in building up “error cascades.” The reason is “cross-contamination.” As losses build up, links and interconnections serve as shock amplifiers, not shock absorbers. While the system is mostly self-repairing, it also exhibits a knife-edge property which under growing homogeneity and complexity bears the danger of collapse (Persaud, 2000). Financial institutions are interconnected, making them ideal candidates for risk

24

CHARILAOS MERTZANIS Chart 1: Global Financial Network: 1985

Chart 2: Global Financial Network: 1995

Chart 3: Global Financial Network: 2005

Fig. 1. The Increasing Complexity of Bank Interactions Over Time. Note: The Global Financial Network over Time. Source: Taken from Haldane (2009).

Complexity Analysis and Risk Management in Finance

25

contamination. The biggest, most complex and best connected ones are systemically important with a high capacity to infect counterparties. Complexity and homogeneity in market behavior undermine the stability of the financial system. While market agents start off by exhibiting heterogeneous behavior, exogenous shocks may eventually drive them to homogenous reaction. In this regard, risk is amplified endogenously and the initially robust financial system turns out to become fragile. The endogeneity of risk and the rising system fragility are the result of the following factors (The Warwick Commission, 2010): the increasing reliance of asset valuation and risk assessment on market prices (mark-to-market valuation); the extensive reliance of funding on leverage; and the tendency of regulators and practitioners to consider risk as one thing, to be treated the same way and measured as the volatility of short-term prices. But risk is not one thing alone, there are different types of risk: credit risk, liquidity risk and market risk, mutually distinct and interacting with each other. Different types of risk should be hedged differently. Credit risks are best hedged by finding uncorrelated or negatively correlated credits. Liquidity risks are best hedged across time: the more time you have before you have to sell an asset, the more you can hold assets that are hard to sell quickly. Market risks, like equity values, are best hedged using a combination of time and diversification. Today, the risk-return trade-off is more easily observed but less easily explored. Stabilizing arbitrage opportunities tend to vanish (Gromb & Vayanos, 2010; Shleifer & Vishny, 1997). Based on the wealth of information readily accessible by everyone, financial assets that appear to offer a slightly higher return than past risk trade-off patterns are identified almost simultaneously by all interested traders. Everyone rushes in at the same time and the asset quickly becomes overvalued, leading to an increase in volatility, which in turn raises the risk profile of the asset, directing risk models to confirm the rise in volatility, thereby inducing everyone to sell, with the result of creating more volatility. The tendency toward the cliff edge is stronger when the homogenizing behavior of markets is coupled with strategic behavior (Persaud, 2003). Strategic behavior can be understood by reference to Keynes’s example of “beauty contest”: market behavior is driven by what investors think about average market beliefs on average market beliefs and so on. Traditional risk models do not capture strategic behavior, since risk calculations are based on Black
Scholes
Merton arbitrage-driven models of asset pricing which treat individual investment behavior as an independent atomistic activity regardless of equilibrium conditions, that is being unrelated to the

26

CHARILAOS MERTZANIS

actions of others. Once strategic behavior is taken into account, asset prices can then be shown to deviate significantly from competitive market prices (Allen, Morris, & Shin, 2006). Complexity and homogeneity make inherent risk-return characteristics of financial assets difficult to infer. As a result, an investor cannot credibly estimate, on the basis of optimization of the mean-variance relationship, the probability of loss of an investment. Risk is not a statistical but a behavioral metric. Thus, professional risk management cannot be driven by personal and detached views of market conditions in an effort to locate assets with better risk-return characteristics.

COMPLEXITY AND RISK Managing Complex Financial Instruments If financial innovation is to be useful in providing flexibility in financial markets, two conditions must be met: (a) underlying assets must have appropriate credit quality with sufficient historical records of defaults and a lucent relationship to macroeconomic developments and (b) complex reengineering structures must be avoided for they lead to more opaqueness and vulnerability to macro shocks and therefore to ambiguity in pricing. Transparency about complex instruments and structures is crucial. However, for some instruments, disclosed information, even if kept up with market needs, would have been futile since instruments are so complex that the required information to appropriately monitor risks may be overwhelmingly large. Excessive information overload may limit the effectiveness of disclosure. This possibility points to a need for controlling complexity and for encouraging appropriate design of disclosure through a good summary of properties and clarity about the assumptions in valuation. Further, disclosed positions must adhere to consistent standards of information aggregation. Complex instruments are largely sold in different private markets characterized by varying information. In stressed situations, the undisclosed positions of risky complex instruments raise counterparty risk and ambiguity with regard to the trading volume deteriorating liquidity, particularly in over-the-counter markets. In the respect, three core policy issues arise, relating to the proper content, extent, and manner of information disclosure.

Complexity Analysis and Risk Management in Finance

27

Proper identification of significant information is required. Transparency rules must identify synthetic risk indicators, based on quantitative metrics which are robust, objective, and backward verifiable. Transparency concerns must focus not only on information per se but also on its proper oversight through standardized risk representation and cost profiles to be monitored by proper allocation of regulatory oversight. The traditional description of financial risks seems inadequate in allowing informed investment decisions and effective risk management in a context where the integration of financial markets, instruments, and participants often makes it difficult to separately analyze different risks. The focus should, instead, be on the measurement and monitoring of the overall synthetic risk profile of complex instruments. Complex financial instrument risk must be classified irrespectively of the ways used for the public offering of such instruments and independently from the heterogeneity in instrument, issuer, distribution channel, cost, and the underlying engineering process. The classification must be strictly linked to the type of the underlying financial structure and subsequently to the concept of liquidability of financial instruments, that is, to the possibility of disinvesting at a specific time without incurring a loss or waiving the benefits in the form of abnormal returns. An interesting way to do this could be to relate financial instruments, particularly non-equity ones, with the underlying financial structures within which they have been created (Minnema, 2011). Financial structures could be effectively classified into three types: risk-target instruments, benchmark instruments, and return-target instruments. Risk-target instruments reflect financial investments aiming at the over-time optimization of a given target in risk exposure. Accordingly, ex ante minimum and maximum thresholds are defined as boundaries of a risk measure and serve as reference point for risk-taking decisions. Benchmark instruments reflect investments anchored to a benchmark for either active or passive asset management. Thus, the composition of the asset portfolio differs from that of the benchmark depending on the manager’s specific objectives, or the instrument is essentially a replica of the benchmark. Return-target instruments reflect financial investments embodying engineering techniques aimed at pursuing a minimum target return. This classification includes all financial instruments resulting from a static/dynamic combination of low- and high-risk assets, the actual choice of which depends on investment policy considerations. Any combinations of protection mechanisms and guarantees that may exist affecting the instruments’ overall risk profile, cost and investment horizon, do not affect the

28

CHARILAOS MERTZANIS

underlying structure and therefore their classification. The latter, by taking into consideration the interaction between the various risk factors and the financial engineering techniques, could make possible not only better instrument riskiness oversight but also lower market fragmentation. Fragmentation results from the different regulation of complex financial instruments with the same underlying financial engineering and the improper allocation of supervisory authority, which often ignores the common financial structures underlying opaque instruments offered by different issuers. Enhanced transparency on the risk profile of financial instruments can obtain on the basis of three stages, corresponding to three synthetic indicators calculated in accordance with the classification above. First, probability scenarios of returns of a financial investment are calculated at the end of a given time horizon, making possible the formation of price of a financial instrument at the subscription date and the provision of clear and concise information on expected investment returns and costs. Comparing this information with the returns of a risk-free asset over the same time period allows for a better assessment of the instrument’s performance risk. Second, a synthetic indicator of the financial instrument’s risk profile can be calculated on the basis of stochastic limit analysis of return volatility, allowing the emergence of each instrument’s initial and evolving degree of risk, consistent with the risk profile reflecting its financial engineering and investment policy. Third, the chosen investment time horizon is calculated as an indication of the optimal investment holding period on the basis of the instrument’s underlying financial structure, related risks, and costs. The resulting time horizon is important not only for transparency but also for suitability purposes as it specifies the time period for investment liquidation based on the probability scenarios. Thus, investor choices in relation to the optimal investment time period implicit in the underlying financial engineering provide the basis for the standardization of risk representation and cost profile of a complex financial instrument, which allows an inference of its future performance. However, identifying the content of information is not enough. The timely and proper collection of information about complex financial instruments and the proper release of aggregate information are important. Both whether and how public information is released matters. Disseminating public information may increase or decrease social welfare depending on agents’ access to information, strategic complementarities in decisionmaking and heterogeneous beliefs (Morris & Shin, 2007). The release of public information is beneficial only if it is sufficiently accurate and comprehensible.

Complexity Analysis and Risk Management in Finance

29

Managing Complex Financial Processes Analytical models that strive to predict the dynamics of complex processes often include a term representing a random (stochastic) factor. In order to explain stylized facts in financial markets, models deploy random terms within an analytical structure involving fundamentals, exogenous rules/constraints, equations, and interactions. A rough taxonomy of the dynamics processes used to reproduce the stylized facts is provided in Fig. 2. Dynamic processes are divided into static and dynamic ones. The dynamic ones are divided into deterministic and non-deterministic ones (Sprott, 2003). Deterministic processes behave according to specified rules or equations that determine the next state of the process based on the current state of the process (i.e., a rule might be to always buy/sell all financial assets included in your portfolio only when all your interacting traders are buying/selling their financial assets: if this rule and the current state of uncertainty in the market are known, then the next state of market uncertainty can be predicted and the suitable risk management practice identified). Deterministic processes can assume either a linear (periodic) or a non-linear (chaotic, non-periodic) form. On the one hand, linear deterministic processes can be simple or complex comprising many interacting sub-units. But every linear process is essentially modular, that is, it can be analyzed by breaking it down into sub-units and measuring each sub-unit’s impact separately. A linear deterministic process is no more or less than the sum of its sub-units. The outcome of linear process operation is regular or periodic, but not cumulative. On the other hand, non-linear deterministic processes are not modular, that is, they cannot be analyzed by being broken Static

Steady state Linear (periodic form) Deterministic Non-linear (chaos) (non-periodic form)

Dynamic Brownian (“mild”) Non-deterministic (random) “Wild” (Mandelbrot)

Fig. 2.

Taxonomy of Dynamic Processes.

30

CHARILAOS MERTZANIS

into separate sub-units. Integral to the process is cooperation among, or competition between, determining forces making the non-linear processes always more or less than the sum of its sub-units. Non-linear processes are associated with the emergence of new forms, patterns or behaviors that did not exist in the initial environment; they are capable of generating nonperiodic patterns whose trajectory is unrepeated. The behavior of nonlinear systems might be predicted only in the very short term. Assuming perfect knowledge of the process’s governing forces at time one, prediction of the process’s evolution can at best be made for time two, but not possibly for time three and beyond. Moreover, the ability to predict declines as the number of iterations of the process increases. Thus in practice the outcome of a fully deterministic process can be (often is) unpredictable. Non-deterministic processes exhibit state-to-state independence. Nothing in the course of the process at time one will determine its course at time two. This state-to-state independence is generally known as “randomness.” Non-deterministic dynamic processes can assume either a mild (Brownian) or a wild (Mandelbrot-like) motion. However, the use of Brownian dynamics in finance has not produced a clear structure or pattern of behavior of financial returns. The stylized facts generated by this type of random dynamics appear geometrically as some sort of smear across space (Mandelbrot, 1997). The randomness term can be modeled in many forms, one of which is through a Markov process. The latter underlies the development of the Efficient Capital Markets Hypothesis. The reliance on random terms might be partly explained by the limitations of available calculation techniques in current modeling. Models often try to explain complex phenomena by including the standard stochastic term (that is, assuming unpredictability) into the equation which generates the sort of “surprising” stylized facts observed in financial market activity. For decades, the standard stochastic terms have performed well (due to heterogeneity and simplicity assumptions underlying the smoothly operating markets). The importance for financial systems of deterministic versus nondeterministic dynamics can be assessed by their ability to generate predictability and their capacity to generate complex forms and patterns. On the one hand, the static and the linear deterministic processes can generate predictable results, while non-linear deterministic processes can do that under strict conditions. The non-deterministic (random) processes cannot generate predictable results. On the other hand, linear processes can produce some interesting behavior but they cannot generate most stylized facts in finance worth studying. Complex forms and patterns can be generated by non-linear deterministic processes.

Complexity Analysis and Risk Management in Finance

31

Modeling financial behavior has largely relied on linear mathematics with the addition of stochastic terms, that is, by utilizing some linearity with some “mild” Brownian randomness. Perhaps, actual financial behavior may be better understood by the use of non-linear chaotic system analysis with a dash of “wild” randomness possessing a fractal quality and adhering to power-law distributions (Mandelbrot, 1997). The “wild” Brownian distribution exhibits both infinite variance and some dependence (long memory) of returns. The relative merits of the different dynamic processes in approximating actual financial outcomes depend on whether financial systems are complex systems properly defined. The issue has been of particular interest to physicists. Borland (2005), Stanley, Amaral, Gopikrishnan, Plerou, and Rosenow (2001), and Johnson, Jefferies, and Pak (2003) argue, among others, that the behavior of financial markets is one of the most vivid examples of complex system dynamics. Financial market processes are assumed to follow “power laws” observed in nature.1 A series of interesting models have emerged, using the invariance principles of statistical physics to deduce scaling properties in tail probabilities (Mandelbrot, 1997; Mantegna & Stanley, 2000) as well as large-deviation theory and extreme-value theory to estimate loss probabilities (Embrecths, Kluppelberg, & Mikosch, 1997; McCulloch, 1996). Physicists have used the universality property of physical processes as their basis for constructing realistic models of financial return distributions. The stylized facts are persistently observed across time and diverse financial markets, which suggests that some other common forces are at work beyond fundamentals. This observation led physicists to believe that these processes can be approximated by the theory of natural phenomena and the stylized facts can therefore be understood as emergent statistical properties of complex financial processes universally explained by power laws. The existence of a powerlaw distribution of financial returns is taken to imply underlying complexity in the financial system that generates this distribution, and be affirmed by the tools provided by the theory of critical phenomena, turbulence, phase transition, and condensed matter physics. These tools provide not only the required mathematical concepts but also the justification of the natural underpinnings of economic phenomena. On the other hand, while not fundamentally denying that financial markets are complex systems, Durlauf (2005) argues that the empirical evidence regarding the application of power laws in financial markets does not produce indisputable results in favor of complexity. Pisarenko and Sornette (2006) show that the use of the power-law model at best provides an

32

CHARILAOS MERTZANIS

approximation of the behavior of financial returns and may not be extended into the unobserved regions of the return distribution’s tail. Brock (1999) stresses that power laws may underdetermine the stochastic processes in financial markets, since one and the same power law should not be compatible with multiple financial return distributions. Newman (2005) argues that power laws can also result from factors other than underlying complexity. It is still debatable whether the theory of critical phenomena and power laws can provide a universal mechanism for explaining the stylized facts in finance under different conditions. Nonetheless, alternative models have emerged that generate probability distributions which better approximate the stylized facts than does the normal (Gaussian) one used in standard risk management models (Chakrabarti, 2010). These models are a promising avenue for exploration by risk management practitioners.

Managing Complex Financial Systems Network Topology The understanding of how complexity affects the behavior of the financial system requires the understanding not only of dynamic behavior but also of the structure of links and interconnections among financial institutions and markets (Latora & Marchiori, 2004). Complex systems research needs to consider the structural features (topology) of financial networks rather than merely focus on the specific form of the non-linear interactions between individual sub-units. Network topology is analyzed by the use of graph theory. A graph is composed of vertices or nodes, and lines (edges) that connect the nodes (Fig. 1). A graph with weighted edges is called a network. Network structures vary considerably, and those most useful for financial analysis are complete, random, scale-free, and hierarchical networks. Network analysis can be effectively used to analyze financial behavior of interconnected institutions and markets in a complex financial system (Soramaki, Bech, Arnold, Glass, & Beyeler, 2007). Financial networks consist of a collections of nodes (financial institutions) and links between nodes (credit and financial relationships: assets and liabilities) affecting the attributes of the nodes (i.e., an institution’s balance sheets is affected by existing links with other institutions), and the structure of the links affects the performance of the financial system as a whole. Network analysis looks at the structure of the links and the manner in which the structure affects the

Complexity Analysis and Risk Management in Finance

33

performance of the financial system as a whole. It includes three main areas of concern: the structural properties of a network (distribution of node degrees, diameter of the graph) so as to produce the appropriate graphs for the various domains in finance (different financial systems); the calculation of measurable quantity of flows within the network (financial asset/liability transfers); and the dynamical properties of network structure. The generation of actual data pattern depends on both the graph structure and the algorithm used for manipulating the graph. Certain common properties shared by large and complex financial networks are of particular interest for financial stability. These are as follows: (a) Financial networks can be scale-free networks, that is, systems where the probability of observing a node (financial institution) with a strong connection (high number of links
financial hub) is very low, while the probability of observing a node with many (weak) connections is very high. Given this structure of connections, a random removal of a node with a strong connection (a failing systemically important financial institution) can spread to other nodes with many (even if weak) connections throughout the system and result in the turning of an initially robust system into a fragile one (“robust yet fragile” thesis). (b) The relative strength and number of the links (weak but many vs. strong but few), in terms of availability/dissemination of information, shapes the topology of the financial network. (c) The character of network “homophily” shows the extent of clustering among nodes (financial institutions). (d) The character of network intermediation structure (“small world phenomenon”) showing the number of links covering the distance between any two nodes, affects the likelyhood of high-low contagion in small financial networks, since the number of affected nodes above which epidemics propagate system-wide is especially low (Bech et al., 2009). Another crucial characteristic of the network structure is the centrality of the nodes (i.e., the relevance of the position of a node in the network) (Borgatti, 2005). Centrality may be measured by the number of links that terminate upon a node (in degree), by the distance from other nodes (closeness), or by the existing connections to central nodes. A measure of centrality particularly suitable for financial networks is the betweenness centrality of a node, defined as the number of shortest paths that go through the node. Centrality might give an insight into which nodes should be considered of systemic importance. However, the impact of network properties and the effectiveness of centrality measures depend on the behavior of the nodes in each case. These properties face limitations for they do not adequately capture all complex behavior.

34

CHARILAOS MERTZANIS

Direct and indirect interlinkages and contagion dynamics among financial institutions, as well as between institutions, markets and infrastructures, can be influenced by three important network characteristics: the degree of connectivity, the degree of concentration, and the size of exposures (Caballero & Simsek, 2009; European Central Bank, 2010; Haldane, 2009). The strength of the various shock amplification mechanisms in the web of financial connections depends on the size of aggregate macroeconomic shocks, asset price volatility, liquidity risk, and financial leverage. Moreover, network analysis can be used to simulate the effect of credit and funding shocks on financial stability by taking into account not only direct balance sheet exposures but also the impact of contingent claims and credit risk transfer techniques. Complexity is manifested through four mechanisms: connectivity, feedback, uncertainty, and innovation. Within a certain range, connections help absorb shocks, but beyond that range connections are shock amplifiers. Connected networks exhibit long tails in the degree distribution, which is the distribution of the number of links per node. Long-tailed distributions are more robust to random disturbances, but more susceptible to targeted attacks. In particular, if a large financial institution is subject to stress, the effects are more likely to spread through the network. Connected networks may exhibit the “small world” property, where few steps exist between any two nodes. A key node can introduce shortcuts, making it more likely that a local problem becomes a global one. These ingredients together make the financial network into a usually robust but potentially fragile network. Under these conditions, the impact of a shock depends on the behavioral responses of agents within the network. A “hide” response (hoarding of liquidity for self-protection) tends to contain the problem locally, whereas a “flight” response (inability to fund their positions leads to fire sales) tends to propagate fragility and aggravate the instability problem. Both responses are rational from the individual institution’s perspective, but can have severe system-wide implications. The ensuing network uncertainty increases counterparty uncertainty. An institution can enter into a contract with a counterparty that it can monitor. But if this counterparty sells the instrument to another institution, then it becomes harder to monitor the creditworthiness of the new counterparty. This becomes even harder once the counterparty has its own counterparties. Uncertainty about the network structure has pricing implications which increase with the expanded dimensionality of the network. Thus, modern risk management should strive to integrate financial network characteristics in the analysis.

Complexity Analysis and Risk Management in Finance

35

Endogenous Risk The standard risk management model divides risk into idiosyncratic and systemic components. Within this model, markets are assumed to operate efficiently
that is, they operate on the basis of a “law of conservation of risk.” The latter states that financial institutions are efficiently allocating risk throughout the system, and risk is neither created nor destroyed but merely shuffled around. Risk follows a dynamic process in which it is destroyed and created in the course of trading activity. Shifting risk may allow for more efficiency in terms of costs to market agents, but what may be lacking in standard risk models is the notion that system-wide risk is more than the sum of its idiosyncratic parts, thus justifying its complex nature. Financial markets today represent an environment in which traders react to what’s happening around them and their reactions shape the realized outcomes. Whenever there is a conjunction of both participants (traders) reacting to their environment (markets) and participants actions affecting their environment, risk is endogenous (Shin, 2008). A sudden, exogenously caused drop in asset prices brings traders closer to their trading limits thereby forcing them to sell, which sets off further downward pressure on asset prices, causing a new round of selling and so on. The downward spiral in asset prices is endogenous, generated within the financial system. Any marketsensitive management of risk will consequently have destabilizing effects. Examples of endogenous risk include the highly turbulent global market reaction after the terrorist 9/11 attacks in the United States; the destabilizing feedback effects on market dynamics of concerted selling pressure arising from mechanical portfolio insurance and dynamic hedging trading rules after the 1987 market crash; the unprecedented asset price movements that followed the collapse of the LTCM of 1998 which forced leveraged traders to face urgent margin calls, leading to a large unwinding of their leveraged positions, thereby causing asset price falls leading to further distress, more margin calls and so on; the large drop of the USD:JPY exchange rate in October 7
8, 1998 that resulted from an initial modest decline in the exchange rate which set off an unprecedented unwinding of JPY carrytrades in place at the time exacerbated by stop-loss orders and the associated unwinding of traders’ hedged positions; and, of course, the recent 2007
2008 financial crisis. In all these episodes, the mutually reinforcing asset sales showed that the harder market participants tried to get away, the more they provoked the self-feeding frenzy. These episodes demonstrate that financial crashes and collapses are not random or deterministic events, nor can they generally be depicted by Markov approximations (Persaud, 2008). Financial crashes are not “once

36

CHARILAOS MERTZANIS

in a thousand-years” events as standard VaR measures would predict, but occur every five to six years during the last three decades. They always follow historically specific, man-made financial booms, which occur because people are making investments that they believe to be “safe” but instead lead to hidden risks, often coupled with excessive leverage, made possible by prevalent monetary conditions. The management of risk of a crisis presupposes the management of the preceding financial boom. The credit mistakes that lead to crashes are not made in the crash, but during the preceding booms. The fundamental problem of crashes is that risks are underestimated in the boom and overestimated in the crash in a cumulative manner. This is not simply a result of investor irrationality but rather an inherent future of how financial systems function. If the underlying uncertainty facing a trader were exogenous, modeling financial risk may be akin to a gambler facing a spin of a roulette wheel, where the bets are placed by him/her and other gamblers do not affect the outcome of the spin. Current risk management practices presuppose a roulette view of uncertainty, whereby the roulette has a large number of outcomes with different probabilities. As long as these probabilities are unaffected by the other gamblers’ actions, the prediction of these outcomes and their respected probabilities can result from applying sophisticated statistical techniques to past outcomes. Current risk management responds by applying more and more refined and sophisticated statistical techniques for tracking the non-linear payoff structures arising from derivative instruments. However, to the extent that the stochastic (random) process assumed to govern asset price reactions depends on what other traders do, the prediction of possible outcomes cannot be made. The uncertainty facing traders is endogenous and depends on the actions of all market participants. Accordingly, endogenous risk is not compatible with the “law of conservation of risk” whereby the total inflow of risk in a financial system must equal the total outflow of risk from the system, plus the change in the risk contained within the system. Endogenous risk means that, under certain conditions, financial risk can be created internally and amplified and not merely transferred from one form or person to another. The endogeneity of risk should alert regulators and risk management practitioners into modifying existing measures of risk so as to make them more robust as well as use stress-testing and back-testing techniques. Regulators are moving into the adoption of a macro-prudential approach to the financial system regulation. Both regulators and risk practitioners must require financial institutions to revise and revalidate their risk models to include scenarios previously considered “extreme” or “unexpected” in

Complexity Analysis and Risk Management in Finance

37

their normal risk calculations. Risk management practitioners need to produce coherent measures or risk (Artzner, Delbaen, Eber, & Heath, 1999). “Conditional VaR” (CoVaR), “extreme value,” “expected shortfall,” “expected regret,” and “maximum drawdown” are some alternative measures of risk devised to account for risk endogeneity and the interdependencies between financial institutions and the financial system (Mertzanis, 2013). However, their implementation in risk management practice is awaiting and their efficiency remains to be proved.

CONCLUSIONS Standard financial risk management practices proved unable to provide an adequate understanding and a timely warning of the financial crisis. In particular, the theoretical foundations of risk management and the statistical calibration of risk models are called into question. Risk management research should strive to identify and address new sources of financial risk, and focus more on diversity (in contrast to homogeneity) and less on mere diversification. Policy makers and practitioners need to take both a micro and macro view of financial risk. Instead of focusing on idiosyncratic risk alone, they should concentrate on accurate price discovery for complex instruments, realistic financial information generation processes, and system-wide risk materializing within complex financial networks. As financial markets satisfy reasonable criteria of being considered complex adaptive systems, complexity analysis can make a useful contribution. However, the methodological suitability of complexity theory for financial systems and by extension for risk management is still debatable. Alternative models drawn from the natural sciences and evolutionary theory are proposed. Today, complexity and homogeneity are important characteristics of financial behavior. Their combination makes financial risk to be endogenously created and therefore not subject to the natural law of conservation. This has important implications for the extension of the laws of nature into the man-made financial activity.

NOTE 1. In statistics, a power law is a functional relationship between two magnitudes, where one magnitude varies as a power of another. A power-law probability

38

CHARILAOS MERTZANIS

distribution is a distribution whose density function (or mass function in the discrete case) has the form: p(x) ≈ L(x)x−α, where α > 1 and L(x) is a slowly varying function which satisfies limx→∞ L(tx)/L(x) = 1 with t constant and t > 0.

REFERENCES Allen, F., Morris, S., & Shin, H. S. (2006). Beauty contests and iterated expectations in asset markets. Review of Financial Studies, 19(3), 719
752. Arora, S., Barak, B., Brunnermeiery, M., & Ge, R. (2009). Computational complexity and information asymmetry in financial products. Princeton, NJ: Princeton University. Arthur, W. B. (1995). Complexity in economic and financial markets. Complexity, 1(1), 20
25. Arthur, W. B. (1999). Complexity and the economy. Science, 284, 107
109. Artzner, P., Delbaen, F., Eber, J. M., & Heath, D. (1999). Coherent measures of risk. Mathematical Finance, 9, 203
228. Bank of International Settlements. (2009). Stocktaking on the use of credit ratings. Basel. Retrieved from http://www.bis.org/press/p090615.htm. Accessed in June. Bech, M. L., Beyeler, W., Glass, R. J., & Soramaki, K. (2009, October 5). Network topology and payment system resilience. Paper presented at Joint D-FS/DG-P Workshop on recent advances in modeling systemic risk using network analysis, ECB, Frankfurt am Main. Borgatti, S. (2005). Centrality and network flow. Social Networks, 27, 55
71. Borland, L. (2005). Long-range memory and non-extensivity in financial markets. Europhysics News, 36, 228
231. Bouchaud, J.-F., & Potters, M. (2003). Theory of financial risk and derivative pricing: From statistical physics to risk management. Cambridge: Cambridge University Press. Brock, W. A. (1999). Scaling in economics: A reader’s guide. Industrial and Corporate Change, 8(3), 409
446. Brunnermeier, M., Crockett, A., Goodhart, C., Persaud, A. D., & Shin, H. (2009). The fundamental principles of financial regulation. Geneva Reports on the World Economy No. 11. Geneva: International Center for Monetary and Banking Studies (ICMB). Caballero, R. J., & Simsek, A. (2009). Complexity and financial panics. Working Paper No. 09-17. MIT Department of Economics. Chakrabarti, B. K. (2010). Fifteen years of econophysics research. Science and Culture, 76(9), 293
296. Christophers, B. (2009). Complexity, finance, and progress in human geography. Progress in Human Geography, 33(6), 807
824. Cont, R. (2001). Empirical properties of asset returns: Stylized facts and statistical issues. Quantitative Finance, 1(2), 223
236. Durlauf, S. (2005). Complexity and empirical economics. Economic Journal, 115(504), 225
243. Embrecths, P., Kluppelberg, C., & Mikosch, T. (1997). Modelling extreme events for insurance and finance. Berlin: Springer. European Central Bank. (2010). Recent advances in modeling systemic risk using network analysis. Retrieved from http://www.ecb.europa.eu/pub/pdf/other/modellingsystemicrisk012010en.pdf. Accessed in January.

Complexity Analysis and Risk Management in Finance

39

Gromb, D., & Vayanos, D. (2010). Limits of arbitrage: The state of the theory. Annual Review of Financial Economics, 2, 251
275. Haldane, A. (2009). Rethinking the financial network. Speech delivered at the Financial Student Association. Amsterdam. Retrieved from http://www.bis.org/review/r090505e.pdf. Accessed in April. International Monetary Fund. (2009). Assessing the systemic implications of financial linkages. Global financial stability report. Washington, DC: IMF. Johnson, N. F., Jefferies, P., & Pak, M. H. (2003). Financial market complexity: What physicists can tell us about market behavior. Oxford: Oxford University Press. Keim, D. B. (2008). Financial market anomalies. In S. N. Durlauf & L. E. Blume (Ed.), The new Palgrave dictionary of economics (2nd ed.). London: Macmillan. Latora, V., & Marchiori, M. (2004). The architecture of complex systems. In M. Gell-Mann & C. Tsallis (Eds.), Nonextensive entropy-interdisciplinary applications. Oxford: Oxford University Press. Lo, A. W., & MacKinlay, C. (1999). A non-random walk down wall street. Princeton, NJ: Princeton University Press. Mandelbrot, B. (1997). Fractals and scaling in finance. New York, NY: Springer-Verlag. Mantegna, R. N., & Stanley, H. E. (2000). Introduction to econophysics: Correlations and complexity in finance. Cambridge: Cambridge University Press. McCulloch, H. (1996). Financial applications of stable distributions. In G. Maddala & C. Rao (Eds.), Handbook of statistics (Vol. 14). Statistical Methods in Finance. Amsterdam: Elsevier Science. Minnema, M. (2011). A quantitative framework to assess the risk-reward profile of non-equity products. London: Risk Books. Mertzanis, Ch. (2013). Risk management challenges after the financial crisis. Economic notes, 42(3), 285
319. Morris, S., & Shin, H. S. (2007). Optimal communication. Journal of the European Economic Association, 5(2
3), 594
602. Newman, M. E. J. (2005). Power laws, Pareto distributions, and Zipf’s law. Contemporary Physics, 46(5), 323
351. Persaud, A. (2000). Sending the herd off the cliff edge: The disturbing interaction between herding and market-sensitive risk management systems. Journal of Risk Finance, 2(1), 59
65. Persaud, A. (2003). Liquidity black holes: Understanding, quantifying and managing financial liquidity. London: Risk Books. Persaud, A. (2008). Valuation, regulation, liquidity. Financial Stability Review, Bank of France, 12, 75
83. Pisarenko, V., & Sornette, D. (2006). New statistic for financial return distributions: Power law or exponential? Physica A, 366, 387
400. Shin, H.-S. (2008). Risk and liquidity, clarendon lectures in finance. Oxford: Oxford University Press. Shleifer, A. (2000). Inefficient markets: An introduction to behavioral finance. Oxford: Oxford University Press. Shleifer, A., & Vishny, R. W. (1997). The limits of arbitrage. Journal of Finance, 52(1), 35
55. Simon, H. (1955). A behavioral model of rational choice. Quarterly Journal of Economics, 69(1), 99
118. Soramaki, K., Bech, M. L., Arnold, J., Glass, R. J., & Beyeler, W. E. (2007). The topology of interbank payment flows. Physica A, 379, 317
333.

40

CHARILAOS MERTZANIS

Sornette, D. (2003). Why stock markets crash: Critical events in complex financial systems. Princeton, NJ: Princeton University Press. Sprott, J. C. (2003). Chaos and time-series analysis. New York, NY: Oxford University Press. Stanley, H. E., Amaral, L. A. N., Gopikrishnan, P., Plerou, V., & Rosenow, B. (2001). Quantifying empirical economic fluctuations using the organizing principles of scale invariance and universality. In H. Takayasu (Ed.), Empirical science of financial fluctuations: The advent of econophysics (pp. 3
11). Berlin: Springer-Verlag. Tett, G. (2009). Fool’s gold. New York, NY: Free Press. The Warwick Commission on International Financial Reform: In Praise of Unlevel Playing Fields. (2010). Report, University of Warwick.

THE EFFECTS OF MACROECONOMIC NEWS ANNOUNCEMENTS DURING THE GLOBAL FINANCIAL CRISIS Pilar Abad and Helena Chulia´ ABSTRACT In this chapter we investigate the response of bond markets to macroeconomic news announcements in the euro area. Specifically, we analyze the impact of (un)expected changes in the interest rate, unemployment rate, consumer confidence index and industrial production index on the returns, volatility and correlations of European government bond markets. Overall, our results suggest that, bond return volatility strongly reacts to news announcements and that the response is asymmetric. However, the influence of macroeconomic news announcements appears insignificant for bond returns. Finally, our results paint a complex picture of the effect of macroeconomic news releases on correlations. Keywords: Monetary integration; bond markets integration; macroeconomic news JEL classifications: E44; F36; G15

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 41
56 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096000

41

42

PILAR ABAD AND HELENA CHULIA´

INTRODUCTION The extent of international bond market linkages merits investigation, as it may have important implications for the cost of financing fiscal deficit, monetary policymaking independence, modeling and forecasting long-term interest rates, and bond portfolio diversification. After the launch of the euro in January 1999, markets virtually priced the debt of different member states as identical. During the period 2003
2007, the spreads were very small and did not reflect the differences in fiscal positions between countries, even when ratings changed. This period was therefore characterized by a significant underpricing of risk, with investors searching for yield in an environment of abundant global liquidity. This progress toward financial integration was interrupted and reversed by the global financial crisis and, more recently by the European sovereign debt crisis. Sovereign bond markets have been dominated by sharp differentiation, especially across borders. The issue of European government bond markets integration has been addressed in the recent literature under different perspectives. One strand of the literature has assessed the relative importance of systemic and idiosyncratic risk in European Monetary Union (EMU) sovereign yield spreads (see Geyer, Kossmeier, & Pichler, 2004; Go´mez-Puig, 2009; Pagano & von Thadden, 2004 among others). Another perspective is given by Christiansen (2007), who assesses volatility spillovers in European bond markets. Finally, a number of papers have studied financial integration exploiting the implications of asset pricing models (see Abad, Chulia´, & Go´mez-Puig, 2010, 2013; Barr & Priestley, 2004; Hardouvelis, Malliaropulos, & Priestley, 2006, 2007). This chapter adopts a different perspective and ties together the market integration and news announcement literature by examining the reaction of European government bond markets returns, volatilities and correlations to macroeconomic news announcements. In a unified bond market, returns, volatilities and correlations of bonds of different countries should respond similarly to the same information. Numerous papers have studied the impact of macroeconomic news releases on financial markets (see Andersen, Bollerslev, Diebold, & Vega, 2007; Fleming & Remolona, 1999a, 1999b; Gu¨rkaynak, Sack, & Swanson, 2005, among others). These studies differ in terms of the panel of economic news considered, the financial instrument, the frequency of observation and the time period examined. Hence, findings regarding which news systematically moves markets, as well as their relative importance, are sometimes conflicting.

The Effects of Macroeconomic News Announcements

43

Our study makes a number of contributions to the relevant literature. First, we analyze the evolution of the correlation between European Government Bond markets and the representative euro area bond in an attempt to assess the evolution of the convergence process. Second, we analyze the effects of macroeconomic news announcements not only on European government bond market returns and volatilities but also on correlations. The prior literature does not consider the effect of news announcements on correlations. Third, our sample includes the financial crisis which enables us to analyze whether the effects of news announcements has changed since the beginning of the crisis and if the global financial turmoil has had a negative effect on the convergence process. To carry out our study, we use the Dynamic Conditional Correlation (DCC) multivariate model of Engle (2002). Our main findings can be summarized as follows. First, we find little impact of macroeconomic news announcements on bond returns On top of that, the crisis has no changed the level of bond returns and their response to macroeconomic news announcements. Second, we confirm that bond volatility responds asymmetrically to news releases. Before and during the crisis, bad news has a meaningful impact on the dynamics of bond market volatility. Finally, our results indicate a significant heterogeneity in the impact of the release of good or bad news announcements on bond correlations. The remainder of the chapter is organized as follows. The section “Data” describes our data. The section “Methodology” lays out the methodology we use. The section “Empirical Results” discusses the empirical results and, finally, the section “Conclusions” concludes.

DATA Bond Data We use daily data covering the period from January 2004 to July 2011. The data consist of the 10-year JP Morgan Government Global Bond Index, in terms of a common currency, the euro, and the sample includes 16 European countries. Our study focuses on 10 EMU EU-15 (Austria, Belgium, France, Germany, Greece, Ireland, Italy, the Netherlands, Portugal and Spain) and 4 non-EMU countries (Denmark, Czech Republic, Poland, and Sweden).1 As a proxy for the entire euro area we use the JP Morgan EMU Government Index. These bond market indices are

PILAR ABAD AND HELENA CHULIA´

44

transformed into returns taking the first difference of the natural log of each bond-price index. Data have been collected from Thompson Datastream.

Announcement Data We examine the effects of four macroeconomic news announcements for the euro area: the industrial production index2 (IPI), the unemployment rate3 (UNE), the consumer confidence index4 (CONF), and the European Central Bank (ECB)’s interest rate decisions (RATE). We obtain the announcement data from Bloomberg. For each macroeconomic announcement, except for the interest rate announcement, we obtain a time series of the realized values as well as market forecasts based on survey expectations. In the empirical analysis we follow the previous literature (see Balduzzi, Elton, & Green, 2001), and use the surprise defined as the standardized news for announcement e (Set): Set =

Aet − Eet σe

ð1Þ

where Aet is the realized value for announcement e at time t, Eet is the corresponding expected value, and, finally, σe is the standard deviation of the announcement surprise (Aet − Eet) across the entire sample. In the case of the interest rate announcement, we use the methodology proposed by Kuttner (2001) to obtain a measure of the surprise in the announcement from the change in the current-month Federal funds futures rate on the day of the announcement. In sum, we compute the unexpected target rate change or the “surprise” SRATE,t as: SRATE;t = ft − ft − 1

ð2Þ

where ft is the current-month futures rate at the end of the announcement day t. Kuttner (2001) uses a scaled version of the one-day change in current-month federal funds future rate as a proxy for the unanticipated component on the day of the policy rate change. This is because in the United States the futures contract’s payoff depends on the monthly average Federal funds rate, and the scaled factor is included to reflect the number of remaining days in the month, which are affected by the change.

The Effects of Macroeconomic News Announcements

45

Given the focus of this study is on the impact of news announcements on European government bond markets, following Bredin, Hyde, Nitzsche, and O’Reilly (2007), we proxy surprise changes in the ECB policy rate by the one-day change in the three-month Euribor futures rate.5 It must be highlighted that, depending on the macroeconomic news announcements, the sign of the surprise will indicate good or bad news. In the case of the interest rate announcement, a positive (negative) surprise indicates that the ECB announced either a surprisingly large rate increase (cut) or a surprisingly small rate cut (increase), this is bad (good) news. For the unemployment announcement, a positive (negative) surprise means that UNE has either increased (decreased) surprisingly much or decreased (increased) surprisingly little with respect to the previous month, this is bad (good) news. In the case of IPI, a positive (negative) surprise indicates that the IPI rate has either increased (decreased) surprisingly much or decreased (increased) surprisingly little with respect to the previous announcement, this is good (bad) news. Finally, for the CONF, a positive (negative) surprise indicates that the CONF rate has either increased (decreased) surprisingly much or decreased (increased) surprisingly little with respect to the previous announcement, this is good (bad) news. Therefore, positive (negative) surprises mean bad (good) news in the case of the interest rate and unemployment rate announcements and good (bad) news in the case of IPI and CONF announcements. Table 1 reports summary statistics for macroeconomic news surprises. According to this table, the number of surprises is very similar for the four macroeconomic announcements considered, showing that there is no a particular type of macroeconomic event that surprises more frequently the market. In the case of UNE and RATE, there are more positive surprises, that is, bad news, CONF shows a higher number of negative surprises which means bad news, and finally, in the case of IPI there are more positive surprises, that is, good news. During the crisis there is a reversal in the sign of the surprises for CONF (higher number of positive surprises) and IPI (lower number of positive surprises).

METHODOLOGY Our methodology to assess the effect of macroeconomic news announcements on conditional returns, returns volatility and the integration or comovement between European government bond markets and the euro

46

Table 1.

Descriptive Statistics for News Announcements Surprises.

Total Sample Period (January 2004 to July 2011) UNE

CONF

IPI

RATE

UNE

CONF

IPI

RATE

88 1.664 1.000 3.949 −0.511

80 −0.040 1.000 3.513 −3.238

86 0.248 1.000 3.587 −2.586

79 0.001 0.045 0.125 −0.155

32 0.899 0.891 3.113 −0.511

27 0.166 1.442 3.513 −3.238

34 −0.140 1.219 3.587 −2.586

31 −0.002 0.052 0.125 −0.115

86 1.713 0.956 3.949 0.046

37 0.720 0.805 3.513 0.069

53 0.864 0.649 3.587 0.083

43 0.029 0.030 0.125 0.005

30 0.990 0.844 3.113 0.046

16 0.982 1.111 3.513 0.069

14 0.989 0.895 3.587 0.083

18 0.033 0.034 0.125 0.005

2 −0.465 0.066 −0.418 −0.511

43 −0.695 0.614 −0.069 −3.238

33 −0.741 0.579 −0.083 −2.586

36 −0.032 0.035 −0.005 −0.155

2 −0.465 0.066 −0.418 −0.511

11 −1.021 0.972 −0.207 −3.238

20 −0.930 0.661 −0.167 −2.586

13 −0.049 0.030 −0.015 −0.115

Notes: N is the number of observations for each macroeconomic announcement surprises. Max and Min refer to maximum and minimum, respectively. We set the start of the financial crisis coinciding with the collapse of Lehman Brothers.

PILAR ABAD AND HELENA CHULIA´

All surprises N Mean Standard deviation Max Min Positive surprises N Mean Standard deviation Max Min Negative surprises N Mean Standard deviation Max Min

Crisis Period (September 2008 to July 2011)

The Effects of Macroeconomic News Announcements

47

area is based on the DCC multivariate model of Engle (2002). The DCC model has the flexibility of univariate GARCH models but does not suffer from the “curse of dimensionality” of multivariate GARCH models. The estimation consists of two steps. First, the conditional variance of each variable is estimated using a univariate GARCH procedure. Second, the standardized regression residuals obtained in the first step are used to model conditional correlations that vary through time. To analyze the response of bond markets to the arrival of European macroeconomic news taking into account the effect of the financial crisis and including the possibility, as suggested in the literature, that the response of asset prices, volatility and correlations depends on the surprise component being positive or negative, we model the evolution of bond returns and the volatility of country i as rti = μei þ ρei rti − 1 þ csrit þ ɛ
it  ɛit Ft − 1 ∼ Nð0; 1Þhit = cshit ωei þ αei ðɛ it − 1 Þ2 þ βei hit − 1

ð3Þ

where Ft − 1 denotes the information set at time t − 1, csrti = ðγ iþ e I tþ e þ γ i− e I t− e ÞS et þ ρi Ct þ ðξiþ e I tþ e þ ξi− e I t− eÞSetCt and cshit = 1 þ ðδiþ e I tþ e þ þe þe − e − e  e − e − e  e δi I t Þ St þ υi C t þ ðθi I t þ θi I t Þ St C t . In the above specification, Ct is a dummy variable taking on the value of one during the financial crisis (from September 15, 2008 onward6) and zero otherwise, Itþ e ðIt− e Þ is a dummy variable taking on the value of one during the event window (t − 1, t + 1) if a positive (negative) surprise macroeconomic event of type e occurred at time t and zero otherwise and, finally, Set are news surprises standardized by their sample standard deviation to control for differences in units of measurement across announcements in the event window. The above specification allows for asymmetric effects of surprises on conditional bond returns and volatility. The coefficient γ iþ e ðγ i− e Þ captures the impact of positive (negative) surprise announcements on the mean return around the announcement dates during the total sample, and γ iþ e þ ξiþ e ðγ i− e þ ξi− e Þ captures the impact of positive (negative) surprise announcements on the mean return around the announcement dates during the financial crisis. Similarly, the dummy coefficient δei þ ðδei − Þ proxy for the impact of positive (negative) absolute standardized macroeconomic news surprises on conditional return variance around the announcement dates during the total sample, and δiþ e þ θiþ e ðδi− e þ θi− e Þ proxy for the impact of positive (negative) absolute standardized macroeconomic news surprises on conditional return variance around the announcement dates during the

48

PILAR ABAD AND HELENA CHULIA´

financial crisis.7 Finally, to give flexibility to the model, ρi and νi enable the crisis to have an effect on the returns and volatility level, respectively. To analyze the impact of news announcements on conditional correlations between each bond market and the entire euro area, the following exponential smoothing function is used:
 qijt = csqit λqijt− 1 þ ð1 − λÞηit − 1 ηjt − 1

ð4Þ

    where csqit = 1 þ ðαi;jþ e Itþ e þ αi;j− e It− e ÞSet  þ βi;j Ct þ ðκi;jþ e Itþ e þ κ i;j− e It− e ÞSet Ct . To deal with the problem suggested by Forbes and Rigobon (2002) that shocks to the conditional covariance among asset returns in proximity to certain macroeconomic announcements may be due to shocks to return volatility, we use residuals standardized as follows: ɛit ɛjt ηit = qffiffiffiffiffiffiffiffiffiffiffiffiffi and ηjt = qffiffiffiffiffiffiffiffiffiffiffiffiffi cshit ⋅hit cshjt ⋅hjt

ð5Þ

In Eq. (4), the coefficient αi;jþ e ðαi;j− e Þ captures the impact of positive (negative) surprise announcements on the conditional covariance between any pair of standardized residuals (countries i and j) around the announcement dates during the total sample, and αi;jþ e þ κiþ e ⋅ðαi;j− e þ κi− e Þ captures the impact of positive (negative) surprise announcements on the covariance around the announcement dates if it takes place during the financial crisis. βi,j enables the crisis to have an effect on the covariance level. In order to estimate the model in Eqs. (3) and (4), a conditional normal distribution for the innovation vector is assumed and the quasi-maximum likelihood method is applied. Bollerslev and Wooldridge (1992) show that the standard errors calculated using this method are robust even when the normality assumption is violated.

EMPIRICAL RESULTS Tables 2
5 display the estimation results. As pointed out by other studies (see Brenner et al., 2009; Christiansen, 2000; Konrad, 2009, among others), we find weak or no evidence of a relation between bond returns and the release of macroeconomic news announcements. Hence, this provides evidence that releases of macroeconomic news are not associated with risk

Table 2.

Response of Bond Markets to the Arrival of European Macroeconomic News Releases: The UNE. Impact on Returns

γ iþ e

γ i− e

Germany 0.092 Austria 0.058 Belgium −0.065 Denmark −0.081 France −0.064 Greece 0.071 Italy The Netherlands −0.054 Portugal 0.000 −0.074 Spain −0.049 Ireland Poland 0.001 0.071 Czech Republic Sweden −0.085

γ iþ e

þ ξiþ e

γ i− e

Impact on Variance þ ξi− e

−0.002 −0.002 −0.003 −0.004 −0.003 −0.003

0.000

0.001

ρi

δei −

0.201 0.000

−0.065 −0.041

−0.002 −0.005 −0.002 −0.008 −0.012 −0.007

δei þ

0.161 −0.041 −0.140 0.000

Note: For ease of reading we only show significant coefficients at the 5% level.

δiþ e

þ θiþ e

0.201 0.199 0.145 0.189 0.189 −0.065 0.083 0.152 0.123 0.122 0.083 0.105 0.146 0.140

δi− e

þ θi− e

−4.213

Impact on the covariance level νi 0.369 0.062 0.044 0.206 0.084 1.285 0.046 0.066 0.074 0.092 0.046

αi;jþ e

−0.014 1.206 −0.108 0.144 0.520 0.458 391.3 0.046 0.598 −0.135 −0.055 −1.096 0.046

1.267 −0.091 0.138

αi;j− e

7.531 0.159

αi;jþ e þ κ iþ e αi;j− e þ κ i− e 0.122 −0.151 0.075 0.144 0.520 0.090 −0.119 0.153 −0.068 0.107 −0.119 0.843 0.281 1.455

391.3

−1.823 −1.096

βi;j −0.097 0.190 −0.080 0.018 −0.019 0.029 0.034

−0.090 0.034 −0.098 0.434 −0.108

Table 3.

Response of Bond Markets to the Arrival of European Macroeconomic News Releases: The CONF. Impact on Returns

γ iþ e Germany Austria Belgium Denmark France Greece Italy The Netherlands Portugal Spain Ireland Poland Czech Republic Sweden −0.002

γ i− e −0.002 −0.001 −0.001 −0.001 −0.001 −0.001 −0.001 −0.001 0.000 −0.002 −0.001

γ iþ e

þ ξiþ e

0.001

0.000

γ i− e

Impact on Variance þ ξi− e

0.000 0.001 0.000 −0.001 0.001 −0.001 0.001 0.001 0.002 −0.002 0.001

ρi

δei þ

1.790 0.000 −0.199 −0.162 −0.240 −0.184 −0.349 −0.245 −0.178 −0.089 −0.221 −0.218

δei − 0.205 0.114 0.135 0.205 0.159

0.140 3.742 0.163 0.209 −0.171 −0.048 0.337 0.000 −0.154

Note: For ease of reading we only show significant coefficients at the 5% level.

δiþ e

þ θiþ e

0.337 0.153 0.173 0.135 0.197 −0.349 0.126 0.156 0.299 0.102 −0.029 0.126 0.089 0.166

δi− e

þ θi− e

0.544 0.259 0.290 0.205 0.291 0.216 0.276 0.011 0.163 0.027 0.029 0.117 0.187

Impact on the Covariance Level νi

αi;jþ e

αi;j− e

0.323 −0.548 3.456 0.075 2.758 0.458 0.048 0.482 0.222 0.217 1.903 0.091 0.254 0.218 1.497 0.120 0.614 0.048 −0.347 −0.103 0.072 9.651 0.625 0.130 −0.214 −0.556 0.103 0.226 −0.294 0.280 11.930 −0.345 0.346 0.131 −0.028 −0.328 0.127 3.700 1.305

αi;jþ e þ κ iþ e αi;j− e þ κ i− e −0.273 0.152 −0.124 −0.156 0.079 −0.282 −0.100 −0.097 0.316 −0.073 4.529 0.838 −0.055 −0.059

2.455 0.097 −0.038 0.961 1.135 −0.137 −0.103 0.236 −0.040 −0.010 −0.118 2.349 0.603 1.305

βi;j −0.040 0.120

−0.059 0.062 0.011 0.055 −0.048 −0.046 −0.029 −0.076 −0.026 0.022

Table 4.

Response of Bond Markets to the Arrival of European Macroeconomic News Releases: The IPI. Impact on Returns

γ iþ e

γ i− e

Germany −0.002 Austria −0.002 Belgium −0.002 Denmark −0.002 France −0.002 Greece −0.002 Italy 0.001 −0.002 The Netherlands −0.002 Portugal 0.000 0.000 Spain −0.003 Ireland −0.002 Poland Czech Republic Sweden −0.001

γ iþ e

þ ξiþ e

γ i− e

Impact on Variance þ ξi− e

0.002

−0.002 0.001 0.000 −0.002 0.000 −0.002 0.000 0.000 0.000 −0.001 −0.002 −0.002

0.002

0.001

0.002

0.002 0.001 0.001

ρi

δei þ

δei −

δiþ e

þ θiþ e

0.000

0.085 61.239 0.105 0.148 −0.052 −0.102

0.000

Note: For ease of reading we only show significant coefficients at the 5% level.

Impact on the Covariance Level þ θi− e

−0.221 0.059

νi

αi;jþ e

αi;j− e

0.237 0.152 0.738 0.243 0.259 0.060

0.338 0.185 6.416 0.078 3.433 1.060 0.060 0.138 0.185 0.217 −0.297 0.097 0.293 0.321 0.338 0.185 6.416 0.046 0.329 0.076 0.841 0.706 0.066 −0.271 −1.281 0.099 0.007 −0.984 0.303 0.553 −0.911 1.362

0.224

0.147

0.189 0.169 0.241 0.216

0.077 0.116 0.104

0.000

δi− e

0.679

1.770

αi;jþ e þ κ iþ e αi;j− e þ κ i− e 0.569 −0.224 0.028 −0.118 −0.079 0.569 0.194 −0.144 0.016 0.202 0.038

1.086

βi;j

0.158 −0.135 −0.026 0.200 0.129 0.158 0.053 0.064 −0.050 0.060 −0.021 1.362

−0.068

1.770

−0.090

0.160 −0.023 −0.089 −0.027

−0.029 −0.084

Table 5.

Response of Bond Markets to the Arrival of European Macroeconomic News Releases: The Interest Rate (RATE). Impact on Returns γ iþ e

γ i− e

Germany Austria Belgium Denmark France Greece Italy The Netherlands Portugal −0.016 Spain Ireland 0.020 Poland −0.049 0.030 Czech Republic Sweden 0.022

γ iþ e

þ ξiþ e

γ i− e

Impact on Variance þ ξi− e

ρi

δei þ

0.000 2.531 2.464 5.085 2.721 2.603 2.855 3.275 3.591 2.674

−0.016 0.020 0.033 0.037 0.022

−0.030 0.030

δei − 6.477 4.010 3.857 9.147 4.712 15.708 4.163 3.709 2.375 5.010 6.211

2.580 −3.522 0.001

Note: For ease of reading we only show significant coefficients at the 5% level.

δiþ e

þ θiþ e

2.531 2.464 5.085 2.721 −9.283 2.603 2.855 −1.407 0.109 −1.220 3.947 2.580 5.154

δi− e

þ θi− e

6.477 4.010 3.857 9.147 4.712 15.708 8.082 3.709 0.119 5.010 12.849 0.643

Impact on the Covariance Level νi

αi;jþ e

αi;j− e

−2.586 4.893 −6.230 −5.565 12.080 −7.914 −5.333 −3.429 −8.556 −5.062 −4.435 −3.631 −3.925 −2.566 −2.783 1.389 62.667 −7.022 −6.018 −3.316 −0.045 −3.023 62.260 0.129 24.125 −4.032 0.274 0.093 0.070 0.544 0.115 1.475 0.064 0.092 0.053 0.122 0.280

αi;jþ e þ κ iþ e αi;j− e þ κ i− e −4.959 −1.301 −3.159 −4.631 2.426 4.102 −0.259 −0.204 6.920 −7.882 −4.807 8.968 −4.657 1.611

28.689 2.141 −2.186 −5.333 31.980 −5.062 38.093 10.691 −6.662 −6.504 −6.018 21.461 13.160 25.137

βi;j −0.013 −0.101 −0.029 0.016 −0.085 −0.042 −0.030 −0.082 0.010 0.037 −0.062 0.027 −0.059

The Effects of Macroeconomic News Announcements

53

premier in the sense of higher returns on announcement days, which is consistent with the findings of Li and Engle (1998) for the US Treasury futures market. The arrival of unemployment news has the greatest impact on bond returns, while interest rates news has the lowest. Regarding volatility, we find that volatility responds asymmetrically to announcement shocks.8 Before and during the crisis, bad news has a meaningful impact on the dynamics of bond market volatility. Releases of the employment situation are especially influential.9 In general, after bad news there is an increase in the volatility, independently of the market and the macroeconomic news announcement. These results are in line with the findings in Jones et al. (1998), Fleming and Remolona (1999b), Christie-David and Chaudhry (1999), Christie-David, Chaudhry, and Khan (2002), Christie-David et al. (2003), Christiansen (2000), and Goeij and Marquering (2006). These authors analyze the response of Treasury bond market volatility to macroeconomic news releases and find significant increases in bond market volatility on announcement days and its persistence. Although there are some differences across countries,10 in general, looking at good news, announcements on the CONF and the interest rates also increase volatility during the crisis. Before the crisis, good CONF news reduces the volatility whereas good news about interest rates increases volatility. We also find that the crisis has increased the volatility of all bond markets. Finally, we find a significant heterogeneity in the effect of macroeconomic news releases on bond correlations.11 Before the crisis, good news on CONF and IPI are more likely to increase the integration of European Government markets. Good unemployment news releases have no effect on correlation and interest rate news releases have a mixed effect depending on the country. Bad news on unemployment, IPI, and CONF before the crisis increases correlation, however, expansionary ECB rate decisions decreases correlation. During the crisis period we find that good unemployment news announcements do not have an effect on correlations and the effect of bad news is mixed. Both, good and bad news releases on IPI increase correlations and CONF and interest rate news releases have a mixed effect.

CONCLUSIONS The analysis of European government bond markets have been recently addressed in the financial literature due to the convergence process that

54

PILAR ABAD AND HELENA CHULIA´

started after the launch of the EMU and the reversal of financial integration during the European sovereign debt market crisis. Our chapter contributes to this literature by providing a comprehensive analysis of the impact of macroeconomic news releases on European Government Bond returns, volatilities and correlations. We study the period from 2004 to 2011, using news releases for the UNE, the IPI, the CONF, and the ECB’s Interest rates decisions. Our setting provides interesting insights into the effects of macroeconomic news announcements on European government bond markets. While the response of bond returns to the release of macroeconomic news announcements is limited, we find that bond return volatility strongly reacts to macroeconomic surprises and that the response is asymmetric. Our estimates paint a complex picture of the effect of macroeconomic news releases on correlations. Bond return comovements most often increase in correspondence with news releases. Finally, we find that the financial crisis does not necessarily increase the sensitivity of bond returns, volatilities and correlations to macroeconomic news announcements.

NOTES 1. The earliest data available for the Czech Republic is November 2004. 2. Eurostat Index of Industrial production (Total industry excluding construction) Data is adjusted for working days. 3. The basis for the calculation of a monthly unemployment rate is the Community Labour Force Survey, where the main statistical objective is to describe the population in three exhaustive and mutually exclusive groups (employed, unemployed, and inactive). The unemployment rate is the number of people unemployed as a percentage of the labor force. 4. This indicator represents the arithmetic average of the answers (balances) to the four questions on the financial situation of households and general economic situation (past and future) together with that on the advisability of making major purchases (Source: European Commission). 5. Bernoth and Von Hagen (2004) find that the three-month Euribor futures rate is an unbiased predictor of euro area policy rate changes. 6. We set the start of the financial crisis coinciding with the collapse of Lehman Brothers. 7. In line with the financial literature, surprise monetary policy news announcements enter the variance equation as absolute values (see Brenner, Pasquariello, & Subrahmanyam, 2009; Christiansen & Ranaldo, 2007, among others). 8. This result is consistent with the analysis of Li and Engle (1998). 9. Christie-David, Chaudhry, and Lindley (2003) examine the effects of 11 macroeconomic news announcements on futures on Treasury bonds and (10-year)

The Effects of Macroeconomic News Announcements

55

Treasury notes finding that the Employment Report has the most pervasive effects on volatility. Goeij and Marquering (2006) also find that releases of the employment situation and producer price index are especially influential in the volatility of US Treasury bonds. 10. Christie-David et al. (2002), using intraday data, observe that the US Treasury Bond, the British Long Gilt, and the German Government Bond futures respond strongly to the US news releases, while the response of the Japanese Government Bond futures is less pronounced and the Italian Government Bond futures display weak responses at best. 11. This result is in line with the findings in Christie-David et al. (2002). These authors uses a completely different methodology to assess the effect of US news announcements on the integration of the bond markets in the United States, Great Britain, Germany, Japan, and Italy.

ACKNOWLEDGMENT This work has been funded by the Spanish Ministry of Economy and Competitiveness (ECO2011
23959 and ECO2012
35584).

REFERENCES Abad, P., Chulia´, H., & Go´mez-Puig, M. (2010). EMU and European government bond market integration. Journal of Banking and Finance, 34, 2851
2860. Abad, P., Chulia´, H., & Go´mez-Puig, M. (2013). Time-varying integration in European government bond markets. European Financial Management, 20(2), 270
290. Andersen, T., Bollerslev, T., Diebold, F., & Vega, C. (2007). Real time price discovery in global stock, bond, and foreign exchange markets. Journal of International Economics, 73(2), 251
277. Balduzzi, P., Elton, E., & Green, C. (2001). Economic news and bond prices: Evidence from the U.S. treasury market. Journal of Financial and Quantitative Analysis, 36, 523
543. Barr, D. G., & Priestley, R. (2004). Expected returns, risk and the integration of international bond markets. Journal of International Money and Finance, 23, 71
97. Bernoth, K., & Von Hagen, J. (2004). Euribor futures market: Efficiency and the impact of ECB policy announcements. International Finance, 7, 1
24. Bollerslev, T., & Wooldridge, J. M. (1992). Quasi-maximum likelihood estimation and inference in dynamic models with time-varying covariances. Econometric Reviews, 11(2), 143
172. Bredin, D., Hyde, S., Nitzsche, D., & O’Reilly, G. (2007). European monetary policy surprises: The aggregate and sectoral stock market response. International Journal of Finance and Economics, 14(2), 156
171. Brenner, M., Pasquariello, P., & Subrahmanyam, M. (2009). On the volatility and comovement of US financial markets around macroeconomic news announcements. Journal of Financial and Quantitative Analysis, 44(6), 1265.

56

PILAR ABAD AND HELENA CHULIA´

Christiansen, C. (2000). Macroeconomic announcement effects on the covariance structure of government bond returns. Journal of Empirical Finance, 7, 479
507. Christiansen, C. (2007). Volatility-spillover effects in European bond markets. European Financial Management, 13(5), 923–948. Christiansen, C., & Ranaldo, A. (2007). Realized bond-stock correlation: Macroeconomic announcement effects. Journal of Futures Markets, 27(5), 439
469. Christie-David, R., & Chaudhry, M. (1999). Liquidity and maturity effects around news releases. Journal of Financial Research, 22(1), 47
67. Christie-David, R., Chaudhry, M., & Khan, W. (2002). News releases, market integration, and market leadership. Journal of Financial Research, 25, 223
245. Christie-David, R., Chaudhry, M., & Lindley, J. T. (2003). The effects of unanticipated macroeconomic news on debt markets. Journal of Financial Research, 26, 319
339. Engle, R. (2002). Dynamic conditional correlation: A simple class of multivariate generalized autoregressive conditional heteroskedasticity models. Journal of Business & Economic Statistics, 20(3), 339
350. Fleming, M. J., & Remolona, E. M. (1999a). Price formation and liquidity in the U.S. treasury market: The response to public information. Journal of Finance, 54(5), 1901
1915. Fleming, M. J., & Remolona, E. M. (1999b). The term structure of announcement effects. Staff Report. Federal Reserve Bank of New York. Forbes, K. J., & Rigobon, R. (2002). No contagion, only interdependence: Measuring stock market comovements. The Journal of Finance, 57(5), 2223–2261. Geyer, A., Kossmeier, S., & Pichler, S. (2004). Measuring systematic risk in EMU government yield spreads. Review of Finance, 8, 171
197. Goeij, P., & Marquering, W. (2006). Macroeconomic announcements and asymmetric volatility in bond returns. Journal of Banking & Finance, 30(10), 2659
2680. Go´mez-Puig, M. (2009). The immediate effect of monetary union over EU-15’s sovereign debt yield spreads. Applied Economics, 41, 929
939. Gu¨rkaynak, R., Sack, B., & Swanson, E. (2005). The sensitivity of long-term interest rates to economic news: Evidence and implications for macroeconomic models. American Economic Review, 95(1), 425
436. Hardouvelis, G. A., Malliaropulos, D., & Priestley, R. (2006). EMU and European stock market integration. Journal of Business, 79, 365
392. Hardouvelis, G. A., Malliaropulos, D., & Priestley, R. (2007). The impact of EMU on the equity cost of capital. Journal of International Money and Finance, 26, 305
327. Jones, C. M., Lamont, O., & Lumsdaine, R. L. (1998). Macroeconomic news and bond market volatility. Journal of Financial Economics, 47(3), 315–337. Konrad, E. (2009). The impact of monetary policy surprises on asset return volatility: The case of Germany. Financial Markets and Portfolio Management, 23(2), 111
135. Kuttner, K. (2001). Monetary policy surprises and interest rates: Evidence from the fed funds futures market. Journal of Monetary Economics, 47, 523
544. Li, L., & Engle, R. F. (1998). Macroeconomic announcements and volatility of treasury futures. Discussion Paper 98
27. Department of Economics, University of California, San Diego, CA. Pagano, M., & von Thadden, E. L. (2004). The European bond markets under EMU. Oxford Review of Economic Policy, 20, 531
554.

PART II RISK AND INTEGRATION IN A POST-CRISIS SETTING

THE PRO-CYCLICAL IMPACT OF BASEL III REGULATORY CAPITAL ON BANK CAPITAL RISK Guoxiang Song ABSTRACT To raise the quality of regulatory capital, Basel III capital rules recognize unrealized gains and losses on all available-for-sale (AFS) securities in Common Equity Tier 1 Capital (CET1). However, by examining the correlations between U.S. GDP growth rate, interest rates and regulatory capital ratios computed using Basel III regulatory capital definition for six U.S. global systemically important banks (G-SIBs) since 2007, this chapter finds that Basel III regulatory capital will enhance the procyclicality of Basel III leverage ratio and Tier 1 capital ratio and their sensitivity to long-term interest rates. Therefore, Basel III capital standards may have significant implications for bank supervision and bank capital risk management in the near future. As banks will hold more high-quality liquid assets (HQLAs) as required by Basel III Liquidity Coverage Ratio (LCR), the weight of unrealized gains and losses arising from fair value accounting will increase in Basel III Tier 1 capital base, the consequent increase of pro-cyclicality in a bank’s regulatory capital ratios may distort the true picture of bank capital adequacy. If an

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 59
81 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096002

59

60

GUOXIANG SONG

expected loss approach (EL) is used as the provisioning model, such capital risk may be increased further. Moreover, as U.S. monetary policy has started tapering quantitative easing, long-term interest rates will increase inevitably. This may increase the negative impact of unrealized gains and losses on AFS securities on bank capital. As a result, it may be difficult for banks to maintain appropriate capital ratios to meet regulatory requirements and support business activities. Keywords: Basel III; regulatory capital; pro-cyclicality; capital risk; quantitative easing

INTRODUCTION During the recent financial crisis, the market focused on tangible common equity rather than regulatory capital as the measure of a bank’s capital adequacy (Tarullo, 2011). Therefore, to raise the quality of regulatory capital, Basel III capital rules recognize unrealized gains and losses on all availablefor-sale (AFS) securities measured at fair value in its Common Equity Tier 1 Capital (CET1). However, as it is argued that fair value accounting may have reinforced the pro-cyclicality of regulatory capital standards for banks (BCBS, 2011; ECB, 2004; IMF, 2008; Song, 2012), Basel III regulatory capital may further enhance the pro-cyclicality of regulatory capital ratios. Consequently, Basel III regulatory capital standards may have significant implications for bank supervision and bank capital risk management. To address this issue, this chapter investigates the pro-cyclical impact of Basel III regulatory capital on regulatory capital ratios of six U.S. global systemically important banks (G-SIBs) since 2007. And the results indicate that Basel III regulatory capital will enhance the pro-cyclicality of bank capital risk. Before Basel III, major components of regulatory Tier 1 capital measured using fair value accounting are trading revenues and realized gains or loess on AFS securities. One factor for excluding unrealized gains or losses on AFS securities from Tier 1 capital is the volatility created by fair value accounting especially that generated by changes in interest rates, as it is not indicative of a bank’s true financial condition (Greenspan, 1990). Another factor is that unrealized gains or losses on AFS securities are not as informative and value relevant as net income to securities price (Ball, Jayaraman, & Shivakumar, 2012; Dong, Ryan, & Zhang, 2014). However, Basel III regulatory capital includes unrealized gains or losses on AFS securities as it is argued that unrealized losses could have an actual

The Pro-Cyclical Impact of Basel III Regulatory Capital

61

impact on a bank’s capital in some specific time period (The Agencies, 2013). But this may add to the pro-cyclicality and volatility to bank capital requirements (ISDA, 2012). In order to examine what difference Basel III regulatory capital definition will make to bank supervision and bank capital risk management, this chapter will compare the impacts of three main components of accounting gains and losses on the Basel III leverage ratio and the Tier 1 capital ratio. The first component is the recurring fair value gains and losses on trading assets, which are currently parts of Tier 1 capital. The second component is the recurring unrealized fair value gains and losses on AFS securities. They are included in Basel III Tier 1 capital but not the Tier 1 capital definition before Basel III. To provide evidence on the relative pro-cyclical and volatile nature of the impact of these accounting gains and losses, realized gains and losses on AFS securities are also examined. The third component is the provisions for loan loss reserves which are the impairments recognized for held-for-investment loans. The increase in these provisions is currently deducted from net income and Tier 1 capital. The behavior of this component is compared with that of fair value accounting so that the relative pro-cyclicality of different accounting rules can be examined as loan loss provision under the current incurred loss approach is argued to be pro-cyclical (BCBS, 2011). Some studies have found that these accounting gains or losses have an obvious impact on bank regulatory capital measures used before Basel III during this financial crisis. For example, SEC (2008) examines 22 U.S. bank failures which were mainly small banks during 2008, and finds that credit losses had played a meaningful role. However, Shaffer (2010) finds that realized and unrealized gains or losses resulting from fair value accounting reduced Tier 1 capital by more than 8% for some large U.S. banks, and unrealized losses on AFS securities reduced tangible common equity ratios by more than 50% for some large U.S. banks. Badertscher, Burks, and Easton (2012) also investigate the impact of unrealized gains and losses on AFS securities. However, they found that fair value changes excluding trading assets have a smaller effect on Tier 1 capital ratio than the provision for loan losses for 150 U.S. Bank Holding Companies (BHCs) in the crisis. Laux and Leuz (2010) recognize the impact of huge losses on trading assets for a few very large BHCs, and Ball et al. (2012) suggest trading losses could have a negative impact on investor sentiment for banks’ shares during the crisis because they may increase information asymmetry. However, there is little empirical evidence on the pro-cyclical impact of Basel III regulatory capital on bank regulatory capital ratios. Many papers

62

GUOXIANG SONG

investigate the pro-cyclicality of bank capital regulation based on simulation exercises (ECB, 2009; IMF, 2008). And they use large samples which mix large and small banks, so there is no clear evidence for G-SIBs. This chapter designs models to measure the direct impact of accounting rules on a bank’s Basel III regulatory capital ratios, and attempts to explore what would happen to U.S. G-SIBs if Basel III regulatory capital definition is applied. Six U.S. G-SIBs, that is, Citigroup (C), JPMorgan Chase (JPM), Bank of America (BAC), Wells Fargo & Company (WFC), The Bank of New York Mellon (BK), and State Street Corporation (STT) are used as cases to investigate the empirical evidence during the period 2007
2013. Among them, C, JPM, BAC, and WFC are the four largest U.S. commercial BHCs. BK and STT are two important clearing and settlement banks. In addition, they are global systemically important financial institutions (G-SIFIs) identified by the FSB and BCBS based on the BCBS methodology using data as of end-2009 (FSB, 2011).1 Why is it necessary to investigate these U.S. G-SIBs? First, the effect of fair value accounting is argued to be more pronounced at large banks than at small banks (Shaffer, 2010). In the sample, the smallest proportion of assets measured at fair value is 24% in the second quarter of 2007 (see Table 1). Second, these banks are G-SIFIs whose failure may have a devastating impact on the financial system (FSB, 2011), and that is why special regulations will be applied to them after the crisis (Tarullo, 2011). Third, the general results from large sample studies which mix large and small banks cannot describe what happens to large banks but may reflect the findings for small banks as there is a big difference in using accounting rules between large banks and small banks. Finally, there are also significant differences among G-SIBs as can be seen from Table 1. For example, STT measured 60% of total assets at fair value whereas WFC measured 24% of total assets at fair value right before this crisis. This chapter investigates the pro-cyclical impact of Basel capital standards by directly examining the correlations between U.S. GDP growth rate and the impact of different accounting gains and losses on regulatory capital ratios estimated using the models in this chapter. It finds that the Basel III leverage ratio and the Tier 1 capital ratio do show stronger correlations with U.S. GDP growth rate than these ratios under capital standards before Basel III, which suggests that including unrealized gains and losses on AFS securities may increase the pro-cyclicality of regulatory capital ratios. The results also indicate that both fair value and loan loss provisions have a significant pro-cyclical impact on Basel III regulatory capital ratios, and these two impacts may reinforce each other. As an expected loss

Assets at Fair Value 30/06/2007

STT C JPM BAC BK WFC Aggregate

Net Loans 30/09/2013

Total of fair value

Trading assets

AFS securities

Total of fair value

Trading assets

AFS securities

60.4 42.23 40.68 27.63 24.18 23.91 36.42

0.76 24.24 30.9 13.83 2.67 1.35 20.24

55.89 11.11 6.58 11.23 20.41 13.37 11.28

48.59 40.46 32.26 31.25 27.55 24.69 32.72

2.21 15.36 15.56 13.16 3.29 4.05 12.05

46.22 14.73 14.25 11.16 20.73 17.46 15.24

30/06/ 2007

30/09/ 2013

10.73 32.98 31.37 48.85 30.18 62.79 38.86

7.18 33.53 28.87 42.98 13.42 53.57 36.48

Notes: Aggregate is estimated using the sum of assets of all six banks. Estimated using data from Form 10-Q and FRY 9-C. AFS, available-for-sale security; C, Citigroup; BAC, Bank of America; JPM, JPMorgan Chase; WFC, Wells Fargo & Company; BK, The Bank of New York Mellon; STT, State Street Corporation.

The Pro-Cyclical Impact of Basel III Regulatory Capital

Table 1. Assets at Fair Value on a Recurring Basis and Net Loans Relative to Total Assets (%).

63

64

GUOXIANG SONG

(EL) approach for loan loss provisioning will use forward-looking information as fair value does, its adoption may increase the positive correlation between the impact of fair value and that of loan loss provision. Moreover, this chapter finds these two impacts also are negatively sensitive to longterm interest rate changes, and Basel III capital ratios have a strong negative sensitivity to both short-term and long-term interest rates. These results have significant implications for bank capital risk management. For G-SIBs which hold a significant share of AFS securities such as STT and BK, the Basel III countercyclical buffer of 2.5% plus the additional loss absorbency requirements of 2.5% for G-SIBs (BCBS, 2013a) may not be enough to mitigate the capital risk generated by an economic recession or the normalization of interest rates from the current low level. This risk is becoming realistic as U.S. monetary policy has started to taper quantitative easing. In addition, the Basel III Liquidity Framework requires banks to hold more high-quality liquid assets (HQLAs) to meet Liquidity Coverage Ratio (LCR) requirement (BCBS, 2013b). This means that banks will have to hold more AFS securities after 2015 and as a result, long-term interest rate changes may have a much greater impact. This chapter contributes to the literature and current debate on bank capital regulation in several aspects: First, the pro-cyclical impact of fair value accounting on Basel III regulatory capital ratios is significant as including unrealized gains and losses on AFS securities increases the portion of Tier 1 capital measured at fair value. Even though proposed new accounting standards for financial instruments by IASB and FASB eliminate the AFS category, still, a category of fair value through other comprehensive income will include AFS securities, so their impact on regulatory capital will not disappear (IASB, 2013a). Therefore, it will be necessary for banks to disclose unrealized gains and losses arising from fair value accounting under new accounting standards so that their impact on regulatory capital can be evaluated separately from realized gains and losses. Second, the increase in the pro-cyclical impact of fair value coupled with the impact of the proposed forward-looking provisioning (BCBS, 2011) suggests that Basel III regulatory capital ratios may be more pro-cyclical. This may not bring bank capital risk management based on regulatory capital measures into line with the actual risk of banks. Third, the strong negative sensitivity of Basel III capital ratios to interest rates may increase bank capital risk significantly when interest rates normalize. The remainder of the chapter proceeds as follows. The section “The Model” introduces the model for estimating the effect of accounting rules on Basel III regulatory capital ratios. The section “The Pro-Cyclicality and

The Pro-Cyclical Impact of Basel III Regulatory Capital

65

Interest Rate Sensitivity of Basel III Regulatory Capital Ratios” examines the empirical evidence for the pro-cyclicality of Basel III regulatory capital ratios and the pro-cyclical impact of accounting rules. The section “Implications for Bank Capital Risk Management” discusses the implications for bank capital risk. The section “Conclusions” concludes.

THE MODEL One of the main differences between Basel III and capital standards before Basel III such as Basel II and Basel I is that unrealized gains or losses on AFS securities are recognized in Basel III Tier 1 capital. Therefore, the impact of accounting gains and losses on Basel III leverage ratio and Tier 1 capital ratio will be different from that on the leverage ratio and Tier 1 capital ratio before Basel III. Here simplified models based on the formulae of the leverage ratio and the Tier 1 capital ratio are developed to estimate the impact of different accounting gains and losses, and show the different impacts they have on Basel III capital standards and capital standards before Basel III. The Effect of the Accounting Gains or Losses on the Leverage Ratio The effect of an accounting gain, which is recognized in Tier 1 capital, on leverage ratio can be used as an example to illustrate the basic model. Assume that a bank has a leverage ratio of L% (0 < L < 100), and original Tier 1 capital is x, and total assets which are used for the leverage ratio calculation are y, then L% = xy. Assume that an accounting gain increases the bank’s Tier 1 capital x by k% (0 < k < 100), thereby increasing total assets by the amount of x*k%, then the new leverage ratio will be þ k%Þx L1 % = ð1y þ k%x . So the effect of this gain on leverage ratio can be calculated using DL = L1 % − L% DL =

ð1 þ k%Þx
L% y þ k%x

=

ð1 þ k%Þ
L% 1=L% þ k%

=

L%k%ð1 − L%Þ 1 þ k%L%

66

GUOXIANG SONG

So DL > 0 because (1 + k%*L%) > 0, k > 0, 1 − L% > 0. Therefore, it can be seen that if this accounting gain increases Tier 1 − L%Þ capital by k%, leverage ratio increases by L%k%ð1 1 þ k%L% . The same analysis can be applied to the case where an accounting loss decreases the bank’s Tier 1 capital x by k% (0 < k < 100), thereby decreasing total assets by the k%Þx amount of x*k%. The new leverage ratio will be ð1y −− k%x . So − L%k%ð1 − L%Þ DL = < 0. This accounting loss will lower the leverage ratio. 1 − k%L% The impact of unrealized gains or losses on AFS securities on the Basel III leverage ratio can be estimated using this model. However, the impact of these gains or losses on the leverage ratio computed using the Tier 1 capital definition before Basel III should be estimated differently as such gains or losses only affect the denominator of the leverage ratio. This impact can be estimated simply by changing the denominator of the leverage ratio by the magnitude of such gains or losses. If 2the magnitude is k% k% (−100 < k < 100) of Tier 1 capital x, then DL = − 1ðL%Þ þ k%L% The Effect of the Accounting Gains or Losses on the Tier 1 Capital Ratio It is difficult to evaluate the effect of the accounting gains or losses on riskweighted capital ratios as there are different approaches to measuring riskweighted assets (RWAs). Here the idea of an internal ratings-based approach under Basel II to calculating the change of RWAs due to the unexpected change of an individual exposure is applied to estimate the change of RWAs (Saunders & Cornett, 2011): First, calculate the capital requirement for an individual exposure. Second, use this capital require1 ment times 8% to obtain RWAs which this new regulatory capital can sup1 port. Here 8% is the asset multiplier for an 8% capital ratio under the Basel Accord. Then the difference between the estimated new risk-weighted capital ratio and the original one is used as an estimate for the change of actual risk-based capital ratio because of this exposure. With this approach, an accounting gain or loss is treated as one new individual exposure.2 This is reasonable because this chapter attempts to investigate the impact of the unrealized gains or losses on bank capital regulation. To simplify, assume that a bank has an original Tier 1 capital ratio of R% (0 ≤ R ≤ 100), and an original Tier 1 capital of x. Then total RWAs which are used for the Tier 1 capital ratio calculation for the base period x according to the Basel Accord can be calculated as RWA = R% . Assume that an accounting gain or loss is k% (−100 ≤ k ≤ 100) of this bank’s Tier 1 capital x, and it changes RWA by the amount of M xk% (0 ≤ M ≤ 8%). 8%

The Pro-Cyclical Impact of Basel III Regulatory Capital

67

Here M is the regulatory capital requirement for the new exposure due to this accounting gain or loss. Before Basel III, for trading book items, the maximum M is 1.1% (FSA, 2009), which is used in this chapter; however, for banking book items, the maximum M is 8% (FSA, 2009). Then the effect of an accounting gain or loss on Tier 1 capital ratio can be estimated using the difference between the new Tier 1 capital ratio and the original ratio R%. Here this difference is denoted as DR. When this accounting gain or loss is recognized in Tier 1 capital, the new Tier 1 capital ratio is ð1 þ k%Þx ð1 þ k%Þx R%ð1 þ k%Þ = x = k%xM k%xM RWA þ 8% 1 þ k%MR% R% þ 8% 8% As a result,

DR =

R%k%ð1 − MR% R%ð1 þ k%Þ 8% Þ − R% = k%MR% k%MR% 1þ 1þ 8% 8%

Whether DR is positive depends on whether k is positive because −100 ≤ k ≤ 100, 0 ≤ R ≤ 100 and 0 ≤ M ≤ 8%. This indicates that an accounting gain will increase Tier 1 capital ratio whereas an accounting loss will decrease the Tier 1 capital ratio. But the magnitude of the change depends on the capital requirement (M) for the new exposure. The model here can used to estimate the impact of accounting gains or losses on trading assets, unrealized and/or realized gains or losses on AFS securities on Basel III Tier 1 capital ratio. However, the impact of the loan loss provision can be estimated simply by changing the numerator of the Tier 1 capital ratio by the magnitude of k%*x if the provision is k% of Tier 1 capital x, and then DR = k%R%. This is because the provision for loan losses will affect Tier 1 capital but not RWA as the carrying value of loan and leases, which does not deduct the allowance for loan and lease, is used for computing RWA. The impact of unrealized gains or losses on AFS securities on the Tier 1 capital ratio before Basel III should also be estimated differently as they only affect the denominator of the capital ratio. Their impact can be estimated simply by changing the denominator of Tier 1 capital ratio by the magnitude of2 M xk% if such gain or loss is k% of Tier 1 capital x. Then 8% ðR%Þ k%M DR = − 8% þ k%MR%.

68

GUOXIANG SONG

THE PRO-CYCLICALITY AND INTEREST RATE SENSITIVITY OF BASEL III REGULATORY CAPITAL RATIOS The Pro-Cyclicality of the Leverage Ratio and the Tier 1 Capital Ratio The pro-cyclical impact of Basel III regulatory capital on bank capital ratios can be seen clearly from the correlations of regulatory capital ratios with real GDP growth rates. The business cycles are defined by real GDP growth rate: in economic expansions, GDP will grow; however, in economic contractions, GDP will decline. The pro-cyclicality of bank capital ratios should present itself as a positive correlation between capital ratios and GDP growth rates. Indeed, the results of Table 2 do show such procyclicality. The pro-cyclicality of Basel III leverage ratio can be seen in Panel A of Table 2. This Panel reports the correlations between leverage ratios in quarter t and GDP growth rates for quarter t − 1, quarter t and quarter t + 1. In terms of Base III leverage ratio, there are both positive and negative correlations with GDP growth rate in quarter t − 1. However, all correlations of six banks are positive with GDP growth rate in quarter t + 1 and their figures are around 20% or more. This clearly implies that high Basel III leverage ratio will be associated with high GDP growth rate in next quarter. Moreover, only one correlation with the GDP growth rate in quarter t is negative but very small (i.e., −0.06 for JPM), indicating that overall, there is a pro-cyclicality in Basel III leverage ratio. However, in terms of the leverage ratio before Basel III, there is no clear pro-cyclicality as there are both positive and negative correlations with GDP growth rates in all quarters. It may be inferred, therefore, that Basel III regulatory capital generates the pro-cyclicality for leverage ratio. The Basel III Tier 1 capital ratio also shows strong pro-cyclicality. All the sample, of banks in quarter t have positive correlations with GDP growth rates for quarter t − 1, quarter t and quarter t + 1 (see Panel B of Table 2). Moreover, the correlation coefficients are around 0.30 or more for quarter t and quarter t + 1. This demonstrates that the pro-cyclicality of the Basel III Tier 1 capital ratio is much stronger than that of Basel III leverage ratios. This is easy to explain: as a higher GDP growth rate may reduce the risk of bank assets, the RWAs of banks may grow less than total assets, therefore, the Tier 1 capital ratio may increase more than the

69

The Pro-Cyclical Impact of Basel III Regulatory Capital

Table 2. Correlations between Leverage Ratios, Tier 1 Capital Ratios in Quarter t and Real GDP Growth Rates in Quarter t − 1, t, and t + 1 (2007: Q1
2013: Q3). Panel A: Leverage ratio GDP (t − 1)

STT C JPM BAC BK WFC

GDP (t + 1)

GDP (t)

Basel III

Before Basel III

Basel III

Before Basel III

Basel III

Before Basel III

0.27 0.11 −0.02 0.11 0.21 −0.28

−0.26 0.02 −0.07 0.06 −0.48 −0.37

0.38 0.36 −0.06 0.41 0.07 0.02

−0.18 0.27 −0.17 0.37 −0.36 −0.13

0.27 0.46 0.19 0.45 0.23 0.23

−0.1 0.4 0.06 0.38 −0.19 0.06

Panel B: Tier 1 capital ratio GDP (t − 1)

STT C JPM BAC BK WFC

GDP (t + 1)

GDP (t)

Basel III

Before Basel III

Basel III

Before Basel III

Basel III

Before Basel III

0.31 0.09 0.1 0.2 0.27 0.36

−0.18 0 0.08 0.17 −0.04 0.36

0.38 0.33 0.27 0.4 0.25 0.51

−0.11 0.25 0.24 0.38 0.15 0.46

0.35 0.44 0.41 0.45 0.41 0.55

0.05 0.39 0.37 0.41 0.34 0.47

Panel C: Basel III leverage ratio

STT C JPM BAC BK WFC

STT

C

JPM

BAC

BK

WFC

1 0.77 0.58 0.69 0.46 0.19

1 0.76 0.83 0.31 0.57

1 0.46 0.19 0.53

1 0.43 0.45

1 0.14

1

Data Sources: 10-Q and FRY 9-C, U.S. Bureau of Economic Analysis. Notes: Estimated using models in this chapter. C, Citigroup; BAC, Bank of America; JPM, JPMorgan Chase; WFC, Wells Fargo & Company; BK, The Bank of New York Mellon; STT, State Street Corporation.

70

GUOXIANG SONG

leverage ratio. In fact, it is found that banks could report strong risk-based capital ratios while they were building up excessive leverage during this crisis (BCBS, 2011). However, the results for Tier 1 capital ratio before Basel III are relatively weaker than those for Basel III as there are two negative correlations for quarter t − 1 and one for quarter t. But the pro-cyclicality of the Tier 1 capital ratio is still stronger than that of the leverage ratios before Basel III in terms of GDP growth rate in quarter t + 1. One possible reason for this could be that when Tier 1 capital ratio is higher, banks may be in a better position to extend loans. The strong pro-cyclicality of the Basel III leverage ratio and the Tier 1 capital ratio can also be demonstrated by the correlations between those ratios for the six banks. This is because if there is such pro-cyclicality, then there should be positive and high correlations between these ratios for different banks. Indeed, Panel C of Table 2 reports that all correlations for Basel III leverage ratio are positive. For Basel III Tier 1 capital ratio, the correlations are stronger. To save space, the results are not reported here.

Sources of the Pro-Cyclicality One source of the pro-cyclicality of Basel III capital ratios is the procyclicality of the impact of accounting rules. Table 3 shows the pro-cyclical impact of both fair value accounting and loan loss provisions on Basel III leverage ratios. The strong pro-cyclicality of the impact of fair value accounting can be seen in Panel A of Table 3. Even though some of the correlations of the impact of fair value in quarter t with GDP growth rate in quarter t − 1 are relatively small and one of them is negative, its correlations with GDP growth rate in quarter t and quarter t + 1 for all six banks are positive, and most of these figures are around 40% or more. The positive correlations between these impacts of all banks (see Panel B of Table 3) also corroborate the strong pro-cyclicality of fair value accounting. Moreover, the strong pro-cyclical impact on the Basel III leverage ratio of unrealized gains and losses on AFS securities for each bank reported in Panel D of Table 3 may have contributed to such pro-cyclicality of fair value accounting. The loan loss provision seems to have a much stronger pro-cyclical impact on Basel III leverage ratio. Panel A of Table 3 shows that all six banks in quarter t have positive correlations with GDP growth rates for all three quarters. Moreover, all figures for quarter t − 1 are around 45% or

71

The Pro-Cyclical Impact of Basel III Regulatory Capital

Table 3. Correlations between the Impacts of Fair Value and Loan Loss Provision on Basel III Leverage Ratio in Quarter t and Real GDP Growth Rates in Quarter t − 1, t, and t + 1(2007: Q1
2013: Q3). Panel A: Impact of fair value and loan loss provision on Basel III leverage ratio GDP (t − 1)

STT C JPM BAC BK WFC

GDP (t + 1)

GDP (t)

Fair value

Loan loss provision

Fair value

Loan loss provision

Fair value

Loan loss provision

0.3 0.11 0.08 −0.03 0.52 0.18

0.72 0.62 0.46 0.72 0.61 0.48

0.67 0.48 0.47 0.21 0.78 0.39

0.35 0.64 0.62 0.64 0.3 0.76

0.39 0.63 0.63 0.57 0.44 0.57

0.13 0.5 0.58 0.43 0.25 0.59

Panel B: Correlations between impact of fair value on Basel III leverage ratio in quarter t

STT C JPM BAC BK WFC

STT

C

JPM

BAC

BK

WFC

1 0.27 0.4 0.32 0.85 0.5

1 0.7 0.3 0.28 0.42

1 0.68 0.48 0.84

1 0.3 0.8

1 0.58

1

Panel C: Correlations between impact of fair value in quarter t and that of loan loss provision in quarter t − 1, t, and t + 1 on Basel III leverage ratio for each bank

Loan loss provision (t − 1) Loan loss provision (t) Loan loss provision (t + 1)

STT

C

JPM

BAC

BK

WFC

−0.22 0.46 0.29

0.14 0.48 0.43

0.23 0.40 0.44

−0.15 0.00 0.21

0.29 0.31 0.53

0.29 0.47 0.42

Panel D: Impact on Basel III leverage ratio of unrealized gains and losses on AFS in quarter t

GDP (t − 1) GDP (t) GDP (t + 1)

STT

C

JPM

BAC

BK

WFC

0.29 0.67 0.4

0.32 0.52 0.41

0.08 0.15 0.32

0.2 0.06 0.21

0.29 0.49 0.44

0.24 0.39 0.54

Data Sources: 10-Q and FRY 9-C, U.S. Bureau of Economic Analysis. Note: Estimated using models in this chapter. C, Citigroup; BAC, Bank of America; JPM, JPMorgan Chase; WFC, Wells Fargo & Company; BK, The Bank of New York Mellon; STT, State Street Corporation.

72

GUOXIANG SONG

more. This indicates that past economic growth has a strong positive impact on loan loss provision, confirming that the incurred loss model may delay recognition of credit losses because credit losses are associated with past economic performance (IASB, 2013b). In addition, these impacts have a strong positive correlation with each other (to save space, the results are not reported here), also indicating the strong pro-cyclicality of historical cost accounting (FSA, 2009). The pro-cyclicality of fair value accounting and historical cost accounting may have reinforced each other through their positive correlations. Panel C of Table 3 provides clear evidence in terms of their impact on the Basel III leverage ratio. Five banks have reported strong and positive correlations between the impacts of fair value and loan loss provision in quarter t. And all the sample of banks report strong and positive correlations between the impact of fair value in quarter t and that of loan loss provision in quarter t + 1. The results for Basel III Tier 1 capital ratio are similar (to save space, they are not reported here). These patterns of correlations also confirm that fair value is informative to investors (SEC, 2008) and regulators as it seems to have a clear correlation with the performance of the economy and loan loss provision. In addition, the introduction of an EL model for forward-looking provisioning, which will recognize changes in expectations of credit losses, may probably enhance the correlation between the impact of fair value and that of loan loss provision on capital regulation. This is because the EL model will use forward-looking information and market information as fair value does (IASB, 2013a, 2013b; Song, 2012). As a result, the pro-cyclicality of Basel III regulatory capital may increase rather than decrease.

Interest Rate Sensitivity As interest rates are found to be pro-cyclical (Sill, 1996), they also may generate some pro-cyclical impact on regulatory capital ratio. In addition, as prices of AFS debt securities are adversely affected by interest rates, Basel III capital ratios may be more sensitive to interest rates than regulatory capital ratios before Basel III. Indeed, Basel III regulatory capital increases the negative sensitivity of regulatory capital ratios to interest rates as can be seen in Table 4. BK’s leverage ratio before Basel III has a strong positive correlation with the yield of both 3-month U.S. Treasury Bills and 10-year U.S. Treasury Bonds (see Panel A of Table 4). However, its Basel III leverage ratio has a

73

The Pro-Cyclical Impact of Basel III Regulatory Capital

Table 4.

Correlations between Leverage Ratios, Tier 1 Capital Ratios, and Interest Rates (2007: Q1
2013: Q3).

Panel A: Leverage ratio 10-Year Treasury Note

STT C JPM BAC BK WFC

3-Month Treasury Bill

Basel III

Before Basel III

Basel III

Before Basel III

−0.47 −0.69 −0.73 −0.45 0.01 −0.43

−0.43 −0.71 −0.67 −0.48 0.66 −0.36

−0.55 −0.62 −0.7 −0.46 −0.02 −0.41

−0.69 −0.67 −0.68 −0.47 0.31 −0.36

Panel B: Tier 1 capital ratio 10-Year Treasury Note

STT C JPM BAC BK WFC

3-Month Treasury Bill

Basel III

Before Basel III

Basel III

Before Basel III

−0.72 −0.74 −0.77 −0.69 −0.62 −0.67

−0.75 −0.75 −0.76 −0.71 −0.84 −0.72

−0.57 −0.66 −0.7 −0.56 −0.5 −0.49

−0.73 −0.7 −0.71 −0.56 −0.75 −0.5

Panel C: The impact of fair value and loan loss provision on Basel III leverage ratio 10-Year Treasury Note

STT C JPM BAC BK WFC

3-Month Treasury Bill

Fair value

Loan loss provision

Fair value

Loan loss provision

−0.08 −0.23 −0.09 −0.1 −0.15 −0.13

−0.18 −0.31 −0.31 −0.32 −0.49 −0.2

0.06 −0.06 0.05 −0.19 0.05 −0.1

0.19 0.18 0.16 0.22 −0.13 0.12

Panel D: The impact of unrealized gains and losses on AFS on Basel III leverage ratio

10-Year Treasury Note 3-Month Treasury Bill

STT

C

JPM

BAC

BK

WFC

−0.06 −0.08

−0.11 0

−0.28 −0.39

−0.16 −0.21

−0.17 0.06

0.02 0.15

Data Sources: 10-Q and FRY 9-C, U.S. Federal Reserve Bank of St. Louis. Note: Estimated using models in this chapter. C, Citigroup; BAC, Bank of America; JPM, JPMorgan Chase; WFC, Wells Fargo & Company; BK, The Bank of New York Mellon; STT, State Street Corporation.

74

GUOXIANG SONG

negative correlation with the short-term interest rate, and a very small positive correlation with the long-term interest rate. In addition, for all other five G-SIBs, Basel III leverage ratios have a strong and similarly negative correlation with both short-term and long-term interest rates. Moreover, both the Basel III Tier 1 capital ratio and the Tier 1 capital ratio before Basel III of all banks have shown significant and similar negative correlations with the yield of both 3-month U.S. Treasury Bills and 10year U.S. Treasury Bonds (see Panel B of Table 4). Such negative sensitivity of the leverage ratio and the Tier 1 capital ratio may to some extent come from the negative sensitivity of the impact of fair value, unrealized gains or losses on AFS securities and loan loss provision on the regulatory capital ratios to long-term interest rates, as shown in Panels C and D of Table 4.3

IMPLICATIONS FOR BANK CAPITAL RISK MANAGEMENT Pro-Cyclicality of Basel III Regulatory Capital Ratios and Bank Capital Risk The pro-cyclical impact of Basel III regulatory capital on regulatory capital ratios has more significant implications for bank capital risk management than those of other Basel capital standards. This is because Basel III includes unrealized gains or losses on AFS securities in its regulatory capital. For banks which have a significant share of assets in AFS securities, the additional capital risk generated may be hard to manage. Indeed, this can be seen from the impact of unrealized gains and losses on AFS securities during the recent recession period and expansion period.4 For example, STT and BK suffered significant unrealized losses on AFS securities during the recession period as they hold over 20% of assets in AFS securities (see Panels A and B of Table 5). Even though capital raising activities and other gains of STT and BK during the recession period may make the impact of these losses relatively smaller, their Basel III leverage ratio could still be decreased by 5.78 and 4.22 percentage points, and Basel III Tier 1 capital ratio could be reduced by 13.7 and 7.26 percentage points. At the beginning of the fourth quarter 2007, the Basel III leverage ratio and the Tier 1 capital ratio estimated for STT are 5.3 and 11.4 percentage points, and those for BK are 6.7 and 9 percentage points. If the negative

Table 5.

The Pro-Cyclical Impact of Basel III Regulatory Capital.

Panel A: The accumulated pro-cyclical impact of fair value gains and losses and loan loss provision on Basel III leverage ratio Accumulated percentage change of starting Basel III leverage ratio Fourth quarter 2007 to second quarter 2009

Trading revenue Loan loss provision AFS: unrealized gains and losses AFS: realized gains and losses Total fair value gains and losses Sum of fair value and loan loss provision

Third quarter 2009 to third quarter 2013

STT

C

JPM

BAC

BK

WFC

STT

C

JPM

BAC

BK

WFC

1.12 −0.05 −5.78

−1.66 −2.91 −0.78

−0.31 −2.23 −0.11

−0.30 −2.85 0.38

2.54 −0.18 −4.22

0.23 −3.96 −1.21

1.45 −0.05 0.49

1.25 −3.48 −0.03

2.06 −2.01 0.83

1.31 −3.03 0.80

1.32 −0.06 2.56

0.50 −3.35 1.78

0.03

0.12

−0.15

−0.16

2.01

0.01

−0.13

−0.37

−0.35

−0.71

1.88

−0.04

−4.69

−2.55

−0.28

0.24

−3.70

−0.99

2.08

1.59

3.25

2.82

2.00

2.31

−4.74

−5.46

−2.50

−2.61

−3.88

−4.95

2.04

−1.88

1.24

−0.21

1.94

−1.03

Panel B: The accumulated pro-cyclical impact of fair value gains and losses and loan loss provision on Basel III Tier 1 capital ratio Accumulated percentage change of starting Basel III Tier 1 capital ratio Fourth quarter 2007 to second quarter 2009

Trading revenue Loan loss provision AFS: unrealized gains and losses AFS: realized gains and losses Total fair value gains and losses Sum of fair value and loan loss provision

Third quarter 2009 to third quarter 2013

STT

C

JPM

BAC

BK

WFC

STT

C

JPM

BAC

BK

WFC

2.44 −0.14 −13.70 0.07 −11.33 −11.47

−2.94 −5.53 −1.51 0.14 −4.59 −10.12

−0.34 −3.45 −0.16 −0.22 −0.28 −3.74

−0.40 −4.22 0.53 −0.25 0.38 −3.84

3.75 −0.30 −7.26 3.02 −6.52 −6.82

0.21 −4.47 −1.19 −0.02 −0.96 −5.43

3.48 −0.10 0.88 −0.33 4.69 4.59

2.42 −6.93 −0.01 −0.74 3.14 −3.79

3.69 −3.56 1.52 −0.64 5.85 2.28

2.22 −5.14 1.35 −1.21 4.78 −0.35

3.14 −0.05 5.41 3.11 5.44 5.39

0.65 −4.43 2.31 −0.05 3.02 −1.41

Table 5.

(Continued )

Panel C: The impact of unrealized and realized gains and losses on AFS on Basel III leverage ratio and Tier 1 capital ratio (first quarter 2013 to third quarter 2013) STT

C

JPM

BAC

BK

WFC

Unrealized Realized Unrealized Realized Unrealized Realized Unrealized Realized Unrealized Realized Unrealized Realized Impact on leverage ratio (percentage point) 2013Q1 0.25 0.00 0.04 0.02 2013Q2 −0.35 0.00 −0.10 0.01 2013Q3 0.05 0.00 0.00 0.00 Impact on Tier 1 capital ratio (percentage point) 2013Q1 0.71 0.01 0.08 0.03 2013Q2 −0.96 −0.01 −0.19 0.02 2013Q3 0.14 0.01 −0.01 0.01

0.26 −0.13 0.01

0.02 0.01 0.00

0.18 −0.18 −0.03

0.00 0.02 0.05

0.39 −0.23 −0.02

0.01 0.01 0.01

0.47 −0.25 0.02

0.01 0.00 0.01

0.48 −0.22 0.01

0.04 0.01 0.00

0.33 −0.32 −0.05

0.00 0.03 0.09

1.16 −0.62 −0.06

0.04 0.03 0.02

0.64 −0.34 0.03

0.01 0.00 0.01

Data Sources: 10-Q and FRY 9-C. Note: Estimated using models in this chapter. C, Citigroup; BAC, Bank of America; JPM, JPMorgan Chase; WFC, Wells Fargo & Company; BK, The Bank of New York Mellon; STT, State Street Corporation. U.S. Interest rate data is from U.S. Federal Reserve Bank of St. Louis

The Pro-Cyclical Impact of Basel III Regulatory Capital

77

impact of unrealized gains and losses on AFS were included, at the end of the recession period, Basel III leverage ratio and Tier 1 capital ratio for STT would be −0.48 and −2.3 percentage points, and those for BK would be 2.48 and 1.74 percentage points. This clearly indicates that these banks would be critically undercapitalized. For other banks, the impact of unrealized gains or losses on AFS securities also depends on these banks’ assets structure. However, such pro-cyclical impact may overstate the actual capital risk. First, the impact of realized gains or losses on AFS securities on Basel III Tier 1 capital is much smaller for most banks. In fact, realized gains and losses on AFS securities for several banks have some positive impact over the recession period (see Panels A and B of Table 5). Therefore, Basel III may introduce substantial volatility to Tier 1 capital and regulatory capital ratios which does not reflect the actual risk. Second, during the recent expansion period, the impact of unrealized gains or losses on AFS securities turned to be positive and again was much greater than that of realized gains or losses on AFS securities for most banks. This confirms that the impact of unrealized gains or losses on AFS is volatile and the probability of it being realized is small. The impact of trading revenue and loan loss provision also shows very clear pro-cyclicality. Trading revenues of C, JPM, and BAC which hold a significant share of assets in trading assets reduced the Basel III leverage ratio and the Tier 1 capital ratio during the recession. However, they increased these ratios in the recent expansion period. Loan loss provisions of C, JPM, BAC, and WFC which hold a significant share of assets in net loans had a much greater negative impact on the Basel III leverage ratio and the Tier 1 capital ratio per quarter during the recession period than during the recent expansion period. And for these banks which hold a significant share of assets in net loans, loan loss provision in general had a much greater impact than that of fair value. However, the impact of fair value gains of JPM is greater than that of loan loss provision during the recent expansion period. If all these pro-cyclical impacts are aggregated, both the Basel III leverage ratio and the Tier 1 capital ratio for all G-SIBs will be significantly reduced during the recession period. For the Basel III leverage ratio, the range of decrease is from 2.5 to 5.46 percentage points, and for the Basel III Tier 1 capital ratio, the range of decrease is from 3.74 to 11.47 percentage points. Therefore, the new capital requirements set by BCBS (2011, 2013a) for G-SIBs may not be enough to mitigate the capital risk for all U.S. G-SIBs.5

78

GUOXIANG SONG

Interest Rate Sensitivity of Basel III Regulatory Capital Ratios and Bank Capital Risk The strong negative sensitivity of Basel III leverage ratio and Tier 1 capital ratio to interest rates may imply high bank capital risk in a changing interest rate environment. The decline of interest rates in general since 2007 may have improved Basel III regulatory capital ratios computed in this chapter. However, as the U.S. economy is recovering, its monetary policy has begun tapering quantitative easing. This suggests that long-term interest rates may increase first and then short-term interest rates will normalize gradually afterward. Such changes in the interest rate environment may reduce Basel III regulatory capital ratios for these G-SIBs. One risk generated by Basel III regulatory capital may come from unrealized gains or losses on AFS securities. For example, in May and June 2013 when U.S. FED Chairman Ben S. Bernanke first discussed the criteria for tapering, the 10-year Treasury yield increased from 1.99% for the first quarter 2013 to 2.71% for the second quarter, and it stays around this level in the third quarter 2013. It may be because of this increase of 0.72 percentage points in the long-term interest rate, that unrealized losses on AFS securities decreased the Basel III leverage ratio and the Tier 1 capital ratio of all G-SIBs in the second quarter of 2013 (see Panel C of Table 5). If the 10-year Treasury yield returns to the level of 4.85% in the first quarter of 2007 before the recent recession, then these G-SIBs may suffer significant losses in Basel III regulatory capital ratios, especially for STT and BK. To illustrate, assume that the impact of unrealized losses on AFS securities generated by the increase of the long-term interest rate remains constant during the tapering period, the increase of 2.14 percentage points in 10-year Treasury yield, that is, from 2.71% to 4.85%, may indicate that Basel III leverage ratio will decrease by 1.34 percentage points for STT and 0.87 percentage points for BK, and Basel III Tier 1 capital ratio will decrease by 3.67 percentage points for STT and 2.36 percentage points for BK. However, such impacts may still underestimate the capital risks generated by the long-term interest rate as banks will hold more HQLA to meet Basel III LCR requirement. In addition, as Basel III capital ratios have a similar negative correlation with the short-term interest rate, the normalization of the short-term interest rate after tapering may further increase bank capital risk. This risk may be much greater as the short-term rate has to increase by 4.94 percentage points from 0.04% in the third quarter 2013 to reach 4.98% which was the level in the first quarter 2007 before this crisis.

The Pro-Cyclical Impact of Basel III Regulatory Capital

79

CONCLUSIONS This chapter examines the pro-cyclical impact of Basel III regulatory capital on regulatory capital ratios and its consequent implications for bank capital risk for U.S. G-SIBs. As Basel III Tier 1 capital recognizes unrealized gains or losses on AFS securities, the pro-cyclical impact of fair value accounting on Basel III leverage ratio and Tier 1 capital ratio increases relative to that before Basel III. As unrealized gains or losses on AFS securities are sensitive to the long-term interest rate, Basel III capital standards also are more sensitive to the long-term interest rate. The magnitude of the impact of Basel III capital standards depends on the asset structure of each bank. For banks which hold a significant share of AFS securities, the impact of unrealized losses on AFS securities on Basel III capital ratios in the recent recession could be significant. And the countercyclical buffer of 2.5% plus the additional loss absorbency requirements of 2.5% for G-SIBs may not be enough to mitigate the capital risk for G-SIBs which could be generated by the business cycle. The normalization of the long-term interest rate during the tapering period may generate another significant capital risk because the impact of AFS securities also has a strong negative sensitivity to long-term interest rate. Moreover, banks may have to hold more HQLA and apply an EL approach for loan loss provisioning soon, so the capital risk may increase further.

NOTES 1. Goldman Sachs and Morgan Stanley are the remaining two U.S. G-SIFIs. They are not included in the sample because they were approved as BHCs at the end of 2008. 2. But for simplicity, Badertscher et al. (2012) estimate the impact of the unrealized losses on Tier 1 capital ratios by assuming that such losses are realized, that is, starting from Tier 1 capital ratios at the end of the reporting period, they add back the effect of such losses on risk-weighted assets to the denominator to estimate the Tier 1 capital ratios which are purged of such unrealized losses. With the model in this chapter, however, the change in risk-weighted assets is estimated as if there is a new exposure when an unrealized gain or loss occurs. The “as if” Tier 1 capital ratios is computed by starting from Tier 1 capital ratios at the beginning of the reporting period. Such results can show a more accurate impact. This is because Tier 1 capital ratios at the end of the reporting period are normally contaminated by many other factors such as capital raising during the reporting period, which may make the impact of the accounting gains or losses appear to be much less important.

80

GUOXIANG SONG

3. The results for Basel III Tier 1 capital ratio are very close to these reported here for Basel III leverage ratios. To save space, they are not reported. 4. According to the U.S. Business Cycle Dating Committee of the National Bureau of Economic Research, the recent recession period started from December 2007, and the recent expansion period started from June 2009. Therefore, in this chapter, the recent recession period is defined as fourth quarter 2007 to second quarter 2009, and the expansion period is defined as third quarter 2009 to third quarter 2013. 5. BCBS (2011) introduces 3% minimum leverage ratio, 6% minimum Tier 1 capital, 2.5% conservation buffer, and 0
2.5% countercyclical buffer. Moreover, BCBS (2013a) introduces additional loss absorbency requirements for G-SIBs, that is, 1% to 2.5% CET1 capital.

ACKNOWLEDGMENT Guoxiang Song acknowledges the helpful comments of Professor Geoff Meeks, Professor Niklas Wagner, and Professor Jonathan Batten.

REFERENCES The Agencies (Department of the Treasury and Federal Reserve System). (2013). Regulatory capital rules. Federal Register, 78(198), 62018
62291. Badertscher, B. A., Burks, J. J., & Easton, P. D. (2012). A convenient scapegoat: Fair value accounting by commercial banks during the financial crisis. The Accounting Review, 87(1), 59
90. Ball, R., Jayaraman, S., & Shivakumar, L. (2012). Mark-to-market accounting and information asymmetry in banks. Chicago Booth Research Paper No. 12-35. BCBS (Basel Committee on Banking Supervision). (2011). Basel III: A global regulatory framework for more resilient banks and banking systems (Revised version). Retrieved from http://www.bis.org/publ/bcbs189.pdf. Accessed in June. BCBS (Basel Committee on Banking Supervision). (2013a). Global systemically important banks: Updated assessment methodology and the higher loss absorbency requirement. Retrieved from http://www.bis.org/publ/bcbs255.pdf. Accessed in July. BCBS (Basel Committee on Banking Supervision). (2013b). Basel III: The liquidity coverage ratio and liquidity risk monitoring tools. Retrieved from http://www.bis.org/publ/ bcbs238.pdf. Accessed in January. Dong, M., Ryan, S. G., & Zhang, X. (2014). Preserving amortized costs within a fair-valueaccounting framework: Reclassification of gains and losses on available-for-sale securities upon realization. Review of Accounting Studies, 19(1), 242–280. ECB (European Central Bank). (2004). Fair value accounting and financial stability. European Central Bank Occasional Paper No. 13. ECB (European Central Bank). (2009). Is Basel II pro-cyclical? A selected review of the literature. Financial Stability Review, December, 143
150.

The Pro-Cyclical Impact of Basel III Regulatory Capital

81

FSA. (2009). The Turner review: A regulatory response to the global banking crisis. Retrieved from http://www.fsa.gov.uk/pubs/other/turner_review.pdf. Accessed in March. FSB (Financial Stability Board). (2011). Policy measures to address systemically important financial institutions. Retrieved from http://www.financialstabilityboard.org/publica tions/r_111104bb.pdf. Accessed in November 4. Greenspan, A. (1990, November 1). Letter to Hon. Richard C. Breeden, Federal Reserve. IASB (International Accounting Standards Board).(2013a). IASB update. Retrieved from http://media.ifrs.org/2013/IASB/November/IASB-Update-November-2013.pdf. Accessed in November. IASB (International Accounting Standards Board). (2013b). Exposure draft ED/2013/3 financial instruments: Expected credit losses. Retrieved from http://www.ifrs.org/Current-Projects/ IASB-Projects/Financial-Instruments-A-Replacement-of-IAS-39-Financial-InstrumentsRecognitio/Impairment/Exposure-Draft-March-2013/Comment-letters/Documents/EDFinancial-Instruments-Expected-Credit-Losses-March-2013.pdf IMF (International Monetary Fund). (2008). Chapter 3: Fair value accounting and procyclicality. Global Financial Stability Report, October 2008. ISDA. (2012). Letter to Mr Stefan Ingves, Chairman Basel Committee and Mr Hans Hoogervorst, Chairman International Accounting Standards Board, 29 June. Laux, C., & Leuz, C. (2010). Did fair-value accounting contribute to the financial crisis? Journal of Economic Perspectives, 24, 93
118. Saunders, A., & Cornett, M. (2011). Financial institutions management: A risk management approach (7th ed.). New York, NY: McGraw-Hill. SEC. (2008). Report and Recommendations Pursuant to Section 133 of the Emergency Economic Stabilization Act of 2008: Study on Mark-To-Market Accounting. Shaffer, S. (2010). Fair value accounting: Villain or innocent victim
exploring the links between fair value accounting, bank regulatory capital and the recent financial crisis. FRB of Boston Quantitative Analysis Unit Working Paper No. 10
01. Sill, K. (1996). The cyclical volatility of interest rates. Business Review, January
February, pp. 15
29. Song, G. (2012). Can accounting rules be made neutral for bank capital regulation? Journal of Governance and Regulation, 1(3), 27
35. Tarullo, D. K. (2011). The evolution of capital regulation, speech at the Clearing House Business Meeting and Conference. New York, NY, November 9.

NONPARAMETRIC EXPECTILE REGRESSION FOR CONDITIONAL AUTOREGRESSIVE EXPECTED SHORTFALL ESTIMATION Marcelo Brutti Righi, Yi Yang and Paulo Sergio Ceretta ABSTRACT In this chapter, we estimate the Expected Shortfall (ES) in conditional autoregressive expectile models by using a nonparametric multiple expectile regression via gradient tree boosting. This approach has the advantages generated by the flexibility of not having to rely on data assumptions and avoids the drawbacks and fragilities of a restrictive estimator such as Historical Simulation. We consider distinct specifications for the information sets that produce the ES estimates. The results obtained with simulated and real market data indicate that the proposed approach has good performance, with some distinctions between the specifications. Keywords: Risk management; expected Shortfall; nonparametric expectile regression; gradient tree boosting

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 83
95 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096003

83

84

MARCELO BRUTTI RIGHI ET AL.

INTRODUCTION In recent years, concerns with financial risk management have increased because of the frequent collapses in the financial system. This worry has spurred the development of risk management tools. A milestone of such tools is the risk measurement, which has been at the center of concepts in risk measures. Despite the fact that volatility has been synonymous with risk for a long time, when the downside risk is of primary concern, as in extreme bad events, the upside and downside movements of returns must receive different treatments. Downside risk measures are also relevant for purposes of regulation and capital requirements, issues that are crucial in modern financial systems. The leading downside risk measure for practical use is currently Value at Risk (VaR). VaR represents the maximum loss given a confidence level during a certain period, that is, the quantile of losses distribution. However, despite its simplicity and popularity, VaR has certain drawbacks, for instance, as the quantile with a given tail probability, it does not consider losses above the quantile of interest, and it depends only on the probability of extreme losses, remaining insensitive to their magnitude. Additionally, VaR is not sub-additive, that is, despite diversification, the VaR of a portfolio could be greater than the VaR of individual assets from the same portfolio. Artzner, Delbaen, Eber, and Heath (1999) note such deficiencies of VaR by demonstrating that it is not a coherent measure1 of risk, and they propose the Expected Shortfall (ES) as a risk measure to overcome such shortcomings. ES is defined as the average loss given that overcomes the VaR. Thus, ES considers the magnitude of losses instead of the quantile of interest alone, and it is coherent, as noted by Acerbi and Tasche (2002), Tasche (2002), and Rockafellar and Uryasev (2002). In addition to the academic discussions of its advantages, ES is becoming increasingly widely used in the financial industry. In that sense, Yamai and Yoshiba (2005) discuss the comparison between VaR and ES from a practical point of view. These risk measures are in essence theoretical concepts. Therefore, for real applications, it is necessary to develop an estimation method to compute numeric values for these risk measures. Despite the emergence of new approaches, this is still a developing field of study because much negligence regarding the estimation of the risk measures remains. Seeking to improve the estimation for risk measures, scholars realize that current and historical information regarding financial markets, economies and even policies can

Nonparametric Expectile Regression for ES Estimation

85

be useful. For estimating VaR, quantile regression, which directly estimates the dynamics of quantiles, is the dominant method. Despite the early approach by Chernozhukov and Umanstev (2001), work on the conditional autoregressive Value at Risk (CAViaR) models (see Engle & Manganelli, 2004), which estimate VaR conditioned to VaR lagged values, is the most prominent in the literature. Other advances for risk management from CAViaR models are presented in Taylor (2008a), Gerlach, Chen, and Chan (2011), Chen, Gerlach, Hwang, and McAleer (2012), Rubia and SanchisMarco (2013), and Fuertes and Olmo (2013). Regarding comparisons of the VaR estimation approaches, Bao, Lee, and Saltoglu (2006), Berkowitz, Christoffersen, and Pelletier (2011), and Mabrouk and Saadi (2012) favor the CAViaR models, emphasizing the quality of the procedures. In the same spirit, it is possible to estimate VaR and ES by the dynamics of expectiles, which depend on both the tail realizations and their probability, whereas only the tail probability determines quantiles. Expectile regression has some advantages over quantile regression. For example, it is computationally friendlier and makes more efficient use of the available data compared to quantiles. This is because expectile estimations rely on the distance to data points, while quantiles only reflect whether an observation is above or below the predictor. Of course, the increased efficiency comes at a price of sensitivity to outliers. Taylor (2008b) and Kuan, Yeh, and Hsu (2009) extend the CAViaR concept to estimate VaR and ES using a new class of expectile models
conditional autoregressive expectiles (CARE). The basic idea is to link VaR and ES directly to conditional expectiles and then to estimate these risk measures through expectile regressions by modeling the conditional expectiles as different parametric functions of the information. However, parametric expectile regression models can be too rigid to fit complex nonlinear relationships for many real applications. Yao and Tong (1996) consider nonparametric expectile regression; however, their model is restricted by the number of explanatory variables. Aiming to improve this type of model, Yang and Zou (2013) propose a method that adopts the gradient tree-boosting algorithm (e.g., Friedman, Hastie, & Tibshirani, 2000; Schapire, 2003) to derive a fully nonparametric multiple expectile regression. This technique enjoys important advantages of tree-based methods: it can fit complex nonlinear relationships and can easily incorporate interaction effects between predictors in the final estimator, reducing the potential for modeling bias. More importantly, by combining many simple tree models adaptively, the technique can give improved predictive performance over single tree models.

86

MARCELO BRUTTI RIGHI ET AL.

With that information in mind, in this chapter, we present a new nonparametric ES estimation procedure using the model of Yang and Zou (2013). This study is a contribution to the literature because our method can estimate the dynamics of downside measures without assumptions about innovation distribution or functional forms, particularly linearity restriction, of variable relationships in a way that is not as sensitive to changes in the estimation window and the sudden or late reaction to market moves as an empirical estimator would be. Thus, nonparametric multiple expectile models can be an interesting and efficient alternative to the ES estimation, once new robust methods enter into the discussion. The remainder of this chapter has the following structure: the section “Proposed ES Estimation Procedure” briefly presents the concept of risk measures, the principles of quantile and expectile regressions and the nonparametric multiple expectile model, and the proposed ES estimator. The section “Empirical Properties of the Proposed Procedure” presents a simulation experiment to verify certain properties of the proposed estimator as well as an empirical illustration with real data to complement the analysis. Finally, the section “Conclusion” summarizes and concludes the chapter.

PROPOSED ES ESTIMATION PROCEDURE We begin by defining ES. To that purpose, let X be the series of returns of a financial position with probability distribution function F. Thus, VaR at a significance level α ∈ (0,1) is the quantile qα of F. Mathematically,   VaRα = qα ðXÞ = inf q : FðqÞ ≥ α

ð1Þ

Based on this definition, we note that VaR does not consider information after the quantile of interest. ES can address this drawback. The ES at significance level α is then the expectation of X, once X is below VaR, ESα = E½XjX < VaRα = qα ðXÞ = α − 1

Z

α

qs ðXÞds

ð2aÞ

0

One can directly estimate the dynamics of the risk measures VaR and ES by expectile regression. We focus here on the ES, although VaR estimation is also possible. We noted the connection between the ES and

Nonparametric Expectile Regression for ES Estimation

87

expectiles. First, we should consider that the population ω-expectile of X is the solution of the following minimization problem over m, τω = argmin E½jω − 1X < m jðX − mÞ2 

ð2bÞ

m

By straightforward algebra, one can show that the solution τ satisfies the expression   1 − 2ω
E ðX − τω Þ1X < τω = τω − EðXÞ ω

ð2cÞ

This suggests a link between ES and expectiles τω . We can rearrange Eq. (2c) as  EðXjx < τω Þ = 1 þ

ω ω EðXÞ Tω − ð1 − 2ωÞFðτω Þ ð1 − 2ωÞFðτω Þ

ð2dÞ

where F is the cdf of X. Because Jones (1994) demonstrates that there is a one-to-one correspondence between expectiles and the quantiles of a related distribution, (2d) actually provides an expression for the ES through expectiles when the α-quantile coincides with the ω-expectiles, that is, VaRα = qα = τω and F(τω) = α. One can rewrite (2d) as  ESα = 1 þ

ω ω EðXÞ τω − ð1 − 2ωÞα ð1 − 2ωÞα

ð2eÞ

If X is defined as a zero mean term E(X) = 0, this further simplifies to the expression,  ESα = 1 þ

ω τω ð1 − 2ωÞα

ð2fÞ

Because the estimation of ESα is equivalent to the estimation of the corresponding expectile , the idea is to estimate the conditional expectile as a function h of past information ΨT. As there is no need for distributional assumptions for X, this type of model is semi-parametric if hω is a complex

88

MARCELO BRUTTI RIGHI ET AL.

function of the linear combination of ΨT; or, it is nonparametric if hω is simply a complex function of ΨT. To estimate the function hω, we use the Asymmetric Least Squares proposed by Newey and Powell (1987): τ^ ω ðXT jΨT Þ = h^ω ðΨT Þ = argminhω

X XT

jω − 1XT < hω ðΨT Þ jðXT − hω ðΨT ÞÞ2

ð3Þ

Based on this framework, Taylor (2008a, 2008b) and Kuan et al. (2009) extend the CAViaR model of Engle and Manganelli (2004) using a CARE approach for ES estimation, that is, lagged values of ES are present in ΨT. In the literature, distinct functional forms for hω in CARE modeling are used. The argument is that these parametric forms can restrict the dynamics of the conditional expectile. In that sense, one can use a nonparametric specification. Considering the canonical bivariate nonparametric model XT = μ(ΨT) + σ(ΨT)·E, where μ and σ are, respectively, the mean and standard deviation operators, and є is a random residual term independent of ΨT. From this approach, it is easy to obtain the relation τω (XT|ΨT = ψ) = hω (ψ) = μ(ψ) + σ(ψ)·τω (E). Yao and Tong (1996) explore such a relation, developing a local linear estimator of the expectile τω (XT|ΨT) when ΨT has one dimension. Although theoretically feasible, in practice, it is very difficult to extend their method to the multiple regression case because local regression suffers from the so-called “curse-of-dimensionality.” To solve this limitation, Yang and Zou (2013) introduce a tree-based boosting estimator for multiple nonparametric expectile regression. The technical details are available in their paper, and we summarize the idea here. They model the conditional expectiles by combining multiple regression trees. The final model is obtained by solving the minimization problem (3) using functional gradient descent. The fundament of a boosting procedure is to combine many prediction models in a way such that the combined model has a superior prediction performance. Gradient boosting uses an iterative procedure that sequentially updates the estimator and then stops after a sufficient number of iterations. These authors provide an implementation of the algorithm in the R package erboost. Thus, with the estimates of conditional expectiles τω (XT|ΨT) obtained through the nonparametric multiple expectile regression, it is possible to compute ES at the significance level α. As we discussed, we consider the direct relation between estimated hconditionali expectiles and ES through a correction factor, conform ESαT = 1 þ ð1 −ω2ωÞα τω ðXT jΨT Þ. We use expectiles

Nonparametric Expectile Regression for ES Estimation

89

as estimators of quantiles. For a fixed value of α, we select the value of such that the proportion of in-sample observations lying below the estimated conditional expectiles τ^ ω ðXT jΨT Þ is α. Once we establish the estimation procedure, the choice remains of which variables to include in the information set ΨT. Under the parametric functional forms, the most used in the literature are the Symmetric Absolute Value (SAV), the Asymmetric Slope (AS), and the Indirect GARCH
Generalized Autoregressive Conditional Heteroscedastic
(IG). The SAV considers that the past information of X equally affects the conditional expectile, while the AS assumes that there is a difference regarding the impact of positive and negative values for previous returns. For its part, the IG forces data to follow a pattern of GARCH specification. Formulations (4
6) define such specifications, respectively. ESαT = CFα ½γ 0 þ γ 1 τω ðXT − 1 jΨT − 1 Þ þ γ 2 jXT − 1 j

ð4Þ

ESαT = CFα ½γ 0 þ γ 1 τω ðXT − 1 jΨT − 1 Þ þ γ 2 ðXT − 1 Þ þ þ γ 3 ðXT − 1 Þ − 

ð5Þ

n 1 ESαT = ð1 − 21α < 0:5 Þ½γ 0 þ γ 1 ðτω ðXT − 1 jΨT − 1 ÞÞ2 þ γ 2 XT2 − 1 2 g

ð6Þ

h i Where CFα = 1 þ ð1 −ω2ωÞα ; ðxÞ þ = maxðx; 0Þ; ðxÞ − = − minðx; 0Þ, and 1p is the indicator function used to ensure the correct sign for the ES over long and short positions once the functional term of IG gives only positive values by definition. Although more lagged information can be included in the sets ΨT, as is usually done in the literature, we restrict it to one lag. We adapt these specifications here for a nonparametric approach by bearing in mind the relation τω (XT|ΨT) = μ(ΨT) + σ(ΨT) · τω(E). Despite the nonparametric functional form, hω is the same for all models, and the information set varies. Thus, keeping the nomenclature, we get that for the SAV, AS, and IG approaches, the information set ΨT is, respectively, {τω (XT− 1|ΨT − 1),| XT − 1|}, {τω (XT − 1|ΨT − 1), (XT − 1)+, (XT − 1)−}, and ½ðτω ðXT − 1 j 1 1 ΨT − 1 ÞÞ2 2 ; ðXT2 − 1 Þ2 g. Hence, it is possible to estimate the ES in a flexible way, without being exposed to the dangers of the HS estimation procedures.

90

MARCELO BRUTTI RIGHI ET AL.

EMPIRICAL PROPERTIES OF THE PROPOSED PROCEDURE In this section, we explore the efficiency of the proposed estimation procedure with regard to risk management matters. A good risk model produces estimates that match the real evolution of financial data. We explore results from simulated and real market data. For best comprehension, we split them into two sub-sections. Simulated Data We assess the efficiency of the proposed ES estimation procedure in the context of a Monte Carlo study. Consider a portfolio where the returns X are drawn from an AR (1)
GARCH (1,1), conform formulations (7
9). XT = 0:50XT − 1 þ ɛT

ð7Þ

ɛ T = σ T zT ; zT e t 8

ð8Þ

σ 2T = 4e − 6 þ 0:10ɛ2T − 1 þ 0:85σ 2T

ð9Þ

Where for period T, XT is the return, σ 2T is the conditional variance, T is the innovation in the conditional mean, and ZT represents a student distributed white noise series. We choose this particular data-generating process to match a realistic representation of equity portfolio returns to contemplate stylized facts such as volatility clusters and heavy tails. Under this specification, we simulate 10,000 processes of length 1,000. Once we know for sure the data-generating process, we also have the real VaR and ES values, which are presented in formulations (10) and (11). VaRαT = 0:50XT − 1 þ σ T t8− 1 ðαÞ

ESαT

 = 0:50XT − 1 þ σ T α

−1

Z 0

α

t8− 1 ðsÞds

ð10Þ ð11Þ

91

Nonparametric Expectile Regression for ES Estimation

For each simulated process, we estimate nonparametric expectile regression, conform explained in the previous section, under the SAV, AS, and IG information sets. We choose this exercise, which is similar to an insample estimation, because the computational cost of using an estimation window and updating parameters at each step would be too costly. We consider the significance level of 1% because it is recommended for regulation by the Basel committee. For all of these sets, we compute the bias and standard error between the estimated and real ES. Moreover, we perform the backtesting for ES estimates of McNeil and Frey (2000), which verifies if returns over a VaR standardized by ES have a mean of zero, with 1,000 bootstraps. The alternative hypothesis of the test is that such returns have a mean larger than zero, that is, it is a one-sided test for risk underestimation. Figure exhibits plots of the densities of the bias, standard error and backtesting p-values for the three specifications used. Table 1 numerically summarizes these results. Fig. 1 and Table 1 indicate, initially, that the bias from the SAV and AS models is slightly negative, with values in the third decimal. The IG specification is more conservative, with a bias around four times larger. This emphasizes a conservative pattern of in-sample estimations, which is corroborated by the p-values obtained for the backtesting approach. The estimates do not come close to the rejection of the null hypothesis for any simulated sample, which would indicate risk underestimation. Regarding standard errors, the dispersion of the bias is very similar for all three specifications. In short, simulated results point to a good performance of the proposed CARE estimation procedure, especially under the SAV and AS specifications. The IG information set can lead to risk overestimation, which, despite being preferable to risk underestimation, implies that the IG set could make, for instance, a financial institution retain an excess of capital that would otherwise be applied to earnings.

Table 1. Mean Bias, Standard Error, and p-Value of the Backtesting for the ES Estimates at the 1% Significance Level for the Simulated Samples. Model Symmetric absolute value Asymmetric slope Indirect GARCH

Bias

Standard Error

p-Value

−0.0062 −0.0061 −0.0225

0.0130 0.0126 0.0129

0.9996 0.9996 1.0000

92

MARCELO BRUTTI RIGHI ET AL.

Fig. 1. Densities of the Bias, Standard Error, and p-Value of the Backtesting for the ES Estimates at the 1% Significance Level for the Simulated Samples.

Real Market Data Any risk model must be able to forecast risk measures in practice. For this empirical illustration, we consider daily data for the S&P500 since its inception in March 4, 1957 to October 25, 2013, totaling 14,262 observations. This is the main index for the equity market, representing the most relevant stocks in the United States. This is an extensive sample, which considers both turbulent and calm periods. Regarding ES computation issues, we apply the same SAV, AS, and IG specifications for the 1% significance level for the data log-returns. We consider an estimation window of 2,000 days, that is, each ES prediction is based on the last four years of data as a result of the updates to the models. To simplify, we use VaR as the empirical 1% quantile of the estimation window. Fig. 2 exhibits plots of the data and estimated ES with the three specifications. Complementing the models,

93

Nonparametric Expectile Regression for ES Estimation Symmetric Absolute Value

Fig. 2.

0.10 0.05 0.00

0.00 –0.20 –0.15 –0.10 –0.05

1960 1970 1980 1990 2000 2010

Indirect GARCH

–0.20 –0.15 –0.10 –0.05

0.00 –0.20 –0.15 –0.10 –0.05

0.05

0.05

0.10

0.10

Asymmetric Slope

1960 1970 1980 1990 2000 2010

1960

1970 1980 1990 2000 2010

Daily Log-Returns and ES Estimates for S&P500 from SAV, AS, and IG Specifications for the CARE Proposed Method.

Table 2. Mean, Standard Deviation, and p-Value of the Backtesting for the ES Estimates at the 1% Significance Level for the S&P500. Model Symmetric absolute value Asymmetric slope Indirect GARCH

Mean

Deviation

p-value

−0.0339 −0.0339 −0.0589

0.0106 0.0106 0.0245

0.2264 0.2003 1.0000

Table 2 numerically summarizes the results, just as Table 1 does for the simulated results. The plots in Fig. 2 visually indicate that the conditional nature of the proposed ES estimation procedure produces estimates that follow the dynamics of the data. It is perceptible that especially in turbulent periods such as Black Monday in 1987 (the worst result in the sample), actual returns can exceed the estimates, and because by definition ES is an expectation, it is reasonable for deviations to occur. The SAV and AS lead to very similar estimates, whereas the IG, conform pointed out in the simulated results, is more conservative. Corroborating this distinction concerning the information sets are the results in Table 2. Regarding dispersion, as emphasized by Danı´ elsson (2002), risk forecasts fluctuate considerably from one period to the next so that if two estimation models give good risk predictions, the less volatile tends to be preferred. Again, the IG presents the more volatile estimates, while the SAV and AS are quite similar. Finally, none of

94

MARCELO BRUTTI RIGHI ET AL.

the models rejects the null hypothesis of the backtest at the usual statistical levels. However, as this is a one-sided test, the IG that has very conservative predictions does not necessarily have the best performance. Thus, as for the simulated samples, the SAV and AS specifications are more recommended, indicating that the proposed procedure leads to good results.

CONCLUSION In this chapter, we present a procedure for ES estimation based on nonparametric multiple expectile regression with tree-based gradient boosting. Such an approach borrows from the flexibility of an empirical estimator, without incurring the risks and fragilities of an HS method. We consider the SAV, AS, and IG specifications for the CARE model, which are often used in the financial literature. The results based on both simulations and real market data indicate that the proposed procedure has had good performance, especially under the SAV and AS information sets. Thus, this approach emerges as an interesting option for ES-based risk management in practical application. For future research, we suggest that the procedure be applied to distinct asset classes and estimation scenarios, such as different significance levels and estimation windows. Moreover, one can also conduct a comparison between the proposed model and other concurring techniques.

NOTE 1. A risk measure is coherent if it simultaneously meets the axioms of translation invariance, positive homogeneity, subadditivity, and monotonicity.

REFERENCES Acerbi, C., & Tasche, D. (2002). On the coherence of expected shortfall. Journal of Banking & Finance, 26, 1487
1503. Artzner, P., Delbaen, F., Eber, J. M., & Heath, D. (1999). Coherent measures of risk. Mathematical Finance, 9, 203
228. Bao, Y., Lee, T.-H., & Saltoglu, B. (2006). Evaluating predictive performance of value-at-risk models in emerging markets: A reality check. Journal of Forecasting, 25, 101
128. Berkowitz, J., Christoffersen, P., & Pelletier, D. (2011). Evaluating value-at-risk models with desk-level data. Management Science, 57, 2213
2227.

Nonparametric Expectile Regression for ES Estimation

95

Chen, C. W. S., Gerlach, R., Hwang, B. B. K., & McAleer, M. (2012). Forecasting value-atrisk using nonlinear regression quantiles and the intraday range. International Journal of Forecasting, 28, 557
574. Chernozhukov, V., & Umanstev, L. (2001). Conditional value-at-risk: Aspects of modeling and estimation. Empirical Economics, 26, 271
292. Danı´ elsson, J. (2002). The emperor has no clothes: Limits to risk modelling. Journal of Banking & Finance, 26, 1273
1296. Engle, R. F., & Manganelli, S. (2004). CAViaR: Conditional autoregressive value at risk by regression quantiles. Journal of Business & Economic Statistics, 22, 367
381. Friedman, J., Hastie, T., & Tibshirani, R. (2000). Additive logistic regression: A statistical view of boosting (with discussion and a rejoinder by the authors). Annals of Statistics, 28, 337
407. Fuertes, A.-M., & Olmo, J. (2013). Optimally harnessing inter-day and intra-day information for daily value-at-risk prediction. International Journal of Forecasting, 29, 28
42. Gerlach, R. H., Chen, C. W. S., & Chan, N. Y. C. (2011). Bayesian time-varying quantile forecasting for value-at-risk in financial markets. Journal of Business & Economic Statistics, 29, 481
492. Jones, M. (1994). Expectiles and m-quantiles are quantiles. Statistical Probability Letters, 20, 149
153. Kuan, C.-M., Yeh, J.-H., & Hsu, Y.-C. (2009). Assessing value at risk with CARE, the conditional autoregressive expectile models. Journal of Econometrics, 150, 261
270. Mabrouk, S., & Saadi, S. (2012). Parametric value-at-risk analysis: Evidence from stock indices. The Quarterly Review of Economics and Finance, 52, 305
321. McNeil, A., & Frey, R. (2000). Estimation of tail-related risk measures for heteroscedastic financial time series: an extreme value approach. Journal of Empirical Finance, 7, 271
300. Newey, W., & Powell, J. (1987). Asymmetric least squares estimation and testing. Econometrica, 55, 819
847. Rockafellar, R. T., & Uryasev, S. (2002). Conditional value-at-risk for general loss distributions. Journal of Banking & Finance, 26, 1443
1471. Rubia, A., & Sanchis-Marco, L. (2013). On downside risk predictability through liquidity and trading activity: A dynamic quantile approach. International Journal of Forecasting, 29, 202
219. Schapire, R. (2003). The boosting approach to machine learning: An overview. Nonlinear Estimation and Classification, Lecture Notes in Statistics, 171, 149–171. Tasche, D. (2002). Expected shortfall and beyond. Journal of Banking & Finance, 26, 1519
1533. Taylor, J. W. (2008a). Using exponentially weighted quantile regression to estimate value at risk and expected shortfall. Journal of Financial Econometrics, 6, 382
406. Taylor, J. W. (2008b). Estimating value at risk and expected shortfall using expectiles. Journal of Financial Econometrics, 6, 231
252. Yamai, Y., & Yoshiba, T. (2005). Value-at-risk versus expected shortfall: A practical perspective. Journal of Banking & Finance, 29, 997
1015. Yang, Y., & Zou, H. (2014). Nonparametric multiple expectile regression via ER-Boost. Journal of Statistical Computation and Simulation. (in press). doi:10.1080/00949655.2013.876024 Yao, Q., & Tong, H. (1996). Asymmetric least squares regression estimation: A nonparametric approach. Journal of Nonparametric Statistics, 6(2
3), 273
292.

THE IMPACT OF EXTERNAL SHOCKS ON STOCK PRICES IN THE EAST ASIAN DOMESTIC BANKING SECTOR Masahiro Inoguchi ABSTRACT This chapter examines the impact of price fluctuations in foreign stock markets on the stock prices of domestic banks in Korea, Malaysia, Singapore, and Thailand. Some studies have argued that the 2007
2009 global financial crisis (GFC) affected domestic banks less in East Asia, even though the supporting evidence is rather limited. Employing a multinomial logit model, we estimate how changes in the United States and Japanese stock markets affected the banking sectors in the sampled countries before the 1997 Asian financial crisis, and before and during the more recent GFC. We interpret the number of banks in a given country that experienced a large price shock on the same day (or “coexceedance”) as shocks to the domestic banking sector. The results suggest that fluctuations in foreign stock market indices exerted a larger impact on the prices of East Asian banking stocks during the 2000s than during the 1990s. In addition, although the shocks brought about by the

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 97
151 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096004

97

98

MASAHIRO INOGUCHI

deterioration of foreign stock markets were significant before the GFC, both increases and decreases in foreign stock prices significantly affected the banking sectors of the respective countries during the crisis. Lastly, we conclude that increasing foreign capital flows and foreign assets and liabilities greatly influenced domestic banking systems in East Asia during the 2000s. Keywords: Domestic banking sector; stock prices; foreign stock market; external influences; Asia JEL classifications: F36; G01; G15; O16

INTRODUCTION Asian monetary authorities have continued to reform their domestic financial systems, including banking, since the 1997 Asian financial crisis. A sound banking system is important for any economy. This is particularly true in East Asia, where a significant number of companies rely on bank loans for financing. Therefore, while financial systems and policies differ throughout Asia, all Asian countries have attempted to consolidate their domestic banking sectors. For instance, Korea eased various restrictions as a means of strengthening the efficiency of its banking system, while Malaysia implemented its 2001 Financial Master Plan to promote the stability and competitiveness of its domestic banking sector. Thai authorities have also relaxed a range of financial market regulations to improve the operation of their financial and banking systems. Lastly, although the Asian financial crisis affected the financial system less severely in Singapore, at least when compared with Korea, Malaysia, and Thailand, Singaporean authorities have also deregulated their financial markets, including the banking sector, and improved the provision of their financial sector infrastructure to the point where it is now Asia’s preeminent international financial center. There is then the argument that the East Asian financial system has become sounder through these restructurings than it was before 1997.1 Certainly, the financial shocks in the aftermath of the 2007
2009 global financial crisis (GFC) do not appear as large as they could have been, especially when compared with the serious damage suffered by many European banks. Unfortunately, it is unclear why these consequences have been so small for banks in the East Asian banking system, although some argue

External Shocks on Stock Prices in East Asian Domestic Banking

99

that they are consistent with improvements in the soundness of the various domestic banking sectors resulting from the reforms undertaken after 1997. The question is then whether the damage to the Asian banking sector from large external shocks was reduced because of bank improvements made before the GFC or because the shocks themselves were not as significant as those previously experienced. One of the reasons why there is no consensus about this is because few studies have investigated the impact of external financial shocks on the banking sector in Asia and how these influences changed after the 1997 Asian financial crisis. Therefore, in this chapter we analyze how shocks in the foreign financial sector have affected the banking sectors of a select number of Asian countries since the 1990s, namely Korea, Malaysia, Singapore, and Thailand.2 In these countries, the influence of external shocks on domestic financial markets, including the banking system, had the potential to be larger after 2007 than before 1997 because in common with most other Asian countries, they removed their de facto dollar-pegged exchange rate systems and relaxed their regulations governing capital flows. For example, Korea and Thailand ceased pegging their currencies to the US dollar and adopted inflation as a monetary policy target. Similarly, Korea eased the ceiling on the foreign ownership of domestic banks in 1998, and as a result, foreign ownership of Korean banks gradually grew. Thailand also relaxed its restrictions on the foreign ownership of domestic banks, while in 1999 Singapore eased its restrictions on both the foreign ownership of domestic banks and the allowable activities of foreign banks. In contrast, Malaysia first imposed controls on foreign capital flows and introduced a dollarpegged exchange rate system in light of the 1997 crisis, but afterward gradually relaxed its capital controls. A number of existing studies have theoretically analyzed the crossborder contagion entailed in banking crises. Freixas, Parigi, and Rochet (2000), for example, describe how the insolvency of one bank can generate systemic risk through the interbank market and the withdrawal of deposits, even if all other banks remain solvent.3 In addition, Freixas et al. (2000) argue that their model allows them to consider the spread of a financial crisis from country to country. Elsewhere, Cifuentes, Ferrucci, and Shin (2005) suggest that contagious failures of banking systems can result when a small liquidity deficit in one bank prompts it to sell assets, thereby depressing market prices and prompting further rounds of sales and price declines. Together, the findings in Freixas et al. (2000) and Cifuentes et al. (2005) imply that banking failures can easily spread between countries, even those without any direct interbank connections.

100

MASAHIRO INOGUCHI

While much empirical literature is evident concerning bank failures via domestic contagion, few studies examine the cross-border effects across banking systems.4 An exception is Gropp, Lo Duca, and Vesala (2009), who use a market-based indicator of bank soundness to estimate crossborder contagion in European banks.5 Their empirical method follows that in Bae, Karolyi, and Stulz (2003), using interest rates, exchange rates, and the stock market indices of individual countries to evaluate contagion in financial markets among regions (Asia, Latin America, the United States, and Europe). Bae et al. (2003) first arbitrarily define an extreme return (“exceedance”) as being either below the 5th or above the 95th percentile of the distribution of daily returns. They then employ the number of countries in a region that experience a large shock on the same day (or “coexceedances”) to capture the coincidence of extreme return shocks across countries within a region and across regions. We also select this approach, as it enables us to measure this influence both across and within regions. Another benefit of this method is that it focuses solely on large fluctuations, and thus is useful for exploring major financial shocks across countries. For instance, in order to examine the cross-border contagion effect, Eichengreen, Rose, and Wyplosz (1996) define a crisis as the extreme values of an index comprising the exchange rate, foreign reserves, and interest rates. Sachs, Tornell, and Velasco (1996) also employ very large changes in exchange rates and bank loans as dependent variables to explain the variations in financial crises across countries. Lastly, Forbes and Warnock (2012) use changes in capital flows more than one standard deviation from the historical average as extreme capital flow movements to identify factors affecting large gross capital flows, and discuss these in the context of financial crises and capital flow volatility. This study also utilizes the method used in Gropp et al. (2009) and Bae et al. (2003) to analyze the effect of external financial shocks on East Asian banking systems, particularly in Korea, Malaysia, Singapore, and Thailand. Our regression analysis employs the daily change in the prices of domestic bank stocks and a multinomial logit model to estimate the influence of Japanese and US financial shocks on domestic banking systems in these four countries. In addition, we examine whether these influences differed before and during the 2007
2009 GFC. Our findings suggest that fluctuations in the US and Japanese stock markets did indeed influence the stock prices of East Asian banks more in the 2000s than in the 1990s and the period before the 1997 Asian financial crisis. One reason for this may be that increasing foreign capital flows and foreign assets and liabilities in these countries exerted a greater influence in

External Shocks on Stock Prices in East Asian Domestic Banking

101

the 2000s. We also find that both upturns and downturns in these foreign stock markets influenced these Asian banking sectors during the 2007
2009 crisis, whereas before 2007
2009 only downturns in global markets had a significant impact on Asian bank stocks. In addition, we find that fluctuations in foreign stock markets had less effect on banks in Singapore than on those in the other Asian countries. The remainder of the chapter is structured as follows. The section “Banking Sector Reforms and Bank Balance Sheets” describes the restructuring of the financial sectors in Korea, Malaysia, Singapore, and Thailand. Section “An Overview of Domestic Bank Stock Prices and Stock Market Indices” graphically illustrates the trends in stock prices and presents data regarding the external assets and borrowings of the banking sector. The section “Regression Analysis” discusses the methods of estimating the effects of external shocks on the domestic banking systems in Korea, Malaysia, Singapore, and Thailand using daily stock prices. The last section provides some concluding remarks.

BANKING SECTOR REFORMS AND BANK BALANCE SHEETS After the 1997 crisis, consolidation took place in the Asian banking sector, and foreign banking sector transactions increased. Both of these developments had the potential to affect how external shocks influence the domestic banking sector. Before analyzing foreign financial shocks to domestic banking systems, in this section we review the banking sector reforms and the foreign assets and liabilities of the domestic banking sectors in Korea, Malaysia, Singapore, and Thailand.

Banking Sector Reforms after 1997 In the most crisis-affected countries, authorities attempted to address the problems of bank capitalization, governance, risk management, and operational inefficiencies in the aftermath of the 1997 crisis. A common response was the closure or consolidation of banks and temporary nationalization. For instance, Korea, Malaysia, Singapore, and Thailand all consolidated their banking sectors based on their plans to reorganize banks, which included the enhanced supervision of the banking system and the relaxation

102

MASAHIRO INOGUCHI

of any existing restrictions on bank ownership. For example, the level of foreign ownership in domestic banks increased, while the regulation of foreign banking was substantially relaxed. In Korea, the Bank Act was revised along with a number of bank closures, capital infusions, and the purchase of nonperforming bank loans. Korea’s government established a restructuring plan to improve and strengthen the soundness and efficiency of its financial system by relaxing the restrictions on foreign ownership and tightening capital adequacy regulations. Foreign ownership of domestic banks increased and remained at a high level. In order to encourage competition between banks and to improve the competitiveness of the domestic banking sector, the Korean authorities eased various other restrictions, including, for example, the gradual liberalization of rules governing the establishment of new bank branches in Korea. In Malaysia, the Bank Negara Malaysia implemented its 2001 Financial Master Plan with the intention of promoting the stability and competitiveness of the Malaysian financial system over the next ten years. Although this plan covered banking and insurance sectors, in the first stage the domestic banking system was encouraged to enhance its stability and capabilities to develop several powerful domestic banks in the initial 2
4 years of the plan. In the second stage, the Malaysian authorities relaxed the restrictions on foreign bank activity in the country and promoted competition between domestic and foreign banks in order to strengthen the competitiveness of the domestic banking sector. The plan is currently in its third stage, with the objective of boosting this competition between domestic and foreign banks even further. Ultimately, Bank Negara Malaysia intends to ease the regulations concerning the entry of new foreign banks into Malaysia and to facilitate the overseas activities of its domestic banks. For instance, in 2009, the authorities relaxed the restrictions on the opening of new branch offices of existing foreign banks and issued new licenses for other foreign banks to operate in Malaysia. In Singapore, the Monetary Authority of Singapore introduced the 1998 Financial Sector Review to promote reforms stimulating competition between domestic and foreign banks. This plan included improving the Singaporean capital markets as well as promoting various deregulations of the financial system in order for Singapore to emerge as Asia’s financial center. Since 1999, further liberalization of the banking sector has eliminated restrictions on foreign ownership of domestic banks, relaxed the regulations governing the activities of foreign banks, and issued new licenses for operations to foreign banks. To consolidate the domestic banking sector, the Singaporean monetary authorities attempted to enhance the

External Shocks on Stock Prices in East Asian Domestic Banking

103

governance of domestic banks, to promote interbank mergers, and to encourage the acquisition of foreign banks by domestic banks. Finally, Thailand improved its financial and banking system in the aftermath of the 1997 crisis by closing and consolidating both its banks and nonbanks, infusing banks with capital, and relaxing existing restrictions on foreign ownership of domestic banks. In 2004, the Bank of Thailand introduced Phase 1 of its Financial Master Plan, first considered in 2002 as a way to consolidate the banking sector even further. This plan included changes in the classification of banks. For instance, some finance companies converted to commercial banks through a modification of the definition of a commercial bank. In 2007, the Thai authorities commenced Phase 2 of the Financial Master Plan, aimed at reducing the operational expenses of financial institutions, enhancing competitiveness in the banking sector, and developing financial system infrastructure. The Thai authorities subsequently announced that they would gradually liberalize the banking sector and grant greater operating leeway to the new financial institutions, both domestically and internationally, as a means of further enhancing the competitiveness of domestic banks in Thailand.

Foreign Assets and Liabilities of the Banking Sector This section describes the foreign assets and liabilities of the various domestic banking sectors in this analysis to highlight the transactional relationships between their respective domestic and foreign banks. Statistical data reveals an increase in the foreign assets and liabilities of the domestic banking sectors of all four countries between 2000 and 2007. The increase from 2004 to 2007 is particularly significant. Although domestic banks in East Asia reduced their foreign assets and liabilities after 1997, capital flows between domestic and foreign banks grew in the 2000s. This implies that the shocks taking place in the foreign banking sectors could have greatly affected the domestic banking system in Asia after 2000. Table 1 provides statistics on the foreign assets and liabilities of the banking systems in Korea, Malaysia, Singapore, and Thailand. Table 1 also details the debt of local banks from foreign (BIS reporting) banks, the ratio of bank debt from foreign (BIS reporting) banks to total foreign debt in each country, and the deposits of foreign (BIS reporting) banks in local banks.6 The external assets and debts of domestic commercial banks in Korea declined after the 1997 crisis until 2001 and then increased during the period from 2002 to 2007
2008.7 The increase in 2006 and 2007 was

Table 1.

Foreign Asset and Liability of Banks. 1990

Korea Foreign assets of domestic commercial banks (millions of US dollars) Foreign debts of domestic commercial banks (millions of US dollars) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars) Malaysia Foreign asset of banks (millions of RM) Foreign asset ratio to all assets (%) Foreign liability of banks (millions of RM) Foreign liability ratio to all liability (%) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars) Singapore Foreign asset of domestic banks (millions of Singapore dollars) Due from banks (abroad) (millions of Singapore dollars) Securities and equities (abroad) (millions of Singapore dollars) Foreign liability of domestic banks (millions of Singapore dollars) Due to banks (abroad) (millions of Singapore dollars)

1991

1992

1993

1994

1995

1996

1997

1998

1999

34,325

53,636

77,199

65,296

53,638

49,867

26,029

36,720

51,229

57,948

50,025

44,468

15,415

20,062

23,398

26,364

37,027

49,949

65,896

58,310

39,637

38,574

53.63

58.57

62.06

63.45

65.42

64.43

65.93

59.36

57.56

59.44

21,520

25,578

39,521

32,407

27,673

1,047

1,980

2,941

5,249

3,865

4,419

6,504

9,904

6,013

3,921

14.41

25.10

34.74

40.30

28.64

26.33

29.25

35.51

28.17

21.45

10,453

14,745

9,605

9,986

8,298

43,859.7 42,088.6 51,643.6 50,401.6 56,956.1 55,321.0 60,302.3 78,686.9 73,781.5 87,802.5 36,767.3 35,625.3 45,787.8 45,115.1 50,741.7 48,910.4 52,557.5 69,714.7 66,935.5 80,191.6 341.6 330.7 287.9 416.2 1,169.8 1,129.2 1,293.9 1,327.5 1,380.5 2,665.6 43,502.1 40,060.2 48,442.5 51,591.9 59,853.7 66,122.5 77,447.4 105,193.6 82,604.9 83,990.2 38,184.8 34,582.4 42,534.6 45,638.0 52,169.6 56,182.1 67,248.2 94,721.9 74,338.3 75,058.8

Table 1. 1990 Ratio to all foreign asset of domestic banks Due from banks (abroad) (%) Securities and equities (abroad) (%) Ratio to all foreign liability of domestic banks Due to banks (abroad) (%) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars) Thailand Foreign asset of commercial banks (millions of US dollars) Due from Banks (abroad) (millions of US dollars) Foreign liability of commercial banks (millions of US dollars) Deposits in foreign currencies (millions of US dollars) Borrowings in foreign currencies (millions of US dollars) Ratio to all foreign asset of commercial banks Due from Banks (abroad) (%) Ratio to all foreign liability of commercial banks Deposits in foreign currencies (%) Borrowings in foreign currencies (%) Ratio to all foreign asset of commercial banks Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars)

(Continued )

1991

1992

1993

1994

1995

1996

1997

1998

1999

83.83 0.78

84.64 0.79

88.66 0.56

89.51 0.83

89.09 2.05

88.41 2.04

87.16 2.15

88.60 1.69

90.72 1.87

91.33 3.04

87.78 106,657

86.33 89,452

87.80 87,943

88.46 105,297

87.16 103,236

84.97 164,634

86.83 156,863

90.05 165,195

89.99 106,308

89.37 76,461

75.85

70.11

62.49

65.94

58.89

85.53

82.89

80.52

76.27

69.58

158,190

164,208

206,116

231,903

231,536

6,722 1,610 4,674

8,232 1,288 6,462

18,868 4,524 13,443

35,163 5,095 32,485

65,268 6,438 63,197

67,572 3,555 65,277

58,462 5,994 53,977

44,508 8,898 38,410

34,038 11,492 25,027

401 3,600

576 5,107

769 11,494

995 29,719

957 60,065

810 62,319

1,502 50,679

1,821 34,915

2,225 21,412

23.95

15.65

23.98

14.49

9.86

5.26

10.25

19.99

33.76

8.58 77.02

8.91 79.03

5.72 85.50

3.06 91.49

1.51 95.04

1.24 95.47

2.78 93.89

4.74 90.90

8.89 85.56

4,489

5,151

6,478

8,890

14,091

25,763

25,904

25,080

15,271

7,312

33.08

26.42

28.26

29.96

32.11

41.01

36.93

37.57

32.04

22.58

9,676

7,083

7,439

9,725

9,389

Table 1. 2000 Korea Foreign assets of domestic commercial banks (millions of US dollars) Foreign debts of domestic commercial banks (millions of US dollars) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars) Malaysia Foreign asset of banks (millions of RM) Foreign asset ratio to all assets (%) Foreign liability of banks (millions of RM) Foreign liability ratio to all liability (%) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars) Singapore Foreign asset of domestic banks (millions of Singapore dollars) Due from banks (abroad) (millions of Singapore dollars) Securities and equities (abroad) (millions of Singapore dollars) Foreign liability of domestic banks (millions of Singapore dollars) Due to banks (abroad) (millions of Singapore dollars)

(Continued )

2001

2002

2003

2004

2005

2006

2007

2008

45,168

37,440

39,244

45,339

50,574

58,388

82,113

108,959

96,982

40,717

33,000

34,814

35,827

39,367

42,572

53,348

63,630

71,068

33,622

30,077

36,679

44,888

46,952

47,041

69,835

104,677

84,510

57.30

56.05

60.28

61.37

64.55

52.71

52.13

49.83

47.50

25,530

26,667

28,077

34,633

47,938

40,507

48,203

63,943

41,302

22,082.8 3.92 23,863.8 4.24 2,990

15,711.0 2.49 25,733.6 4.08 3,703

34,874.6 4.58 37,954.7 4.99 7,632

27,036.7 3.06 41,088.4 4.64 9,975

42,794.0 4.16 31,728.5 3.09 9,544

81,847.2 7.14 50,126.6 4.37 11,876

37,652.3 2.94 51,438.0 4.02 9,472

22,255.2 21,014.1 4.34 3.97 16,247.7 15,104.2 3.17 2.85 3,780 3,072 18.15

14.13

14.75

15.14

26.55

29.36

25.13

25.51

24.81

11,789

9,698

8,053

9,607

21,328

14,052

22,175

38,939

11,640

82,362.6 98,057.6 73,939.8 87,180.8 3,390.8 3,858.5

92,948.6 79,372.8 5,619.7

89,891.8 101,048.1 117,688.1 168,357.3 185,190.6 222,655.6 76,192.3 81,240.3 94,501.0 132,434.8 134,904.7 164,517.7 5,884.7 8,040.8 9,602.2 12,919.4 13,594.3 13,888.9

93,835.9 103,771.0 101,214.7 101,776.4 110,602.3 115,768.0 151,479.4 179,639.1 208,505.8 83,699.9 89,805.8

87,841.5

88,357.3

96,138.6

97,838.3 126,588.5 148,295.1 166,121.0

Table 1. 2000 Due from banks (abroad) (%) Securities and equities (abroad) (%) Due to banks (abroad) (%) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars) Thailand Foreign asset of commercial banks (millions of US dollars) Due from Banks (abroad) (millions of US dollars) Foreign liability of commercial banks (millions of US dollars) Deposits in foreign currencies (millions of US dollars) Borrowings in foreign currencies (millions of US dollars) Due from Banks (abroad) (%) Deposits in foreign currencies (%) Borrowings in foreign currencies (%) Debt of local banks from foreign (BIS reporting) banks (outstanding in millions of US dollars) Ratio of bank debt to country foreign debt from foreign (BIS reporting) banks (%) Deposits of foreign (BIS reporting) banks to local bankings (outstanding in millions of US dollars)

(Continued )

2001

2002

2003

2004

2005

2006

2007

2008

89.77 4.12 89.20 70,143

88.91 3.93 86.54 65,051

85.39 6.05 86.79 56,452

84.76 6.55 86.82 53,596

80.40 7.96 86.92 64,559

80.30 8.16 84.51 65,723

78.66 7.67 83.57 74,786

72.85 7.34 82.55 90,536

73.89 6.24 79.67 79,276

70.20

69.41

66.08

64.07

77.97

62.68

54.06

49.91

49.23

254,842

253,547

264,832

272,943

297,158

296,077

343,457

436,073

417,970

30,702 13,241 18,018 1,923 14,363 43.13 10.67 79.71 5,744

26,925 13,221 14,290 1,775 11,072 49.10 12.42 77.48 4,668

23,193 8,951 12,207 1,504 9,404 38.59 12.32 77.04 3,282

24,163 9,265 10,944 1,627 7,805 38.34 14.87 71.32 3,813

24,652 10,326 11,349 1,795 7,427 41.89 15.82 65.44 4,127

25,443 11,541 11,726 2,668 6,134 45.36 22.75 52.31 5,175

32,600 19,609 11,334 2,977 5,160 60.15 26.27 45.53 6,326

33,998 20,910 10,816 2,575 4,456 61.50 23.81 41.20 6,426

24,800 10,325 13,334 3,500 5,716 41.63 26.25 42.87 6,495

21.56

19.65

18.75

21.40

24.40

24.23

28.22

31.37

29.33

10,997

11,768

7,534

14,116

16,395

23,753

29,264

32,008

17,481

108

MASAHIRO INOGUCHI

significant. The information in the Bank for International Settlements (BIS) data shows that the foreign debt of domestic banks in Korea increased from 2002 to 2007.8 The ratio of this debt to national foreign debt rose during the period from 2002 to 2004 and fell after 2005. Together, these statistics indicate that the external assets and liabilities of domestic banks in Korea reached a high level in the latter half of the 2000s. However, the foreign debt of these banks fell compared with that of the other sectors after 2005. The foreign assets and liabilities of banks in Malaysia increased dramatically in 2007. The BIS data in Table 1 indicate a decline in the foreign debt of domestic banks in 1998, with a low level sustained until 2003. The trend was upward from 2004 to 2007. The ratio of this debt to national debt was also at a high level after 2004; however, it was lower than during the mid1990s. The foreign assets and liabilities of Malaysian banks were greater in the late 2000s than in the early 2000s. The fluctuations in the value of domestic bank foreign assets and those due from foreign banks in Singapore exhibited an upward trend from 1990 to 2008, and then increased significantly after 2004. The foreign liabilities of domestic banks and those due to foreign banks increased from 1990 to 1997, declined in 1998, and increased thereafter, especially from 2004. In contrast, the percentage of assets due from foreign banks to all domestic bank foreign assets did not rise, and tended to decrease after 2000, while the percentage of foreign securities and equity assets increased slowly. In addition, the percentage of foreign liabilities due to foreign banks to all domestic bank foreign liabilities has declined slightly since 2005. The BIS data show that the foreign debt of domestic banks increased from 2004 to 2007, though the ratio to national foreign debt declined continuously. Together, the external assets and liabilities of Singaporean domestic banks expanded in the 2000s. However, the foreign assets and debts of domestic banks due from foreign banks declined compared with those from other sectors. For commercial banks in Thailand, foreign assets due from foreign banks increased from 1997 to 2000, declined heavily in 2002, and increased again from 2003 to 2007. The ratio of these assets to total foreign assets increased during the period 1997
2001 and again from 2004 to 2007. In contrast, the foreign liabilities of domestic banks remained at a low level after 2003, with a pronounced downward trend from 1997 to 2003, despite a slight increase in both 2004 and 2005. Borrowings in foreign currencies and their share of the total foreign liabilities of the commercial banks declined from 1997 to 2007. However, deposits in foreign currencies exhibited an upward trend. The debt of domestic banks from BIS reporting banks decreased after the

External Shocks on Stock Prices in East Asian Domestic Banking

109

1997 crisis and increased after 2003, with the share of this debt to national debt growing since 2003. Thus, the information in these data indicates that the assets due from foreign banks and debt from BIS reporting banks expanded in the period from 2003 to 2007 for Thai banks. In general, these statistics indicate that the foreign assets and liabilities of domestic banks in our selected Asian countries (Korea, Malaysia, Singapore, and Thailand) have tended to increase in the 2000s, which implies that the capital flows between foreign and local banks expanded and that the relationship has recently become more significant. Although external liabilities are still larger in Korea and Singapore, the share of total foreign liabilities in Malaysia and Thailand expanded in the 2000s before the GFC.

AN OVERVIEW OF DOMESTIC BANK STOCK PRICES AND STOCK MARKET INDICES This section graphically illustrates the fluctuations in the prices of bank stocks and stock market indices in Korea, Malaysia, Singapore, and Thailand. In general, we find that the prices of most domestic bank stocks typically move together, despite some differences in the magnitude of change. Stock market indices fluctuate similarly across most countries. Fig. 1 depicts the fluctuations in the monthly average rates of change in bank stock prices and stock market indices from January 1993 to December 1996 and in the 2000s. The prices of the Korean bank stocks generally rose in the latter half of 1995 and declined thereafter. The trend in 1996 was generally downward. Most share prices of the domestic bank stocks in Korea increased in the first half of 2002, declined soon afterward, and then began to increase again from 2003. The trend was then upward until 2007
2008, but stock prices fell in mid-2004 and in 2006. The increases and declines generally occur in similar periods in the 1990s and the 2000s. However, the rates of change in the prices of Korean bank stocks differ for each bank. The fluctuations in the Korea Composite Stock Price Index (KOSPI) are similar to those of the individual domestic banks in Korea. Most share prices of the domestic banks in Malaysia increased in 1993 and again in 1996. In the 2000s, the prices of the Malaysian bank stocks generally fell until the first half of 2001, rose in the first half of 2002, declined thereafter, and began to increase again from the second half of 2003. The trend was generally upward until 2007, despite declines in mid2004 and 2005. The rates of change in the prices of specific domestic bank

110

MASAHIRO INOGUCHI Korea: change rates of stock prices (January 1993–December 1996) Woori Finance Holdings

Hana Financial Group

Shinhan

Korea Exchange Bank

Daegu Bank

Pusan Bank

Jeonbuk Bank

Bank of Cheju

Puren Mutual Svg. Bank

30 25 20 15

%

10 5 0 –5 –10 –15 Oct-96

Dec-96

Aug-96

Apr-96

Jun-96

Feb-96

Oct-95

Dec-95

Aug-95

Apr-95

Jun-95

Feb-95

Oct-94

Dec-94

Aug-94

Apr-94

Jun-94

Feb-94

Oct-93

Dec-93

Aug-93

Apr-93

Jun-93

Feb-93

–20

Korea: change rates of stock prices (January 2000–October 2009) Woori Finance Holdings

Hana Financial Group

Shinhan

Korea Exchange Bank

Daegu Bank

Pusan Bank

Jeonbuk Bank

Bank of Cheju

Puren Mutual Svg. Bank

80

60

%

40

20

0

–20

Feb-00 May-00 Aug-00 Nov-00 Feb-01 May-01 Aug-01 Nov-01 Feb-02 May-02 Aug-02 Nov-02 Feb-03 May-03 Aug-03 Nov-03 Feb-04 May-04 Aug-04 Nov-04 Feb-05 May-05 Aug-05 Nov-05 Feb-06 May-06 Aug-06 Nov-06 Feb-07 May-07 Aug-07 Nov-07 Feb-08 May-08 Aug-08 Nov-08 Feb-09 May-09 Aug-09

–40

Fig. 1.

Change Rates of Stock Prices (Monthly).

111

External Shocks on Stock Prices in East Asian Domestic Banking Malaysia: change rates of stock prices (January 1993–December 1996) Affin Bank

Hong Leong bank

Public Bank

BIMB

EON Bank

Malayan Banking Berhad (Maybank)

RHB Bank

CIMB

60 50 40

%

30 20 10 0 –10

Oct-96

Dec-96

Aug-96

Apr-96

Jun-96

Feb-96

Oct-95

Dec-95

Aug-95

Apr-95

Jun-95

Feb-95

Oct-94

Dec-94

Aug-94

Apr-94

Jun-94

Feb-94

Oct-93

Dec-93

Aug-93

Apr-93

Jun-93

–30

Feb-93

–20

Malaysia: change rates of stock prices (January 2000–October 2009) Affin Bank

Hong Leong bank

Public Bank

BIMB

EON Bank

Malayan Banking Berhad (Maybank)

RHB Bank

CIMB

Ambank

30

20

10

%

0

–10

–20

–40

Feb-00 May-00 Aug-00 Nov-00 Feb-01 May-01 Aug-01 Nov-01 Feb-02 May-02 Aug-02 Nov-02 Feb-03 May-03 Aug-03 Nov-03 Feb-04 May-04 Aug-04 Nov-04 Feb-05 May-05 Aug-05 Nov-05 Feb-06 May-06 Aug-06 Nov-06 Feb-07 May-07 Aug-07 Nov-07 Feb-08 May-08 Aug-08 Nov-08 Feb-09 May-09 Aug-09

–30

Fig. 1.

(Continued )

DBS Bank

Fig. 1.

(Continued )

Oversea-Chinese Banking Corporation Apr-96

Singapore: change rates of stock prices (January 2000–October 2009)

United Overseas Bank

40

30

20

10

0

–10

–20

–30 Dec-96

Oct-96

Aug-96

Jun-96

Oversea-Chinese Banking Corporation

Feb-96

Dec-95

Oct-95

Aug-95

Jun-95

Apr-95

Feb-95

Dec-94

Oct-94

Aug-94

Jun-94

%

DBS Bank

Apr-94

Feb-94

Dec-93

Oct-93

Aug-93

Jun-93

Apr-93

Feb-93

–20

Feb-00 May-00 Aug-00 Nov-00 Feb-01 May-01 Aug-01 Nov-01 Feb-02 May-02 Aug-02 Nov-02 Feb-03 May-03 Aug-03 Nov-03 Feb-04 May-04 Aug-04 Nov-04 Feb-05 May-05 Aug-05 Nov-05 Feb-06 May-06 Aug-06 Nov-06 Feb-07 May-07 Aug-07 Nov-07 Feb-08 May-08 Aug-08 Nov-08 Feb-09 May-09 Aug-09

%

112 MASAHIRO INOGUCHI

Singapore: change rates of stock prices (January 1993–December 1996) United Overseas Bank

50

40

30

20

10

0

–10

113

External Shocks on Stock Prices in East Asian Domestic Banking Thailand: change rate of stock prices (January 1993–December 1996) Bangkok Bank Public Company Limited

Krung Thai Bank Public Company Limited

Siam Commercial Bank Public Company Limited

Kasikornbank Public Company Limited

TMB Bank Public Company Limited

Bank of Ayudhya Public Company Ltd.

Siam City Bank Public Company Limited

Tisco Bank Public Company Limited

40

30

%

20

10

0

–10

–20

Oct-96

Dec-96

Aug-96

Apr-96

Jun-96

Feb-96

Oct-95

Dec-95

Aug-95

Apr-95

Jun-95

Feb-95

Dec-94

Oct-94

Aug-94

Apr-94

Jun-94

Feb-94

Oct-93

Dec-93

Aug-93

Apr-93

Jun-93

Feb-93

–30

Thailand: change rate of stock prices (January 2000–October 2009) Bangkok Bank Public Company Limited

Krung Thai Bank Public Company Limited

Siam Commercial Bank Public Company Limited

Kasikornbank Public Company Limited

TMB Bank Public Company Limited

Bank of Ayudhya Public Company Ltd.

Siam City Bank Public Company Limited

Tisco Bank Public Company Limited

50 40 30 20

%

10 0 –10 –20 –30

–50

Feb-00 May-00 Aug-00 Nov-00 Feb-01 May-01 Aug-01 Nov-01 Feb-02 May-02 Aug-02 Nov-02 Feb-03 May-03 Aug-03 Nov-03 Feb-04 May-04 Aug-04 Nov-04 Feb-05 May-05 Aug-05 Nov-05 Feb-06 May-06 Aug-06 Nov-06 Feb-07 May-07 Aug-07 Nov-07 Feb-08 May-08 Aug-08 Nov-08 Feb-09 May-09 Aug-09

–40

Fig. 1.

(Continued )

114

MASAHIRO INOGUCHI Stock index: change rates (January 1993–December 1996) KOSPI (Korea)

KLCI (Malaysia)

STI (Singapore)

SET (Thailand)

Nikkei225 (Japan)

DJIA (U.S.)

PHLXBKX (US bank sector) 20

15

10

%

5

0

–5

Oct-96

Stock index: change rates (January 2000–October 2009) KOSPI (Korea)

KLCI (Malaysia)

STI (Singapore)

SET (Thailand)

Nikkei225 (Japan)

DJIA (U.S.)

PHLXBKX (US bank sector)

30

20

%

10

0

–10

–30

Feb-00 May-00 Aug-00 Nov-00 Feb-01 May-01 Aug-01 Nov-01 Feb-02 May-02 Aug-02 Nov-02 Feb-03 May-03 Aug-03 Nov-03 Feb-04 May-04 Aug-04 Nov-04 Feb-05 May-05 Aug-05 Nov-05 Feb-06 May-06 Aug-06 Nov-06 Feb-07 May-07 Aug-07 Nov-07 Feb-08 May-08 Aug-08 Nov-08 Feb-09 May-09 Aug-09

–20

Fig. 1.

(Continued )

Dec-96

Jun-96

Aug-96

Apr-96

Feb-96

Oct-95

Dec-95

Aug-95

Apr-95

Jun-95

Feb-95

Oct-94

Dec-94

Aug-94

Apr-94

Jun-94

Feb-94

Oct-93

Dec-93

Aug-93

Apr-93

Feb-93

–15

Jun-93

–10

External Shocks on Stock Prices in East Asian Domestic Banking

115

stocks differ for each bank, while the increases and decreases occur in similar periods. Fluctuations in the Kuala Lumpur Composite Index (KLCI) are similar to those of the domestic bank stocks. The prices of the three Singaporean bank stocks generally rose in the latter half of 1993 and fell in the first half of 1994. In the 2000s, the prices of these three bank stocks fell in 2001, increased soon afterward, and then declined from 2002 until the first half of 2003. Thereafter, the trend was upward until mid-2007. The rates of change in the prices of Singaporean bank stocks differ for each bank, but they tend to increase and decrease in similar periods. The price fluctuations in these stocks are similar to those in the Straits Times Index (STI). Most prices of domestic bank stocks in Thailand expanded in the latter half of 1993 and then declined in the first half of 1994. The trend was upward from the latter half of 1994 to the first half of 1996 and then downward until the end of 1996. The prices of the Thai bank stocks fluctuated in the 2000s. While the rates of change in the prices of the individual Thai bank stocks differ for each bank, many stock prices tended to fall until the first half of 2000, rose in 2003, and remained at a relatively high level until the first half of 2008. The fluctuations in the Stock Exchange of Thailand’s (SET) market index are similar to the price fluctuations indicated in the domestic Thai banks. Figures also illustrate the fluctuations and rates of change in the stock market index for the Tokyo Stock Exchange in the form of the Nikkei 225, the US Dow Jones Industrial Average (DJIA), and the Philadelphia Stock Exchange (PHLX) KBW Bank Sector Index (BKX) in addition to KOSPI, KLCI, STI, and SET. The rates of change in the indices increased and decreased in similar periods. Most of the indices did not expand from the latter half of 2002 to the first half of 2003 and then generally increased until 2007, with only slight falls in mid-2004. In addition, the price movements of the domestic bank stocks in each country generally have more in common than the price movements across the stock market indices in the different countries.

REGRESSION ANALYSIS Data and Terms We consider three periods in which to compare the influence of external shocks on the domestic banking sector: a first period during the 1990s

116

MASAHIRO INOGUCHI

before the Asian financial crisis, a second period during the 2000s before the GFC, and a third period during the 2000s after the GFC.9 The specific periods analyzed are January 1993
December 1996, January 2000
May 2007, and June 2007
October 2009, respectively.10 We use daily data for bank stock prices, the stock market indices, and interest rates. All stock prices are from either the Datastream or Taiwan Economic Journal (TEJ) databases. We obtain all prices for the Asian and US stock indices from the CEIC database, while the interest rates are from the official government websites for each country and the CEIC database. We calculate the slope of the yield curve as the difference in yields between 10-year and one-year sovereign bonds in Korea, Singapore, and Thailand. Given the lack of comparable data for Malaysia, we use interbank interest rates and do not include the variable proxying the shape of the yield curve in the analysis of the period January 1993
December 1996. We employ data on listed domestic commercial banks in Malaysia, Singapore, and Thailand and listed commercial, regional, and specialized banks in Korea, and remove some banks from the sample because they either merged or failed after the 1997 crisis. Further, some additional banks are included in the regression for the period following the 1997 crisis because of the emergence of several new banks. For the 2000s, we analyze nine banks in each of Korea and Malaysia, three in Singapore, and eight in Thailand. For the period January 1993
December 1996, our analysis includes six banks in Korea, eight in Malaysia, three in Singapore, and seven in Thailand.

Methodology In our analysis, we estimate how changes in the US and Japanese market indices affect the prices of domestic bank stocks in order to assess the influence of external shocks on the Korean, Malaysian, Singaporean, and Thai banking sectors. Following Gropp et al. (2009) and Bae et al. (2003), we count the number of banks in a given country that experience a large shock on the same day (or coexceedances) as shocks to the domestic banking sector. We arbitrarily define large shocks as an extreme return
an exceedance
below the 5th or above the 95th percentile of the marginal return distribution of the daily percentage change in each bank’s stock price. We treat positive and negative extreme returns separately. In addition, we categorize the number of coexceedances into the following four categories: “zero,” “one,” “two,” and “three or more.” This approach

External Shocks on Stock Prices in East Asian Domestic Banking

117

enables us to measure the influence of foreign shocks on the entire banking sector of an individual country. We then specify these coexceedances as a function of the relevant domestic stock price index, the US and Japanese stock market indices, and the changes in the yield curve in each country.11 With regard to the domestic stock market variables, we construct two kinds of indicators measuring shocks in each country’s stock market. The first is a dummy variable that takes a value of one if the stock market index experiences a shock large enough to rank below the 5th or above the 95th percentile of the distribution of daily returns (exceedances). This exceedance includes observations from the positive and negative tail of the distribution, with exceedances from the positive or negative tail used in regression when the dependent variable is from the positive or negative return coexceedances. The second shock indicator is the daily change in the volatility of each country’s stock market index.12 Following Gropp et al. (2009) and Bae et al. (2003), we define volatility as the conditional variance of the stock market index when estimated using a generalized autoregressive conditional heteroskedasticity (GARCH) model.13 We employ the number of exceedances of the stock market indices in the United States (DJIA) and Japan (Nikkei 225), the exceedances of the stock market index in the United States (DJIA), or the exceedances of the PHLX Bank Index (BKX) as the variables determining the foreign shocks. We first construct an indicator that takes a value of one if there are exceedances in the stock market index of a foreign country, otherwise zero. We calculate the dummy variable representing shocks in both the Japanese and US stock markets as the sum of the exceedances in the Nikkei 225 and the DJIA. Therefore, the value of the coexceedances in this variable is from zero to two. In addition, exceedances in the DJIA are a measure of shocks to the US stock market, whereas exceedances in the BKX indicate shocks to the foreign banking sector. Each regression model specifies one of these dummy variables as an explanatory variable. We include the positive (negative) return coexceedances in the regression where the dependent variable is from the positive (negative) return coexceedances.14 The yield curve variable is the daily change in the absolute value of the slope of the yield curve. As discussed, we calculate these variables as the differences in yields between 10-year and one-year sovereign bonds, and as such they are a proxy measure of economic growth and monetary policy expectations. One interpretation of banking is that commercial banks transform short-term liabilities (deposits) into long-term assets (loans). A flat yield curve indicates an increase in the interest rate that banks must

118

MASAHIRO INOGUCHI

pay on short-term liabilities, with no corresponding increase in the interest rates charged on loans by the banks. Any large or small difference between these yields would then increase or decrease the price of a bank’s stock, respectively. Therefore, the number of coexceedances should relate to the shape of the yield curve as indicated by the spread between 10-year and one-year bond yields. Due to the lack of suitable data, we employ interbank interest rates in the regression for Malaysia, but omit this variable from the test for the period January 1993
December 1996. Following Gropp et al. (2009) and Bae et al. (2003), we employ a multinomial logit model to estimate coexceedances because the number of coexceedances is a count variable and marginal effects at each count are not the same.15 According to Bae et al. (2003), Gropp and Moerman (2004), and Gropp et al. (2009), the multinomial logit model should be the primary estimation method. The regression equation is as follows: Pr½Y = j =

e½αj Ct − 1 þ βj S þ γj F þ δj R PJ 1 þ j = 1 e½αj Ct − 1 þ βj S þ γj F þ δj R

ð1Þ

where j: number of banks in the tail of the distribution for the daily percentage change in domestic bank stock prices (coexceedance) in a given country (j = 1, 2, 3), Ct − 1: lagged number of coexceedances in domestic banks in a given country, S: variable concerning the domestic stock market, being the exceedance of the stock price index in a given country (this variable takes a value of one for exceedance, zero otherwise) or the price volatility of the stock market index in a given country, F: variable measuring shocks in foreign stock markets, being either the sum of the exceedances of the DJIA and the Nikkei 225, or the exceedance of the DJIA, or the exceedance of the PHLX BKX, R: yield curve in Korea, Singapore, and Thailand, and interbank rate in Malaysia (the latter is not used in the regression equation for the period January 1993
December 1996). As per convention, we define Y = 0 (coexceedances equal zero) as the base category to remove the indeterminacy associated with the model. Because the estimated coefficients from the multinomial logit model are difficult to interpret, it is useful to calculate the marginal effects of the

External Shocks on Stock Prices in East Asian Domestic Banking

119

regressors. We obtain these by differentiating the above equation with respect to the explanatory variables. For instance, the marginal effect of F is as follows: ∂Pr½Y = j = γj ∂F


  Pr½Y = j  1 − Pj

ð2Þ

Regression Results Tables 2
4 provide the estimated regression results for the periods June 2000
May 2007 (before the GFC), January 1993
December 1996 (before the Asian financial crisis), and June 2007
October 2009 (after the GFC).16 The left-hand sides of the tables focus on the positive return or “upper tail” exceedances (Models 1
6), while the right-hand sides focus on the negative return or “lower tail” exceedances (Models 7
12). All tables report the estimated coefficients alongside the marginal probabilities.17 Results in the 2000s before the GFC Table 2 shows the estimated results for the period January 2000
May 2007 in Korea, Malaysia, Singapore, and Thailand. As shown in Table 2, the coefficients and marginal effects for the foreign stock market shock (DJIA and Nikkei 225), the US stock market shock (DJIA), and the US banking sector shock (PHLX BKX) are significantly statistically positive in the lower tail category of two and three or more coexceedances in Korea. For the upper tail coexceedances, these foreign stock market shock coefficients are significant in the category of two coexceedances. Most of the estimated coefficients for the domestic stock market are significant, and the signs for domestic index volatility are positive (negative) in the upper (lower) tail coexceedances. All coefficients and marginal effects for the yield curve are significant for the upper tail coexceedances. In Malaysia, the coefficients and the marginal effects for shocks occurring in foreign stock market index (DJIA and Nikkei 225), the DJIA, and the PHLX BKX, are significant and positive for the lower tail coexceedances. In contrast, few coefficients representing the foreign stock market index are significant for the upper tail coexceedances. However, the coefficients and marginal effects for shocks in Malaysia’s KLCI are significant, while the sign for volatility in the KLCI is positive (negative) in the upper (lower) tail coexceedances.

Table 2.

Regression Results in the 2000s before the Global Financial Crisis. Upper Tails

Model 1 Coefficient

Korea Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal effect

0.390 *** 1.819 ***

0.040 *** 0.165 ***

0.168

0.016

Model 2 Coefficient

Marginal effect

0.442 ***

0.043 ***

0.043 *** 0.142

0.004 *** 0.013

Model 3 Coefficient

0.390 *** 1.861 ***

0.149

Marginal effect

0.040 *** 0.164 ***

0.016

Model 4 Coefficient

Marginal effect

0.443 ***

0.043 ***

0.044 ***

0.004 ***

0.124

0.012

53.205 *** −2.364 ***

5.304 ***

54.849 *** −2.356 ***

5.332 ***

53.466 *** −2.359 ***

5.315 ***

55.182 *** −2.352 ***

5.356 ***

0.601 *** 2.056 ***

0.015 *** 0.054 *

0.662 ***

0.014 ***

0.594 *** 2.299 ***

0.015 *** 0.073 **

0.659 ***

0.014 ***

0.001 *** 0.012 *

0.001 ***

0.018 **

0.066 *** 0.542 *

0.070 ***

0.663 **

0.244

0.006

0.321 67.536 ** −4.030 ***

1.630 **

0.404 *** 3.695 ***

0.008 ** 0.292 ***

0.063

0.001

71.390 *** −4.142 ***

1.464 **

0.504 ***

0.008 ***

0.094 *** 0.326

0.002 *** 0.0055

70.058 *** −3.986 *** 0.405 *** 3.721 ***

−0.013 106.413 *** 2.213 *** −4.409 *** −1188.168 0.086

103.564 *** 1.759 *** −4.466 *** −1170.685 0.099

0.010 1.733 **

0.008 ** 0.287 ***

−0.0010

106.906 *** 2.218 *** −4.408 *** −1190.228 0.084

73.442 *** −4.107 ***

1.518 **

0.506 ***

0.008 ***

0.096 ***

0.002 ***

0.330

0.007

104.241 *** 1.761 *** −4.460 *** −1172.308 0.098

Model 5 Coefficient

0.391 *** 1.824 ***

Marginal effect

0.040 *** 0.164 ***

Model 6 Coefficient

Marginal effect

0.444 ***

0.044 ***

0.044 ***

0.004 ***

0.329 53.538 *** −2.369 ***

0.037 5.319 ***

0.426 54.885 *** −2.367 ***

0.046 5.324 ***

0.590 *** 2.296 ***

0.015 *** 0.077 **

0.659 ***

0.014 ***

0.069 ***

0.001 ***

0.283 70.876 *** −3.987 *** 0.417 *** 3.632 ***

0.007 1.760 **

0.008 ** 0.271 ***

0.451 0.011 106.019 *** 2.188 *** −4.430 *** −1189.587 0.085

0.373 73.412 *** −4.113 ***

0.008 1.521 **

0.513 ***

0.008 ***

0.094 ***

0.002 ***

0.927 ** 0.023 104.661 *** 1.748 *** −4.508 *** −1169.719 0.100

Table 2.

(Continued ) Upper Tails

Model 1 Coefficient

Malaysia Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Log-likelihood Pseudo R2

Marginal effect

0.182 ** 2.297 ***

0.020 * 0.036

0.335 *

0.052 **

Model 2 Coefficient

Marginal effect

0.161 **

0.012 *

0.132 *** 0.217

0.002 *** 0.025

Model 3 Coefficient

0.184 ** 2.339 ***

0.172 3.259 −1.668 *** 0.404 *** 3.825 *** −0.211

−6.733 −2.750 ***

0.616

0.019 *** 0.225 *** −0.015

−0.376

0.692 *** 5.265 ***

0.014 *** 0.408 ***

0.064

0.000

1.367 −1.596 ***

0.766

0.390 ***

0.004 ***

0.256 *** −0.286

0.001 *** 0.012

−12.461 −2.923 ***

0.286

0.714 ***

0.001 ***

0.418 *** 0.294

0.000 *** 0.0016

3.420 −1.648 *** 0.406 *** 3.817 ***

−28.387 ** 0.057 * −4.873 *** −1302.074 0.190

0.021 * 0.038

0.032 0.644

0.019 *** 0.214 ***

−0.191

−0.010

−6.597 −2.762 ***

−0.370

0.681 *** 5.344 ***

−0.956 −14.019 −0.309 −3.745 *** −1454.747 0.097

Marginal effect

0.013 *** 0.420 ***

−0.0142 **

−15.113 −0.330 −3.683 *** −1455.527 0.096

Model 4 Coefficient

Marginal effect

0.162 **

0.020 *

0.134 ***

0.017 ***

0.083

0.014

1.488 −1.582 ***

0.320

0.389 ***

0.013 ***

0.255 ***

0.008 ***

−0.165 −12.417 −2.941 ***

Coefficient

0.182 ** 2.323 ***

0.322 3.327 −1.653 *** 0.404 *** 3.751 ***

Marginal effect

0.020 * 0.042

0.048 0.627

0.019 *** 0.206 ***

Model 6 Coefficient

Marginal effect

0.160 **

0.020 *

0.133 ***

0.017 ***

0.352 1.434 −1.594 ***

0.052 0.312

0.388 ***

0.013 ***

0.254 ***

0.008 ***

−0.006 −0.444

0.709 ***

0.003 ***

0.418 ***

0.002 ***

−0.318

Model 5

0.407 −6.658 −2.788 *** 0.699 *** 5.305 ***

0.021 −0.373

0.014 *** 0.421 ***

0.503 −12.316 −2.976 ***

0.018 −0.440

0.717 ***

0.003 ***

0.418 ***

0.002 ***

−0.001

−29.081 ** −0.132 ** −4.791 *** −1303.550 0.189

−0.320 −0.008 −14.155 −0.310 −3.732 *** −1455.407 0.096

−0.128 −0.001 −28.814 ** −0.130 ** −4.811 *** −1302.312 0.190

Table 2. (Continued ) Upper Tails Model 1 Coefficient

Singapore Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

0.405 *** 2.940 *** −0.154

Marginal effect

0.026 *** 0.279 *** −0.010

Model 2 Coefficient

0.523 *** 0.080 *** −0.357

Marginal effect

0.019 ***

Model 3 Coefficient

0.399 *** 2.897 ***

Marginal effect

0.025 *** 0.268 ***

0.0030 *** −0.013 −0.013

−0.001

Model 4 Coefficient

Marginal effect

0.517 ***

0.019 ***

0.079 ***

0.003 ***

−0.221

Model 5 Coefficient

0.390 *** 2.866 ***

0.327 4.407 *** −0.011

1.418 **

0.003 0.181 *** 0.0000

45.102 *** −3.933 *** 0.460 0.128 *** −0.068

1.651 ***

0.001

22.762 ** −3.163 *** 0.340 4.452 ***

1.414 **

44.371 *** −3.936 ***

0.003 0.191 ***

0.489 *

0.001

0.128 ***

0.0003 ***

0.0003 *** 0.000 −0.439

−0.003

−0.409

1.637 ***

22.161 ** −3.166 *** 0.330 4.391 ***

0.527 **

102.291 *** −7.627 ***

0.234 **

0.624 ** 6.513 ***

0.001 0.192 ***

0.800 **

0.0001

0.069

0.000

0.188 *** −0.125

0.00002 0.0000

56.667 ** −5.728 *** 0.630 ** 6.509 ***

0.183

0.517 **

101.719 *** −7.611 ***

0.0008 0.192 ***

0.790 **

0.0001

0.187 ***

0.00002

0.240

0.000

0.0003

0.232 ***

57.058 ** −5.757 *** 0.590 ** 6.458 ***

0.659 46.138 0.058 −7.713 *** −652.554 0.199

124.214 *** 0.013 −11.052 *** −566.718 0.305

46.203 0.058 −7.720 *** −652.461 0.200

Marginal effect

0.504 ***

0.019 ***

0.079 ***

0.003 ***

0.020

0.180

0.007

1.374 **

43.523 *** −3.937 ***

1.609 ***

0.003 0.185 ***

0.474 *

0.001

0.127 ***

0.0003 ***

0.074

0.0002

−0.001 0.104

57.887 ** −5.766 ***

0.025 *** 0.268 ***

Coefficient

−0.007 0.280

22.854 ** −3.154 ***

Marginal effect

Model 6

124.815 *** 0.013 −11.104 *** −567.283 0.304

0.001 0.520 *

0.001 0.188 ***

0.001

45.738 0.057 −7.735 *** −652.12941 0.200

101.406 *** −7.625 ***

0.232 ***

0.780 **

0.0001

0.186 ***

0.00002

0.474

0.0001

122.954 ** 0.013 −11.041 *** −567.497 0.304

Table 2.

(Continued ) Upper Tails

Model 1 Coefficient

Thailand Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal effect

0.259 *** 3.052 ***

0.026 *** 0.047

0.034

0.004

Model 2 Coefficient

Marginal effect

0.310 ***

0.023 ***

0.257 *** −0.223

0.019 *** −0.017

Model 3 Coefficient

0.259 *** 3.065 ***

−0.405 23.849 *** −2.636 ***

2.343 ***

29.719 *** −2.911 ***

0.195 4.720 ***

0.004 0.160 ***

0.296 *

−0.035

−0.001

0.416 *** −0.217

2.237 ***

0.002

0.377 *** 5.912 *** −0.105

0.619 *

0.010 *** 0.579 *** −0.003

34.485 *** −5.155 ***

−0.038

0.538 ***

0.002 ***

0.520 *** −0.141

0.002 *** −0.001

44.036 *** 0.200 *** −5.961 *** −937.37144 0.2959

0.313 ***

0.024 ***

0.256 ***

0.019 ***

−0.640

0.192 4.719 ***

0.004 0.163 ***

0.295 *

0.002

0.414 ***

0.003 ***

25.778 ** −4.144 0.385 *** 5.886 ***

−0.005 0.617 *

0.010 *** 0.571 ***

0.009

30.733 *** 0.794 ** −4.344 −1139.6529 0.1439

−0.331 34.249 *** −5.149 ***

Model 5 Coefficient

0.258 *** 3.052 ***

Marginal effect

0.025 *** 0.052

Model 6 Coefficient

Marginal effect

0.309 ***

0.023 ***

0.255 ***

0.019 ***

−0.209 29.810 *** −2.920 ***

−0.015 2.253 ***

−0.039 **

29.933 *** −2.909 ***

0.003 *** −0.002

0.276 **

Marginal effect

2.386 ***

0.227 30.633 *** 0.797 ** −4.314 *** −1140.452 0.143

0.026 *** 0.051

Coefficient

24.282 *** −2.626 ***

−0.207 25.824 ** −4.152 ***

Marginal effect

Model 4

2.252 ***

0.001 23.907 *** −2.634 ***

−0.003 2.359 ***

0.190 4.713 ***

0.004 0.164 ***

0.293 *

0.002

0.413 ***

0.003 ***

−0.002 0.275 **

0.534 ***

0.002 ***

0.522 ***

0.002 ***

0.359

0.002

43.932 *** 0.193 *** −6.019 *** −935.648 0.297

−0.199 26.128 ** −4.151 *** 0.381 *** 5.874 ***

−0.006 0.630 *

0.010 *** 0.568 ***

0.737 * 0.030 28.261 ** 0.715 ** −4.325 *** −1138.8809 0.145

−0.284 34.802 *** −5.158 ***

−0.002 0.281 **

0.539 ***

0.002 ***

0.523 ***

0.002 ***

0.931 ** 0.007 41.046 *** 0.172 ** −6.030 *** −934.514 0.298

Table 2. (Continued ) Lower Tails Model 7 Coefficient Korea Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal effect

0.265 *** 1.727 ***

0.023 *** 0.068

0.384 *

0.033

Model 8 Coefficient

Marginal effect

0.310 ***

0.027 ***

−0.052 *** 0.101

−0.005 *** 0.008 *

Model 9 Coefficient

0.270 *** 1.804 ***

0.371 28.497 * −2.358 ***

2.643

0.474 *** 3.257 ***

0.011 *** 0.151 ***

0.775 ***

0.018 **

42.400 −4.048 ***

0.983

0.569 *** 4.191 ***

0.013 *** 0.383 ***

0.837 ***

0.019 ***

34.489 ** −2.412 ***

3.087 *

0.535 ***

0.007 ***

−0.117 *** 0.257

−0.002 *** 0.003

58.586 * −4.646 ***

0.761 *

0.651 ***

0.009 ***

−0.123 *** 0.517 **

−0.002 *** 0.008 *

29.136 * −2.348 *** 0.479 *** 3.401 ***

42.213 0.571 −4.546 *** −1045.227 0.168

0.024 *** 0.062

0.031 2.710

0.011 *** 0.157 ***

0.959 **

0.033

41.639 −4.021 ***

0.967

0.573 *** 4.358 ***

0.989 ** 19.433 0.364 −4.058 *** −1106.688 0.119

Marginal effect

0.013 *** 0.408 ***

0.033

18.215 0.336 −4.020 *** −1109.098 0.117

Model 10 Coefficient

Marginal effect

0.311 ***

0.027 ***

−0.052 ***

−0.005 ***

0.113 34.678 ** −2.410 ***

3.108 *

0.007 ***

−0.117 ***

−0.002 ***

57.928 * −4.644 ***

0.259 *** 1.800 ***

Marginal effect

0.009 ** 0.043

Coefficient

Marginal effect

0.300 ***

0.027 ***

−0.052 ***

−0.005 ***

0.465 29.285 * −2.352 *** 0.434 *** 3.373 ***

0.039 1.666

0.311 34.277 * −2.414 ***

0.027 3.080 *

0.003 *** 0.042 ***

0.493 ***

0.006 ***

−0.116 ***

−0.001 ***

1.121 *** 56.283 * −4.681 ***

0.024 * 0.711 *

0.008 0.751 *

0.653 ***

0.009 ***

−0.126 ***

−0.002 ***

0.698 *

Coefficient

Model 12

0.008

0.536 ***

0.491

Model 11

1.290 *** 41.576 −4.045 *** 0.551 *** 4.397 ***

0.025 ** 0.744

0.003 *** 0.057 ***

0.617 ***

0.009 ***

−0.127 ***

−0.002 ***

0.014

40.691 0.542 −4.535 *** −1045.701 0.167

0.898 ** 0.018 20.649 0.748 −4.021 *** −1107.173 0.118

0.927 ** 0.019 40.437 0.531 −4.555 *** −1042.585 0.170

Table 2. (Continued ) Lower Tails Model 7 Coefficient Malaysia Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Log-likelihood Pseudo R2

Marginal effect

0.246 *** 2.049 ***

0.032 *** 0.048

0.409 **

0.047 *

Model 8 Coefficient

Marginal effect

0.237 ***

0.032 ***

−0.132 *** 0.235

−0.017 *** 0.029

Model 9 Coefficient

0.245 *** 2.085 ***

0.491 −0.963 −1.649 ***

−0.150

0.200 3.491 ***

0.007 0.204 ***

0.859 ***

0.037 ***

−0.659 −1.633 *** 0.176 −0.269 *** 0.624 **

−0.092

−1.231 −1.630 ***

0.004

0.193 3.562 ***

0.044

0.421 *** 4.776 ***

0.009 *** 0.388 ***

1.012 ***

0.022 ***

0.292 −3.324 ***

0.014

0.325 **

0.002 *

−0.415 *** 0.637 **

−0.002 *** 0.004 *

0.312 −2.954 *** 0.412 *** 4.869 ***

0.048 −0.187

0.007 0.206 ***

−3.598 −0.022 −4.976 *** −1279.716 0.190

0.071 * 0.025

0.009 *** 0.401 ***

Coefficient

0.046 *

1.195 0.034 −3.860 *** −1430.103 0.095

Marginal effect

0.236 ***

0.032 ***

−0.132 ***

−0.017 ***

0.281 −0.796 −1.622 *** 0.167

−0.109

0.004 −0.008 ***

0.880 **

0.037 *

−0.097 −3.300 ***

0.002

0.312 **

0.002

0.988 **

Model 11 Coefficient

0.240 *** 2.037 ***

Marginal effect

0.011 *** 0.047

Model 12 Coefficient

Marginal effect

0.230 ***

0.032 ***

−0.131 ***

−0.017 ***

0.904 *** −0.740 −1.652 ***

0.145 ** −0.105

0.033

−0.272 ***

−0.417 *** 1.318 ***

0.669 0.020 −3.889 *** −1427.728 0.096

0.032 *** 0.042

−0.008 *** 0.018 ** 1.135 ***

0.730 −2.992 ***

Marginal effect

Model 10

0.940 *** −1.005 −1.658 *** 0.192 3.569 ***

1.196 *** 1.017 −2.973 *** 0.412 *** 4.871 ***

0.051 ** 0.736

0.006 0.052 ***

0.033 * 0.423

0.003 *** 0.060 ***

−0.002 ***

0.163

0.004

−0.272 ***

−0.008 ***

1.124 *** 0.514 −3.322 ***

0.042 * 0.021

0.312 **

0.002 *

−0.419 ***

−0.002 ***

0.008

−3.658 −0.022 −4.964 *** −1279.937 0.190

1.434 *** 0.023 ** 2.362 0.264 −3.896 *** −1426.369 0.097

1.038 ** 0.007 −3.060 −0.018 −4.973 *** −1276.325 0.192

Table 2. (Continued ) Lower Tails Model 7 Coefficient Singapore Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal effect

0.478 *** 2.348 ***

0.029 *** 0.213 ***

0.208

0.012

Model 8 Coefficient

Marginal effect

0.621 ***

0.023 ***

−0.077 *** −0.373

−0.003 *** −0.014

Model 9 Coefficient

0.474 *** 2.384 ***

2.617 ***

0.553 ** 3.834 ***

0.007 ** 0.220 ***

0.534 *

0.007

57.146 *** −4.223 *** 0.652 *** −0.111 *** 0.003

2.101 ***

0.003 **

−0.074 ***

−0.003 ***

0.024

−0.152

−0.006

43.647 *** −3.549 ***

2.607 ***

55.483 *** −4.197 ***

0.532 ** 3.906 ***

0.007 ** 0.221 ***

0.000 *** 0.000

56.326 *** −5.435 ***

0.714 ***

95.304 *** −7.079 ***

−0.297 5.248 ***

0.000 0.084 **

−0.892

−0.0002

−0.144 *** 0.557

−0.00003 0.0001

0.002

0.399 ***

116.667 ** 0.021 −10.399 *** −574.695 0.305

0.022

55.335 *** −5.431 ***

0.694 ***

−0.307 5.590 ***

−0.0005 0.112 ***

1.811 *** 21.727 0.024 −7.233 *** −678.640 0.180

Marginal effect

0.023 ***

1.055 **

1.237 ***

0.028 *** 0.207 ***

Coefficient

0.610 ***

0.373 43.848 *** −3.553 ***

Marginal effect

Model 10

0.0058

17.286 0.018 −7.144 *** −678.399 0.180

2.085 ***

0.619 **

0.003 **

−0.110 ***

0.000 ***

0.607

Coefficient

0.444 *** 2.351 ***

Marginal effect

0.027 *** 0.194 ***

0.747 ** 42.759 *** −3.555 ***

0.057 * 2.541 ***

0.480 ** 3.914 ***

0.006 * 0.219 ***

Model 12 Coefficient

Marginal effect

0.581 ***

0.006 ***

−0.073 ***

0.000 ***

0.408 53.747 *** −4.187 ***

0.019 0.422 ***

0.611 **

0.001 **

−0.110 ***

0.000 ***

0.672 92.399 *** −7.041 ***

0.004 0.112 ***

−0.756

0.000

−0.150 ***

0.000

0.003

93.518 *** −7.086 ***

0.386 ***

−1.020

0.000

−0.147 *

0.000

1.477 **

Model 11

1.191 *** 54.485 *** −5.423 ***

0.025 0.683 ***

−0.389 5.750 ***

−0.001 0.133 ***

0.001

115.473 *** 0.018 −10.515 *** −574.212 0.306

1.537 ** 0.004 15.028 0.016 −7.061 *** −678.002 0.180

0.237 0.000 124.268 ** 0.016 −10.498 *** −575.937 0.304

Table 2.

(Continued ) Lower Tails

Model 7 Coefficient Thailand Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal effect

0.130 2.875 ***

0.013 0.080 *

0.808 ***

0.083 ***

13.255 ** −2.373 ***

1.305 *

0.218 3.863 ***

0.006 0.106 ***

0.938 ***

0.027 ***

Model 8 Coefficient

Marginal effect

0.246 **

0.019 **

−0.271 *** 0.574 ***

−0.021 *** 0.045 ***

25.795 *** −3.166 ***

2.010 ***

0.359 **

0.005 **

−0.363 *** 0.722 **

−0.005 *** 0.010 **

Model 9 Coefficient

0.123 2.921 ***

0.729 **

45.618 *** −5.519 ***

0.613 ***

0.241 5.826 ***

0.005 0.570 ***

0.411 **

0.002 **

1.064 ***

0.021 ***

−0.480 *** 0.915 ***

−0.002 *** 0.004 **

69.547 *** 0.325 *** −7.224 *** −942.430 0.301

Marginal effect

0.236 **

0.019 **

−0.273 ***

−0.021 *** 0.076 *

0.116 **

0.754 **

13.369 ** −2.341 ***

1.314 **

25.724 *** −3.150 ***

1.997 ***

0.006 0.105 ***

0.349 **

0.005 **

−0.366 ***

−0.005 ***

0.209 3.923 ***

24.511 ** −3.960 *** 0.226 5.897 ***

1.067 ** 23.761 * 0.474 * −4.508 *** −1153.610 0.144

0.012 0.076

Coefficient

0.912 ***

0.991 ** 24.252 ** −4.006 ***

Marginal effect

Model 10

0.037

0.921 ** 45.650 *** −5.498 ***

0.613 ***

0.005 0.581 ***

0.401 **

0.002 **

−0.484 ***

−0.002 ***

24.732 * 0.505 * −4.458 *** −1160.473 0.139

1.092 **

Coefficient

0.126 2.932 ***

0.406 13.202 ** −2.310 *** 0.213 3.936 ***

Marginal effect

0.013 0.074

Model 12 Coefficient

Marginal effect

0.241 **

0.019 **

−0.274 ***

−0.022 ***

0.398 25.567 *** −3.130 ***

0.036 1.992 ***

0.356 **

0.005 **

−0.368 ***

−0.005 ***

0.014 0.741 **

0.617 44.708 *** −5.464 ***

0.011 0.602 ***

0.005 0.584 ***

0.406 **

0.002 **

−0.486 ***

−0.002 ***

0.048 1.297 *

0.006 0.104 ***

0.017

0.743 **

0.028

Model 11

0.432 24.325 ** −3.925 *** 0.225 5.914 ***

0.008

70.091 *** 0.330 *** −7.195 *** −944.630 0.299

0.285 0.005 25.308 ** 0.525 * −4.420 *** −1166.109 0.135

0.716 0.004 68.848 *** 0.327 *** −7.146 *** −947.421 0.297

Notes: The period in regression covers January 2000
May 2007. In explanation variables, *, **, and *** indicates that the statistic is significant at the 10%, 5%, and 1% levels, respectively. The independent variables include the lagged number of coexceedances in domestic banks (Coexceedance lagged), the exceedance of domestic stock market index (Domestic stock index), the volatility of the general stock index (Volatility), the sum of exceedance of stock market indices in Japan and the United States (Foreign stock index), the exceedance of the US stock market index (US stock index), the exceedance of the Philadelphia Stock Exchange Bank Index (Philadelphia bank index), and the yield curve (Yield curve).

Table 3.

Regression Results in the 1990s. Upper Tails

Model 1 Coefficient Korea Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

0.627 *** 1.121 *** −0.125

Marginal effect

0.063 *** 0.086 −0.011

Model 2 Coefficient

Marginal effect

0.674 ***

0.066 ***

0.052 *** −0.176

0.005 *** −0.016

Model 3 Coefficient

0.625 *** 1.123 ***

−0.220 −2.174 *** 0.107 2.823 *** −0.359

−2.179 *** 0.000 0.132 ** −0.006

0.218 0.127 *** −0.369

0.424 * 2.621 ***

0.002

−0.262

−0.005

0.100 2.821 ***

0.570 **

0.007 *

0.146 *** −0.384

0.0019 * −0.005

0.425 * 2.641 ***

−4.249 *** −558.778 0.096

Coefficient

Marginal effect

0.670 ***

0.066 ***

0.052 ***

0.005 ***

−0.311

0.00027 0.131

−0.006

0.209

0.002

0.127 ***

0.002 ***

−0.545

−0.0128

−3.781 *** −581.879 0.058

0.573 **

0.0065 *

0.147 ***

0.002 ***

−1.016

Coefficient

0.619 *** 1.103 ***

Marginal effect

0.064 0.113

Model 6 Coefficient

Marginal effect

0.662 ***

0.066

0.051 ***

0.005

0.599 * −2.222 ***

0.077

0.540 −2.226 ***

0.066

0.080 2.786 ***

0.000 0.150

0.184

0.001

0.125 ***

0.002

−0.0052

−4.227 *** 0.008 0.146178 ***

Model 5

−0.027

−2.181 ***

−4.027 ***

−0.806 −3.785 *** −582.125 0.058

−0.019

0.002 *** −0.004

−4.222 *** 0.008 0.145 ***

0.063 *** 0.085

−2.175 ***

−0.497 −4.019 ***

Marginal effect

Model 4

0.659 −4.085 ***

0.013

0.542 −4.275 ***

0.008

0.419 * 2.605 ***

0.001 0.030

0.571 **

0.0013

0.146 ***

0.000

−0.009

−4.251 *** −558.405 0.096

−35.428 −0.025 −3.759 *** −579.246 0.063

−33.795 −0.014 *** −4.230 *** −556.078 0.100

Table 3.

(Continued ) Upper Tails

Model 1 Coefficient Malaysia Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

Marginal effect

0.530 *** 1.932 ***

0.061 *** 0.130 **

0.036

0.004

Model 2 Coefficient

Marginal effect

0.468 ***

0.045 ***

0.114 *** −0.194

0.011 *** −0.019

Model 3 Coefficient

0.531 *** 1.936 ***

0.170 −2.044 ***

−2.180 ***

0.279 3.157 ***

0.006 0.168 ***

0.113

0.003

0.185 0.168 *** −0.146

0.002

0.003 0.296 ***

0.205

0.001

0.358 0.284 *** −0.128

0.275 3.167 ***

0.0003

0.567 ** 5.363 ***

0.006 0.166 ***

0.037

Marginal effect

0.461 ***

0.045 ***

0.113 ***

0.011 ***

−0.140

0.0030 0.298 ***

0.0048

−5.344 *** −597.083 0.115

Model 5 Coefficient

0.532 *** 1.940 ***

Marginal effect

0.061 *** 0.131 **

Model 6 Coefficient

Marginal effect

0.462 ***

0.045 ***

0.113 ***

0.011 ***

−0.014 0.024 −2.043 ***

0.186

0.002

0.167 ***

0.003 ***

0.400

0.009

0.284 3.169 ***

0.111 −3.469 ***

−3.927 ***

0.0003 * −0.0001

−6.858 *** −521.538 0.227

Coefficient

−2.189 ***

−3.523 ***

0.658 −5.328 *** −598.069 0.114

0.014

0.003 *** −0.002

−3.887 ***

0.575 ** 5.347 ***

0.061 *** 0.130 **

−2.050 ***

0.862 −3.473 ***

Marginal effect

Model 4

0.359

0.0003

0.283 ***

0.0003 *

0.142

0.000

−6.874 *** −521.430 0.227

0.532 * 5.372 ***

0.001

0.007 0.170 ***

0.003

0.003 0.295 ***

1.098 0.011 −5.375 *** −597.376 0.115

−0.089 −2.192 ***

−0.009

0.182

0.002

0.167 ***

0.003 ***

−0.022 −3.899 ***

−0.0003

0.348

0.0003

0.283 ***

0.0003 *

0.725 0.001 −6.918 *** −521.498 0.227

Table 3.

(Continued ) Upper Tails

Model 1 Coefficient Singapore Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

Marginal effect

0.198 1.953 ***

0.013 0.192 ***

0.327

0.023

Model 2 Coefficient

Marginal effect

0.131

0.007

0.063 *** 0.118

0.0037 *** 0.007

Model 3 Coefficient

0.210 1.963 ***

0.622 −2.605 ***

−2.749 ***

0.751 *** 3.009 ***

0.010 *** 0.118 **

0.139

0.002

0.716 *** 0.103 *** −0.168

0.005 **

0.0004 0.127 *

0.458

0.002

−0.084 0.170 *** 0.033

0.746 *** 3.047 ***

−0.0001

0.117 4.200 ***

0.010 *** 0.120 **

−0.004

Marginal effect

0.138

0.008

0.063 ***

0.004 ***

0.230

0.015

0.699 ***

0.005 **

0.104 ***

0.0007 ***

−0.931

0.0004 0.135 ***

−0.088 0.171 ***

0.00052

−5.442 *** −412.337 0.102

−0.438

Model 5 Coefficient

0.200 2.008 ***

Marginal effect

0.014 0.233

Model 6 Coefficient

Marginal effect

0.132

0.007

0.064 ***

0.004

0.177 −2.581 ***

0.013

−0.202 −2.729 ***

−0.011

0.755 *** 3.018 ***

0.010 0.131

0.709 ***

0.005

0.102 ***

0.0007

0.012

−0.005 −4.991 ***

0.0001

0.0001 0.031

−0.125

−0.005 0.645 −4.461 ***

−4.968 ***

0.00009 0.00001

−7.509 *** −369.456 0.195

Coefficient

−2.750 ***

−4.407 ***

0.160 −5.483 *** −412.810 0.101

0.057

0.0007 *** −0.001

−4.977 ***

0.105 4.129 ***

0.014 0.190 ***

−2.611 ***

−0.307 −4.432 ***

Marginal effect

Model 4

−0.00005

0.126 4.211 ***

0.00009

0.174 ***

−0.00002 0.00002

0.000

−7.494 *** −368.841 0.197

−32.532 −0.005 ** −5.385 *** −412.480 0.101

−30.865 −0.0005 −7.499 *** −368.232 0.198

Table 3.

(Continued ) Upper Tails

Model 1 Coefficient Thailand Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

0.359 *** 1.580 * 0.173

Marginal effect

0.035 *** −0.073 ** 0.023

Model 2 Coefficient

Marginal effect

0.358 ***

0.031 ***

0.094 *** 0.012

0.008 *** 0.003

Model 3 Coefficient

0.361 *** 1.569 *

0.445 −2.169 *** 0.137 4.628 *** −1.554

−2.352 *** 0.002 0.148 *** −0.044

0.144 0.168 *** −1.569

0.001

0.017 *** 0.704 ***

0.106

0.004

0.570 *** 0.241 *** −0.063

0.134 4.566 ***

0.002 **

0.550 *** 5.957 ***

0.001 0.087

−0.033 ***

Marginal effect

0.359 ***

0.031 ***

0.094 ***

0.008 **

0.197

0.020

0.127

0.001

0.168 ***

0.001

−13.711

0.017 0.765

0.0117

−3.690 *** −614.8816 0.1695

Coefficient

0.359 *** 1.605 *

0.154 4.479 ***

Marginal effect

0.034 *** −0.071 **

−0.010

0.003 0.138 ***

Model 6 Coefficient

Marginal effect

0.356 ***

0.031 ***

0.095 ***

0.008 ***

−0.248 −2.340 ***

−0.020

0.145

0.001

0.165 ***

0.002 ***

−0.012 *** 0.235 −3.563 ***

0.568 ***

0.002 *

0.241 ***

0.001 **

−0.155

Model 5

−0.043 −2.149 ***

−4.354 ***

0.001 *** −0.0002

−5.779 *** −499.88385 0.3248

Coefficient

−2.361 ***

−3.496 ***

0.342 −3.682 *** −615.710 0.168

0.056

0.002 *** −0.016

−4.327 ***

0.549 *** 5.970 ***

0.035 −0.070

−2.177 ***

−13.620 −3.469 ***

Marginal effect

Model 4

0.550 *** 5.972 ***

0.006

0.016 *** 0.709 ***

−0.101 −4.379 ***

−0.001

0.590 ***

0.002 **

0.242 ***

0.001 ***

−0.001

−5.778 *** −499.035 0.326

0.878 0.042 −3.733 *** −617.02082 0.167

0.667 0.003 −5.879 *** −500.994 0.323

Table 3.

(Continued ) Lower Tails

Model 7 Coefficient

Korea Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

0.669 *** 1.244 *** −0.125 **

Marginal effect

0.072 0.105 −0.013

Model 8 Coefficient

Marginal effect

0.785 ***

0.080

−0.074 *** −0.150

−0.007 −0.014

Model 9 Coefficient

0.681 *** 1.226 ***

−0.549

Marginal effect

0.074 0.126

−0.046

Model 10 Coefficient

0.79886 ***

Marginal effect

0.081

−0.0743 ***

−0.007

−0.6044

−0.049

Model 11 Coefficient

0.667 *** 1.249 ***

−0.083 −2.117 ***

−2.190 ***

0.376 3.066 ***

0.008 0.282

0.146

0.005

0.651 ** −0.180 *** −0.154

−2.108 *** 0.008

0.415 * 2.986 ***

−2.1819 *** 0.005 0.153

−0.003 −0.002

0.68112 *** −0.1792 ***

−14.145

−0.032 ***

−13.984

0.925 *** 3.249 *** −34.136

−4.199 *** 0.000 0.004 −0.012

1.230 *** −0.197 *** −32.023

−3.511 *** 0.000

0.914 *** 3.246 ***

0.005

1.20167 *** −0.1911 ***

0.0000 −0.006 −14.084

−0.010 ***

−13.13

0.367 3.057 ***

−5.533 *** −542.374 0.151

−4.797 *** −584.512 0.085

Coefficient

0.78635 ***

Marginal effect

0.079 ***

−0.0745 ***

−0.007 ***

−0.2244

−0.021

−2.1947 *** 0.008 0.246 ***

0.63773 ** −0.1809 ***

0.008 ** −0.003 ***

−0.016 *** 0.016

−3.586 *** 0.003

0.877 *** 3.325 ***

0.13764

0.003

−4.2224 *** 0.007 ** 0.103 **

0.000

1.18112 ***

0.005 **

−0.1957 ***

−0.001 ***

−0.4678

−0.002

−0.005 ** −0.074

−4.759 *** −585.693 0.083

−0.011

−0.001

−4.1582 *** 0.004 0.064

0.070 *** 0.083 **

−2.125 ***

0.432 −3.579 ***

Marginal effect

Model 12

−5.5133 *** −541.637 0.152

−0.001

−4.848 *** −587.080 0.081

−5.5885 *** −544.352 0.148

Table 3.

(Continued ) Lower Tails

Model 7 Coefficient

Malaysia Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

Marginal effect

0.464 *** 1.954 ***

0.051 *** 0.134 **

0.532 **

0.060 **

Model 8 Coefficient

Marginal effect

0.566 ***

0.058 ***

−0.087 *** 0.267

−0.009 *** 0.029

Model 9 Coefficient

0.457 *** 1.967 ***

0.627 *

Marginal effect

0.050 *** 0.133 **

0.084

Model 10 Coefficient

Marginal effect

0.564 ***

0.058 ***

−0.087 ***

−0.009 ***

0.227

Model 11 Coefficient

0.465 *** 1.950 ***

−2.152 ***

0.479 ** 3.347 ***

0.012 * 0.193 ***

0.188

0.003

−2.080 ***

0.566 **

0.008 *

−0.163 *** −0.314

−0.002 *** −0.006

0.475 ** 3.353 ***

0.236

−2.136 *** 0.012 * 0.190 ***

0.003

−3.623 ***

−4.072 ***

0.813 *** 5.386 ***

0.004 ** 0.270 ***

1.064 *

0.005

0.801 *** −0.226 *** 0.557

0.002

0.789 *** 5.418 ***

0.571 **

0.008 **

−0.163 ***

−0.002 ***

−0.414

−0.006

−4.081 *** 0.004 ** 0.278 ***

−0.0005 ** 0.001

0.788 *** −0.227 ***

1.300

0.012

0.721

0.479 ** 3.370 ***

−6.201 *** −531.654 0.202

−5.554 *** −581.610 0.127

Marginal effect

0.568 ***

0.059 ***

−0.088 ***

−0.009 ***

0.002

0.820 *** 5.336 ***

0.034

0.005

−2.125 *** 0.012 * 0.195 ***

−0.017

−3.588 ***

0.568 **

0.008 *

−0.166 ***

−0.002 ***

−1.335

−0.012 **

−4.073 *** 0.004 ** 0.265 ***

0.818 *** −0.225 ***

0.000 **

0.002 * −0.001 **

0.002 1.059

−5.621 *** −580.485 0.129

0.038

−2.061 ***

−0.774 −3.631 ***

0.052 *** 0.135 **

Coefficient

0.026 0.289

−2.107 ***

Marginal effect

Model 12

−6.174 *** −532.210 0.201

0.010

−5.538 *** −581.990 0.127

0.386

0.001

−6.123 *** −531.542 0.202

Table 3. (Continued ) Lower Tails Model 7 Coefficient

Singapore Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

Marginal effect

0.020 2.230 ***

0.001 0.292 ***

0.099

0.008

Model 8 Coefficient

0.176 −0.076 *** −0.197

Marginal effect

0.011

Model 9 Coefficient

0.010 2.292 ***

Marginal effect

0.000 0.311 ***

−0.005 *** −0.013 −0.151

−0.014

Model 10 Coefficient

0.167

Marginal effect

0.011

−0.077 ***

−0.005 ***

−0.530

−0.029

Model 11 Coefficient

0.025 2.227 ***

0.158 −2.376 ***

−2.536 ***

0.822 *** 2.574 ***

0.009 ** 0.054

0.433

0.005

−2.361 ***

1.029 **

0.006 *

−0.104 *** −0.074

−0.001 *** 0.000

0.872 *** 2.473 ***

−2.530 *** 0.010 ** 0.046

1.105 *** −0.104 ***

1.036

0.019

0.429

−5.126 ***

0.849 ** 24.779

0.000 0.128 **

−0.915

0.000

0.699 −0.344 *** −9.261 *

−4.690 *** 0.000

0.828 ** 18.457

0.006 **

0.121 −0.220 ***

0.000 0.000 −0.416

0.000

−2.393

0.795 ** 2.775 ***

−15.346 *** −357.908 0.227

−19.825 −407.402 0.120

Coefficient

0.181

Marginal effect

0.012

−0.075 ***

−0.005 ***

−0.114

−0.007

−2.543 *** 0.009 ** 0.070 *

1.023 **

0.005 *

−0.113 ***

−0.001 ***

−1.099

−0.004

0.003 −0.004

−4.577 *** 0.00000

0.951 ** 24.516

−5.221 *** 0.000 0.100 **

−0.00001

0.297

0.00001

−0.221 ***

−0.00001

−0.723

−0.00002

0.000 0.527

−26.031 *** −407.487 0.120

0.015

−0.001 ***

−5.248 *** 0.000 0.121 **

0.001 0.297 ***

−2.375 ***

−0.335 −4.644 ***

Marginal effect

Model 12

−9.890 *** −360.702 −360.702

0.000

−26.162 *** −408.227 0.118

−10.059 *** −361.676 0.219

Table 3.

(Continued ) Lower Tails

Model 7 Coefficient

Thailand Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Constant Log-likelihood Pseudo R2

Marginal effect

0.183 2.930 ***

0.016 0.054

0.262

0.020

Model 8 Coefficient

0.180 −0.126 *** 0.062

Marginal effect

0.012

Model 9 Coefficient

0.180 2.935 ***

−2.635 ***

0.139 4.180 ***

0.003 0.110 **

1.035 ***

0.029 ***

0.137 −0.160 *** 0.853 **

0.013 ** 0.616 ***

0.814 **

0.024 **

0.355 *

0.152 4.239 ***

−0.238 *** 0.617

0.038

0.434 *** 5.785 ***

0.012

0.151

0.517

0.013 *** 0.624 ***

0.375 * −0.240 ***

0.018

−3.670 *** −604.318 0.151

−0.035

Model 11 Coefficient

0.194 2.928 ***

Marginal effect

0.017 0.045

−0.008 ***

Model 12 Coefficient

0.183 −0.126 ***

Marginal effect

0.012 −0.008 ***

−0.001

0.002

0.403 −2.225 ***

0.042

0.190 4.284 ***

0.005 0.122

−0.002 ***

0.076 −2.633 *** 0.169

0.005

0.002

−0.164 ***

−0.002 ***

−0.002

0.009

−4.293 ***

−0.001 *** 0.003

−5.488 *** −482.570 0.322

0.000

−0.162 ***

0.561 −3.738 *** −601.557 0.154

0.003 0.113 **

−3.590 *** 0.002

0.183

Marginal effect

−2.630 ***

−0.002 *** 0.011 *

−4.365 ***

0.421 *** 5.733 ***

0.033

−2.219 *** 0.002

Coefficient

−0.126 ***

0.942 −3.670 ***

0.016 0.046

−0.008 *** 0.003 0.376

−2.225 ***

Marginal effect

Model 10

0.002

0.137 −3.546 ***

0.002

−0.149 −4.269 ***

0.454 *** 5.768 ***

0.014 0.615

0.375 **

−0.001 ***

−0.239 ***

0.002 −0.001 ***

0.000

−5.436 *** −484.552 0.319

0.620 0.023 −3.680 *** −604.933 0.150

0.203 0.001 −5.442 *** −484.831 0.319

Notes: The period in regression covers 1993
1996. In explanation variables, *, **, and *** indicates that the statistic is significant at the 10%, 5%, and 1% levels, respectively. The independent variables include the lagged number of coexceedances in domestic banks (Coexceedance lagged), the exceedance of domestic stock market index (Domestic stock index), the volatility of the general stock index (Volatility), the sum of exceedance of stock market indices in Japan and the United States (Foreign stock index), the exceedance of the US stock market index (US stock index), the exceedance of the Philadelphia Stock Exchange Bank Index (Philadelphia bank index), and the yield curve (Yield curve).

Table 4.

Regression Results during the Global Financial Crisis. Upper Tails

Model 1

Korea Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Coefficient

Marginal Effect

0.128 1.807 **

0.011 0.074

0.388

0.027

Model 2 Coefficient

Marginal Effect

0.142

0.012

0.017 *** 0.342

0.001 ** 0.026

Model 3 Coefficient

0.109 1.926 ***

0.851 22.336 −2.403 *** −0.023 1.951 ** 0.813

1.936

−0.002 0.038 0.027

22.934 −2.365 *** 0.001 0.021 ** 0.730

1.898

−0.001

Marginal Effect

0.009 0.047

0.076

22.253 −2.413 *** −0.001 2.212 **

1.937

−0.001 0.036

0.001 ** 0.024 0.980

0.038

Model 4 Coefficient

Marginal Effect

0.127

0.011

0.017 ***

0.001 **

0.744

0.075

22.714 −2.370 ***

1.871

0.025

0.000

0.023 **

0.001 **

0.837

0.034

Model 5 Coefficient

0.134 1.873 ***

0.490 20.545 −2.390 *** −0.002 2.101 **

1.039 23.382 −3.368 ***

0.778

0.181 3.942 ***

0.007 0.480 ***

1.243 ***

0.046 ***

24.980 −3.358 ***

0.776

22.446 −3.350 ***

0.290

0.005

0.241 4.396 ***

0.067 *** 1.013 **

0.001 *** 0.0188 * 1.484 **

0.752

0.009 0.570 ***

0.0899

24.552 −3.340 ***

0.760

0.357 *

0.006 *

0.073 ***

0.001 ***

1.091 *

0.027

19.426 −3.346 *** 0.373 ** 4.388 ***

−0.548 8.418 0.197 −3.430 *** −396.537 0.108

41.093 * 0.747 −4.178 *** −386.897 0.130

4.001 −3.379

0.028 −397.852 0.105

43.106 * 0.737 * −4.241 *** −388.254 0.127

Marginal Effect

0.011 0.040

0.047 1.793

−0.001 0.028

0.057 0.632

0.015 ** 0.583 ***

−0.021

4.159 0.048 −3.303 *** −399.434 0.101

Model 6 Coefficient

Marginal Effect

0.146

0.012

0.017 ***

0.001 **

0.469 21.207 −2.353 ***

0.043 1.746

0.017

0.000

0.023 **

0.001 **

1.009

0.051

21.454 −3.344 ***

0.642

0.453 **

0.008

0.074 ***

0.001 ***

−0.308

−0.007

44.778 * 0.777 * −4.215 *** −388.932 0.125

Table 4.

(Continued ) Upper Tails

Model 1 Coefficient

Malaysia Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Log-likelihood Pseudo R2

Marginal Effect

0.348 *** 2.104 ***

0.047 *** 0.026

0.549 *

0.069

Model 2 Coefficient

Marginal Effect

0.390 ***

0.051 ***

0.094 *** 0.396

0.012 *** 0.050

Model 3 Coefficient

0.346 *** 2.115 ***

0.617 9.805 −1.826 ***

1.188

0.205 3.706 ***

0.006 0.189 **

0.656

0.023

24.888 −3.346 ***

1.000

0.528 ** 5.421 ***

0.008 * 0.455 ***

1.787 ***

0.030 ***

3.379 −1.768 ***

0.338

9.648 −1.804 ***

0.269

0.004

0.178 3.727 ***

0.182 *** 0.542

0.003 *** 0.010

0.065 1.177

0.004 0.174 **

Marginal Effect

0.384 ***

0.050 ***

0.095 ***

0.012 ***

0.548

0.072

3.242 −1.756 ***

0.320

0.232

0.003

0.184 ***

0.003 ***

0.062

1.282 *

0.039

0.969

27.935 −3.951 ***

0.535

0.700 **

0.002

0.272 ***

0.001 **

1.852 **

0.015

27.726 −3.925 ***

0.546

24.457 −3.345 ***

0.622 **

0.002

0.593 ** 5.461 ***

0.272 *** 1.786 ***

0.001 ** 0.0053 *

48.684 0.148 −6.180 *** −389.559 0.226

0.047 *** 0.014

Model 4 Coefficient

1.191 *

1.845 *** 12.689 0.176 −4.462 *** −442.210 0.121

Marginal Effect

0.010 ** 0.489 ***

0.0700

9.273 0.128 −4.268 *** −446.261 0.113

43.025 0.153 −5.921 *** −392.385 0.220

Model 5 Coefficient

0.345 *** 2.060 ***

0.611 8.354 −1.793 *** 0.205 3.647 ***

0.740 23.042 −3.307 *** 0.677 *** 5.364 ***

Marginal Effect

0.046 ** 0.012

0.089 0.982

0.005 0.168 **

0.032 0.923

0.013 ** 0.493 ***

0.735 0.015 12.076 0.203 −4.207 *** −448.726 0.108

Model 6 Coefficient

Marginal Effect

0.391 ***

0.051 ***

0.094 ***

0.012 ***

0.362 2.902 −1.747 ***

0.052 0.274

0.293

0.005

0.180 ***

0.003 ***

0.331 26.355 −3.867 ***

0.006 0.530

0.813 ***

0.003 *

0.268 ***

0.001 **

0.305 0.001 40.842 0.158 −5.791 *** −394.891 0.215

Table 4.

(Continued ) Upper Tails

Model 1 Coefficient

Singapore Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal Effect

0.338 4.159 ***

0.010 0.411

1.000 **

0.030

Model 2 Coefficient

Marginal Effect

0.496 *

0.006

0.046 *** 0.596

0.0005 *** 0.007

Model 3 Coefficient

0.344 4.304 ***

1.297 * 44.143 −4.413 ***

1.337

0.291 5.168 ***

0.002 0.297

2.065 ***

0.017

81.262 ** −5.663 ***

0.965 **

0.559

0.001

0.059 *** 1.622 ***

0.0001 0.004

46.303 −4.434 0.416 5.362 ***

2.053 **

Marginal Effect

0.010 0.352

0.063

Model 4 Coefficient

Marginal Effect

0.523 *

0.006

0.047 ***

0.001 ***

0.714

0.011

1.381

84.499 ** −5.722 ***

0.004 0.332

0.701 *

0.002

0.061 ***

0.0002 *

1.041

0.004

0.053

0.973 **

Model 5 Coefficient

0.543 ** 4.216 ***

−17.902 −4.977 ***

−0.158

−33.486 6.170 ***

−0.001 0.004

2.289 ***

0.000

37.413 −6.824 *** −34.426 0.118 *** 1.161

0.079

−16.678 −4.729

−0.181

0.0000

−13.752 6.746 ***

−0.0095 0.087

0.00000 0.0000 3.157 ***

33.585 0.001 −6.152 *** −153.792 0.343

203.993 *** 0.000 −13.960 *** −130.277 0.443

0.0116

21.886 0.014 −5.989 *** −158.217 0.324

49.272 −6.738 *** −12.761

Coefficient

Marginal Effect

0.016 0.392

0.654 **

0.008

0.047 ***

0.001 ***

−0.550 52.740 −4.442 ***

−0.014 1.604

0.614 * 5.121 ***

0.006 0.338

1.134

0.022

0.125

−16.943 −4.615 ***

−0.202

0.0000

−32.204 6.123 ***

−0.002 0.005

0.121 ***

0.00000

1.908

0.000

211.701 *** 0.000 −14.205 *** −133.497 0.430

Model 6

Marginal Effect

−0.759 87.853 ** −5.711 ***

−0.006 1.020 **

0.789 **

0.002

0.060 ***

0.0002 *

0.888

0.0036

44.226 −6.672 *** −35.622 0.116 ***

1.859 * 0.000 −1.790 0.000 −5.296 *** −161.12808 0.312

0.112

0.0000 0.00000

0.996 0.0000 186.359 ** 0.000 −13.321 *** −133.622 0.429

Table 4.

(Continued ) Upper Tails

Model 1 Coefficient

Thailand Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal Effect

0.246 1.245 *

0.018 0.020

0.536

0.041

Model 2 Coefficient

Marginal Effect

0.275

0.021

0.083 *** 0.481

0.006 *** 0.036

Model 3 Coefficient

0.243 1.337 *

0.595 50.382 *** −3.175 ***

4.037 ***

0.512 ** 2.950 ***

0.015 ** 0.166 **

0.520

0.014

44.518 ** −3.110 ***

3.416 **

50.224 *** −3.151

0.583 ***

0.009 **

0.489 ** 3.042 ***

0.176 *** 0.740

0.003 *** 0.011 1.114

67.158 ** −4.654 ***

1.956 **

0.496 ** 3.811 ***

0.013 ** 0.367 ***

0.967 **

0.025 **

66.599 ** −5.183 ***

1.011 **

0.713 ***

0.003 *

0.272 *** 1.561 ***

0.001 ** 0.006 *

69.214 ** −4.709 0.511 ** 3.993 ***

0.575

Marginal Effect

0.018 0.018

0.052 4.054 ***

0.014 ** 0.159 **

0.050 1.988 **

0.014 ** 0.409 ***

0.016

Model 4 Coefficient

Marginal Effect

0.269

0.020

0.084 ***

0.006 ***

0.599

0.052

43.871 ** −3.079

3.376 **

0.551 **

0.008 **

0.178 ***

0.003 ***

1.389 *

0.038

66.901 ** −5.201 ***

0.989 **

0.673 ***

0.003 *

0.268 ***

0.001 **

1.284

0.010

Model 5 Coefficient

0.245 1.326 *

42.580 0.151 −6.267 *** −315.37195 0.2391

4.546 −0.076 −3.784 *** −367.39722 0.1136

23.481 0.092 −5.665 *** −317.297 0.235

0.018 0.020

Model 6 Coefficient

Marginal Effect

0.275

0.021

0.085 ***

0.006 ***

0.301 49.899 *** −3.129 ***

0.010 4.042 **

0.424 43.996 ** −3.070 ***

0.027 3.418 **

0.518 ** 3.040 ***

0.014 ** 0.157 **

0.643 ***

0.009 **

0.180 ***

0.003 ***

1.570 ** 70.319 ** −4.805 *** 0.503 ** 3.981 ***

1.315 ** 13.883 0.189 −4.021 *** −366.066 0.117

Marginal Effect

0.084 1.936 **

0.013 ** 0.402 ***

0.056

8.903 0.051 −3.924 *** −365.27399 0.119

1.994 *** 62.551 ** −5.250 ***

0.073 0.872 *

0.794 ***

0.003 *

0.272 ***

0.001 **

2.295 ***

0.029

19.378 0.063 −5.827 *** −313.760 0.243

Table 4.

(Continued ) Lower Tails

Model 7 Coefficient

Korea Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal Effect

0.303 ** 2.499 ***

0.028 * 0.061

0.688 *

0.063

Model 8 Coefficient

Marginal Effect

0.410 ***

0.039 ***

−0.028 *** 0.508

−0.003 *** 0.049

Model 9 Coefficient

0.309 ** 2.664 ***

0.781 2.836 −2.244 ***

0.180

0.325 3.820 *

0.011 0.170 *

0.734

0.024

2.103 −2.263 ***

0.091

0.578 **

0.011 **

−0.066 *** 0.348

−0.001 *** 0.006

1.631 −2.221 *** 0.313 3.968 ***

1.058

Marginal Effect

0.029 ** 0.051

0.070 0.081

0.010 0.154 *

0.044

Model 10 Coefficient

Marginal Effect

0.413 ***

0.039 ***

−0.029 ***

−0.003 ***

0.672 0.935 −2.246 ***

0.324 ** 2.638 ***

−0.015

0.010 **

−0.067 ***

−0.001 ***

1.312 ** −4.600 −2.208 *** 0.344 3.963 **

0.160

0.411 * 4.990 ***

0.012 * 0.493 ***

1.233 ***

0.036 **

23.858 −4.166 *** 0.767 *** −0.097 *** 0.822 *

0.485

0.005 **

3.891 −3.376 *** 0.431 ** 5.326 **

0.119

0.012 * 0.541 ***

−0.001 *** 0.006

22.169 −4.161 *** 0.777 *** −0.099 ***

1.753 ***

0.101

1.348 *

0.450

0.005 **

−3.955 −3.349 *** 0.486 ** 5.288 ***

73.649 ** 0.540 ** −5.737 *** −346.830 0.233

17.624 0.555 −3.811 *** −390.965 0.136

0.132 −0.458

0.011 0.162 **

0.056 −0.132

0.013 ** 0.526 ***

−0.001 ***

Marginal Effect

0.423 ***

0.040 ***

−0.028 ***

−0.003 ***

1.149 ** −4.429 −2.235 ***

0.151 −0.529

0.571 **

0.011 **

−0.066 ***

−0.001 ***

1.090 17.533 −4.137 *** 0.819 *** −0.098 ***

0.026 0.372

0.005 ** −0.001 ***

0.016 2.353 ***

22.267 0.708 −3.838 *** −390.439 0.137

0.030 ** 0.055

Model 12 Coefficient

0.020 1.437 *

5.227 −3.395 ***

Marginal Effect

0.077

0.555 **

0.810

Model 11 Coefficient

68.837 ** 0.482 * −5.744 *** −346.840 0.233

0.150 **

0.492 0.038 −3.773 *** −387.213 0.144

1.948 ***

0.027

56.185 * 0.384 * −5.731 *** −343.990 0.240

Table 4. (Continued ) Lower Tails Model 7

Malaysia Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Interest rate Constant Log-likelihood Pseudo R2

Coefficient

Marginal Effect

0.097 1.977 **

0.010 0.033

0.473

0.048

Model 8 Coefficient

0.096 −0.095 *** 0.124

Marginal Effect

0.011

Model 9 Coefficient

0.112 2.068 ***

Marginal Effect

0.011 0.013

−0.010 *** 0.013 0.290

0.016

Model 10 Coefficient

0.106

Marginal Effect

0.012

−0.096 ***

−0.011 ***

−0.063

−0.009

Model 11 Coefficient

0.111 2.053 ***

−0.273

0.207 3.432 ***

0.007 0.145 *

1.066 **

0.035 **

5.181 −2.025 *** 0.208 −0.190 *** 0.664

0.466

−0.714 −1.847 ***

0.003

0.271 3.734 ***

0.913

0.272 5.018 ***

0.006 0.494 ***

1.278 ***

0.028 **

46.771 −4.616 *** 0.385 −0.311 *** 1.128 **

0.793

24.080 −3.526 ***

0.001

0.331 5.356 ***

0.020 0.895

0.007 0.524 ***

−0.001 * 0.002

96.538 ** 0.180 −7.519 *** −344.737 0.260

5.258 −2.019 *** 0.252 −0.196 ***

1.934 *** 24.229 0.552 −4.158 *** −408.460 0.124

0.009 0.163 **

−0.003 *** 0.011 0.611

25.393 −3.600 ***

−0.293

0.100

20.596 0.445 −4.187 *** −409.694 0.121

0.166 45.529 −4.573 *** 0.440

0.011 0.010

Model 12 Coefficient

0.093 −0.095 ***

1.116 ** −0.382 −1.866 ***

Marginal Effect

0.483

−4.034 −1.862 ***

0.004

0.292 3.728 ***

0.131 −0.649

0.010 0.158 *

−0.003 ***

Marginal Effect

0.010 −0.010 ***

0.890 *

0.124

2.378 −2.039 ***

0.159

0.256 −0.198 ***

0.004 −0.003 ***

0.003 0.772

0.001

−0.325 ***

0.000

1.938 ***

0.008

88.233 ** 0.135 −7.659 *** −343.874 0.262

1.367 ** 20.073 −3.551 *** 0.420 * 5.391 ***

0.055 0.759

0.009 0.534 ***

2.251 *** 0.105 * 12.227 0.268 −4.174 *** −406.949 0.127

1.235 * 42.954 −4.647 ***

0.028 0.705

0.511 *

0.001

−0.323 ***

−0.001 *

2.469 *** 0.013 81.922 * 0.136 −7.578 *** −342.584 −342.584

Table 4.

(Continued ) Lower Tails

Model 7 Coefficient

Singapore Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

−0.136 3.221 *** 0.971 **

Marginal Effect

−0.005 0.178 ** 0.036 **

Model 8 Coefficient

0.207 −0.041 *** 0.641

Marginal Effect

0.003

Model 9 Coefficient

−0.033 3.497 ***

0.666

0.615 4.916 ***

0.003 0.123 *

1.390 **

0.006

−0.001 0.189 **

−0.001 *** 0.011 0.420

17.849 −3.661 ***

Marginal Effect

42.232 −4.815 ***

0.698

17.413 −3.579 ***

1.137 **

0.001

0.570 5.157 ***

−0.068 *** 1.110 **

0.000 0.001 2.146 **

9.557 −5.897 ***

0.040 ***

78.941 −8.852 ***

0.049

0.292 5.937 ***

0.002 0.420 ***

0.835 *

0.0006

1.015 *

0.006

−0.077 *** 0.663

−0.00006 0.0005

−17.232 −5.531 *** 0.384 6.255 ***

0.215

Model 10 Coefficient

0.361

Marginal Effect

0.006

−0.043 ***

−0.001 ***

0.018

−0.055

−0.001

0.673

43.202 −4.803 ***

0.707

1.194 ***

0.001

−0.069 ***

0.000

1.680 *

0.002

45.155 −8.418 ***

0.026

0.976 **

0.0007

−0.079 ***

−0.00006

0.003 0.117 *

0.029 −0.082

0.0024 0.466 ***

0.0011

−0.441

Model 11 Coefficient

−0.045 3.481 ***

126.507 ** 0.092 −9.309 *** −144.353 0.377

49.770 0.303 −6.117 *** −162.604 0.298

−0.002 0.184 **

Model 12 Coefficient

0.323 −0.042 ***

0.947 14.564 −3.576 *** 0.754 ** 5.279 ***

2.192 ** −23.534 −5.513 *** 0.351 6.228 ***

0.052 0.557

0.003 0.128 **

0.028 −0.106

0.002 0.459 ***

Marginal Effect

0.005 −0.001 ***

0.528 39.347 −4.770 ***

0.011 0.643

1.373 ***

0.001

−0.069 ***

0.000

1.832 * 46.224 −8.484 ***

0.003 0.026

0.914 *

0.0006

−0.080 ***

−0.00005

−0.151

−0.0001

0.000 0.662

47.728 0.279 −6.183 *** −161.444 0.303

Marginal Effect

132.158 ** 0.094 −9.359 −144.287 0.377

0.005

44.599 0.274 −6.054 *** −162.598 0.298

133.520 ** 0.089 −9.450 *** −144.494 0.376

Table 4.

(Continued ) Lower Tails

Model 7 Coefficient

Thailand Coexceedance = 1 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance = 2 Coexceedance lagged Domestic stock index Volatility Foreign stock index US stock index Philadelphia bank index Yield curve Constant Coexceedance ≥ 3 Coexceedance lagged Domestic stock index Volatility Foreign stock index

Marginal Effect

0.254 3.455 ***

0.023 0.077

0.439

0.038

Model 8 Coefficient

Marginal Effect

0.337 *

0.028 *

−0.102 *** 0.237

−0.008 *** 0.019

Model 9 Coefficient

0.303 ** 3.504 ***

0.036

Marginal Effect

0.027 * 0.066

−0.007

Model 10 Coefficient

Marginal Effect

0.385 **

0.032 **

−0.102 ***

−0.008 ***

−0.189

−0.017

Model 11 Coefficient

0.274 3.504 ***

0.346 −1.322 −2.298 ***

−0.183

0.299 4.901 ***

0.010 0.218 **

0.633

0.022

−0.131 −2.452 *** 0.390 −0.244 *** 0.620

−0.021

−3.352 −2.247 ***

0.004

0.301 4.946 ***

−0.359

0.010 0.201 **

−0.003 *** 0.006

−1.053 −2.424 *** 0.353 −0.249 ***

0.885

0.045

1.170

−0.097

−3.591 −2.249 ***

0.003

0.335 4.975 ***

0.792

0.286 6.693 ***

0.004 0.517 ***

1.305 ***

0.018 **

12.637 −4.945 *** 0.321 −0.427 *** 1.412 **

0.136

18.640 −3.703 ***

0.000

0.308 6.824 ***

0.000 0.000

0.740

0.004 0.550 ***

11.789 −4.973 *** 0.271 −0.431 ***

Coefficient

0.025 0.061

0.341 *

0.028 *

−0.103 ***

−0.008 ***

0.027 −0.363

0.012 0.202 **

0.293 −1.314 −2.430 *** 0.365 −0.248 ***

−0.002 ***

Marginal Effect

0.025 −0.116

0.004 −0.003 ***

0.021 0.439

20.223 −3.736 ***

Model 12

Marginal Effect

0.122

16.945 −3.659 ***

0.000

0.293 6.911 ***

0.000

0.015 0.688

0.004 0.555 ***

1.058 7.837 −4.873 *** 0.112 −0.458 ***

0.018 0.084

0.000 0.000

Table 4.

(Continued ) Lower Tails

Model 7 Coefficient

US stock index Philadelphia bank index Yield curve Constant Log-likelihood Pseudo R2

Marginal Effect

Model 8 Coefficient

Marginal Effect

Model 9 Coefficient

1.509 *

−19.820 −0.307 −4.213 *** −337.392 0.177

−99.207 −0.032 −6.875 *** −283.910 0.308

Marginal Effect 0.046

−25.973 −0.413 −4.034 *** −339.128 0.173

Model 10 Coefficient

2.149 **

Marginal Effect

Model 11 Coefficient

Marginal Effect

1.745 **

0.057

Model 12 Coefficient

Marginal Effect

0.002

−108.79 * −0.036 −6.669 *** −284.122 0.307

−34.875 −0.528 −3.953 *** −338.798 0.174

3.454 ***

0.005

−129.01 ** −0.026 −6.881 *** −282.362 0.311

Notes: The period in regression covers June 2007
October 2009. In explanation variables, *, **, and *** indicates that the statistic is significant at the 10%, 5%, and 1% levels, respectively. The independent variables include the lagged number of coexceedances in domestic banks (Coexceedance lagged), the exceedance of domestic stock market index (Domestic stock index), the volatility of the general stock index (Volatility), the sum of exceedance of stock market indices in Japan and the United States (Foreign stock index), the exceedance of the US stock market index (US stock index), the exceedance of the Philadelphia Stock Exchange Bank Index (Philadelphia bank index), and the yield curve (Yield curve).

External Shocks on Stock Prices in East Asian Domestic Banking

145

A few coefficients for the foreign stock market shock (the United States and Japan) and the US stock market shock, are significantly positive for Singapore. In contrast, the coefficients for shocks to the domestic stock market are significant, and the sign of the coefficient for the volatility of the STI is positive (negative) in the upper (lower) tail coexceedances. While all coefficients and marginal effects for the yield curve are significant for the categories of one and two coexceedances in the upper tail coexceedances, many coefficients and marginal effects in the three coexceedances category are insignificant. In Thailand, the coefficients and marginal effects for shocks in the foreign stock market (the United States and Japan) and the US stock market are significant and positive in the lower tail coexceedances, whereas those for the US banking sector are insignificant. Most of the coefficients representing the foreign stock market index are insignificant for the upper tail coexceedances. The coefficients and marginal effects for the volatility of SET and for the yield curve are significant for Thailand. The sign of the coefficient for SET volatility is positive (negative) in the upper (lower) tail coexceedances. These findings suggest that declines in the foreign stock market index and the US stock market index did indeed affect the price of domestic bank stocks in Korea, Malaysia, and Thailand. The regression results also indicate that shocks to the US banking sector affected the domestic banking sector in Korea and Malaysia during the period January 2000
May 2007 (before the GFC). Results in the 1990s before the Asian Crisis Table 3 shows the results for the period January 1993
December 1996 in Korea, Malaysia, Singapore, and Thailand. The coefficients for shocks occurring in the foreign stock market (the United States and Japan), the US stock market (DJIA), and the US banking sector are not significant in Korea. In Malaysia, only a few coefficients for the foreign stock market shock (the United States and Japan) are significant and positive for the lower tail coexceedances. The estimated coefficients for the PHLX BKX are not significant. The coefficients for shocks occurring in the foreign stock market (the United States and Japan), the US stock market (DJIA), and the US banking sector are not significant in Singapore. In Thailand, the coefficients and marginal effects of the foreign stock market shock (the United States and Japan) are significant and positive for the lower tail category of two and three or more coexceedances. The coefficients for foreign stock market index are not significant for the upper tail coexceedances.

146

MASAHIRO INOGUCHI

In contrast, most of the coefficients and marginal effects for domestic stock markets are significant in Korea, Malaysia, Singapore, and Thailand. In addition, the sign of the coefficient for volatility in each country’s domestic index is positive (negative) for the upper (lower) tail coexceedances. These results suggest that declines in foreign stock market index affected the price of Thai domestic bank stocks in the 1990s. The regression results also indicate that the effects of foreign stock market shocks on the prices of domestic bank stocks were smaller during the pre-crisis period in the 1990s than in the pre-crisis period in the 2000s. Results in the 2000s during the GFC Table 4 shows the results for the period June 2007
October 2009 in Korea, Malaysia, Singapore, and Thailand. In Korea, the coefficients for the foreign stock market shock (DJIA and Nikkei 225), the US stock market shock (DJIA), and the US banking sector shock are significantly positive for the lower tail category of three or more coexceedances. In addition, the coefficients for the foreign stock market shock (the United States and Japan) and the US stock market shock are also significant for the upper tail category of three or more coexceedances. Most of the coefficients and marginal effects for the domestic stock markets are also significant, with the signs on the volatilities of the domestic index being positive (negative) for the upper (lower) tail coexceedances. The estimated coefficients for the yield curve are never significant for the category of one and two coexceedances, but some coefficients in the three or more coexceedances category are significant for the upper tail coexceedances. In Malaysia, some coefficients for the shocks in the foreign stock market (the United States and Japan) and the US stock market are significantly positive for the category of three or more coexceedances. Regarding the PHLX BKX, some of the estimated coefficients and marginal effects are significant and positive, but only for the lower tail coexceedances. The estimated coefficients and marginal effects for domestic stock market shocks are significant, and the sign of the coefficient for volatility in the KLCI is positive (negative) for the upper (lower) tail coexceedances. The results indicate that some of the coefficients representing foreign stock market shock (the United States and Japan), the US stock market shock, and the US banking sector shock are significantly positive for Singaporean bank stocks. The shock coefficients for the STI are also significant, while the sign of the coefficient for volatility is positive (negative) for the upper (lower) tail coexceedances. Although some of the yield curve coefficients are significant for the upper tail coexceedances, all of the

External Shocks on Stock Prices in East Asian Domestic Banking

147

estimated coefficients in the two coexceedances category are statistically insignificant. In Thailand, many of the coefficients for the foreign stock market shock (the United States and Japan), the US stock market shock, and the US banking sector shock are significant and positive for the three or more coexceedances category. Those in the category of one and two coexceedances are not significant. The coefficients and marginal effects for the SET are significant, and the sign of the coefficient for volatility in the SET is positive (negative) in the upper (lower) tail coexceedances. While the yield curve coefficients and marginal effects are significant in the upper tail category of one and two coexceedances, many coefficients in the remaining categories are not significant. Together, these results suggest that shocks in foreign stock market, the US stock market, and the US banking sector affected the prices of domestic bank stocks in Korea, Malaysia, Singapore, and Thailand during the GFC. However, there was no positive foreign stock market shock effect in Singapore, and the negative shocks in the US banking sector had no influence on Thailand in the 2000s before the GFC. In addition, our findings imply that both increases and decreases in the foreign stock market index (the United States and Japan) and the US stock market index affected the banking sector in East Asia during the crisis, while the negative shocks were mostly significant in pre-crisis Malaysia and Thailand in the 2000s. The results also suggest that changes in the yield curve affected domestic bank stocks less during the GFC than before.

CONCLUSION This chapter analyzes the impact of fluctuations in foreign stock market indices on the prices of domestic bank stocks to investigate the influence of external shocks on the banking sector in Korea, Malaysia, Singapore, and Thailand in the 2000s. Few previous studies have examined how the external shocks in the 2000s affected the banking sectors of East Asia. Because we can compare the magnitude of the impacts in different periods, our regression analysis focuses on three sample periods: the 1990s before the Asian financial crisis, the 2000s before the GFC, and the 2000s after the GFC. The dependent variable in the test is the number of bank coexceedances representing the shocks to the domestic banking sector. We arbitrarily defined large shocks as returns below the 5th or above the 95th percentile of

148

MASAHIRO INOGUCHI

the marginal return distribution of the daily percentage change in the bank stock price. We then estimated how the domestic stock market, foreign stock markets (Japan and the United States), and the domestic yield curve affected these coexceedances. Using this approach, we were able to quantify the influence of foreign shocks on the banking sector of each country. The regression result showed that the shocks in foreign stock markets impacted significantly on the prices of Asian bank stocks during the 2000s, more so than they did in the period before the Asian financial crisis. In addition, both increases and decreases in foreign stock markets were influential during the 2007
2009 GFC, whereas only price decreases were significant before the GFC, and then only in Malaysia and Thailand. The suggestion is that in Korea, Malaysia, Singapore, and Thailand, external shocks exerted a greater influence on the domestic banking sector in the 2000s than before 1997. While these countries have since consolidated their banking systems, increasing foreign capital flows may have had an influence. We conclude that if East Asian domestic banks had not improved in light of the 1997 crisis, the damage to their banking systems would have been considerably greater during the GFC. In addition, fluctuations in foreign stock markets had smaller effects on the bank sector in Singapore than on those in other Asian countries. It is possible that this is a result of differences in the capital flows between foreign and local banks for our sample countries. For example, foreign assets and debts of domestic banks declined in comparison with other sectors in Singapore, while the foreign assets and liabilities of domestic banks were larger in Korea and Singapore than in Malaysia and Thailand. The proportion of foreign assets and liabilities of domestic banks also increased in the 2000s in both Malaysia and Thailand. Lastly, in Korea, the level of foreign ownership of banks is relatively high. Our results also suggest that shocks from domestic stock markets also influenced the prices of Asian bank stocks.18 Regarding the volatility of domestic stock market indices, the prices of bank stocks increased when the volatility was high.19 In addition, changes in the yield curve affected the prices of domestic bank stocks less during the GFC than before.20

NOTES 1. For instance, some previous studies have described the fall in the number of nonperforming loans in the banking system and the growth in security markets in East Asian countries, and have thereby reported many notable improvements.

External Shocks on Stock Prices in East Asian Domestic Banking

149

2. Because the size of the external impact on banks does not necessarily imply the degree of soundness of a banking system, we do not directly discuss whether the various domestic banking sectors have become sounder. 3. Allen and Gale (2000) suggest that liquidity shocks causing a crisis in one region can spread through contagion, even with an incomplete interbank market. 4. Hartmann, Straetmans, and Vries (2005) assess risk in the European and US banking systems. While they maintain the relative importance of cross-border bank spillovers compared with domestic bank spillovers in Europe, they suggest that bank spillover risk in Europe generally appears to be lower than in the United States. 5. Gropp et al. (2009) employ distance-to-default as an indicator of bank soundness. Gropp and Moerman (2004) also use this variable and Monte Carlo simulation to show that conventional distributional assumptions cannot replicate the pattern of the tails in the data. 6. Those data are from the Bank for International Settlements (BIS) data for reporting banks. 7. That information includes only domestic commercial banks in Korea. 8. The BIS data include all types of banks in Korea. 9. We do not include the period during and after the Asian financial crisis in 1997
1999, because this particular shock originated in Asian countries and then affected the economies of other regions in the world. 10. The subprime crisis commenced in mid-2007 when two hedge funds managed by US investment bank Bear Stearns reported heavy losses arising from defaulting subprime loans. A number of other hedge funds exposed to subprime loans went bankrupt. In July 2007, ratings agencies Moody’s and Standard & Poor’s dramatically reduced the ratings on securitized bonds backed by subprime mortgages. 11. We do not use exchange rates as an explanatory variable in the regressions because the prevailing exchange rate regime before the 1997 crisis in most Asian countries was a de facto US dollar-pegged system: Malaysia operated a dollarpegged exchange rate system from 1998 to 2005. 12. Gropp et al. (2009) and Bae et al. (2003) found this variable to be important when explaining emerging market coexceedances. 13. We employ a GARCH (1,1) model. 14. When Asian markets are open (closed), US markets are closed (open). Because the shocks that occur in the US market could influence the Asian markets the following day, we lag the US variables by one day. 15. This implies that the effect on the dependent variables varies in size when changing from one to two independent variables compared with changing from two to three independent variables. While the ordered logit model restricts these marginal effects so that they remain the same, the multinomial logit model allows complete flexibility. 16. In the regression for the category of three coexceedances in Singapore, there is the possibility of a small-sample bias in the results. 17. We also estimate the model using ordered logit to check for robustness. The test using ordered logit shows that some coefficients for foreign stock markets are significant in the 2000s, even though the same coefficients are not significant in the multinomial logit regression. Other than this, there are not very many differences between the results of multinomial logit and ordered logit.

150

MASAHIRO INOGUCHI

18. We consider only the direct channel for external shocks from foreign stock markets to the domestic banking sector, and not the influence of foreign stock markets on the domestic stock market as a whole. Accordingly, our analytical approach may underestimate the impact of external shocks on the domestic banking sector. However, examination of this path would not be easy, as it is difficult to extract the foreign stock market shocks on domestic stock markets from the many other sources of domestic and external shocks. 19. We are unable to comment directly on this, as the regression does not analyze the relationship between volatility and stock returns. 20. We hypothesize that one reason for this is that the monetary policy changes had less influence during the crisis given the presence of large external shocks.

ACKNOWLEDGMENTS Masahiro Inoguchi would like to thank Kaliappa Kalirajan, Jenny Corbett, Hal Hill, Takatoshi Ito, Masahiro Kawai, Ryuzo Miyao, Sanae Ohno, and Takeshi Hoshikawa for useful suggestions and discussion. Thanks also to attendees at the 2009 Modern Monetary Economics Conference held in Kobe, the 26th International Conference of the American Committee for Asian Economic Studies held in Kyoto in 2010, the June 2010 Japanese Economic Association Meeting, the Australia
Japan Research Centre 30th Anniversary Conference held at the ANU in March 2011, and the 40th Australian Conference of Economists in 2011 for their helpful comments. All errors in this chapter are solely mine. This research was supported in part by the Japanese MEXT KAKENHI Grant Numbers: 21730267 and 25380413.

REFERENCES Allen, F., & Gale, D. (2000). Financial contagion. Journal of Political Economy, 108, 1
33. Bae, K.-H., Karolyi, G. A., & Stulz, R. M. (2003). A new approach to measuring financial contagion. Review of Financial Studies, 16, 717
763. Cifuentes, R., Ferrucci, G., & Shin, H. S. (2005). Liquidity risk and contagion. Journal of the European Economic Association, 3, 556
566. Eichengreen, B., Rose, A. K., & Wyplosz, C. (1996). Contagious currency crises: First tests. The Scandinavian Journal of Economics, 98, 463
484. Forbes, K. J., & Warnock, F. E. (2012). Capital flow waves: Surges, stops, flight, and retrenchment. Journal of International Economics, 88, 235
251. Freixas, X., Parigi, B. M., & Rochet, J.-C. (2000). Systemic risk, interbank relations and liquidity provision by the central bank. Journal of International Money and Finance, 32, 611
640.

External Shocks on Stock Prices in East Asian Domestic Banking

151

Gropp, R., Lo Duca, M., & Vesala, J. (2009). Cross-border bank contagion in Europe. International Journal of Central Banking, 5, 97
139. Gropp, R., & Moerman, G. (2004). Measurement of contagion in bank equity prices. Journal of International Money and Finance, 23, 405
459. Hartmann, P., Straetmans, S., & Vries, C. G. D. (2005). Banking system stability: A crossAtlantic perspective. NBER Working Paper No. 11698. Sachs, J., Tornell, A., & Velasco, A. (1996). Financial crises in emerging markets: The lessons from 1995. Brookings Papers on Economic Activity, 1, 147
198.

MEASURING FINANCIAL INTEGRATION: EVIDENCE FROM TEN INDUSTRIES IN A “US-EMERGING WORLD” Michael Donadelli ABSTRACT This chapter measures financial integration in 10 industries over 4 different periods. We use two robust measures of integration: (i) the Pukthuanthong and Roll (2009)’s multi-factor R-square and (ii) the Volosovych (2011)’s integration index. Both measures, based on PCA, indicate that the difference between the level of integration over the period 2009
2012 (“Post-Lehman” era) and the level of integration over the period 1994
1998 (“Post-Liberalizations” era) is relatively high. In addition, the level of financial integration across international equity markets decreased during the late 1990s. This suggests that de jure integration does not necessarily improve de facto integration. Overall, our findings give rise to a “diversification benefits-insurance benefits trade-off.” Keywords: Financial integration; industries; R-square; integration index; PCA JEL classifications: F3; F4; G1

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 153
178 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096005

153

154

MICHAEL DONADELLI

INTRODUCTION Changes in the degree of integration across international equity markets affect cross-country diversification benefits as well as households consumption-smoothing motive. On the one side, higher levels of financial integration tend to decrease international portfolio diversification benefits. In other words, highly integrated international financial markets induce strong positive cross-country equity return correlations, and, therefore, lower diversification benefits (Donadelli, 2013; Goetzmann, Li & Rouwenhorst, 2005; Kearney & Lucey, 2004, among others). On the other side, a higher level of integration across international equity markets improves risk-sharing. In other words, highly integrated financial markets allow for larger insurance benefits, and therefore, improve households consumption smoothing (Colacito & Croce, 2013; Jappelli & Pistaferri, 2011; Suzuki, 2014, among others). In addition, financial integration provides short- and long-run welfare benefits (Colacito & Croce, 2010). For all these reasons, the evolution of the global integration process has received an enormous amount of attention in the literature, much of it devoted to assessing a range of possible integration measures.1 However, the debate on the proper measure of integration is still open. In fact, the literature provides a large number of integration measures. Traditional proxies to measure integration include barriers to international investments (e.g., legal restrictions), price-based measures, quantity-based measures, cointegration- and correlation-based measures. Price-based measures are based on the interest parity or purchasing parity conditions. The literature refers to this as “direct measures” in that they invoke the law of one price, that is, assets with identical cash flows should command the same return. As known, the law of one price holds only in equilibrium. Therefore, it does not specify the process toward the equilibrium. It turns out that it cannot provide a full description of the integration process (Lewis, 1999; Tesar & Werner, 1995). Quantity-based measures rely on the stocks of external assets and liabilities and the international capital flows’ volume (i.e., they rely on the concept of international capital market completeness). The literature refers to this as “indirect measures.”2 Cointegration measures try to capture the degree of integration across markets by means of short- and long-run linkages (Aggarwal, Lucey, & Muckley, 2004; Arshanapalli & Doukas, 1993; Chan, Gup & Pan, 1992, 1997; Gallagher, 1995; Gilmore & McManus, 2002; Hatemi-J, 2012; Kasa, 1992; Kenourgios & Samitas, 2011; Manning, 2002; Voronkova, 2004, among others). Correlation-based measures examine international equity markets integration from the perspective

Measuring Financial Integration

155

of changes in the level of co-movements between their returns over time (Bekaert, Hodrick, & Zhang, 2009; Chambet & Gibson, 2008; Kuper & Lestano, 2007; Quinn & Voth, 2008; Yang, 2005; Yu, Fung, & Tam, 2010, among many others).3 Other measures instead rely on the time-varying nature of equity risk premia (Bekaert & Harvey, 1995; de Jong & de Roon, 2005; Donadelli & Prosperi, 2012; Panchenko & Wu, 2009). Nevertheless, the degree of robustness of all these measures has recently been questioned (Bekaert et al., 2009; Carrieri, Errunza, & Hogan, 2007; Pukthuanthong & Roll, 2009; Volosovych, 2011; Yu et al., 2010). It is also worth noting that there is no a general consensus on whether advanced and emerging equity markets are fully integrated as well as on the shape of the financial integration process (Bekaert, Harvey, Lundblad, & Siegel, 2011). In other words, the empirical evidence on financial integration is mixed. We stress that this is motivated by several factors. First, some empirical studies have been conducted in a static context. This does not allow for a full understanding of the evolution of the level of integration across international stock markets. Second, most of the existing empirical studies have employed pre-2005 data. However, international stock markets have been heavily influenced by the emerging systemic banking crisis of the late 1990s and early 2000s. Third, other studies have focused only on the dynamics of the financial integration process across countries belonging to the same region. Fourth, there is a high degree of heterogeneity (across existing studies) in the set of countries employed to examine both global and regional financial integration. This might produce different financial integration patterns. Last, as mentioned above, existing works have employed different approaches to measure integration. This chapter improves the existing literature in four main directions. First, we take a step back from the more traditional measures. To this end, we employ two newly introduced robust integration measures: (i) the Pukthuanthong and Roll (2009)’s alternative measure and (ii) the Volosovych (2011)’s integration index. Both measures rely on the principal component analysis (PCA). The first integration measure is represented by the adjusted R-square obtained from a multi-(artificial) factor model.4 In the spirit of Pukthuanthong & Roll (2009), as global risk factors, we use the first ten principal components extracted from a set of 39 regional industry portfolio excess returns. The second measure is represented by the proportion of total variation in individual excess returns explained by the first principal component. Second, in contrast to existing empirical studies, we do not focus on the level of financial integration across equity markets belonging to the same

156

MICHAEL DONADELLI

regions (Claus & Lucey, 2012; Hatemi-J, 2012; Kenourgios & Samitas, 2011; Kim, Kim, & Wang, 2006; Phylaktis & Ravazzolo, 2002; Schotman & Zalewska, 2006; Volosovych, 2011; Voronkova, 2004; Yu et al., 2010; among others). Our sample includes 20 emerging equity markets and, as benchmark, the equity market of the United States. Third, to capture the dynamics of the financial integration process, both integration measures are computed in four “ad-hoc” periods: (I) January 1994
December 1998, namely, the “Post-liberalizations” era; (II) January 1999
December 2003, namely, the “Post-Crises” era: (III) January 2004
December 2008, namely, the “Rising Rates” era; and (IV) January 2009
July 2012, namely, the “Post-Lehman” era. Fourth, differently from previous studies, which have mainly focused on national equity markets (i.e., country equity indices), in this chapter the national market (in each region or country) is divided in 10 different industries. Therefore, financial integration is measured in 10 different industrial equity markets: Basic Materials, Consumer Goods, Consumer Services, Financials, Healthcare, Industrials, Oil & Gas, Technology, Telecommunications, and Utilities. The choice of using industry level data is motivated by several factors: (i) it allows to capture shocks in specific industries (e.g., IT bubble); (ii) it allows to examine whether there is heterogeneity in the integration dynamics across industries, thus, to exploit crossindustry diversification benefits; (iii) it reflects standard financial industry’s investment strategies focusing on sector rather than country equity indices (e.g., a private/institutional investor might be interested in investing only in stocks belonging to specific sectors). To the best of our knowledge, there is in the literature only one other study by Donadelli & Persha (2014) that covers such an extensive range of emerging markets as well as industry equity market indices. In addition, this is the first study that measures the level of integration in 10 different industrial equity markets across countries via two newly introduced robust integration measures: (i) the Pukthuanthong and Roll (2009)’s multi-factor R-square and (ii) the Volosovych (2011)’s index of integration. The main results of this chapter can be summarized as follows. First, our empirical findings suggest that the level of integration in the aftermath of the Lehman Brothers’ collapse (i.e., 2009
2012) is higher than the level of integration in the aftermath of equity market liberalizations (i.e., 1994
1998). This result holds across industries and suggests that de jure integration does not necessarily improve de facto integration (see also Claus & Lucey, 2012; Donadelli, 2013). Second, we observe that, in most

Measuring Financial Integration

157

industries, financial integration slows down between the first and second eras. We argue that this drop has been mainly caused by the presence of systemic banking crisis across emerging economies in the late 1990s. We stress that both measures produce similar “industry-by-industry” integration patterns, that is, “Volosovych (2011) meets Pukthuanthong and Roll (2009)”. Third, we observe that financial integration grows faster as financial and trade openness grow faster.5 We conclude by arguing that the empirical findings of this chapter give rise to a “diversification benefitsinsurance benefits trade-off.” On the one side, the increasing degree of integration in all industries across international economies reduces both cross-country and cross-industry diversification benefits. On the other side, the higher level of integration produces a more efficient international risk-sharing environment, that is, improves consumption smoothing (i.e., insurance benefits against bad times). In addition, the relatively high level of financial integration observed in the last two periods suggest that a financial autarky regime or a one-traded bond world embodied in standard international business cycle models might represent an unrealistic international capital markets structure.6 The rest of the chapter is organized as follows. The section “Data Description and Summary Statistics” describes the data. The section “On the Financial Integration Measures” presents the employed methodology. The section “Results” examines the evolution of the financial integration process. The section 5 “Concluding Remarks” concludes.

DATA DESCRIPTION AND SUMMARY STATISTICS Industry Equity Indices This study employs monthly industry equity indices for 20 emerging equity markets, namely, Argentina, Brazil, Chile, China, Colombia, Czech Republic, India, Israel, Hungary, Mexico, Malaysia, Pakistan, Peru, Philippines, Poland, Russia, Sri Lanka, Taiwan, Thailand, and Turkey. These markets are classified as emerging because of their low- or middle-income and low investable market capitalization/GDP ratio status (see International Finance Corporation [IFC], 1999). As representative developed market, we use monthly data for the United States. The US stock market is included here because it represents the largest stock market in the world, it is a

158

MICHAEL DONADELLI

leading indicator of international equity market returns’ movements, and it may be expected to have strong effects on emerging equity markets (Donadelli & Persha, 2014; Donadelli & Prosperi, 2012; Graham, Kiviaho, & Nikkinen, 2012; Hatemi-J, 2012; Narayan & Narayan, 2012). We employ monthly data instead of weekly or daily data to avoid a set of common high-frequency data issues: (i) presence of zero returns, (ii) noise, and (iii) non-synchronicity. Our sample goes from December 1993 to July 2012. All industry equity market indices are obtained from level 2 of Datastream Global Equity Indices (DGEI) database. In this dataset, stock data are classified by industry and sector type. For example, financials is an industry within which a number of sectors are included such as banks, life assurance, and real estate. Level 2 of DGEI divides the market into the following 10 industries: Basic Materials, Consumer Goods, Consumer Services, Industries, Health Care, Financials, Oil & Gas, Technology, Telecommunications, and Utilities.7 To get a homogeneous dataset, all indices are total return indices (TRIs) denominated in US dollars. Equity indices expressed in this form include reinvested dividends, retain only US inflation (i.e., no currency risk), and are widely used in the international finance literature (Bilson, Brailsford, & Hooper, 2001; Chambet & Gibson, 2008; de Jong & de Roon, 2005; Donadelli & Persha, 2014; Donadelli & Prosperi, 2012; Ferson & Harvey, 1994; Grootveld & Salomons, 2003; Harvey, 1995; Lee, Chen, & Chang, 2013; Pukthuanthong & Roll, 2009; Yu et al., 2010, among many others).

Regional Industry Portfolios and Excess Returns In line with standard asset management strategies, we focus on three regional equally weighted industry equity indices, namely, Asia, Eastern Europe and Middle East, and Latin America, and on the US industry equity indices. Regional industry portfolios are constructed according to the geographic distribution reported in Table 1. Formally, RIEIiR;t =

XN n=1

wn DGEIin;t

ð1Þ

where DGEIin;t is the DGEI of industry i in country n at time t, wn = 1/N denote weights, and N represents the total number of countries in each regions R (for the United States, N = 1). Notice that data for some

159

Measuring Financial Integration

Table 1. Asia China India Malaysia Pakistan Philippines Sri Lanka Taiwan Thailand

Geographic Distribution of the Countries in the Sample. EU + ME

Latin America

Czech Republic Hungary Israel Poland Russia Turkey

Argentina Brazil Chile Colombia Mexico Peru

Advanced United States

industries are not available since December 1993. As industry data become available a specific country n is added to the portfolio. Regional industry equity excess returns are computed as follows,

ExRiR;t

=

RIEIiR;t RIEIiR;t-1

! − 1 − Rf ;t

ð2Þ

where Rf,t is the one-month Treasury-bill rate. Our aggregation strategy gives rise to 39 excess returns.8 These regional industry equity excess returns represent the most important elements of our analysis. The analysis is based on four different sub-periods: (I) January 1994
December 1998, namely, the “Post-liberalizations” era; (II) January 1999
December 2003, namely, the “Post-Crises” era; (III) January 2004
December 2008, namely, the “Rising Rates” era; and (IV) January 2009
July 2012, namely, the “Post-Lehman” era.9 The choice of these four specific sub-periods is motivated by two main factors. First, each sub-period includes emerging or US financial shocks that have heavily affected international equity markets. In the spirit of Bloom (2009), we rely on uncertainty shocks. Second, shocks are heterogeneous across periods. While the first two sub-periods have been mainly characterized by financial shocks in emerging countries, sub-periods III and IV are mainly driven by US/EU financial shocks. The postliberalizations sample is aimed at capturing the degree of financial integration in industrial equity markets across countries in the aftermath of the emerging equity market liberalizations of the late 1980s and early 1990s.10 The second era allows to measure integration in the aftermath of the systemic banking crisis in Indonesia, Malaysia, Philippines, and Thailand.

160

MICHAEL DONADELLI

The third sub-period allows to measure integration in a period characterized by a sharp increase in international trade, financial linkages and fed funds rate levels.11 The last period is employed to capture financial integration in the aftermath of the Lehman Brothers’ collapse, an era characterized by high economic policy uncertainty both in the United States and Euro Area (see Baker, Bloom, & Davis, 2013; Donadelli & Persha, 2014). Uncertainty shock dates are reported in Fig. 1.12 Summary statistics, computed over the full period and over the four sub-periods, suggest that: (i) emerging equity markets tend to deliver higher average excess returns than the US equity market; (ii) the performance of emerging industrial equity markets have been heavily affected by the systemic banking crisis of the late 1990’s; (iii) the subprime crisis has mainly affected the performance of the US equity market (see also Donadelli & Persha, 2014).13

Fig. 1. Sub-Periods and Uncertainty Shocks. Key events (date): Trade world concerns (Jan95); financial deregulation in China and state-owned enterprises reforms (Apr95); tax rebates and tariff policy in China (Sep95); systemic banking crisis in Indonesia, Malaysia, Philippines, and Thailand (Jan98); Russian financial crisis (Aug98); LTCM default (Sep98); financial crisis in Turkey (Nov00); US terrorist attacks (Sep01); China is a new WTO member (Dec01); debt crisis in Argentina (Dec01); US accounting scandals (Jul02); II Gulf War (Feb03); German federal election (Sep05); Northern Rock financial support stimulus (Aug07); stimulus debate and large interest rate cuts (Dec07); Lehman Brothers Chapter 11 and Troubled Asset Relief Program (Sep08); EU recovery plan for growth and jobs (Nov08); EU Sovereign Debt Crisis (2009:IVQ); Greek government requested an initial loan of h45 billion from the EU and IMF (Apr10); Standard and Poor’s downgraded Italian debt from A + to A (Aug11); China slowdown fears and disorderly political transition (May12).

Measuring Financial Integration

161

ON THE FINANCIAL INTEGRATION MEASURES Correlation-Based versus PCA-Based Measures: A Review As discussed in the introduction of this chapter, most studies have examined integration across international equity markets by means of correlation-based or cointegration-based measures.14 While it is popularly agreed that correlation is one of the most important element of the meanvariance optimization process, the debate on whether correlation might be used as proxy to measure integration is still open. Bekaert et al. (2009) confirm that correlations are important ingredients in the analysis of international diversification benefits and global market integration. However, they argue that correlations neither measure international diversification benefits nor global market integration. Other international finance works raise “robustness issues.” Wilcox (2005) and Huber and Ronchetti (2009) argue that the sample correlation is not a robust statistics in the presence of outliers or a heavy-tailed distribution. Other studies suggest that the presence of conditional heteroskedasticity of market returns as well as the hypothesis that cross-country market return correlations depend on market volatility might lead to biased conclusions about integration (Boyer, Gibson, & Loretan, 1999; Forbes & Rigobon, 2002; Longin & Solnik, 2001). Volosovych (2011) points out that a high correlation of economic or financial series cannot be used as evidence of substantial integration. His measure, based on PCA, accounts for both country-specific and global shocks as well as is immune to outliers. Pukthuanthong and Roll (2009) argue that the cross-country correlation of equity index returns do not represent a robust measure of integration. In particular, they show that two countries can be highly integrated even if their equity market returns are negatively correlated. This occurs because of countries’ sensitivities to common global factors are different. In other words, perfect integration implies that a set of common factors explains 100% of the broad index returns in both countries, but if country indices differ in their sensitivities to these factors, they do not exhibit perfect correlation. Similarly, Yu et al. (2010) argue that the concept of integration is based on whether the markets are affected by common factors rather than the price convergence.

PCA: The “Multi-Factor R-Square” and the “1st Component” The PCA is a non-parametric empirical strategy used to reduce the original dimension of a set of variables. Its ultimate goal is to capture common

162

MICHAEL DONADELLI

features across variables variations. In practice, the PCA transforms an original set of variables into a new subset of variables. These transformed variables, namely principal components, are represented by linear combinations of the original set of variables, and are aimed at capturing a large part of the variation in the original set of variables. Formally, let X be a vector of p variables x variables, x1 ; x2 ; …; þ xp . Then, a linear combination of these variables can be represented as follows: Z = Ω⋅X

ð3Þ

where the first row in Eq. (3) takes the form z1 = w11 x1 þ w12 x2 þ ⋯ þ w1p xp , and Ω is the loading-matrix. Weights in Ω are obtained in a way to guarantee the maximum sample variance of z1. The first principal component, z1, is represented by a linear function that has the maximum possible variance. The second principal component, z2, is the linear function with maximum possible variance subject to being uncorrelated with the first principal components, and the third principal component, z3, is the linear function with maximum possible variance subject to being uncorrelated with the first and the second principal components, and so on. Principal components can be extracted by using either the covariance matrix or the correlation matrix. It is standard practice to use the covariance matrix when the variable scales are similar and the correlation matrix when variables are expressed in different scales. By using the correlation matrix, data are standardized. Theoretically, if all the employed series are expressed in the same scale (e.g., equity asset returns), then a correlation-based PCA might throw out a relevant amount of information. However, if the covariance matrix is used, the variables with the highest variance tend to dominate the first principal component. To overcome this issue, we employ the correlation matrix.15 The “Multi-Factor R-Square” The adjusted R-square measures how well equity market excess returns can be explained by common factors. In other words, global integration relies on whether the equity markets are affected by common risk factors rather than the equity price index convergence. In this chapter, global risk factors are represented by artificial risk factors. In the spirit of Pukthuanthong & Roll (2009), the first 10 principal components, which generally capture 90% of excess returns’ variation, are our artificial risk factors.16 In this exercise, principal components are extracted from the dataset composed by

163

Measuring Financial Integration

39 regional portfolio industry excess returns (see “Data Description and Summary Statistics”). To capture changes in the level of integration in all industrial equity markets, the analysis is carried out in period I, II, III, and IV. Therefore, as in Pukthuanthong & Roll (2009),17 principal components are extracted in different sub-periods.18 Details on the proportion of total variation in individual excess returns explained by each principal component in each sub-period are presented in Table 2. The adjusted R-square, Table 2. PCA: Sub-Period Results. Number

Value

Difference

Prop

Cum Value

Cum Prop

Panel A: PCA (Sample: 1994M01-1998M12) 1 15.78 11.13 2 4.65 1.02 3 3.64 0.19 4 3.45 2.05 5 1.39 0.15 6 1.24 0.26 7 0.99 0.09 8 0.89 0.02 9 0.87 0.24 10 0.63 0.04

0.40 0.12 0.09 0.09 0.04 0.03 0.03 0.02 0.02 0.02

15.78 20.44 24.07 27.52 28.91 30.16 31.14 32.04 32.91 33.54

0.40 0.52 0.62 0.71 0.74 0.77 0.80 0.82 0.84 0.86

Panel B: PCA (Sample: 1999M01-2003M12) 1 16.63 12.46 2 4.17 0.70 3 3.47 0.62 4 2.85 1.36 5 1.49 0.37 6 1.12 0.07 7 1.05 0.09 8 0.96 0.21 9 0.75 0.06 10 0.68 0.06

0.43 0.11 0.09 0.07 0.04 0.03 0.03 0.02 0.02 0.02

16.63 20.80 24.27 27.12 28.62 29.74 30.78 31.74 32.49 33.17

0.43 0.53 0.62 0.70 0.73 0.76 0.79 0.81 0.83 0.85

Panel C: PCA (Sample: 2004M01-2008M12) 1 27.86 25.42 2 2.44 0.97 3 1.47 0.50 4 0.97 0.16 5 0.81 0.09 6 0.72 0.18 7 0.54 0.12 8 0.42 0.01 9 0.40 0.05 10 0.36 0.02

0.71 0.06 0.04 0.02 0.02 0.02 0.01 0.01 0.01 0.01

27.86 30.31 31.78 32.74 33.55 34.27 34.81 35.23 35.63 35.99

0.71 0.78 0.81 0.84 0.86 0.88 0.89 0.90 0.91 0.92

164

MICHAEL DONADELLI

Table 2. Number

Value

Difference

Panel D: PCA (Sample: 2009M01-2012M07) 1 27.17 24.55 2 2.61 0.90 3 1.71 0.55 4 1.16 0.28 5 0.88 0.24 6 0.64 0.08 7 0.56 0.01 8 0.55 0.11 9 0.44 0.02 10 0.42 0.05

(Continued ) Prop

Cum Value

Cum Prop

0.70 0.07 0.04 0.03 0.02 0.02 0.01 0.01 0.01 0.01

27.17 29.78 31.49 32.65 33.53 34.17 34.73 35.28 35.72 36.14

0.70 0.76 0.81 0.84 0.86 0.88 0.89 0.90 0.92 0.93

Notes: This table reports the results of the PCA. PCA results are reported for each sub-period: (I) January 1994
December 1998 (Panel A); (II) January 1999
December 2003 (Panel B); (III) January 2004
December 2008 (Panel C); and (IV) January 2009
July 2012 (Panel D).

which serves as measure of integration, is obtained from standard OLS estimations. According to our aggregation strategy, the adjusted R-square is estimated for each industry (i) in each region. Formally, i;I I I I I I I I I I I ExRi;R;tI =αi;I R;t þC1;t þC2;t þC3;t þC4;t þC5;t þC6;t þC7;t þC8;t þC9;t þC10;t þɛ R;t i;II II II II II II II II II II II ExRi;R;tII =αi;II R;t þC1;t þC2;t þC3;t þC4;t þC5;t þC6;t þC7;t þC8;t þC9;t þC10;t þɛ R;t i;III III III III III III III III III III III ExRi;R;tIII =αi;III R;t þC1;t þC2;t þC3;t þC4;t þC5;t þC6;t þC7;t þC8;t þC9;t þC10;t þɛ R;t i;IV IV IV IV IV IV IV IV IV IV IV ExRi;R;tIV =αi;IV R;t þC1;t þC2;t þC3;t þC4;t þC5;t þC6;t þC7;t þC8;t þC9;t þC10;t þɛ R;t

ð4Þ

where ExRi;j R;t is the regional industry portfolio excess return in period j, j j are constants, C1;t ; …; C10;t ; are the first ten principal components extracted in each sub-period j, and j = I, I, III, IV.

αi;j R;t

The “1st Principal Component” As discussed in Volosovych (2011), in a PCA-based analysis, the “1st principal component” captures most of the variation of the original data. Therefore, if international equity markets are highly integrated, the

165

Measuring Financial Integration

proportion of total variation in individual excess returns explained by the first component should be close to one. To be consistent with the previous analysis, we first compute the integration index by using all the 39 regional portfolio industry excess returns. This gives rise to a “global measure of integration.” We then extract the principal component at the industry level, that is, we compute the level of integration in each industry (in each period) in a “1st principal component” context. In order to have more than four series per industry (i.e., the four regional industry portfolio excess returns), we use data at the country level. To have a homogeneous dataset, we select only those countries for which industry data are available since December 1993.19 As for the adjusted R-square, the “1st principal component” is extracted in period I, II, III, and IV.20

RESULTS Evidence from the “Multi-Factor R-Square” Fig. 2 reports the evolution of financial market integration in each industry across regions. The integration index is captured by the adjusted R-square of a multi-(artificial) factor regression. As discussed in the section “On the Financial Integration Measures,” the 10 artificial global risk factors are represented by the first 10 principal components extracted from the set of variables composed by our 39 regional portfolio industry excess returns. The analysis is conducted for each period and for each region. The average level of integration (in each industry) is then represented by the average adjusted R-square, that is, in each sub-period the adjusted R-square is averaged across the four regions (Asia, Eastern Europe + Middle East, Latin America, United States). Results suggest that the average percentage of variation in regional industry excess returns explained by the first 10 principal components (i.e., R-square) in periods III and IV is higher than the percentage explained in the period I (on average, 0.9 vs. 0.8). In other words, the level of financial integration across regions in each industry sharply increased during the last 5
10 years. At the country level, Pukthuanthong & Roll (2009), Yu et al. (2010), and Donadelli (2013) obtain similar results. It is also worth noting that, in most industries, the level of financial integration in the “Post-Liberalizations” era (i.e., first sub-period) is higher than in the “Post-Crises” era (i.e., second sub-period). Exceptions are the technology and telecommunications industries.21 We argue that the dynamics

0.92

Basic Materials

0.90 0.88 0.86 0.84 0.82

Consumer Goods

0.90

0.92

Mean Adjusted R-square

Mean Adjusted R-square

0.94

0.88 0.86 0.84 0.82 0.80

1994-1998

0.93

1999-2003

2004-2008

0.78

2009-2012

1994-1998

0.940

Consumer Services

1999-2003

2004-2008

2009-2012

2004-2008

2009-2012

2004-2008

2009-2012

Financials

0.91

Mean Adjusted R-square

Mean Adjusted R-square

0.92

0.90 0.89 0.88 0.87 0.86 0.85 0.84 0.83

1994-1998

2004-2008

0.880 0.860 0.840 0.820 0.800 1999-2003

2004-2008

Industrials

0.920

0.880 0.860 0.840

1994-1998

1999-2003 Technology

Oil & Gas 0.880 Mean Adjusted R-square

Mean Adjusted R-square

1999-2003

0.900

0.820

2009-2012

0.900 0.890 0.880 0.870 0.860 0.850 0.840

0.860 0.840 0.820 0.800 0.780

0.830 1994-1998

0.940

1999-2003

2004-2008

0.760 2009-2012

1994-1998

0.950

Telecommunications Mean Adjusted R-square

0.920 Mean Adjusted R-square

0.840

0.900

0.910

0.820

0.860

1994-1998

Healthcare

1994-1998

0.880

0.940

0.900

0.780

0.900

0.820

2009-2012

Mean Adjusted R-square

Mean Adjusted R-square

0.920

1999-2003

0.920

0.900 0.880 0.860 0.840 0.820

1999-2003

2004-2008

2009-2012

2004-2008

2009-2012

Utilities

0.900

0.850

0.800

0.750

0.800 0.780

1994-1998

1999-2003

2004-2008

2009-2012

0.700 1994-1998

1999-2003

Fig. 2. Indicator of Global Market Integration by Industry: The “Multi-Factor R-Square.” Notes: This figure reports the adjusted R-square estimated for each industry in each individual region, and then averaged across regions (Asia, Eastern Europe and Middle East, Latin America, and the United States). The adjusted R-square from a regression of industry index excess returns on global risk factors captures financial market integration. Global risk factors (in each period) are represented by the first 10 principal components. Principal components are extracted as described in the section “Evidence from the “Multi-Factor R-Square.”” Principal components are extracted using monthly data over historic sub-periods. Sub-periods are defined as in Fig. 1. Regional industry equally weighted portfolios are constructed as defined in Eq. (1). Adjusted R-squares are obtained via standard OLS estimations. Constant is included. Standard errors are Newey & West (1987, 1994). The full sample goes from January 1994 until July 2012.

167

Measuring Financial Integration

of financial market integration between the first and second sub-period has been mainly driven by the presence of systemic banking crisis across emerging economies in the late 1990s (see Fig. 1). It turns out that financial market integration tends to be stronger during recession periods.

Evidence from the “1st Principal Component”: Volosovych (2011) Meets Pukthuanthong and Roll (2009) Fig. 3 reports the percentage of variance explained by the first principal components across the 39 regional industry excess returns. In this exercise, the first principal component represents the first global risk factors in Eq. (4). Therefore, values reported in Fig. 3 correspond to entries in Table 2 (see first line (column 4) in Panels A
D). Not surprisingly, we find that the level of global integration in the “Rising Rates” and “PostLehman” eras is higher than in the “Post-Liberalizations” and “PostCrises” eras. It is also worth noting that financial market integration raised by 30% between the second and third sub-periods. 0.8 71.4%

69.7%

2004-2008

2009-2012

Global Integration Index

0.7 0.6 0.5 0.4

40.5%

42.6%

0.3 0.2 0.1 0 1994-1998

1999-2003

Fig. 3. Global Financial Integration Index: The “1st Component.” Notes: This figure reports the proportion of total variation in individual excess returns explained by the first principal component. The principal component, which corresponds to the first global risk factors in Eq. (4), is extracted from the set of data composed by the 39 regional industry portfolio excess returns. Principal components are extracted using monthly data over historic sub-periods.

168

MICHAEL DONADELLI

Fig. 4 reports the evolution of the financial integration index in the 10 industrial equity markets. As in Volosovych (2011), market integration is captured by the proportion of variation in individual excess returns explained by the first principal component. As discussed in the section “PCA: The “Multi-Factor R-Square” and the “1st Component,”” to have a larger set of variables for each industry, we use industry equity indices at the country rather than regional level. As for the adjusted R-square, the percentage of variation explained by the first principal component is estimated in each sub-period. Overall, we find that the adjusted R-square and the proportion of variation explained by the first principal component follow similar patterns. It turns out that the common principal component approach of Pukthuanthong and Roll (2009) and the first principal component approach of Volosovych (2011) give rise to similar financial integration patterns. We find differences only in their order of magnitude. However, if the adjusted R-squared is obtained from a multi-factor regression with only three or four global risk factors, then the two measures (i.e., the two percentages) do not display only similar patterns, but also similar values. In this case the adjusted R-square is significantly lower (see Pukthuanthong & Roll, 2009).22

Market Openness versus Financial Integration The evidence provided so far shows that the last two eras are characterized by a higher proportion of equity market returns’ variation explained by common global risk factors (i.e., higher adjusted R-square) as well as by a larger proportion of variation attributed to a single important factor than ever before. Results suggest also that financial market integration sharply increased over the period 2002
2008. This can be informally observed also by looking at the evolution of the mean adjusted R-square (see right-hand side of Fig. 4) reported in Pukthuanthong & Roll (2009). Using country equity market indices, Donadelli (2013) finds a similar result. Both the international finance and international business cycle literature have shown that much of the increase in the level of global market integration might be attributed to an increase in the degree of trade and financial openness (Colacito & Croce, 2013; Donadelli, 2013; Imbs, 2006; Pretorius, 2002, among others). Fig. 5, which plots the evolution of the total value of WORLD stocks traded (black line) and international trade of goods and services (gray line) over the period 1995
2012, confirms these findings. In particular, it shows that the 2002
2008 period has been characterized by a

169

Measuring Financial Integration 0.6

Financial Integration Index

Financial Integration Index

0.5 0.45 0.4 0.35 0.3 0.25 0.2

0.35 0.3 0.25 0.2 1994-1998

1994-1998

1999-2003

2004-2008

2009-2012

2004-2008

2009-2012

2004-2008

2009-2012

2004-2008

2009-2012

2004-2008

2009-2012

Financials

0.5 0.45 0.4 0.35 0.3 0.25 0.2

2009-2012

1994-1998

0.55

1999-2003 Industrials

0.5 Financial Integration Index

Financial Integration Index

2004-2008

Healthcare

0.4 0.35 0.3 0.25

1994-1998

0.5

1999-2003

2004-2008

0.45 0.4 0.35 0.3 0.25 0.2

2009-2012

1994-1998

0.7

Oil & Gas

1999-2003 Technology

0.65

0.45 Financial Integration Index

Financial Integration Index

1999-2003

0.45

0.4 0.35 0.3 0.25 0.2

0.6 0.55 0.5 0.45 0.4 0.35 0.3 0.25

1994-1998

0.55

1999-2003

2004-2008

0.2

2009-2012

Financial Integration Index

0.5

0.4 0.35 0.3 0.25 1994-1998

1999-2003

2004-2008

1994-1998

0.5

Telecommunications

0.45

0.2

0.2

0.55

0.4

0.15

0.3 0.25

0.6

0.45

0.2

0.4 0.35

2009-2012

Consumer Services

0.5

Financial Integration Index

2004-2008

1999-2003

Financial Integration Index

Financial Integration Index

0.5

0.15

0.45

0.15 1994-1998

Consumer Goods

0.5

Basic Materials

0.55

2009-2012

1999-2003 Utilities

0.45 0.4 0.35 0.3 0.25 0.2

1994-1998

1999-2003

Fig. 4. Indicator of Global Market Integration by Industry in Four Different Periods: The “1st Component.” Notes: Market integration in each industrial equity market is captured by the proportion of total variation in individual industry equity excess returns explained by the first principal component. The number of countries included to extract the first principal component corresponds to the number of industry equity indexes available since December 1993 (see note 15). The first principal component is extracted over period I, II, III, and IV.

170

MICHAEL DONADELLI Trade and Financial Openness 30

Stocks traded (% of GDP)

160

28

140 26

120 100

24

80 60

22

40

Stocks traded (left)

20

International traded (right)

0

1996

1998

2000

2002

2004

2006

2008

2010

International traded (% of GDP)

180

20 2012

Fig. 5. Financial and Trade Openness. Notes: This table reports the evolution of the total value of stocks traded (black line) and international trade of goods and services (gray line) over the period 1995
2012. Both series are measured as percentage of GDP. The stocks traded series is from the World Development Indicators and refers to the total value of shares traded during the period. The international trade (sum of imports and exports) series is from the OECD database. Shaded areas denote official NBER-dated recessions.

steep increase in the level of trade and financial openness.23 Overall, we find evidence, at the industry level, that financial integration tend to be accompanied by increasing levels of trade and financial openness. In other words, international trade of assets and goods provides a channel for financial integration.

Some Robustness Checks 1. In our analysis, principal components are extracted by employing the correlation matrix. We investigate whether the use of the covariance matrix to extract principal components affects our results. In practice, we re-compute our 10 global risk factors (i.e., the 10 principal components extracted from the set of 39 regional portfolio excess returns) by using the covariance rather than the correlation matrix. We find that the Pukthuanthong & Roll (2009)’s integration measure and the Volosovych

Measuring Financial Integration

171

(2011)’s integration index display similar dynamics. We also replicate the exercise in Fig. 4 using the covariance matrix. Needless to say, market integration patterns are similar. 2. Do we really need 10 global risk factors? Do fewer principal components produce similar R-square patterns? To address this issue, we compute the R-square using just the first one and the first three of the ten principal components. Using three factors instead of ten produces a similar integration pattern. However, the R-squares are slightly lower. With one factor, integration patterns are also similar. We just find much lower R-square values. Therefore, factors one through 10 are indeed contributing something to the measured level of integration. 3. Why do we use principal components as global risk factors? Do different global risk factors generate similar results? We have employed principal components rather than other variables to be consistent with the analysis of Pukthuanthong & Roll (2009). However, adjusted R-squares can be obtained also by regressing international equity market returns on standard macroeconomic and financial risk factors (i.e., large market indices, global liquidity measures, US/EU industrial production, economic policy uncertainty index (UI), among many others). For example, Yu et al. (2010) estimate the R-square by regressing Asian stock market returns on the following common components: cross-economy averages of currency return, excess equity return, dividend yield, and forward premia. Similarly, at the industry level, Donadelli & Persha (2014) show that the R-square obtained from a world CAPM (i.e., the world excess return is used as unique global risk factor) is increasing over time. As a robustness check, based on existing empirical works (Bilson et al., 2001; Donadelli & Prosperi, 2012; Ferson & Harvey, 1994), we have recomputed our industry average R-squares by using the following four global risk factors: (i) the world excess returns (WORLD); (ii) the rate of change of the CBOE volatility index (VIX); (iii) the rate of change of the US consumer confidence index; and (iv) the weighted average of the US and EU economic policy UI.24 Not surprisingly, these four global risk factors provide almost the same integration pattern over time for each industry as we have seen earlier based on principal components. It turns out that the evolution of financial market integration is robust to the choice of factors. 4. The sub-periods employed in our analysis rely on economic, financial or political shocks. Do different sub-periods give rise to different results? To account for this possible issue, we estimate both the “multi-factor R-square” and the “1st principal component” in a rolling-window

172

MICHAEL DONADELLI

framework. In practice, both measures have been re-estimated using a rolling window of 60 months (5 years). We find a very similar pattern for both the adjusted R-square and the proportion of total variation in individual excess returns explained by the first principal component (i.e., the difference between the level of integration in period IV and the level of integration in period I is relatively high). 5. Do a country-level analysis provide similar adjusted R-square patterns? As an additional check, we have used a larger set of industry equity indices. In other words, we have extracted the first ten principal components by using the original set of variables (120 industry equity indices). To have a homogeneous dataset, we have used only those industries (i.e., industry equity indices) for which data are available since December 1993 (see note 15). The excess return of each industry in each country is then regressed on the “new 10 principal components.” The adjusted R-square is then averaged across countries. We observe that industry average R-square patterns are similar to those reported in Fig. 2.25 6. As benchmark developed market, this chapter employs the United states. What about other developed markets? Using a different market or a portfolio of advanced equity markets, rather than the US market, we obtain similar integration patterns. This is due to the high degree of comovement between the excess return of the US equity markets and the excess return of the other develop equity markets over the four analyzed sub-periods (i.e., the average correlation ranges from a minimum of 0.72 (period I) to a maximum of 0.88 (period IV)).26

CONCLUDING REMARKS This chapter examines the level of integration in 10 industrial equity markets in a “US-Emerging world.” Financial integration in each industry is captured via two robust integration measures. The first measure, Pukthuanthong & Roll (2009)’s integration index, is represented by the adjusted R-square of a multi-(artificial) factor model. The second measure, Volosovych (2011)’s integration index, corresponds to the proportion of total variation in individual excess returns explained by the first principal component. Our main empirical findings are as follows. First, in each industry, we observe that the level of integration in the aftermath of the Lehman’s collapse (i.e., 2009
2012) is higher than in the aftermath of emerging equity market liberalizations (i.e., 1994
1998). Second, we

173

Measuring Financial Integration

observe that financial integration slows down between the first and second sub-periods. We argue that this has been mainly caused by the emerging systemic banking crises of the late 1990s. This evidence holds for all industries and is supported by both measures. Therefore, “Volosovych (2011) meets Pukthuanthong and Roll (2009).” Third, we observe that a steep increase in the level of financial integration is associated with a steep increase in the level of trade and financial openness. Overall, the empirical findings of this chapter give rise to a “diversification benefits-insurance benefits trade-off.” On the one hand, a higher level of integration (at the industry and country level) reduces both crosscountry and cross-industry diversification benefits. On the other hand, stronger financial market integration produces a more efficient international risk-sharing environment, that is, improves consumption smoothing (i.e., insurance benefits against bad times).

NOTES 1. A survey of this literature can be found in Kearney & Lucey (2004). 2. Indirect integration measures can be found in Portes & Rey (2005) and Bekaert, Harvey, & Lumsdaine (2003). 3. The three-dimensional analysis of wavelet coherency can be included in this class (see Graham, Kiviaho, & Nikkinen, 2012). 4. A similar approach, namely, common component approach, can be found in Yu et al. (2010). 5. See, for details, Pretorius (2002), Imbs (2006), Colacito & Croce (2013), and Donadelli (2013). 6. A two-country model with financial autarky or just one-traded bond can be found in Heathcote & Perri (2002) and Benigno & Thoenissen (2008). 7. DGEI break down into six levels. Level 1 is the Market Index. This covers all the sectors in each region or country. Level 2 divides the market into 10 industries and covers all the sectors within each group in each region or country. Levels 3
6 subdivide the level 2 classifications into sector classifications in increasing detail. Source: Datastream. 8. Due to lack in data availability we are not able to build the Latin America Technology Index. 9. Throughout the chapter we use the terms era, sub-period, and period interchangeably. 10. Date of first stock market liberalization (Country): November 1989 (Argentina), March 1988 (Brazil), May 1989 (Chile), April 1991 (China), December 1991 (Colombia), June 1986 (India), June 1987 (Korea), May 1987 (Malaysia), May 1989 (Mexico), May 1986 (Philippines), May 1986 (Taiwan), and January 1988 (Thailand). Further details on equity market liberalization dates can be found in Henry (2000).

174

MICHAEL DONADELLI

11. Note that the fed funds rate moves from 1.00% (as of January, 2004) to 5.26% (as of July, 2007). 12. For additional details on uncertainty shocks, see http://www.policyuncertainty.com. 13. Mean, standard deviation, skewness and kurtosis values, and the Sharpe ratio are available upon request. 14. See, for example, Goetzmann, Li & Rouwenhorst (2005), Obstfeld & Taylor (2003), Quinn & Voth (2008), and Yu et al. (2010), among others. 15. Note that the two approaches give rise to very different results only if variables are expressed in different scales. 16. Yu et al. (2010) use a common component approach to measure financial market integration in Asia. As in Pukthuanthong & Roll (2009), the adjusted R-square obtained from a multi-factor regression serves as measure of integration. In contrast to Pukthuanthong & Roll (2009), the authors do not employ principal components as global risk factors. Instead, they use the following four common factors: cross-economy averages of currency return, excess equity return, dividend yield, and forward premia. 17. Differently from Pukthuanthong & Roll (2009), this chapter employs monthly data and in-sample principal components. In their work, principal components are estimated from returns in the subsequent year, that is, the eigenvectors obtained from the year t − 1 covariance matrix are applied to the same set of returns during year t. Overall, they extract principal components for 34 years. 18. Donadelli (2013) extracts the first 10 principal components (i.e., global risk factors) from a set of variables composed by 19 national stock market excess returns over the period January 1988
December 2011, that is, principal components are extracted only once. The 10 global risk factors are then regressed on the country equity index returns in a rolling-window context. 19. List of employed countries to extract the “1st principal component”: Basic Materials (Argentina, Chile, Colombia, Mexico, China, India, Malaysia, Pakistan, Philippines, Taiwan, Thailand, Czech Republic, Hungary, Israel, Turkey, the United States); Consumer Goods (Argentina, Chile, Colombia, Mexico, China, India, Malaysia, Pakistan, Philippines, Sri Lanka, Taiwan, Thailand, Czech Republic, Hungary, Turkey, the United States); Consumer Services (Argentina, Chile, Colombia, Mexico, China, Malaysia, Pakistan, Philippines, Sri Lanka, Taiwan, Thailand, Israel, Turkey, the United States); Financials (Argentina, Chile, Colombia, Mexico, China, India, Malaysia, Pakistan, Philippines, Sri Lanka, Taiwan, Thailand, Hungary, Israel, Turkey, the United States); Healthcare (Chile, India, Pakistan, Thailand, Hungary, Israel, the United States); Industrials (Chile, Mexico, Peru, China, India, Malaysia, Pakistan, Philippines, Sri Lanka, Taiwan, Thailand, Czech Republic, Israel, Turkey, the United States); Oil & Gas (Argentina, Chile, Colombia, India, Malaysia, Pakistan, Philippines, Thailand, Czech Republic, Israel, Turkey, the United States); Telecommunications (Argentina, Chile, Mexico, India, Malaysia, Philippines, Thailand, Israel, Turkey, the United States); Technology (India, Thailand, Israel, Turkey, the United States); Utilities (Chile, Colombia, India, Malaysia, Pakistan, Philippines, Czech Republic, Turkey, the United States). 20. Using only our four regional portfolios, at the industry level, we obtain almost identical financial integration patterns. We find differences only

Measuring Financial Integration

175

in the order of magnitude of the proportion of variation in individual excess returns (in each industry) explained by the first principal component. This is due to a difference in the total number of variables employed in the two original sets. 21. Note that the technology and telecommunications industries display a constantly increasing integration index. This has been driven by the IT bubble (see also Brooks & Del Negro, 2004). 22. Using only the four regional portfolios for each industry (i.e., in each industry the first principal component is extracted by using four regional portfolio industry excess returns), we obtain similar integration patterns. Results are available upon request. 23. Note that (as of December 2001) China became a WTO member. Therefore, China’s market openness has influenced global market openness. 24. Details are available upon request but the bottom line is that the results are hardly distinguishable. 25. Compared to the original procedure in which only the 39 regional equity industry excess returns are employed, this procedure produces: (i) a lower proportion of total variance explained by the first 10 principal components and (ii) a slightly lower average R-squares. We stress that the evolution of the integration measure across periods does not changes. 26. However, the US equity market has been used a benchmark in several studies (see Donadelli & Persha, 2014; Graham et al., 2012; Hatemi-J, 2012, among others).

ACKNOWLEDGMENT Michael Donadelli thanks the editor (Jonathan Batten), Christian Schlag and Paolo Vitale. All remaining errors are the authors own.

REFERENCES Aggarwal, R., Lucey, B., & Muckley, C. (2004). Dynamics of equity market integration in Europe: Evidence of changes over time and with events. Working Paper n. 19/04. Institute for International Integration Studies. Trinity College Dublin. Arshanapalli, B., & Doukas, J. (1993). International stock market linkages: Evidence from the pre- and post-october 1987 period. Journal of Banking and Finance, 17(1), 193
208. Baker, S., Bloom, N., & Davis, S. (2013). Measuring economic policy uncertainty. Working Paper. Stanford University. Bekaert, G., & Harvey, C. R. (1995). Time-varying world market integration. Journal of Finance, 50, 403
444. Bekaert, G., Harvey, C. R., & Lumsdaine, R. (2003). Dating the integration of world equity markets. Journal of Financial Economics, 65, 203
248.

176

MICHAEL DONADELLI

Bekaert, G., Harvey, C. R., Lundblad, C., & Siegel, S. (2011). What segments equity markets? Review of Financial Studies, 24(12), 3841
3890. Bekaert, G., Hodrick, R., & Zhang, X. (2009). International stock return comovements. Journal of Finance, 64(6), 2591
2626. Benigno, G., & Thoenissen, C. (2008). Consumption and real exchange rates with incomplete markets. Journal of International Money and Finance, 27, 926
948. Bilson, C. M., Brailsford, T. J., & Hooper, V. J. (2001). Selecting macroeconomic variables as explanatory factors of emerging stock market returns. Pacific-Basin Finance Journal, 9, 401
426. Bloom, N. (2009). The impact of uncertainty shocks. Econometrica, 77(3), 623
685. Boyer, B. H., Gibson, M. S., & Loretan, M. (1999). Pitfalls in tests for changes in correlations. International Finance Discussion Paper No.597R. Federal Reserve Board, Washington, DC. Brooks, R., & Del Negro, M. (2004). The rise in comovement across national stock markets: Market integration or IT bubble? Journal of Empirical Finance, 11, 659
680. Carrieri, F., Errunza, V., & Hogan, K. (2007). Characterizing world market integration through time. Journal of Financial and Quantitative Analysis, 42, 915
940. Chambet, A., & Gibson, R. (2008). Financial integration, economic instability and trade structure in emerging markets. Journal of International Money and Finance, 27, 654
675. Chan, K. C., Gup, B. E., & Pan, N. (1992). An empirical analysis of stock prices in major Asian markets and the United States. The Financial Review, 27, 289
307. Chan, K. C., Gup, B. E., & Pan, N. (1997). International stock market efficiency and integration: A study of 18 countries. Journal of Business, Finance and Accounting, 24, 803
813. Claus, E., & Lucey, B. M. (2012). Equity market integration in the Asia Pacific region: Evidence from discount factors. Research in International Business and Finance, 26(2), 137
163. Colacito, R., & Croce, M. (2010). The short- and long-run benefits of financial integration. American Economic Review: P&P, 100, 527
531. Colacito, R., & Croce, M. (2013). International asset pricing with recursive preferences. Journal of Finance, 68(6), 2651
2686. de Jong, F., & de Roon, F. A. (2005). Time varying market integration and expected returns in emerging markets. Journal of Financial Economics, 78, 583
613. Donadelli, M. (2013). Global integration and emerging stock market excess returns. Macroeconomics and Finance in Emerging Market Economies, 6(2), 1
36. Donadelli, M., & Persha, L. (2014). Understanding emerging market equity risk premia: Industries, governance and macroeconomic policy uncertainty. Research in International Business and Finance, 30, 284
309. Donadelli, M., & Prosperi, L. (2012). On the role of liquidity in emerging markets stock prices. Research in Economics, 66(4), 320
348. Ferson, W. E., & Harvey, C. R. (1994). Sources of risk and expected returns in global equity markets. Journal of Banking and Finance, 18, 775
803. Forbes, K. J., & Rigobon, R. (2002). No contagion, only interdependence: Measuring stock market comovements. Journal of Finance, 57, 2223
2261. Gallagher, L. (1995). Interdependencies among the Irish, British and German stock markets. Economic and Social Review, 26(2), 131
147. Gilmore, C. G., & McManus, G. M. (2002). International portfolio diversification: US and Central European equity markets. Emerging Markets Review, 3(1), 69
83.

Measuring Financial Integration

177

Goetzmann, W. N., Li, L., & Rouwenhorst, K. G. (2005). Long-term global market correlations. Journal of Business, 71, 1
38. Graham, M., Kiviaho, J., & Nikkinen, J. (2012). Integration of 22 emerging stock markets: A three-dimensional analysis. Global Finance Journal, 23, 34
47. Grootveld, H., & Salomons, R. (2003). The equity risk premium: Emerging vs. developed markets. Emerging Markets Review, 4(2), 121
144. Harvey, C. R. (1995). Predictable risk and returns in emerging markets. Review of Financial Studies, 8, 773
816. Hatemi-J, A. (2012). Is the UAE stock market integrated with the USA stock market? New evidence from asymmetric causality testing? Research in International Business and Finance, 26(3), 273
280. Heathcote, J., & Perri, F. (2002). Financial autarky and international business cycles. Journal of Monetary Economics, 49, 601
627. Henry, P. (2000). Stock market liberalization, economic reform and emerging market equity prices. Journal of Finance, 55, 529
564. Huber, P. J., & Ronchetti, E. M. (2009). Robust statistics (2nd ed.). Wiley Series in Probability and Statistics. New York, NY: Wiley. Imbs, J. (2006). The real effects of financial integration. Journal of International Economics, 68, 296
324. International Finance Corporation. (1999, July). The IFC indexes: Methodology, definitions and practices. World Bank emerging markets data base, Washington, DC. Jappelli, T., & Pistaferri, L. (2011). Financial integration and consumption smoothing. The Economic Journal, 121, 678
706. Kasa, K. (1992). Common stochastic trends in international stock markets. Journal of Monetary Economics, 29(1), 95
124. Kearney, C., & Lucey, B. M. (2004). International equity market integration: Theory, evidence and implications. International Review of Financial Analysis, 13, 571
583. Kenourgios, D., & Samitas, A. (2011). Equity market integration in emerging Balkan markets. Research in International Business and Finance, 25, 296
307. Kim, S., Kim, S. H., & Wang, Y. (2006). Financial integration and consumption risk sharing in East Asia. Japan and the World Economy, 18, 143
157. Kuper, G. H., & Lestano, L. (2007). Dynamic conditional correlation analysis of financial market interdependence: An application to Thailand and Indonesia. Journal of Asian Economics, 18, 670
684. Lee, C., Chen, M., & Chang, C. (2013). Dynamic relationships between industry returns and stock market returns. The North American Journal of Economics and Finance, 26, 119
144. Lewis, K. (1999). Trying to explain the home bias in equities and consumption. Journal of Economic Literature, 37, 571
608. Longin, F., & Solnik, B. (2001). Extreme correlation of international equity markets. Journal of Finance, 56, 649
676. Manning, N. (2002). Common trends and convergence? South East Asian equity markets, 1988
1999. Journal of International Money and Finance, 21(2), 183
202. Narayan, P. K., & Narayan, S. (2012). Do US macroeconomic conditions affect Asian stock markets? Journal of Asian Economics, 23(6), 669
679. Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55(3), 703
708.

178

MICHAEL DONADELLI

Newey, W. K., & West, K. D. (1994). Automatic lag selection in covariance matrix estimation. Review of Economic Studies, 61(4), 631
654. Obstfeld, M., & Taylor, A. M. (2003). Globalization and capital markets. In M. D. Bordo, A. M. Taylor, & J. G. Williamson (Eds.), Globalization in historical perspective. Chicago, IL: University of Chicago Press. Panchenko, V., & Wu, E. (2009). Time-varying market integration and stock and bond return concordance in emerging markets. Journal of Banking and Finance, 33, 1014
1021. Phylaktis, K., & Ravazzolo, F. (2002). Measuring financial and economic integration with equity prices in emerging markets. Journal of International Money and Finance, 21, 879
903. Portes, R., & Rey, H. (2005). The determinants of cross-border equity flows. Journal of International Economics, 65(2), 269
296. Pretorius, E. (2002). Economic determinants of emerging stock market interdependence. Emerging Markets Review, 3, 84
105. Pukthuanthong, K., & Roll, R. (2009). Global market integration: An alternative measure and its application. Journal of Financial Economics, 92(2), 214
232. Quinn, D. P., & Voth, H.-J. (2008). A century of global equity market correlations. American Economic Review, 98, 535
540. Schotman, P., & Zalewska, A. (2006). Non-synchronous trading and testing for market integration in Central European emerging markets. Journal of Empirical Finance, 13, 462
494. Suzuki, Y. (2014). Financial integration and consumption risk sharing and smoothing. International Review of Economics and Finance, 29, 585
598. Tesar, L., & Werner, I. (1995). Home bias and high turnover. Journal of International Money and Finance, 14, 467
492. Volosovych, V. (2011). Measuring financial market integration over the long run: Is there a U-shape? Journal of International Money and Finance, 30, 1535
1561. Voronkova, S. (2004). Equity market integration in Central European emerging markets: A cointegration analysis with shifting regimes. International Review of Financial Analysis, 13, 633
647. Wilcox, R. R. (2005). Introduction to robust estimation and hypothesis testing. San Diego, CA: Academic Press. Yang, S. Y. (2005). A DCC analysis of international stock market correlation: The role of Japan on the Asian four tigers. Applied Financial Economics Letters, 1(2), 89
93. Yu, I.-W., Fung, K.-P., & Tam, C.-S. (2010). Assessing financial market integration in Asia
Equity markets. Journal of Banking and Finance, 34, 2874
2885.

PART III MONETARY POLICY AND CREDIT IN A POST-CRISIS SETTING

WILL QUANTITATIVE EASING ENHANCE OR DRAIN THE AVAILABILITY OF FUNDS TO FINANCIAL MARKETS? Yasushi Suzuki ABSTRACT This chapter challenges a commonly accepted view such that the increase in monetary supply aiming to lower the market rate and/or aiming to ease liquidity conditions would encourage the banks as financial intermediaries or the investors as fund providers to provide more funds, which results in stimulating the macro-economy. This chapter suggests that there is no clear-cut mechanism in the economic theory for underpinning the commonly accepted view upon which the Quantitative Easing policy is based. This theoretical analysis suggests that there may exist an appropriate level of market reference rate, which can encourage the investors to absorb the relatively wider range of credit risk in the bond market. Extremely higher market rate would discourage the borrowers to raise funds, while lower market rate would drain “risk” funds in the bond market. In this context, the appropriate level of market rate may stand on a narrow range of the kind of “knife-edge,” though the level

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 181
192 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096006

181

182

YASUSHI SUZUKI

per se does not always guarantee the optimal allocation of financial resources. Keywords: Abenomics; liquidity trap; market reference rate; quantitative easing; risk funds

INTRODUCTION A commonly accepted view insists that the increase in monetary supply would lower the market rate easing liquidity and/or credit conditions, encouraging the enterprises as borrowers to invest in their projects since their funding cost is expected to be lowered. This view implies that the increase in monetary supply would encourage the banks as financial intermediaries or the investors as fund providers to provide more funds, which results in stimulating the macro-economy. To what extent is this view held? To predict the impact of the US monetary policy in a series of Quantitative Easing Policy (QEP) and Japan’s recent QEP in the so-called “Abenomics,” it is very important to re-examine the theories underpinning the view. Quantitative Easing (QE) refers to changes in the composition and/or size of a central bank’s balance sheet that are designed to ease liquidity and/or credit conditions (Blinder, 2010). According to Bernanke and Reinhart (2004), when the size corresponds to expanding the balance sheet, while keeping its composition unchanged, the policy is narrowly defined quantitative easing. On the other hand, when the composition corresponds to changing the composition of the balance sheet, while keeping its size unchanged by replacing conventional assets with unconventional assets, they narrowly define the policy as credit easing. In practice, given constraints on policy implementation, central banks have combined the two elements of their balance sheet, size, and composition, to enhance the overall effects of unconventional policy. In this context, broadly defined quantitative easing, often used in a vague manner, better fits as a package of unconventional policy measures making use of both the asset and liability sides of the central bank balance sheet, designed to absorb the shocks hitting the economy (Shiratsuka, 2010). As is argued later, several empirical studies on QEP are conducted, suggesting the limited effect of QEP on raising aggregate demand and prices. However, the existing debate seems not to sufficiently explain the reason of the limited effect of QEP. Besides suggesting the limited instrumental

Quantitative Easing and Available Funds for Financial Markets

183

rationality in the economic theory underpinning the QEP, this chapter raises a hypothesis that the increase in monetary supply may drain “risk” funds in the financial market, by a theoretical approach from the supplyside (fund providers’) perspective, which approach is yet to be adequately argued in the existing literature. “Risk” funds are here meant by the funds or capital that are provided by the investors who are willing to directly undertake and absorb the risk of borrowers (or issuers in the bond market), often appearing in the direct financing route. The section “Knife-Edge Hypotheses on Market Reference Rate” suggests the limited instrumental rationality in the economic theory underpinning the QEP and argues that the appropriate level of market rate may stand on a narrow range of the kind of “knife-edge.” The section “Linkage between Hypotheses and Empirical Data” aims to link the hypothesis to the empirical data with reference to the effect of Japan’s QEP. The final section puts concluding comments.

KNIFE-EDGE HYPOTHESES ON MARKET REFERENCE RATE Let us begin with suggesting the limited instrumental rationality in the economic theory underpinning the QEP. For the banks as financial intermediaries in the indirect financing route, their nominal net profit from the “floating rate” lending is not affected by the change in market rate (the reference rate or base rate for the banks), so far as the spread margin as risk premium toward the borrowers remains unchanged and the loan exposure remains the same. In other words, the banks’ net profit from the floating rate lending is affected only if (1) the banks consider the borrowers’ lowered funding cost to reasonably lower their probability of default (to increase their probability of success), then the banks are willing to increase the loan exposure toward the borrowers when higher risk-adjusted returns are expected, and (2) the borrowers increase the demand of fund-raising for their investment. The above (1) is related to the banks’ subjective judgment of screening and monitoring, while the above (2) depends on the borrowers’ subjective sentiment of investment. Even though the market rate is lowered, the borrowers do not necessarily increase the demand of fund-raising (lower funding cost does not always ease the pessimistic sentiment of borrowers when the other factors, for instance, uncertainties about the product/technology obsolescence

184

YASUSHI SUZUKI

under the severe competition are more significant. Using the Keynes’ term, lowering the rate of interest does not always stimulate investment if lowering the schedule of the marginal efficiency of particular capital assets can offset the effect of stimulating investment). Even though the borrowers increase the demand of fund-raising, the banks do not necessarily respond sufficiently to the demand, because their credit policy is determined at each bank’s discretion. Furthermore, even though the banks expect higher risk-adjusted returns and increase the loan exposure, they do not necessarily maintain the exposure when the spread margin is adjusted (diminished) with lower credit risk profile under the competitive credit market. There is no clear-cut/no a priori mechanism in the relationship between the monetary policy and the loan exposure (consequently affecting the effective demand as stimulus to the macro-economy). In short, how the monetary policy can affect the macro-economy depends the lenders’ subjective judgment of credit risk screening and the borrower’s subjective sentiment of investment. Furthermore, it would make sense to raise the hypothesis that the increase in monetary supply may drain “risk” funds provided by investors in financial markets. In a simple general equilibrium model of Arrow
Debreu as to the direct and indirect financial market, if we accept the unrealistic assumption that there were zero monitoring cost, the only possible general equilibrium would be one where all risk-adjusted interest rates are equal (Freixas & Rochet, 1997; Suzuki, 2011). In the simplistic framework, the coupon rate on bonds (denoted by r) and the lending rate (denoted by rL) for the firms should be perfect substitutes. If one of the two rates is lower than the other, firms would prefer to raise all the funds there, resulting in the potential disappearance of the other. In reality, each borrower has distinctive credit risks related to their type and the type of activity in which they engage. Therefore, investors face significant and borrower-specific information and monitoring costs for screening and monitoring this credit risk. It is extremely difficult and costly for individual investors (particularly, households) who are not professionals in monitoring to evaluate the credit risk of, in particular, small and middle-sized enterprises (SMEs), although it may be somewhat easier to do this for internationally reputable large firms. Consider the floating rate notes (FRNs) purchased by the investors in the corporate bond market to compare the floating rate lending by the banks in the credit market. FRNs are bonds that have a variable coupon, equal to an interbank money market reference rate, like LIBOR (London Interbank Offered Rate) or federal funds rate, plus a quoted spread (margin). The spread is a rate that remains constant. Almost all FRNs

Quantitative Easing and Available Funds for Financial Markets

185

have quarterly coupons, that is, they pay out interest every three months. At the beginning of each coupon period, the coupon is calculated by taking the fixing of the reference rate for that day and adding the spread. A typical coupon would look like three months USD LIBOR + 1.5% (or 150 basis points or 150 bps). The coupon rate (r) of the FRNs is composed of the money market reference rate (denoted by rIBOR) and the spread margin (denoted by bps). Basically, the spread margin is reflected in the credit risk of each borrower. r = rIBOR þ bps We consider that each investor has its own benchmark which makes them expect the satisfactory profit taking into account the associated risk, bringing the prospective yield (denoted by rQ) of the investment. The prospective yield is based on the investor’s subjective judgment of screening risk and return. In theory, if rQ > r, the investor would not engage in the investment. On the contrary, if rQ < r, it means that the coupon rate would be attractive for the investor. In other words, the minimum condition for engaging in the investment, rQ should be equal to r. rQ ≦ r substitute r = rIBOR + bps, then we can obtain the following conditions: rQ − rIBOR ≦ bps or rQ − bps ≦ rIBOR Assume that the spread (bps) as risk premium for each borrower remains unchanged at least in the short-term period. When the market rate (rIBOR) decreases, if the investors seek for the fixed prospective yield, they may have less incentives to hold the bond in the same credit risk category. Some of them may have an incentive to engage in the other bond offering higher spread. However, the bond offering higher spread is associated with higher credit risk. Particularly for the risk-averse investors who are not professionals in monitoring to evaluate the credit risk of unknown or SMEs, rQ would be closer to ∞ for engaging in the high-yield junk bonds.

186

YASUSHI SUZUKI

Rather, when the market rate (rIBOR) decreases, some investors may have an incentive to leave for the other “low-risk and low-return” type bond which is on the same indifference curve of the risk-return preference. To illustrate this, assume that there are two bonds; one is “high-risk and high-return” type bond (Bond A, in which the spread is, for instance, 2%), while the other is “low-risk and low-return” type bond (Bond B, in which the spread is 0.5%). When the market rate stays at 5%, the coupon rate (r) of Bond A becomes 7% (5 + 2), while that of Bond B is 5.5% (5 + 0.5). When the market rate decreases to 1%, the coupon rate of Bond A would be 3% (1 + 2), while that of Bond B would be 1.5% (1 + 0.5). As the market rate (rIBOR) decreases, the weight of rIBOR in the coupon rate would decrease (5/7 (71.4%) to 1/3 (33.3%) in Bond A, 5/5.5 (90.9%) to 1/1.5 (66.7%) in Bond B) if the spread remains unchanged. For the cash-rich investors who do not have to borrow the funds for purchasing the bond, the weight of rIBOR in the coupon rate functions as a buffer or cushion for absorbing the issuer’s credit risk which is reflected in the spread margin. In the above case, the weight of buffer in the coupon rate of Bond A would decrease if the market rate decreases. On the other hand, the weight of spread in the coupon rate of Bond B would increase (from 9.1% to 33.3%) even though the associated credit risk remains unchanged. As a result, some risk-averse investors would leave for Bond B, because they feel that the weight of buffer (1/3) is not enough to absorb the 200 bps risk and the weight of spread (1/3) is attractive to absorb the 50 bps risk. We hypothesize that as the market rate decreases, more investors would lose the incentive to absorb higher credit risk, in other words, lose the incentive to provide “risk” funds because the market rate functions as a buffer for absorbing credit risk for the cash-rich and risk-averse investors. If this hypothesis is not rejected, it may explain a dimension of the cause of “liquidity trap” (Krugman, 1998), which is yet to be focused in the academic literature. For the banks being engaged in the floating rate lending, the lending rate (rL) is determined by each bank’s funding cost (the base rate covering the funding cost, denoted by rBR) and the spread (bps) for each borrower. The funding cost is reflected in, for instance, the deposit rate (denoted by rD), the money market reference rate (rIBOD), the propensity to pay dividends (funding cost from equities, denoted by div) which is reflected in the policy for capital adequacy requirement, and the operating and administration cost (denoted by op). rL = rBR þ bps;

Quantitative Easing and Available Funds for Financial Markets

187

where rBR = f ðrD ; rIBOD ; div; opÞ As mentioned earlier, in a simple general equilibrium model of Arrow
Debreu as to the direct and indirect financial market, the coupon rate on securities (r) and the lending rate (rL) for the firms should be perfect substitutes. If this condition is held, rIBOD þ bps = rBR þ bps therefore, we can reach to the following condition for the equilibrium. rIBOD = rBR In reality, this condition is not held. This is because the other factors such as deposit rate (rD) and propensity to pay dividends (div) in each bank may vary in the determination of the base rate (rBR). One of the implication is that (1) if the bank relies only on the interbank money market for funding, and (2) if the bank has no obligation to keep the capital adequacy requirement (or if the bank has the same liability composition and cost of deposit and equity as the reference bank) and also (3) if the operating cost is identical to that of the reference bank which offers the money market reference rate, it is possible to hold the condition for a partial equilibrium in financial market. In practice, we often observe that r is lower than rL particularly for prominent borrowers. It implies that the funding cost in the average banks is higher than that in the reference bank. On the other hand, there are few capital/bond markets for SMEs because the risk-averse investors would not engage in the high-yield junk bond even though the high spread (bps) is offered, where the weight of base rate (rBR) would become less meaningful for the investors. As a result, only the debt (bank loan) market is available for SMEs and marginally creditworthy borrowers. As an alternative explanation leading to the situation where r < rL for prominent borrowers, we hypothesize that the risk preference of the investors (particularly households) may swing to a “risk-neutral/loving” or “euphoric” position toward the investment in prominent issuers. When the money market reference rate (rIBOR) provides an adequate buffer for undertaking the associated risk, their subjective judgment of rQ < r will possibly lower the coupon rate through accepting the lower spread (bps) applied for

188

YASUSHI SUZUKI

the bond in the long run. On the contrary, when the money market reference rate decreases, the reduced credit risk buffer may lead the risk preference of the investors swing back to a “conservative” or “risk-averse” strategy for the investment, consequently draining “risk” funds in financial market. It is considered that the subjective prospective yield (rQ) from the investment in a particular FRN is determined by the money market reference rate, the spread and the risk preference (denoted by θ). When the risk preference lies in the risk-neutral position, it is considered to be θ = 0. The risk-loving preference would have an effect on accepting the lower yield, while the risk-averse preference would require higher risk premium or keep the investors completely away from the investment. It is worth noting that the risk preference or sentiment even in an individual may possibly swing all the times. rQ = f ðrIBOR ; bps; θÞ The coupon rate (r) and the floating lending rate (rL) are considered to have a certain realistic upper limit. This is because higher rate would discourage the issuers (or the borrowers) to raise funds. The spread margin (bps) is also considered to have a certain realistic upper limit. This is because the extremely high spread based upon less creditworthy non-rated firms would not attract any investor nor any conservative bankers. We hypothesize that the portion of the money market reference rate (rIBOR) functions as a buffer for the investors to absorb the issuer’s credit risk. An adequate level of buffer may shift the investor’s risk preference to a risk-neutral or risk-loving direction, lowering the prospective yield (rQ) and possibly providing more “risk” funds in the bond market. Under this situation, it is likely that the coupon rate (r) would be lower than the floating lending rate (rL) for the firms in the same and certain range of prominent/acceptable credit risk category. When the market reference rate decreases, the reduced buffer may shift the investor’s risk preference to a risk-neutral or risk-averse direction, requiring relatively higher prospective yield against the coupon rate, resulting in the situation where rQ = r or rQ > r. This situation would discourage the investors to engage in the bond, or encourage them to leave for the low-risk-low-return type bond as mentioned earlier. From another perspective, as the market reference rate decreases, the convergence of debt and bond markets is accelerated in a sense that the banks would regain the comparative advantage in mediating funds, though the banks do not necessarily mediate funds.

Quantitative Easing and Available Funds for Financial Markets

189

LINKAGE BETWEEN HYPOTHESES AND EMPIRICAL DATA Several empirical studies on QE are conducted. For instance, Ugai (2006) surveys the empirical analyses that examine the effects of the Bank of Japan (BOJ)’s QE policy, which was implemented from March 2001 through March 2006. Ugai (2006) points out that the QE effect on raising aggregate demand and prices was often limited, generally this effect, if any, was smaller than that stemming from the commitment, due largely to the progressing corporate balance sheet adjustment, as well as the zero bound constraint on interest rates. Joyce, Lasaosa, Stevens, and Tong (2010) attempt to assess the impact of the Bank of England (BOE)’s QE policy on asset prices which began in March 2009. They estimate the reaction of gilt (UK government securities) prices to the program and suggest that QE may have depressed gilt yields. But they point out that the initial impact of GE was muted, and the wider impact on other asset prices is more difficult to disentangle from other influences. Shiratsuka (2010) re-examines the BOJ’s experience of QEP from 2001 to 2006, pointing out that the credit spreads for nonfinancial businesses, measured as the differences between the credit product indicators across ratings and the TB rate in three-month contracts, declined but with certain time lags after the introduction of the QEP (rather, the credit spreads for the BB-rated firms were enlarged during around one year after the introduction, see Fig. 1). In spite of a general reduction in the external financing premium in a middle-term range, the level of lending by private Japanese banks continued to be shrunken. The outstanding loans toward SMEs had dropped sharply
from ¥344.9 trillion in December 1998 to ¥260.3 trillion in December 2003, then to ¥253.1 trillion in December 2009 (SMEA, 2005, 2010; Suzuki, 2011, p. 5). Several empirical studies point out the limited effect of QE, however, the existing debate seems not to sufficiently explain the reason of the limited effect of QE. Shiratsuka (2010), a staff of the BOJ, raises several policy implications from the analysis on the BOJ’s QEP from 2001 to 2006. In particular, he points out that QE is a temporary policy response. “The increase in size and the change in composition of the central bank balance sheet simply buy time until certain progress can be made in balance-sheet adjustments at financial institutions, such as disposal of nonperforming assets and recapitalization. The increase in size and the change in composition of the central bank balance sheet do not directly lead to the early restoration of the financial intermediation function” (Shiratsuka, 2010, p. 99). In addition,

190

YASUSHI SUZUKI Basis points

200 ZIRP

AA

QEP

A

150

BBB BB

100

50

0

–50 1999

2000

01

02

03

04

05

06

07

08

09

Fig. 1. Nonfinancial Business: Bloomberg Fair Market Value Index for Companies with rating AA, A, BBB, and BB. Source: Shiratsuka (2010, p. 90).

he points out that QE is likely to produce side effects such as the corollary of public intervention in private financial transactions, potentially distorting incentives and resource allocation in the private sector. My analysis suggests that there may exist an appropriate level of market reference rate, which can encourage the investors to absorb the relatively wider range of credit risk in the bond market. Extremely higher market rate would discourage the borrowers to raise funds, while lower market rate would drain “risk” funds in the bond market. In this context, the appropriate level of market rate may stand on a narrow range of the kind of “knife-edge,” though the level per se does not always guarantee the optimal allocation of financial resources. My hypothesis suggests that there is no clear-cut mechanism in the economic theory for underpinning the commonly accepted view upon which the QEP is based.

CONCLUDING COMMENTS Keynes mentioned in the final chapter
Concluding notes on the social philosophy toward which the General Theory might lead
of the General Theory; “Thus it is to our best advantage to reduce the rate of interest to that point relatively to the schedule of the marginal efficiency of capital at

Quantitative Easing and Available Funds for Financial Markets

191

which there is full employment” (Keynes, 1936, p. 375). However, the scale of investment is not always promoted by a low rate of interest as Keynes assumed in the General Theory. Keynes suggested that we aim in practice at “an increase in the volume of capital until it ceases to be scarce, so that the functionless investor will no longer receive a bonus” (Keynes, 1936, p. 376). On the other hand, despite the availability of sufficient funds, screening and monitoring activities still matter because the failure of monitoring by (functional) lenders and investors would exacerbate the principal-agent problem or the general uncertainty from which lenders suffer, and thereby restrict the optimal allocation of risk funds. In my view, lenders and investors need a certain buffer or cushion for providing risk funds, in other words, for responding to the fundamental uncertainty. Under the unprecedented QEP in several developed countries, we should review the function of rate of interest as an incentive for lenders and investors to provide risk funds to the future potentials. The developed economies such as the United States and Japan continue to rely heavily on the QE as an instrument to continuously stimulate the economies. We should take more care of the above side effects that become more obvious as the duration of quantitative easing is prolonged. Knifeedge hypotheses on market reference rate should be argued to examine the side effects or by-products of the QEP.

REFERENCES Bernanke, B., & Reinhart, V. R. (2004). Conducting monetary policy at very low short-term interest rates. American Economic Review, 94(2), 85
90. Blinder, A. S. (2010, November/December). Quantitative easing: Entrance and exit strategies. Federal Reserve Bank of St. Louis Review, 92(6), 465
479. Freixas, X., & Rochet, J.-C. (1997). Microeconomics of banking. Cambridge, MA: The MIT Press. Joyce, M., Lasaosa, A., Stevens, I., & Tong, M. (2010). The financial market impact of quantitative easing. Working Paper No. 393. Bank of England. Keynes, J. M. (1936). The general theory of employment, interest and money (7th ed.). Cambridge: Macmillan, Cambridge University Press. Krugman, P. (1998). It’s Baaack: Japan’s slump and the return of the liquidity trap. Brookings Papers on Economic Activity, 29(2), 137
205. Shiratsuka, S. (2010, November). Size and composition of the central bank balance sheet: Revisiting Japan’s experience of the quantitative easing policy. Monetary and Economic Studies, Institute for Monetary and Economic Studies, Bank of Japan, 79
106.

192

YASUSHI SUZUKI

SMEA (Small and Medium Enterprise Agency). (2005). Nendo Chusho Kigyo Hakusho (2005 White Paper on Small and Medium Enterprises in Japan), Small and Medium Enterprise Agency, Ministry of Economy, Trade and Industry of Japan. SMEA (Small and Medium Enterprise Agency). (2010). Nendo Chusho Kigyo Hakusho (2010 White Paper on Small and Medium Enterprises in Japan), Small and Medium Enterprise Agency, Ministry of Economy, Trade and Industry of Japan. Suzuki, Y. (2011). Japan’s financial slump, collapse of the monitoring system under institutional and transition failures. Basingstoke: Palgrave Macmillan. Ugai, H. (2006). Effects of the quantitative easing policy: A survey of empirical analyses. Bank of Japan, Working paper series, 06-E-10.

MONEY DEMAND CAUSALITY FOR TEN ASIAN COUNTRIES: EVIDENCE FROM LINEAR AND NONLINEAR CAUSALITY TESTS Ahdi Noomen Ajmi and Nicholas Apergis ABSTRACT This chapter estimates causality properties between real money demand and a number of determinants, that is, real output, the lending rate and the real exchange rate, across 10 Asian economies through linear and nonlinear causality methodologies spanning the period 1990
2012. The results document both bidirectional and unidirectional causality between monetary aggregates (M1 and M2) and their determinants for different country groups. The empirical findings exemplify the role of the demand for money as a policy tool and can provide useful policy recommendations to the Asian monetary authorities in their vision of forming a future monetary union. Keywords: Real demand for money; linear and nonlinear causality tests; Asian countries JEL classification: E41

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 193
210 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096007

193

194

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

INTRODUCTION The literature on money demand function estimation is concerned with the existence of a stable money demand function. It can provide a framework which helps to distinguish between explicit changes in money explained by developments in macroeconomic variables and changes that are specific to the situation at hand. At the same time, a stable money demand function forms the core in the conduct of monetary policy as it enables a policydriven change in monetary aggregates to have predictable influences on output, interest rate and the price level. Therefore, the existence of a wellspecified and modeling specific relationship between money and a number of variables can be seen as prerequisite for the use of monetary aggregates in the conduct of monetary policy. Monetary policy has played a crucial role and become so successful in taming inflation. Indeed, central banks have changed their strategy of the conduct of the monetary policy “lean against the wind” to maintain the macroeconomic and financial stability. In this respect the stability of money demand function is a natural starting point for comprehensive realizing of monetary policy strategy, where nominal shocks mainly originate from instability of money demand function. However, much of the recent literature on monetary economics seems to de-emphasize the importance of money demand (Duca & VanHoose, 2004). Despite the consensus that money demand function has little role under an interest-rate-based (Taylor-rule type) monetary policy, it is still believed that money demand is important for monetary policy. In addition, financial innovations and ongoing improvements in information processing technology have affected payment and portfolio allocation behavior and the central bank could lose the control over aggregate demand (McCallum, 2003). Even though the interest-rate-based policy has reduced the importance of monetary aggregates, demand for money remains relevant. This is especially relevant for countries like the Asian countries, where monetary policy does not work only through the interest rate channel, and that money demand can provide useful information about portfolio allocations. The goal of this chapter is to examine for the first time causality issues between real money demand and a number of determinants across 10 Asian countries. The results are expected to be of great significance given the plan these countries to form a monetary and economic union. Based on this plan, they will have to beg their currencies, and thus they will be committed to striving toward the eventual adoption of a common currency, upon fulfillment of certain conditions, including ceilings for inflation and long-term interest rates, budget deficits and government debt and exchange

Money Demand Causality for Ten Asian Countries

195

rate stability, yielding high importance to money demand analysis to become more relevant as money is deemed to be highly relevant to detect risks to price stability. An additional novelty of the chapter is that it uses the methodological approaches of linear and nonlinear Granger causality. Given that most economic and financial series are characterized by nonlinearities due to the presence of structural breaks, while the linear tests are sensitive to causality in the conditional mean and may not be sufficient to detect nonlinear effects on the conditional distribution, the nonlinear causality test recommended by Hiemstra and Jones (1994) is capable of detecting nonlinear causal relationships based only on the correlation integral. The Granger causality test (Granger, 1969) is based on the assumption of linear relationships between variables. However, this test suffers from many shortcomings. First, in some cases it may have no economic interpretation. Second, Baek and Brock (1992) and Hiemstra and Jones (1994) show that linear Granger causality tests can wrongly presume that causality implies that a change in one series causes a reaction in another series. In fact, one of the series may react first without necessarily causing a reaction in another series. Also, as shown by Baek and Brock (1992), these tests generally have low power against nonlinear relationships. In this line, the major novelty of the chapter is to investigate short-run causal relationships between real money demand and a number of its determinants by adopting a nonlinear approach. Baek and Brock (1992) recommend a nonparametric statistical method for uncovering these relationships. We apply the nonlinear causality test in the sense of Hiemstra and Jones (1994). The most two important features of this test are that: (i) through Monte Carlo simulation, Baek and Brock (1992) show that the forecasting performance of linear models decreases in the presence of nonlinearities, while the forecasting performance of nonlinear models is better vis-a`-vis that of linear models, and (ii) the test is capable of detecting nonlinear causal relationships while avoiding problems arising from model misspecification. Hiemstra and Jones (1994), through a simulation exercise, show that the modified version of their test is robust to a number of model misspecifications. To this end, the nonlinear Granger causality test has been widely employed in many fields, including macroeconomics and finance (Hiemstra & Jones, 1994). The rest of this chapter is structured as follows. The section “Literature Review” provides a brief review of the relevant literature. The section “The Methodological Approach” presents the details of the methodological approach, while the section “Empirical Results” reports the econometric results on causality. Finally, the section “Conclusions and Policy Implications” concludes and offers some policy implications.

196

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

LITERATURE REVIEW Relatively few studies were conducted on developing countries. However, it has been increasing in recent years, primarily triggered by the concern among central banks and researchers around the world on the impact of moving toward flexible exchange rates regimes, globalization of capital markets, ongoing financial liberalization and innovation in domestic markets, and the country-specific events on the demand for money (Sriram, 1999). Studies that have explored the demand for money in developing countries include, among other, studies by Gupta and Moazzami (1989), Bahmani-Oskooee and Malixi (1991), Simmons (1992), and Sriram (2002). Bahmani-Oskooee and Malixi (1991) estimate the demand for money function in 13 developing countries as a function of inflation, real income and the real effective exchange rate. They conclude that, ceteris paribus, a depreciation in real effective exchange rate results in a fall in the demand for domestic currency. However, they did not include the interest rate spread to capture the general process of financial asset substitution. Agenor and Khan (1996) estimate a dynamic currency substitution model incorporating forward-looking rational expectations for a group of 10 developing countries. They also allude to the view that the foreign interest rate and the expected rate of depreciation of the parallel market exchange rate play a crucial role in the choice between holding domestic money or switching to foreign currency held abroad. Simmons (1992) employs an error-correction model to estimate the demand for money in five African economies. His empirical results indicate that the domestic interest rate is an important determinant of the demand for money. He also finds that in four out of five cases inflation plays an important role in determining the demand for money. Another strand has considered the general process of financial asset substitution and justified the use of an exchange rate and a foreign interest rate in the analysis (Bahmani-Oskooee & Rhee, 1994; Chowdhury, 1995). These studies are clearly in favor of both the currency substitution and capital mobility hypotheses. Therefore, it is very important to include the real effective exchange rate in the money demand function. Tang (2007) finds that real M2 aggregate, real expenditure components, the exchange rate, and the inflation rate are cointegrated for Malaysia, the Philippines, and Singapore. Using narrow money demand for Indonesia, as it described in Hossain (2007), the empirical results suggest that real income, inflation (a proxy for expected inflation) and the return on foreign financial assets are the major determinants of a narrow money demand function. Hamori

197

Money Demand Causality for Ten Asian Countries

(2008) analyzes the money demand function in Sub-Saharan African and his empirical results reveal that there exists a steady relationship of the money demand function in the Sub-Saharan African region. In other words, there is a close relationship between the money supply and the real economy, while monitoring money supply promises to play an important role in stabilizing the level of prices in this region. Inoue and Hamori (2009) indicate that when money supply is represented by M1 and M2, a cointegrating vector is detected among real money balances, interest rates, and output in India. In contrast, it is found that when money supply is represented by M3, there is no long-run equilibrium relationship in the money demand function. Narayan, Narayan, and Mishra (2009) estimate a money demand function for a panel of five South Asian countries. They find that the money demand and its determinants, namely real income, real exchange rate and short-term domestic and foreign interest rates are cointegrated both for individual countries as well as for the panel, and panel long-run elasticities provide robust evidence of statistically significant relationships between money demand and its determinants.

THE METHODOLOGICAL APPROACH Linear Granger Causality The Granger causality test (Granger, 1969) is designed to detect causal direction between two stationary time series x and y in terms of predictability. More precisely, a time series yt Granger-causes another time series xt if series xt can be predicted better by using past values of yt than by using only the historical values of xt. Testing for causal relations between the two series involves estimating a p-order linear vector autoregressive model, VAR(p), as follows:

yt = α0 þ

p X

α1i yt − i þ

i=1

xt = β 0 þ

p X i=1

p X

α2i xt − i þ ɛ1t

ð1Þ

β2i xt − i þ ɛ2t

ð2Þ

i=1

β1i yt − i þ

p X i=1

198

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

Where ɛt = (ɛ1t, ɛ2t) is a white noise process with zero mean and covariance matrix Σ and p is the lag order of the process. In the empirical section, the Akaike information criteria test is used to select the optimal lag order p. α0 and β0 are constants and, αs and βs are parameters. {xt} does not Grangercause linearly {yt} is equivalent to say that α2i = 0, for all j = 1, 2, …, p. Similarly, {yt} does not Granger-cause linearly {xt} if, and only if, β1i = 0, for all j = 1, 2, …, p.

Nonlinear Granger Causality: The Hiemstra
Jones Test One important problem with the linear approach to causality testing is that such tests can have low power detecting some kinds of nonlinear causal relations. However, it is now widely recognized that most economic and financial series are characterized by nonlinearities due to the presence of structural breaks. In addition, the linear test is only sensitive to causality in the conditional mean and may not be sufficient to detect nonlinear effects on the conditional distribution (Baek & Brock, 1992). Various nonparametric tests have been proposed in the literature. The most prominent one perhaps is developed by Hiemstra and Jones (1994), which is a modified version of the test recommended by Baek and Brock (1992). The Hiemstra and Jones (1994) test is a nonparametric statistical method to detect nonlinear causal relationships based on the correlation integral. To define nonlinear Granger causality, assume that there are two strictly and weakly dependent time series {Xt} and {Yt}, t = 1,2,3,…, T. Let m-length lead vector of Xt be designated by Xm t , and the Lx-length and LyLy length vectors of Xt and Yt, respectively, by XLx t − Lx and Yt − Ly . For given values of m, Lx and Ly ≥ 1 and for all e > 0, {Yt} does not strictly Granger {Xt} if:   Lx Ly Ly m Lx Pð‖Xm t − Xs ‖ < e‖Xt − Lx − Xs − Lx ‖ < e; ‖Yt − Ly − Ys − Ly ‖ < eÞ  Lx m Lx  = Pð‖Xm t − Xs ‖ < e ‖Xt − Lx − Xs − Lx ‖ < eÞ;

ð3Þ

where Pð⋅Þdenotes probability and ‖⋅‖denotes the maximum norm. Eq. (3) states that the conditional probability that two arbitrary m-length lead vectors of {Xt} are within distance e, given that the corresponding lagged Lx-length lag vectors of {Xt} are e-close, is the same as when one also conditions on the Ly-length lag vectors {Yt} of being e-close. The test based on

Money Demand Causality for Ten Asian Countries

199

Eq. (3) can be implemented by expressing the conditional probabilities in terms of the corresponding ratios of joint probabilities: C1ðm þ Lx; Ly; eÞ C3ðm þ Lx; eÞ = C2ðLx; Ly; eÞ C4ðLx; eÞ

ð4Þ

where C1, C2, C3, and C4 are the correlation integral estimators of the joint probabilities which are discussed in detail by Hiemstra and Jones (1994). With an additional index n, Hiemstra and Jones (1994) show that under the assumption that {Xt} and {Yt}are strictly stationary, weakly dependent, if {Yt} does not strictly Granger cause {Xt} then,   pffiffiffi C1ðm þ Lx; Ly; e; nÞ C3ðm þ Lx; e; nÞ a n − ∼ Nð0; σ 2 ðm; Lx; Ly; eÞÞ ð5Þ C2ðLx; Ly; e; nÞ C4ðLx; e; nÞ where n = T + 1 − m − max(Lx,Ly). See the appendix of Hiemstra and Jones (1994) for both definitions and an estimator of σ2(m,Lx,Ly,e). It is also shown that this test has a very good power against a variety of nonlinear Granger causal and non-causal relations (Hiemstra & Jones, 1994; Ma & Kanas, 2000). To test for nonlinear Granger causality between {Xt} and {Yt}, the test in Eq. (5) is applied to the estimated residual series from the bivariate VAR model. The null hypothesis is that Yt does not nonlinearly strictly Granger cause Xt, and Eq. (5) holds for all m, Lx, Ly ≥ 1 and e > 0. By removing the linear predictive power from a linear VAR model, any remaining incremental predictive power of one residual series for another can be considered as nonlinear predictive power (Baek & Brock, 1992).

EMPIRICAL RESULTS Data The empirical analysis makes use of annual data on money supply defined as M1 and M2 (M1, M2), real GDP at 2005 prices (Y), the lending rate (RL), the GDP deflator (P), and the real exchange rate (REER) defined as the effective rate for 10 Asian countries (i.e., China, Fiji, Hong Kong, India, Indonesia, Japan, Malaysia, Philippines, Singapore, South Korea)

200

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

spanning the period 1990
2012. All variables are expressed in logarithms. Data are obtained from the Bloomberg database.

Unit Root Tests First, we check the stationarity of the variables by using the efficient Ng and Perron (2001) unit root test. We find evidence of nonstationarity for the variables in levels and stationarity for the variables in first differences. Next, we begin with linear Granger causality tests to detect the causal relationships between the variables under investigation (Table 1).

Linear Granger Causality Results Using the series in first differences, the VAR models between the variables under study are constructed. The results reported in Table 2 are mixed across countries. In particular, for China, there exists unidirectional causality running from M1/P and M2/P to real GDP and the real exchange rate as well as from the lending rate to M1/P. A bidirectional causality between M1/P and the lending rate can be also found for Malaysia. For Philippines, the linear Granger causality test detects unidirectional causality running Table 1. Country

M1/P

Levels

China Japan Malaysia Philippines Singapore Indonesia Hong Kong Fiji Korea India a

0.287 0.422 1.637 1.293 1.432 1.718 2.072 0.660 0.592 0.411

Ng
Perron Unit Root Results. M2/P

Lending Rate

Real Effective Exchange Rate

1st Levels 1st Levels 1st Levels 1st Levels 1st differences differences differences differences differences −21.53a −14.11a −15.42a −15.56a −15.36a −15.04a −13.39a −15.25a −14.61a −11.94a

1.668 0.528 1.173 0.753 0.346 −0.281 −4.659 0.081 0.117 −2.538

−13.34b −13.43b −36.84a −15.76a −14.57a −12.80b −9.519b −12.89b −12.74b −10.40b

−3.711 0.259 0.748 −1.346 −1.413 −3.297 −2.804 0.618 −1.020 −2.495

−18.64a −28.27a −15.81a −15.15a −13.32b −46.05a −15.68a −15.88a −18.81a −15.94a

indicates the rejection of the null hypothesis at the 1%, level. indicates the rejection of the null hypothesis at the 5% level. NP statistics are −13.8, −8.1, and −5.7 at 1%, 5%, and 10%, respectively.

b

Real GDP

−4.246 −0.195 −2.719 2.058 −1.357 0.700 0.722 0.657 −0.198 −0.562

−22.52a −12.73b −15.40a −12.22b −14.76a −14.06a −14.06a −17.39a −14.95a −17.08a

−0.775 −2.483 −0.693 −3.009 −3.689 −2.153 −1.890 −0.975 −1.750 −2.524

−15.97a −19.29a −15.63a −15.97a −9.91a −15.32a −15.06a −13.58b −15.43a −10.19b

201

Money Demand Causality for Ten Asian Countries

Table 2.

Linear Causality Results.

Hypothesis

p-Value

Hypothesis

p-Value

Hypothesis

p-Value

China M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.0481b 0.0074a 0.0090a 0.4992

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.3117 0.0486b 0.4127 0.0150b

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.5008 0.2923 0.5365 0.5539

Japan M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.5681 0.2457 0.8452 0.5721

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.1293 0.4137 0.8623 0.4457

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.4823 0.5526 0.6412 0.2769

Malaysia M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.4837 0.0674c 0.3949 0.9158

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.8807 0.1627 0.4050 0.1483

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.0409b 0.2770 0.9050 0.3190

Philippines M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.6634 0.3868 0.8800 0.0934c

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.1726 0.7213 0.3246 0.5078

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.0257b 0.7626 0.0507c 0.3127

Singapore M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.3769 0.4635 0.2160 0.3450

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.4231 0.0089a 0.7798 0.7578

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.3118 0.3076 0.5961 0.9793

Indonesia M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.5397 0.1242 0.9951 0.7671

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.2408 0.3578 0.1872 0.6010

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.2480 0.0061a 0.0279b 0.0844c

Hong Kong M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.0854c 0.1161 0.0077a 0.7856

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.4561 0.4294 0.7622 0.0576c

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.7006 0.8304 0.9726 0.8001

Fiji M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.9391 0.5847 0.9156 0.2512

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.8200 0.1232 0.8299 0.6723

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.5768 0.7178 0.0746c 0.8247

202

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

Table 2.

(Continued )

Hypothesis

p-Value

Hypothesis

p-Value

Hypothesis

p-Value

Korea M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.3149 0.0202b 0.7720 0.0083a

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.0049a 0.1461 0.0015a 0.1744

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.5982 0.0756c 0.6450 0.0482b

India M1\P ↛ Y RL ↛ M1\P M2\P ↛ Y RL ↛ M2\P

0.4825 0.3175 0.6382 0.3520

Y ↛ M1\P M1\P ↛ REER Y ↛ M2\P M2\P ↛ REER

0.2561 0.4442 0.3201 0.4258

M1\P ↛ RL REER ↛ M1\P M2\P ↛ RL REER ↛ M2\P

0.3293 0.7681 0.7033 0.7473

Figures denote p-values. a indicates the rejection of the null hypothesis at the 1% level. b indicates the rejection of the null hypothesis at the 5% level. c indicates the rejection of the null hypothesis at the 10% level.

from M1/P and M2/P to the lending rate. A unidirectional causality relationship from M1/P to the real effective exchange rate exists for Singapore, while for the case of Indonesia the results support the presence of unidirectional relationships running from the real effective exchange rate to M1/P and M2/P and from M2/P to the lending rate as for Fiji. For Hong Kong, there exists unidirectional causality running from M1/P and M2/P to real GDP. Unidirectional causality running from real GDP, the lending rate and the real effective exchange rate to M1/P and M2/P is detected for Korea. Finally, for the remaining countries, that is, Japan and India, no causal relationships are found.

Hiemstra
Jones Nonlinear Causality Results In terms of the Hiemstra and Jones (1994) test, we fix the values for the head length at m = 1 and for the common scale parameter at e = 1.5. This lag length runs from 1 to 5 denoting the number of lags on the residuals series used in the test. Table 3 reports the results of the Hiemstra
Jones nonlinear causality test. The existence of nonlinear causality between M1/P and the real effective exchange rate can be found in three countries, that is, China, Malaysia and India, as well as between M2/P and the real effective exchange rate for four countries, that is, China, Japan, Fiji, and India. We also find significant bidirectional causality between M1/P and real GDP for

Lags 1 2 3 4 5 Lags 1 2 3 4 5 Lags 1 2 3 4 5 Lags 1 2 3 4 5

Hiemstra
Jone’s Nonlinear Causality Tests.

China M1\P ↛ Y −1.3871 −19.8617 0.0000 0.0000 0.0000

Y ↛ M1\P −2.9904 30.7637a 2.2699b 1.5406 −3.2564

M1\P ↛ RL −1.4018 −1.6718 −1.0250 −1.1456 −1.2898

RL ↛ M1\P −2.1962 0.2208 −0.7683 −0.8570 −0.9176

M1\P ↛ REER −2.4959 −3.3929 3.9143a 0.9333 −1.5973

REER ↛ M1\P −4.0991 −7.4318 40.9563a 4.5160a 0.7430

China M2\P ↛ Y −2.4366 −5.7867 −12.9454 78.1780a 7.4554a

Y ↛ M2\P −1.9359 −1.7332 −0.3335 −0.7200 −1.3137

M2\P ↛ RL −2.7817 −2.2457 −5.6425 35.1906a 0.2750

RL ↛ M2\P −2.2686 −1.1844 0.0000 0.0000 0.0000

M2\P ↛ REER −2.6595 −6.0786 16.2854a 0.9489 −0.8274

REER ↛ M2\P −3.7587 −3.0891 8.1481a −0.7338 −0.5251

Japan M1\P ↛ Y −5.7174 16.9435a −3.9361 −1.2182 −1.6660

Y ↛ M1\P −5.9295 −24.6166 4.2357a −2.6843 −4.2210

M1\P ↛ RL −24.6056 −1.5873 −2.5649 0.0000 0.0000

RL ↛ M1\P 0.1110 −3.9765 −9.0274 0.0000 0.0000

M1\P ↛ REER −0.1751 −5.0000 0.0000 0.0000 0.0000

REER ↛ M1\P −5.2124 0.0000 0.0000 0.0000 0.0000

Japan M2\P ↛ Y −2.3589 −1.4943 −2.2592 −3.5834 −1.3450

Y ↛ M2\P −1.1731 −1.4663 −1.5069 −2.0646 −1.1864

M2\P ↛ RL −4.0721 84.8511a −1.5283 −3.4898 0.0000

RL ↛ M2\P −1.1842 97.4754a 1.3359 −1.4347 −0.7239

M2\P ↛ REER −3.2995 −0.5324 −4.3773 45.6515a −5.5302

REER ↛ M2\P −2.0405 2.5946a −5.0422 45.2451a −5.5461

Money Demand Causality for Ten Asian Countries

Table 3.

203

Lags 1 2 3 4 5 Lags

Lags 1 2 3 4 5 Lags 1 2 3 4 5

Malaysia M1\P ↛ Y −1.7315 −3.5035 26.1452a 0.6109 −0.9639

Y ↛ M1\P −1.9084 −2.9875 18.3707a 0.2078 0.0000

M1\P ↛ RL −1.6319 −0.1826 −1.6217 −1.9084 −2.8957

RL ↛ M1\P −1.9548 −1.1417 −1.3598 −1.3371 −2.2543

M1\P ↛ REER −1.5017 2.5370b 12.0918a 0.8226 −0.4854

REER ↛ M1\P −2.2116 −3.8317 10.2068a −0.0326 −0.1036

Malaysia M2\P ↛ Y −1.4309 −1.0696 −1.0846 −1.6627 −2.5722

Y ↛ M2\P −1.2711 −0.8971 −0.7818 −0.1686 0.0000

M2\P ↛ RL −3.9363 −1.2542 −4.4954 91.2558a 5.1038a

RL ↛ M2\P −0.4308 −0.1611 −1.6217 14.4574a 0.5003

M2\P ↛ REER −0.9956 −0.5251 −1.0171 −1.4439 −2.6315

REER ↛ M2\P −0.5402 −1.2383 −1.3535 −2.3251 −2.9279

Philippines M1\P ↛ Y 0.2598 −0.2401 0.1226 0.4246 1.4838

Y ↛ M1\P −3.0582 −3.3757 −4.5237 −6.7005 −7.1498

M1\P ↛ RL −4.3047 −1.9556 −7.5036 −4.6904 0.0000

RL ↛ M1\P −8.4934 7.0547a 0.0000 0.0000 0.0000

M1\P ↛ REER −9.4925 −0.0561 0.6666 0.5060 0.0000

REER ↛ M1\P −4.8423 −3.6461 −3.4188 −7.4309 0.0000

Philippines M2\P ↛ Y −2.9469 −2.4631 −3.7743 −2.884 −7.6229

Y ↛ M2\P −1.9767 0.1385 0.6045 −0.8953 6.8430a

M2\P ↛ RL 0.4221 0.0000 0.0000 0.0000 0.0000

RL ↛ M2\P −2.4759 −4.6904 0.0000 0.0000 0.0000

M2\P ↛ REER −3.5441 −2.1777 −0.7606 0.3411 1.4903

REER ↛ M2\P −4.3676 −3.3083 −4.4759 −0.1294 0.2241

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

1 2 3 4 5

204

Table 3. (Continued )

1 2 3 4 5 Lags 1 2 3 4 5 Lags 1 2 3 4 5 Lags 1 2 3 4 5

Singapore M1\P ↛ Y −18.5195 −1.8124 −2.8908 −3.5219 −5.0704

Y ↛ M1\P −15.2683 0.1337 1.6285 1.9647b 2.8285a

M1\P ↛ RL 87.4413a −3.6179 0.0000 0.0000 0.0000

RL ↛ M1\P 67.1957a −5.1134 8.5044a −0.1652 −5.0990

M1\P ↛ REER −3.6817 1.1921 1.2798 −1.1595 0.0000

REER ↛ M1\P −3.3168 −0.4050 −1.6975 −1.9697 −5.0990

Singapore M2\P ↛ Y −56.4445 1.7445c −1.2273 0.1909 0.0000

Y ↛ M2\P 174.3737a 6.2362a −1.4710 −7.9749 1.6537c

M2\P ↛ RL −5.1591 −1.4191 −1.1581 3.3626a −2.4697

RL ↛ M2\P 79.7746a −1.1543 −6.3696 12.7892a 0.0000

M2\P ↛ REER −6.1652 −5.1310 −1.7144 0.3839 0.0486

REER ↛ M2\P −1.8642 −11.2646 −0.1444 −0.6708 −2.0021

Indonesia M1\P ↛ Y −4.7995 −4.9101 43.2278a −0.2312 −1.6895

Y ↛ M1\P 3.5721a −2.1947 −4.4270 4.1410a 2.5778a

M1\P ↛ RL −3.2708 −7.5728 −3.6827 −3.7121 0.0000

RL ↛ M1\P −2.8222 −1.3486 3.9168a −0.4771 6.1028a

M1\P ↛ REER −0.9371 −0.8955 −2.6693 −4.7654 −5.5520

REER ↛ M1\P −3.6041 −8.6193 11.5871a 4.8081a 7.4252a

Indonesia M2\P ↛ Y −0.9605 −1.9836 0.06509 0.0320 −0.5829

Y ↛ M2\P −4.5316 −10.3997 7.4821a −0.2383 −2.6960

M2\P ↛ RL 3.0507a 9.7562a −1.8485 0.0000 0.0000

RL ↛ M2\P −7.1404 −15.5354 1.9914b −0.5395 −0.6047

M2\P ↛ REER −1.7388 −2.9175 −1.0759 −5.5193 −1.6328

REER ↛ M2\P −2.7920 −5.7291 −4.7864 11.9010a 0.0000

Money Demand Causality for Ten Asian Countries

Lags

205

Lags 1 2 3 4 5 Lags

Lags 1 2 3 4 5 Lags 1 2 3 4 5

Hong Kong M1\P ↛ Y −4.2377 84.8003a −1.6996 −2.5495 −5.0990

Y ↛ M1\P −5.1736 −12.6100 0.0000 0.0000 0.0000

M1\P ↛ RL −11.3368 −0.3271 −3.2609 0.0000 0.0000

RL ↛ M1\P −15.0169 −7.6134 −11.8393 0.0000 0.0000

M1\P ↛ REER −0.2960 38.5702a −8.5438 0.0000 0.0000

REER ↛ M1\P −4.5240 −41.4269 −5.0990 0.0000 0.0000

Hong Kong M2\P ↛ Y −6.5761 0.2666 −0.1097 −0.1342 0.0000

Y ↛ M2\P −29.9876 0.6370 −0.7460 −1.3716 −1.7039

M2\P ↛ RL −2.4340 −0.8230 1.1833 −1.0206 −1.4408

RL ↛ M2\P −3.7999 0.0332 0.2591 −1.2729 −1.5963

M2\P ↛ REER −2.1829 −8.3103 0.0053 −0.2605 3.0072a

REER ↛ M2\P −6.0619 0.7491 −0.3859 0.0000 0.0000

Fiji M1\P ↛ Y −6.5235 −3.4938 −4.3551 −18.5812 −5.0990

Y ↛ M1\P −18.1759 0.1246 −0.3225 −1.9774 0.7478

M1\P ↛ RL −3.0929 −1.3702 −0.9174 −0.8605 −0.9991

RL ↛ M1\P −4.2903 −23.1099 1.4298 −0.1059 −0.5858

M1\P ↛ REER −6.1705 6.1167a −0.8563 1.4030 −1.0633

REER ↛ M1\P −3.5937 0.92801 −2.4717 −0.7548 −1.0633

Fiji M2\P ↛ Y −3.1100 −1.4554 −0.6650 −1.0106 −4.0311

Y ↛ M2\P −5.8542 −6.6479 −7.3095 −5.0261 −4.0311

M2\P ↛ RL −1.3220 108.1486a −1.8737 −0.8519 −1.3803

RL ↛ M2\P −7.4503 −220.1786 6.6971a −1.3284 −1.9322

M2\P ↛ REER 2.0216b −13.6894 −5.0098 −3.5361 −4.5581

REER ↛ M2\P −6.1953 10.2606a 3.0912a −1.1509 −1.6739

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

1 2 3 4 5

206

Table 3. (Continued )

Lags

Lags 1 2 3 4 5 Lags 1 2 3 4 5 Lags 1 2 3 4 5

Y ↛ M1\P −3.2214 −6.8593 3.0312a −4.7958 0.0000

M1\P ↛ RL −2.3326 −3.7819 0.8282 −4.0115 −3.7463

RL ↛ M1\P −1.4716 −0.0327 −2.1984 5.5349a 2.1488b

M1\P ↛ REER 6.8002a −0.3634 0.0000 0.0000 0.0000

REER ↛ M1\P −4.0449 −2.5415 −2.2458 1.5348 −4.7958

Korea M2\P ↛ Y −8.3626 0.3654 −0.4540 0.0000 0.0000

Y ↛ M2\P −3.4547 15.9777a 1.3502 −4.7958 0.0000

M2\P ↛ RL −1.8975 −3.5718 −6.1860 −13.6808 −8.3040

RL ↛ M2\P −2.3868 −2.8625 −3.5125 −5.7540 −2.4494

M2\P ↛ REER −3.0740 −6.0623 −0.0711 0.0000 0.0000

REER ↛ M2\P −5.8463 −16.2250 −3.6322 3.2635 −4.7958

India M1\P ↛ Y −2.0824 −1.5196 −5.0816 −0.5105 0.0000

Y ↛ M1\P −2.1219 −1.1390 −6.5596 −3.9111 0.0000

M1\P ↛ RL −2.3271 −5.4663 −0.6573 −11.6286 3.1158a

RL ↛ M1\P −3.0913 −1.8527 −9.7222 −7.2504 −6.6915

M1\P ↛ REER −3.5623 −0.8957 −1.7461 −4.6824 3.7409a

REER ↛ M1\P −1.1949 −5.3910 −6.6804 −19.1348 17.1508a

India M2\P ↛ Y −1.6426 −7.5247 −4.5858 0.0000 0.0000

Y ↛ M2\P −4.8169 −14.3124 0.0000 0.0000 0.0000

M2\P ↛ RL −7.2399 −0.1028 −0.6767 −0.3009 0.0000

RL ↛ M2\P −26.3859 2.5775a −1.1572 −1.6346 −1.7758

M2\P ↛ REER −1.1681 5.1617a 0.1951 −1.8470 −2.6969

REER ↛ M2\P −5.8705 4.9868a 0.3747 0.0996 −1.7963

207

This table reports the standardized test statistic of Hiemetra
Jones (1994). “Lags” denotes the number of lags in the residual series used in the test. a indicates the rejection of the null hypothesis of absence of causality at the 1% level. b indicates the rejection of the null hypothesis of absence of causality at the 5% level. c indicates the rejection of the null hypothesis of absence of causality at the 10% level.

Money Demand Causality for Ten Asian Countries

1 2 3 4 5

Korea M1\P ↛ Y −2.2454 3.5803a −10.2895 −6.1905 0.0000

208

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

Japan, Malaysia, Indonesia and Korea, and between M2/P and real GDP for Singapore. Moreover, the results support the presence of a bidirectional causality between M2/P and the lending rate for Japan, Malaysia, Singapore, Indonesia, and Fiji, and between M1/P and the lending rate for Singapore. Next, the empirical findings show significant unidirectional causality running from real GDP to M1/P for China and Singapore, and from real GDP to M2/P for Philippines, Indonesia, and Korea. In the reverse sense, we find unidirectional causality for Hong Kong running from M1/P to real GDP and for China running from M2/P to real GDP, while both M1/P and M2/P seem to cause the lending rate for India and China, respectively. However, the lending rate causes M1/P for Philippines, Indonesia and Korea, and M2/P for India. For the cases of Hong Kong, Fiji and Korea, a significant unidirectional causality seems to be running from M1/P to the real effective exchange rate. Finally, for the case of Indonesia a unidirectional relationship running from the real effective exchange rate to both M1/P and M2/P is also detected.

CONCLUSIONS AND POLICY IMPLICATIONS This study examined the dynamic causality interactions between real money demand and a number of determinants, that is, real output, the lending rate and the real exchange rate, across 10 Asian countries through the methodology of linear and nonlinear causality over the period 1990
2012. The presence of causalities running from the determinants to real money demand is important for the conduct of monetary policy since it can provide useful policy guidelines to central banks in their quest for price stability. Since causality is established relative to both the M1 and M2 monetary aggregates, a viable option for monetary authorities of corresponding countries to conduct money targeting either with respect to M1 or to M2 for their monetary policy implementation is present. The fact that the real exchange rate plays a vital role in causality indicated that these countries substitute domestic for foreign currency, implying that they may lose money from seigniorage, which might reduce the monetary authorities’ capacity to maintain full monetary control. Thus, the presence of the currency substitution effect suggests that the economy is vulnerable to both domestic and external shocks which may undermine the ability of the central bank to exert control over money supply. Therefore, a policy suggestion is that the liquidity requirements should be broadened to

Money Demand Causality for Ten Asian Countries

209

include foreign currency deposits held at the commercial banks to ensure effective control over money supply. Moreover, the empirical findings supported that under current economic circumstances in those Asian countries, M1 and M2 monetary aggregates still can be maintained as intermediate target of central banks to control inflation. However, in the future, these central banks might seek a new intermediate target policy (such as interest-rate targeting or inflation targeting) since the alteration in economy-operating environment, variations in interest rate regime and exchange rate regime might lead to the further instability of money demand function. Finally, the potential future formation of a monetary union is expected to include all countries considered in the analysis, since money targeting is documented as a feasible monetary policy target. .

REFERENCES Agenor, P. R., & Khan, M. (1996). Foreign currency deposits and the demand for money in developing countries. Journal of Development Economics, 50, 101
118. Baek, E., & Brock, W. (1992). A general test for non-linear Granger causality: Bivariate model. Working Paper. Iowa State University and University of Wisconsin, Madison, WI. Bahmani-Oskooee, M., & Malixi, M. (1991). Exchange rate sensitivity of the demand for money in developing countries. Applied Economics, 23, 1377
1383. Bahmani-Oskooee, M., & Rhee, H. J. (1994). Long-run elasticities of the demand for money in Korea: Evidence from cointegration analysis. International Economic Journal, 8, 83
93. Chowdhury, A. R. (1995). The demand for money in a small open economy: The case of Switzerland. Open Economies Review, 6, 31
44. Duca, J. V., & VanHoose, D. D. (2004). Recent developments in understanding the demand for money. Journal of Economics and Business, 56, 247
272. Granger, C. W. J. (1969). Investigating causal relations by econometrics models and cross spectral methods. Econometrica, 37, 424
438. Gupta, K. L., & Moazzami, B. (1989). Demand for money in Asia. Economic Modelling, 6, 467
473. Hamori, S. (2008). Empirical analysis of the money demand function in Sub-Saharan Africa. Working Paper No. 35. Kobe University. Hiemstra, C., & Jones, J. D. (1994). Testing for linear and nonlinear granger causality in the stock price-volume relation. Journal of Finance, 49, 1639
1664. Hossain, A. A. (2007). The narrow money demand behavior in Indonesia, 1970
2005. ASEAN Economic Bulletin, 24, 320
338. Inoue, T., & Hamori, S. (2009). An empirical analysis of the money demand function in India. Economics Bulletin, 29, 1224
1245. Ma, Y., & Kanas, A. (2000). Testing nonlinear relationship among fundamentals and exchange rates in the ERM. Journal of International Money and Finance, 19, 135
152.

210

AHDI NOOMEN AJMI AND NICHOLAS APERGIS

McCallum, B. T. (2003). Multiple-solution indeterminacies in monetary policy analysis. Journal of Monetary Economics, 50, 1153
1175. Narayan, P. K., Narayan, S., & Mishra, V. (2009). Estimating money demand functions for South Asian countries. Empirical Economics, 36, 685
696. Ng, S., & Perron, P. (2001). Lag length selection and the construction of unit root tests with good size and power. Econometrica, 69, 1519
1554. Simmons, R. (1992). An error-correction approach to demand for money in five African developing countries. Journal of Economic Studies, 19, 29
48. Sriram, S. S. (1999). Survey literature on demand for money: Theoretical and empirical work with special reference to error-correction models. IMF Working paper No. 99/64. International Monetary Fund, Washington, DC. Sriram, S. S. (2002). Determinants and stability of demand for M2 in Malaysia. Journal of Asian Economics, 13, 337
356. Tang, T. C. (2007). Money demand function for Southeast Asian countries: An empirical view from expenditure components. Journal of Economics Studies, 34, 476
496.

PRODUCT MARKET COMPETITION AND INFLATION PERSISTENCE Amr Sadek Hosny ABSTRACT A number of studies have examined the role of typical demand supply factors in explaining inflation. The contribution in this chapter is to investigate the possible role of product market imperfections on inflation and inflation persistence, especially among emerging market economies. Using a backward-looking Phillips curve framework in a dynamic panel framework of 105 countries over the 2008
2011 period, our findings suggest that product market competition, as measured by the World Economic Forum’s measure of goods market imperfections do have a significant impact on inflation persistence. On average, higher competition and efficiency in product markets reduces the inflation persistence effect especially in the MENA region and countries at lower stages of development. Keywords: Inflation; competition; inflation persistence JEL classification: E5

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 211
219 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096008

211

212

AMR SADEK HOSNY

INTRODUCTION A central objective of a country’s central bank is to achieve price stability. Such a task requires a thorough understanding of the factors driving the inflationary process, in both the short-run and the long-run. This objective becomes more important in countries where inflation has been high and persistent. A number of demand and supply factors have been typically discussed in the literature to explain inflation. In addition to these standard determinants of inflation, market competition imperfections, or product market institutions more generally, have received much attention recently. Attempts at explaining inflation from such an angel has only been investigated in the cases of advanced countries. Examples include Correa-Lopez, Garcia-Serrador, and Mingorance-Arnaiz (2010) for a panel of OECD countries, while studies by Jaumotte and Morsy (2012), Biroli, Mourre, and Turrini (2010), and Andersson, Masuch, and Schiffbauer (2009) have been conducted for the Euro Area. Przybyla and Roma (2005) took the analysis a step further and studied the importance of such a channel at a sectoral level in a sample of EU countries.1 To the best of our knowledge, there has been no attempt to study the role of product market imperfections on inflation in other regions or even in Emerging Market Economies. Therefore, this chapter re-investigates the determinants of inflation using a set of variables that have been identified in the recent literature. The contribution, however, is to explain the role of product market competition in inflation and inflation persistence. For this purpose, we use a measure of goods market imperfections in a cross-section of countries for the period 2008
2011 to fill this gap in the empirical literature. Our objective is to draw some policy lessons, especially for emerging economies as it has been repeatedly argued that product market imperfections manifested in the monopolistic actions in the supply chain have been responsible for the continued inflationary pressures in such countries. Using a dynamic panel data methodology, our findings indicate that product market inefficiencies have a significant impact on inflation persistence in our country sample. On average, higher competition and efficiency in product markets reduces the inflation persistence effect especially in the MENA region and countries in lower stages of development. This chapter is structured as follows. After this introduction, the econometric model and methodology are presented in the second section, while the results are in the third section. The fourth section concludes.

Product Market Competition and Inflation Persistence

213

THE MODEL AND METHODOLOGY The first step is to identify a measure of market competition. Ideally, one would use such an indicator in a time-series framework such as a VECM to investigate its effect and inter-relation with other variables that may influence inflation. Unfortunately, such a measure does not exist for a long period of time for our sample countries, especially not on a high monthly or quarterly frequency. Therefore, we use the “goods market efficiency” index from the “Global Competitiveness Report (GCR)” (see WEF, Many issues). The GCR has been publishing a number of competitiveness indicators in a cross-section of countries since 2008. Since the latest report was published in 2011, this indicator is only available for four years on an annual basis; we therefore decided to perform a panel data analysis covering a cross-section of emerging market economies in an attempt to draw some policy lessons. We use a backward-looking Phillips curve framework in a panel data context to formally study the role of goods market imperfections in explaining inflation in a set of developing and emerging market economies. We follow the specification laid out by earlier literature that studied this channel of inflation in a number of European countries as listed above in the introduction. We use a panel of 105 countries over the 2008
2011 period, where the dependent variable; inflation, is regressed on its own lag, nominal effective exchange rates and a measure of fiscal and monetary policy in an attempt to use a model specification that is as close to identified in previous literature. A number of other explanatory variables are introduced to the model as will be explained below. The variable of interest here is the goods market competitiveness index, which is allowed to affect inflation (i) directly through its own level effect and (ii) indirectly through its effect on inflation persistence. The full specification takes the following form: inf it = β0 þ β1 inf it − 1 þ β2 ðcomptit inf it − 1 Þ þ β3 comptit þ β4 dneerit þ β5 dm1it þ β6 dgovcit þ β7 ITdummyit þ µi þ λt þ ɛit The dependent variable (infit) is inflation of country i at time t, while the independent variables are lagged inflation (infit − 1) to capture inflation persistence, annual percentage change of nominal effective exchange rate (dneerit), annual broad money growth rate (dm1it), and annual growth rate of government consumption expenditure (dgovcit). These explanatory

214

AMR SADEK HOSNY

variables could be considered as the fundamental sources of inflation. We use the goods market imperfection index (comptit) in levels as well as the (comptit*infit − 1) interaction variable to investigate the role of market imperfections on inflation through their direct and indirect effects on inflation, respectively. The goods market index is constructed such that a higher value implies more market efficiency and competition. Finally, the (ITdummyit) indicator takes a value of one if the country formally adopts IT, and zero otherwise. The ITdummy is constructed using information from Roger (2010) and various Central Banks, while the goods market competitiveness index is taken from the GCR as mentioned above. All other data is collected from the IFS database. A dynamic panel methodology is used to estimate the effect of product market imperfections on inflation and inflation persistence. Introducing a lagged dependent variable as in the above equation would render the least square estimator biased and inconsistent as it will be correlated with the error term. To overcome this problem, Anderson and Hsiao (1981, 1982) suggested using lags of the RHS regressors as instruments to yield consistent estimators and Arellano and Bond (1991) further suggested using a generalized method of moments (GMM) estimator. Specifically, we use the most recent and commonly used System-GMM estimator (SGMM) developed by Arellano and Bover (1995) and Blundell and Bond (1998).

THE RESULTS This section presents results of estimating the above regressions, as well as a number of robustness checks. Dynamic panel results are reported in Table 1. Model (1) in Table 1 presents results of the most basic specification. It only includes the fundamentals of inflation, that is, growth rates of money, nominal effective exchange rate and government consumption expenditure, as well as lagged inflation and an interaction term between lagged inflation and the goods market imperfection score. Most coefficients show the expected signs. Lagged inflation is positive and statistically significant indicating the presence of a persistence effect. A depreciation of local currencies is associated with an increase in inflation in our sample of countries, as expected. Most importantly, it appears that goods market imperfections do affect the persistence of inflation. Specifically, higher product market efficiency and competition reduces the inflation persistence effect. This is

215

Product Market Competition and Inflation Persistence

Table 1.

Dynamic Panel Estimation: The Results. Model (5)

Model (6)

1.419*** 0.740*** (2.77) (2.65) comptit*infit − 1 −0.390*** −0.204*** (3.41) (2.87) comptit 0.009 (0.53) dneerit −0.058*** −0.048 −0.057*** −0.044 −0.052*** (2.75) (1.40) (2.76) (1.28) (2.75) dm1it −0.099*** −0.130*** −0.098*** −0.135*** −0.089*** (4.21) (3.56) (4.27) (3.63) (3.60) dgovcit −0.007 −0.008 −0.007 −0.005 −0.006 (0.37) (0.39) (0.36) (0.26) (0.29) ITdummyit −0.003 −0.021 −0.001 (0.34) (1.79) (0.13) cons 0.062*** 0.071*** 0.063*** 0.080*** 0.021 (10.20) (7.69) (8.17) (6.25) (0.27) Year dummies N Y N Y N Country dummies N Y N Y N

1.350*** (2.64) −0.375*** (3.28) 0.019 (0.91) −0.035 (1.03) −0.134*** (3.65) −0.006 (0.30) −0.021* (1.81) 0.001 (0.01) Y Y

infit − 1

Model (1)

Model (2)

Model (3)

0.796** (2.44) −0.221** (2.57)

1.484*** 0.780** (2.91) (2.35) −0.401*** −0.218** (3.49) (2.50)

Model (4)

Source: Author’s estimations. * Significant at 10%, ** Significant at 5%,*** Significant at the 1% significance level. Z-statistics (in absolute value) are in parenthesis. Bold values are simply the variables that are statistically significant.

seen from the negative and statistically significant coefficient on the interaction term (comptit*infit − 1) shown in the table. Model (2) adds year and country dummies. Results are largely unchanged. We then add the ITdummy to the model to assess the effect of adopting inflation targeting on inflation in our sample (see models 3 and 4). The effect seems negligible although it shows the expected negative sign. Adding the goods market imperfection index (compt) in its level does not change results, and doesn’t seem to add any explanatory power to the model (see models 5 and 6). The ITdummy carries the correct negative sign in models (3) through (6), but becomes statistically significant in model (6) only. Our results hold to a number of robustness checks. We performed a number of robustness checks and report them in Table 2. In Panel A, we experimented with different variables using (m2) instead of (m1), nominal exchange rate (xr) instead of nominal effective exchange rate (neer), government fixed capital formation (govf) instead of government consumption expenditure (govc) as well as all possible combinations of these variables.

216

AMR SADEK HOSNY

Table 2. Dynamic Panel Estimation: Robustness Checks. Panel A Variables

m1, neer, govc

m1, xr, govc

m1, neer, govf

m1, xr, govf

m2, neer, govc

m2, xr, govc

m2, neer, govf

m2, xr, govf

infit − 1

0.74*** (2.65) −0.21*** (2.87)

1.06** (2.34) −0.29** (2.50)

0.93*** (3.10) −0.24*** (3.13)

1.12** (2.56) −0.31*** (2.70)

0.718** (2.16) −0.19** (2.07)

1.08** (2.44) −0.3*** (2.56)

0.88*** (2.56) −0.21** (2.33)

1.12** (2.53) −0.31*** (2.64)

comptit*infit − 1

Panel B Region

All

SSA

EAP

ECA

LAC

MNA

SAS

infit − 1

0.74*** (2.65) −0.21*** (2.87)

2.17 (0.95) −0.574 (0.89)

0.0544 (0.05) −0.052 (0.18)

1.43 (0.95) −0.366 (1.00)

0.468 (0.93) −0.176 (1.22)

1.576* (1.70) −0.361* (1.81)

0.6404 (0.13) −0.1754 (0.16)

comptit*infit − 1

Panel C Stage of Development infit − 1 comptit*infit − 1

All

LMIC

UMIC

FD

ED

ID

0.74*** (2.65) −0.21*** (2.87)

3.849 (1.42) −1.027 (1.43)

1.172*** (4.14) −0.316*** (3.84)

5.35** (2.45) −1.55** (2.48)

1.155* (1.68) −0.312* (1.65)

1.235 (0.29) −0.248 (0.28)

Source: Author’s estimations. * Significant at 10%, ** Significant at 5%,*** Significant at the 1% significance level. Z-statistics (in absolute value) are in parenthesis. In Panel A: neer: nominal effective exchange rate, xr: nominal exchange rate, govc: government consumption expenditure, govf: government fixed capital formation. In Panel B: SSA: Sub-Saharan Africa, EAP: East Asia and the Pacific, ECA: Europe and Central Asia, LAC: Latin America and the Caribbean, MNA: Middle East and North Africa, SAS: South Asia. In Panel C: LMIC: lower middle income countries, UMIC: upper middle income countries, FD: factordriven economies, ED: in transition to efficiency and efficiency-driven economies, ID: in transition to innovation and innovation-driven economies. Bold values are simply the variables that are statistically significant.

Using these different variable definitions does not seem to change our basic results. Coefficients on both lagged inflation and the interaction term show the same sign and significance as our baseline model. Higher product market efficiency does lead to lower inflationary persistence and the effect seems to be in the range of (−0.19) to (−0.31) depending on the model specification. The sign on (xr) coefficients (not reported) is now positive since a higher value indicates a depreciation of local currency, as opposed to using (neer), where a higher value indicated an appreciation and thus we got a negative coefficient in Table 1.

Product Market Competition and Inflation Persistence

217

We also report results by region and by stage of development in Panels B and C in Table 2, respectively. We use model (5) specification from Table 1 in carrying this analysis, and re-report its results in the first column in Table 2 for comparison purposes with other models. Focusing on Panel B, it seems that goods market efficiency has no impact on inflation persistence, except in the case of MENA countries. The impact is close to the range reported in Tables 1 and 2, Panel A. Two classifications regarding different stages of development are used in Panel C of Table 2. The first is the Atlas method of the World Bank,2 while the second is the GCR’s classification. Regarding the World Bank classification, we only report results of lower and upper middle income countries since there are very few observations in the other two groups (low and high income countries). The market efficiency interaction term appears insignificant in the case of lower middle income countries, but is statistically significant and within the range reported earlier in case of upper middle income countries. Finally, we use the GCR classification which distinguishes between five different stages of development; namely, factor-driven (FD), efficiencydriven, and innovation-driven economies as well as economies in transition from one stage to the other. In the case of FD economies, the goods market interaction term seems very effective in reducing inflation persistence, while it is statistically insignificant in the case of (ID) countries that are either in transition to innovation or are innovation-driven economies. The product market interaction term is within the earlier reported range of our basic model in the case of (ED) countries that are either in transition to or are efficiency-driven economies. In sum, our results suggest that imperfections in the goods market do have an effect on inflation persistence in our selected sample of countries. This is especially true for countries that are at lower stages of development.

CONCLUSION The purpose of this research is to explain the determinants of inflation in a panel of countries focusing on the role of inefficient monopolistic actions of suppliers along the supply chain in affecting inflation and inflation persistence. We use the WEF’s measure of goods market imperfections and inefficiencies in a cross-section of countries for the period 2008
2011. Findings indicate that product market imperfections do have a significant

218

AMR SADEK HOSNY

impact on inflation persistence. On average, higher competition and efficiency in product markets reduces the inflation persistence effect especially in the MENA region and countries at lower stages of development.

NOTES 1. Bowdler and Nunziata (2007) studied the role of labor market regulations in explaining inflation in the Euro area. 2. Countries are classified into different groups according to their GNI per capita as follows: low income, $1,005 or less; lower middle income, $1,006
$3,975; upper middle income, $3,976
$12,275; and high income, $12,276 or more.

ACKNOWLEDGMENT The views expressed in this chapter belong solely to the author. Nothing contained in this chapter should be reported as representing IMF policy or the views of the IMF, its Executive Board, member governments, or any other entity mentioned herein.

REFERENCES Anderson, T. W., & Hsiao, C. (1981). Estimation of dynamic models with error components. Journal of the American Statistical Association, 76(375), 589
606. Anderson, T. W., & Hsiao, C. (1982). Formulation and estimation of dynamic models using panel data. Journal of Econometrics, 18(1), 47
82. Andersson, M., Masuch, K., & Schiffbauer, M. (2009). Determinants of inflation and price level differentials across the Euro Area countries. ECB Working Paper No. 1129. European Central Bank, Frankfurt, Germany. Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies, 58(2), 277
297. Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68(1), 29
51. Biroli, P., Mourre, G., & Turrini, A. (2010). Adjustment in the Euro Area and regulation of product and labour markets: An empirical assessment. Economic Papers 428. Brussels, Belgium: European Commission. Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(1), 115
143.

Product Market Competition and Inflation Persistence

219

Bowdler, C., & Nunziata, L. (2007). Inflation adjustment and labour market structures: Evidence from a multi-country study. Scandinavian Journal of Economics, 109(3), 619
642. Correa-Lopez, M., Garcia-Serrador, A., & Mingorance-Arnaiz, C. (2010). Product market competition and inflation dynamics: Evidence from a panel of OECD countries. BBVA Working Paper No. 10/25. Banco Bilbao Vizcaya Argentaria, Argentina. Jaumotte, F., & Morsy, H. (2012). Determinants of inflation in the Euro Area: The role of labor and product market institutions. IMF Working Paper No. 12/37. International Monetary Fund, Washington, DC. Przybyla, M., & Roma, M. (2005). Does product market competition reduce inflation? Evidence from EU countries and sectors. ECB Working Paper No. 453. European Central Bank, Frankfurt, Germany. Roger, S. (2010). Inflation targeting turns 20. Finance and Development, (March), 47(1), 46
49. WEF (World Economic Forum). (Many issues). The global competitiveness report. Geneva, Switzerland: World Economic Forum.

UNLOCKING CREDIT Ike Mathur and Isaac Marcelin ABSTRACT Pledging collateral to secure loans is a prominent feature in financing contracts around the world. Existing theories disagree on why borrowers pledge collateral. It is even more challenging to understand why in some countries collateral coverage exceeds, for example, 300% of the value of a loan. This study looks at the association between collateral coverage and country-level governance and various institutional proxies. It investigates the economic implications of steep collateral coverage and sketches policy options to lower ex-ante asymmetric information and ex-post agency problems. Within this framework, should a lender collect the debt forcibly on default and liquidated assets fetch prices below outstanding loan values, the lender’s loss is covered through credit insurance, which would significantly reduce the need for steep collateral coverage. This proposal may increase level of private credit, investment and growth; particularly, in a number of developing countries where collateral spread is the main inhibitor of finance. Keywords: Collateral spread; credit; institutions; law and finance JEL classifications: G21; D43; E32

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 221
252 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096009

221

222

IKE MATHUR AND ISAAC MARCELIN

INTRODUCTION Although factors such as asymmetric information, adverse selection, exante private information accumulated through relational lending, ex-post frictions such as moral hazard, and borrowers’ specific risk characteristics have been essential to our understanding of the prominence and persistence of collateral in financing contracts, they sometime may be insufficient. Many important country-level governance factors, assumed to play an important role in the perpetuation of excessive collateral, have been inadequately investigated. For instance, why do more than half of the firms in emerging markets have no access to credit
a percentage that reaches about 80% in the Middle East and in Sub-Saharan Africa?1 Why has collateral coverage topped 204% in Ethiopia, over 308% in Laos; or why do firms in Zimbabwe, Burundi, and Benin have to, respectively, pledge assets worth up to 261%, 267%, and 306% of the value of the loan they seek?2 This pattern of overcollateralization may be economically (socially) non-optimal since it may thwart efficient resource allocation. The provision of collateral to access credit has been a prominent feature in financing contracts for a long time. Inadequate pre-loan investigation drives lenders’ skepticism about projects’ quality and borrowers’ trust- and credit-worthiness. This has turned into a mistrust, which has culminated into credit rationing. Scholars have proposed many theories ascribing the use of collateral to the monitoring of debtors’ misbehavior. In practice, not only does collateral allow lenders to discipline borrowers’ actions, it should ensure greater commitments to projects’ success while improving the probability of repayments. The persistence of collateral in credit markets, especially in loans to relatively creditworthy borrowers, remains puzzling. Whereas insufficient collateral coverage is one of the primary reasons firms are rejected when applying for bank credit,3 the World Bank Enterprise Surveys show that about 79.45% of all loans granted in over 100 countries during the years of 2002 through 2013 were backed by collateral.4 Economists have long been addressing the question of secured lending. Existing theories assert that collateral decreases moral hazard (Smith & Warner, 1979), bankruptcy costs (Scott, 1979, 1986), adverse selection (Chan & Thakor, 1987; Stiglitz & Weiss, 1981), risky behavior and asset substitution (Scott, 1986; Stulz & Johnson, 1985). A strand of research by Berger and Udell (1990, 1992), Booth (1992), and Angbazo, Mei, and Saunders (1998) investigates the relationship between credit risk, risk premium, and the provision of collateral. Earlier studies look at credit rationing (Bester, 1985; Stiglitz & Weiss, 1981). Bester (1985) and Stulz and

Unlocking Credit

223

Johnson (1985) concur that collateral enhances social welfare by limiting adverse selection and moral hazard problems. An incipient body of literature looks into credit risk and liquidity and how they relate to financial stability (Acharya, Shin, & Yorulmazer, 2010; Acharya & Viswanathan, 2011; Birge & Ju´dice, 2013; Gatev, Schuermann, & Strahan, 2009; Gorton & Metrick, 2011; He & Xiong, 2012; Berger, Imbierowicz, & Rauch, 2014); Jime´nez & Saurina, 2004). A study by McAleer, Jimenez-Martin, and Perez-Amaral (2013) examines how different risk management strategies performed during the 2008
2009 global financial crisis. Gnabo and Moccero (Forthcoming) suggest that risk in financial markets is a more powerful driver of monetary policy regime changes than variables typically suggested in the literature. Previously, Shin (2009) postulates that if securitization enables credit expansion through higher leverage of the financial system as a whole, by itself, however, it may not enhance financial stability if the imperative to expand assets drives down lending standards. A widespread opinion before the credit crisis of 2007/2008 was that securitization enhances financial stability by dispersing credit risk (Shin, 2009). Shim and von Peter (2007) analyze what triggers distress selling, why asset prices deteriorate, and how falling prices generate additional rounds of selling. Recently, Berger, Imbierowicz, and Rauch (2014) investigate the relationship between two major sources of bank default risk (1) liquidity risk and (2) credit risk for a sample of U.S. banks over 1998
2010 and find that both types of risk increase the probability of banks’ default. Kirkwood (2010) finds a negative indirect effect on banks’ net interest margin induced by mortgage originators competing with banks on the Australian market prior to the onset of the financial crisis. A related strand of literature looks at hypothecation and rehypothecation using financial assets as collateral in the dealers’ market (see Oehmke, 2014; Sakurai & Uchida, 2013). Specifically, Oehmke (2014) shows that creditor structure in repo lending involves a fundamental trade-off between risk sharing and inefficient “rushing for the exits” by competing sellers of collateral. An important unanswered question in the debate on secured lending is why collateral to loan value is so high in some countries. Understanding what drives high collateral coverage requires an examination of lenders’ perception of risk as part of their lending market. Lending environments are shaped by institutions and contracts enforcement mechanisms. The financial development literature holds that countries with weak enforcement mechanisms are financially less developed (Laeven & Majnoni, 2005; Marcelin & Mathur, 2014). The impetus for high collateral coverage seems

224

IKE MATHUR AND ISAAC MARCELIN

to be a reaction to weak contract enforcement environments, a predominant feature of financially less developed markets. Effective institutions and legal enforcement increase the likelihood that the lender can forcibly collect on default. Lenders view collateral as a purveyor of rights to discipline risky borrowers; the riskier is the lending environment, the greater the demand for this right. Thus, the higher is the collateral coverage. In other words, lenders choose a level of collateral coverage that affords adequate insurance against nonpayment risk and the liquidated value of seized assets. Contributions to the literature show that collateral requirements have reached prohibitive levels. While Hanedar, Broccardo, and Bazzana (2014) argue that pledging collateral is often an efficient solution to easing access to credit, an earlier research by Beck, Demirgu¨c¸-Kunt, and Laeven (2006) points out that collateral is one of the most significant obstacles to finance. Insufficient collateral is among the top reasons for difficulty in access to finance (Alvarez de la Campa, 2010; Beck et al., 2006). In fact, Menkhoff, Neuberger, and Rungruxsirivorn (2012) dispute the notion of borrowers’ lack of collateral in developing countries. In practice, high collateral coverage prevents owners of many valuable assets from accessing credit for the expansion of their enterprise to reap extra benefits from their capital stock holding. Empirical evidence identifies country-level governance, characterized by institutional efficiency, strength of legal rights, creditor rights protection, the legal infrastructure and the judiciary, as instrumental in enforcing financing contracts.5 Modigliani and Perotti (2000) theorize that when contract enforcement is weak, lenders emphasize more on collateral. In fact, market and institutional failures may impede firms’ access to formal finance. Beck and Demirgu¨c¸-Kunt (2006) maintain that well-defined property rights, effective contract enforcement, and firm access to finance characterize a business environment that is conducive to competition and private commercial transactions. Despite great strides in reforming collateral laws in recent years, collateral to loan value appears to be the most single roadblock to finance. Understanding how country-level governance drives secured credit practices may be useful to policymakers designing legal systems permitting/ preventing collateral use while seeking to unlock private credit for greater investment, growth, and economic development. It may be impractical to grasp the issue of prohibitive collateral outside of a country’s institutions and enforcement machinery, factors influencing, and perpetuating the needs for overcollateralization. Evidence reveals that improving access to

Unlocking Credit

225

finance may be achieved through enhancing institutional and legal environments (Beck et al., 2006; Beck, Demirgu¨c¸-Kunt, & Levine, 2003; La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1997; La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1998; Laeven & Majnoni, 2005; Marcelin & Mathur, 2014). Currently, investigations into collateral spreads are rather scarce. This study seeks to pinpoint whether there are gaps and limitations in the extant literature on collateral that pertain to country-level governance percolating through collateral practices. More specifically, the study adds to the extant literature by looking at the influence of institutions, legal rules, social, cultural, and national traits on collateral requirements across countries; by asking whether excessive collateral requirements are a reverberation of country-level governance. This inquiry allows us to identify some of the key national features with the potential to unlock private credit, facilitate or simply make possible some economic transactions for better allocation of credit. Broadly, the study looks at the empirical evidence on collateral and analyzes it within the context of country governance practices and legal framework. This study documents the existence of a funding gap, reflected in high collateral coverage, facing borrowers in countries with poor contract enforcement. This gap is wider in countries with high prevalence of corruption; particularly, French civil law countries, where the recovery rate in case of bankruptcy is lower and collateral coverage has a stronger negative effect on investment. The evidence suggests that higher collateral to loan ratio is not only a hindrance to external finance but also an indication of poor institutions and weak governance. Effective institutions ensure contract enforcement and risk mitigation, leading to greater likelihood of businesses’ success and lenders’ willingness to expand credit. High collateral coverage may lead to dead capital, which may result in low levels of private credit, investment, growth, and credit misallocation, which, in turn, may accelerate economic decline. This study proposes the adoption of a central of balance sheets (CBS) to reduce information asymmetries and provide credit insurance to lenders. In the CBS context, the lender would incur reduced costs of pre-loan investigation, which may prove very useful especially in countries where information institutions are lacking or ineffective. The rest of this study is structured as follows. The next section surveys the related literature on collateral. It looks at the association between firms’ ease of access to finance and collateral requirements. The section “National Attributes and Collateral” discusses the importance of institutional

226

IKE MATHUR AND ISAAC MARCELIN

development to overcome constraints to external finance, that is, steep collateral requirements and the risk effects of collateral on lending firms. The section “Power Information Theory of Credit and Collateral” looks at the power information theory of credit and collateral requirements. The section “New Perspectives and Policy Recommendations” proposes a new structure to deal with information asymmetric and ex-post agency problems along with credit insurance for credit providers. The section “Conclusion” concludes.

THEORY AND EVIDENCE ON COLLATERAL Access to Finance One of the most salient features of the secured credit regime is that collateral paves the way to finance while being an effective deterrent to borrowers’ risky behaviors, including asset-substitution and ex-post moral hazard as the fear that the pledged collateral will be grabbed and liquidated. As a result, lenders often require borrowers to pledge collateral to limit their potential losses due to inherent risk involved in financing contracts. In recent years, investigations into collateral have grown considerably in volume and in substance. Collateral is often studied as a palliative to information asymmetric, adverse selection and ex-post moral hazard problems, compounded with other obstacles to financing contracts such as pre-loan screening, monitoring, and enforcement. While collateral helps reduce these agency problems, it also has been the main hindrance to borrowers in many countries with poor institutions and weak contract enforcement. There are some agreements that collateral induces risky borrowers to exert greater efforts to avoid default lest they lose their collateral (Boot, Thakor, & Udell, 1991; Jime´nez, Salas, & Saurina, 2006). Aghion and Bolton (1992) and La Porta et al. (1998) concur that collateral can be used as an instrument to discipline borrowers. Although collateral is a widely observed debt contracting feature, the underlying motivation for collateral is not well understood (Berger, Espinosa-Vega, Frame, & Miller, 2011). In fact, there is no consensus regarding whether collateral proxies for borrower quality. Economists hold diverging views on why borrowers put up collateral to access finance.

Unlocking Credit

227

Whereas Rajan and Winton (1995) suggest that collateral, along with covenants, improves the creditor’s incentive to monitor, and that lenders require more collateral from observably riskier borrowers; Besanko and Thakor (1987), Boot et al. (1991), and Jime´nez et al. (2006) postulate that high-quality borrowers will pledge more collateral to signal their creditworthiness so that they can be separated from low-quality borrowers. Conversely, Booth and Booth (2006) argue that in the presence of asymmetric information, low-quality borrowers are more likely to offer collateral as a risk-reducing contractual feature. Other studies backing this view include (Angbazo et al., 1998; Berger & Udell, 1990; Berger, Espinosa-Vega, et al., 2011; Berger & Udell, 1990, 1995; Boot et al., 1991; Brick & Palia, 2007; Leeth & Scott, 1989; Orgler, 1970; Swary & Udell, 1988). Mann (1997) warns that given the likelihood that creditworthy companies are financially sophisticated and given the benefits attributed to secured transactions, in some contexts, any useful discussion of the pattern of secured credit must provide a coherent explanation for the general dearth of secured credit among companies with excellent credit ratings. Over several decades, researchers have been intrigued by the question of why some loans are secured and others unsecured. Since Jackson and Kronman (1979), it has been argued that the more able monitors will extend unsecured credit to capitalize on their comparative advantage, while less efficient monitors will take collateral to reduce monitoring burdens. Highlighting the inadequacy of signaling or screening explanations, Scott (1986) raises the possibility that information asymmetries still explain the persistent use of secured lending. Earlier, Stulz and Johnson (1985) predict that profitable projects will not be undertaken by a firm that can use only equity or unsecured debt to finance them but will be undertaken if they can be financed with secured debt. Later, Manove, Padilla, and Pagano (2001) put forward the lazy-bank theory suggesting that overcollateralization weakens a bank’s incentive to evaluate the profitability of an investment project. Furthering existing theories, Steijvers, Voordeckers, and Vanhoof (2010) conduct an inquiry into why some loans are granted without collateral and others are secured with business collateral or even personal collateral/commitment and report that relationships are significant determinants in deciding whether to pledge collateral or personal commitments in credit contracts. To the contrary, Levmore (1982) argues that the better monitors will take collateral as compensation for the tendency of less efficient creditors to “free ride” on their policing efforts. Previously, Stiglitz and Weiss (1981)

228

IKE MATHUR AND ISAAC MARCELIN

hypothesize a negative relationship between collateral and borrower risk because wealthier borrowers are assumed better able to pledge collateral and are likely less risk averse than poorer borrowers. In Mann’s (1997) model, creditors react to the provision of collateral by imposing higher interest costs on borrowers. Recent empirical evidence by Gottesman and Roberts (2007) shows that non-collateralized loans are associated with lower spreads. Using a sample of 500 Japanese borrowing firms, Ono, Sakai, and Uesugi (2012) find that borrowers with observably high riskiness are more likely to pledge collateral. The extant literature, though scanty, has diverse views on why borrowers grant collateral. Booth and Booth (2006) ask: Since collateral is traditionally viewed as risk-reducing contractual feature, why firms do not pledge collateral to the point where they have no more assets to pledge? The question of why borrowers would always secure their debts or pledge collateral to the limit of their wealth has been debated. For example, in an early study, Schwartz (1981) argues that rational borrowers would secure their debt to the greatest degree possible. By contrast, Kripke (1985) contends that in “the factual world,” in no uncertain terms, firms that can avoid giving secured debt elect to do so. In another study Aghion and Bolton (1992) argue that in the best of all worlds, firms would choose to obtain funding with “no-strings” attached; and that contractual incompleteness and wealth constraints are the source of the problem. Later, Mann (1997) suggests that the borrower will grant collateral to secure a loan only when the borrower believes that the net benefits of the most favorable secured transaction will be greater than or equal to the net benefits of the most favorable unsecured transaction. Mann further argues that after all, a borrower receives no direct benefit from an arrangement that enhances a lender’s ability to force the borrower to repay a loan. What undermines Schwartz and Booth and Booth’s view, however, is that in practice, it would be inconsistent with risk management and diversification principles for a firm to tie-up all of its valuable assets to secure a loan to purchase a project because beyond signaling effect and moral hazard, there are number of factors that could derail a project. Peculiar stresses including macroeconomic business cycles, political and policy uncertainties, and even sheer bad luck may undermine the borrower’s efforts in exploiting a venture and these events may overtake even the most committed and industrious management. In a rational world, if collateral eases access to finance, borrowers would stack-up unsecured debts until lenders start demanding collateral, which would be pledged in subsequent debt transactions up to the frontier of borrowers’ wealth.

229

Unlocking Credit

300

Regarding borrowers’ quality, one view is that collateral offers financing opportunities to high-risk borrowers who would not otherwise qualify for credit and these low-quality borrowers ought to provide more collateral for credit (Besanko & Thakor, 1987; Bester, 1985; Boot et al., 1991; Chan & Kanatas, 1985). Boot et al. (1991) surmise that because private information may either accentuate or retard the positive relationship between collateral requirements and borrower risk encountered with just moral hazard, higher collateral may be posted by either safer or riskier borrowers. Previous studies by Stiglitz and Weiss (1981, 1983) predict that in a competitive credit market, observably high-risk borrowers will choose a combination of high interest rate and low collateral; while high-quality borrowers will choose a combination of low interest rate and high collateral. Stiglitz and Weiss’ prediction has been confirmed in recent data (e.g., Berger, Espinosa-Vega, et al., 2011; Wang, 2010). Despite the growing number of studies on the use of collateral, there is no consensus on the degree to which collateral excludes borrowers from credit markets. Fig. 1 provides an overview of the relationship between collateral to loan value, which reaches over 300% in some countries (e.g., Laos and Benin) and firms’ access to finance. Although falling in a region with high collateral requirements, fewer than 20% of firms in Laos identify access to finance as a major constraint. In Benin, however, where

Lao PDR

Benin

Burundi Zimbabwe Yemen, Rep. Togo Dominican Republic Central African Republic Paraguay Namibia Vietnam Rwanda Jamaica Ethiopia Uruguay Sri Lanka Congo, Dem. Rep. Mauritania Bosnia and Herzegovina Mali Nicaragua Macedonia, FYR Bolivia Ecuador Costa Rica CambodiaMexico Georgia AlgeriaCameroonNiger Morocco ArgentinaUganda Ukraine Panama Peru Malawi Kyrgyz ElRepublic Salvador Hungary Russian Albania Federation Chile Honduras Zambia Latvia Guatemala Bulgaria Ireland Croatia Poland Colombia Burkina Faso Romania Armenia Chad Nigeria Korea, Rep. Greece Thailand Botswana Slovak Republic Spain Ghana Jordan Slovenia Senegal Germany Azerbaijan India Uzbekistan Belarus Syrian Arab Tanzania Republic Kazakhstan Kenya Turkey Portugal Lithuania Czech Republic Egypt, Arab Rep. Madagascar South Africa Angola Mozambique

Nepal

100

200

Philippines

Pakistan Malaysia

Lesotho Sierra Leone

Brazil

MongoliaCongo, Rep.

Côte d'Ivoire

0

Indonesia

0

20 40 60 Percent of firms identifying access to finance as a major constraint

80

Value of collateral needed to secure a loan

Fig. 1.

Collateral Requirements and Firms’ Access to Finance.

230

IKE MATHUR AND ISAAC MARCELIN

collateral to loan value has topped over 306%, more than 75% of firms identify access to finance as a major constraint.6 However, Fig. 2 reveals that the proportion of firms using banks to finance investments is about the same as in Laos and in Benin. From a public policy standpoint, countries should strive to fall in the third region of Fig. 1 to relieve some burdens on firms seeking financing while facilitating better resource redeployment.

Collateral and Investment

300

A collateral channel has been identified in studies as far back as Fisher (1933). A decline in the value of the firm’s collateral value reduces its creditworthiness, debt, and investment capacity (Kashyap, Scharfstein, & Weil, 1990; Kiyotaki & Moore, 1997). Gan (2007) points to a feedback loop between collateral and credit cycles caused by a decline in the value of collateralized assets. In general, the benefits generated by pledging collateral include increased financial resources to finance firms’ expansion, but Berger, Espinosa-Vega, et al. (2011) stress that the common application of collateral may also have macroeconomic consequences. A negative association between collateral requirement and the percent of firms using bank loans to finance investment has been detected and

LaoBenin PDR

100

200

Burundi Zimbabwe Nepal Yemen, Rep. Philippines Togo Dominican Republic Central African Republic Paraguay Namibia Vietnam Rwanda Jamaica Ethiopia Uruguay Congo, Dem. Rep. Mauritania Sri Lanka Mali Bosnia and Herzegovina Nicaragua Macedonia, FYR Bolivia Costa Rica Ecuador Georgia Cambodia Algeria Uganda Cameroon Mexico Argentina NigerMorocco Peru Lebanon Ukraine Panama Malawi Kyrgyz Republic El Salvador Hungary Russian Federation Chile Honduras Albania Zambia Bulgaria Latvia Ireland Colombia Poland Croatia Guatemala Burkina Faso Romania Nigeria Armenia Chad Korea, Rep. Greece SlovakBotswana Republic Ghana Spain Jordan Senegal Uzbekistan Azerbaijan Germany India Slovenia Belarus Tanzania Syrian Arab Republic Kenya Kazakhstan Turkey Portugal Lithuania Czech Republic Egypt, Arab Rep. Madagascar South Africa Angola Mozambique Pakistan Sierra Leone Côte d'Ivoire Congo, Rep.

Lesotho

Thailand

Brazil Malaysia

Mongolia

0

Indonesia

0

10 20 30 40 Percent of firms' investment finance with bank credit

50

Value of collateral needed to secured a loan

Fig. 2.

Collateral Requirements and Investment Financing.

231

Unlocking Credit

illustrated in Fig. 2. Indeed, Safavian (2008) emphasizes that lack of collateral is an important factor in financial exclusion as numerous credit applications are rejected by banks for lack of collateral to secure the loan while many firms never apply for loans with the key deterrent being the collateral requirements to obtain loans. In addition, Engelhardt and Regitz (2008) concur that collateral requirements often preclude businesses from participating in credit contracts: Businesses without sufficient collateral will not be able to access bank finance for many types of investment loans, and even for short-term working capital lines of credit. This inability to access finance should hinder firms’ investment in their expansion. Using Chinese data, however, Allen, Qian, and Qian (2005) find that high collateral requirements have not hampered the private sectors’ ability to grow. The ability to pledge collateral may lead to better outcomes when leading to higher credit volume, enhanced loan and capital allocations. A study by Almeida and Campello (2007) supports a similar hypothesis stressing that the ability to pledge collateral leads to more borrowing, facilitating additional investment in those assets when borrowers have imperfect access to credit. Alvarez de la Campa (2010) finds that constrained access to finance remains among the top limitations on private sector’s growth in the developing world while removing barriers to a wide range of financial services can unleash private enterprise productivity. Overcollateralization can lead to inefficient credit allocation, credit rationing, underinvestment and lower economic growth. Still, in the Steijvers et al. (2010) framework, underinvestment problems originate where investment projects with a low positive net present value (NPV) and low risk are rejected because only unsecured debt financing is available. However, Wang (2010) predicts an inefficient investment outcome due to collateral constraints. Martin (2009) offers a framework in which a fall in aggregate investment is associated with collateralization. Bester (1985) considers the role of collateral for screening in environments of adverse selection with indivisibilities in investment. Unlike in Besanko and Thakor (1987), in which adverse selection leads to overinvestment, Bernanke and Gertler (1989, 1990) and Kiyotaki and Moore (1997) expect the collateral channel to lead to underinvestment due to collateral losses.

Collateral and Risk The key advantage associated with the granting of collateral is the enhancement of the lender’s ability to restrain the borrower from engaging in risky

232

IKE MATHUR AND ISAAC MARCELIN

conduct that (in the lender’s view) decreases the borrower’s ability to repay the loan (Mann, 1997). However, to the extent that collateral ties up assets that might otherwise be put to more productive uses, it imposes opportunity costs on borrowers. Whereas collateral helps solve agency issues between borrowers and lenders including risk-shifting (Aghion & Bolton, 1997; Holmstrom & Tirole, 1997), capital goods’ prices may fall well below their initial value after they have been pledged as collateral. Scholars have looked into the potential negative impacts of collateral. Chofaras (2004) stresses that apart from the likelihood of a decrease in the value of collateral, credit risk is always present. Later, Gai, Kondor, and Vause (2006) postulate that if the collateral asset is also used in production, a feedback loop between aggregate output and the value of collateral emerges. The borrower loses valuable economic assets when transferring them to the lender to secure a loan, which otherwise might enhance his ability to generate returns necessary to pay back the lender (Haselmann, Pistor, & Vig, 2010). Particularly, Wang (2010) argues that the pledge of collateral can possibly influence an enterprise’s production decisions or even its survival chances. Berger, Frame, and Ioannidou (2011) find that the risk-collateral channels depend on the economic characteristics and types of collateral. The collateral channel may provide some insights on why collateral coverage is so high in some countries. Lenders can better control borrower risk if they know they will be able to seize collateralized assets, or credibly threaten to take these assets, ex-post, in default (Qian & Strahan, 2007). When a debt is secured by an asset whose value falls below the debt outstanding value; if the project financed by this obligation cannot generate adequate cash flows to meet repayments (capital and interests) schedules; and the borrower is unable to restructure the debt or raise finance through other channels; then pledged collateral is likely to be seized and auctioned off on the spot market, and the creditor may only have the right to a deficiency judgment. In Krishnamurth (2003), the collateral channel implies that bad times for the economy will also be times when the liquidation value of collateral will be low since potential buyers of these assets will be cash-strapped. As a result, this can lead to low debt capacity, which can then turn bad times into even worse times as limited financing forces firms to curtail production, exacerbating the bad times, causing collateral values to fall further. In some scenarios, such an externality may result in a broader self-fulfilling financial crisis. Whereas Alvarez de la Campa (2010) argues that welldesigned secured transactions systems contribute to robust financial systems by promoting credit diversification, and allowing financial institutions

233

Unlocking Credit

to make informed credit decisions on collateral lending; Houben and Slingenberg (2013) stress that the increased collateral reliance in financial transactions has important implications for the structure of the financial system as it heightens interconnectedness within the system. Kiyotaki and Moore (1997) present a theory on how common shocks to credit constrained firms are amplified through changes in collateral values and transmitted as fluctuations in output. Introducing the theory of incomplete hedging, Krishnamurth (2003) extends the Kiyotaki and Moore’s (1997) model by arguing that the supply of hedging available in the economy is constrained by the aggregate value of collateral. Particularly, Krishnamurth (2003) argues that when this same collateral is a productive input for business, then aggregate conditions have a direct effect on the operations of business by altering collateral values and therefore the debt capacity of individual firms. Berger, Espinosa-Vega, et al. (2011) concur with Krishnamurth’s view by arguing that the borrower may also suffer fluctuations in their credit availability as the values of their assets vary. Williamson (1988) highlights the link of redeployability of productive assets through the channel of debt capacity and liquidation value of assets as a good substitute for debt finance. However, when mismanaged, the borrowing firm will be unable to repay the debt, and creditors will take the assets away and redeploy them. Shleifer and Vishny (1992) pinpoint an illiquidity conduit that operates through assets specialization and non-redeployability whereby assets fetch prices well below their values when liquidated. The authors further argue that when firms have trouble meeting debt payments and sell assets (or liquidate them), the highest value potential buyers of these assets are likely to be other firms in the same industry, which may have trouble making their debt payments at the time assets are put up for sale as long as the shock that causes the seller’s distress is industry- or economy-wide. Liquidity of the collateral asset is shown to play a key role in amplifying the financial cycle (Gai et al., 2006).

Borrowers’ Attributes There is abundant evidence on the degree to which borrowers’ characteristics such as size, age, and legal form of the firm affect the use of collateral. Models related to asymmetric information, adverse selection, and moral hazard expect smaller firms to more likely pledge collateral to access finance as these firms are more opaque because as Menkhoff et al. (2012) suggest, lenders hold less information on these firms’ investment

234

IKE MATHUR AND ISAAC MARCELIN

opportunities and managerial capabilities. The opacity problem is exacerbated for startups since they have no prior relationships with lending practitioners. To circumvent this inconvenience, small and young firms along with startups may be more inclined to pledge collateral to signal their quality and ensure lenders of reduced risk of moral hazard and adverse selection. As a risk-reducing feature, high collateral coverage may lock this class of firms out of credit markets and effectively hinder their growth prospects. Demand for collateral is a fundamental cost of financing in many models of financial constraints (Modigliani & Perotti, 1997a, 1997b). Berger and Udell (1998) confirm that small businesses have greater difficulty to communicate reliable information to lenders about their real status and performance. Using a sample of 10,000 firms from 80 countries, Beck et al. (2006) find that older, larger, and foreign-owned firms face less financing obstacles. A survey of the empirical evidence on SMEs’ access to finance by Beck and Demirgu¨c¸-Kunt (2006) reveals that, with fixed transaction costs and information asymmetries, small firms, typically more opaque with less collateral to offer, face higher transaction costs along with higher risk premiums. This explains, to some extent, why the relationship between SMEs and banks is often characterized by asymmetric information, adverse selection, and moral hazard problems (Steijvers et al., 2010). Although borrower-specific variables are more important than country specific variables in determining collateral requirements on loan contracts (Hanedar et al., 2014), Beck, Demirgu¨c¸-Kunt, and Maksimovic (2008) show that small firms and firms in countries with poor institutions use less external finance. Still, if, as maintained in the extant literature, lenders adjust interest rates, maturity, and loan size to borrowers’ perceived riskiness, then the widespread high collateral coverage inhibiting access finance, across countries, appears to impose on borrowers some burdens unexplained in the variation of borrowers’ specific risk. Since wealthier and older firms, less opaque, instill more confidence to credit providers, borrower’s organizational structure, and size, rather than financial sophistication, may drive the need to pledge collateral as a multiple of loan value. This gap has yet to be filled in empirical research on collateral and institutions.

Collateral, Loan Characteristics, and Lending Relationship Collateral goes in tandem with the interest rate, maturity, size, and possible covenants (Jime´nez et al., 2006). One way to reduce lenders’ risk exposure, suggested in the literature, is through debt covenants. Because secured lenders can focus on particular assets, it is cost-effective for borrowers to

Unlocking Credit

235

allow those lenders to impose stringent, specific covenants that are effective in protecting the particular assets on which the lender has a lien; the covenants operate at a business-wide level, however, they do not prevent the borrower from engaging in several other types of risky activity that can reduce a lender’s chances of repayment (Mann, 1997). Many small and young firms with limited collateral may build a relationship with lenders through short-term loans. In fact, Steijvers and Voordeckers (2009) highlight that by entering into short-term loans the borrower allows the lender to generate information about the firm without engaging in any long-term contracts. Granting credit on a short-term basis would allow lenders ample time to collect information on small and new borrowers over time to reduce ex-ante information asymmetric problems. Borrowers, however, would need to frequently renegotiate their loan, giving lenders an opportunity to identify potential issues with borrowers’ ability to remain creditworthy. Obviously, revolving credit is riskier for borrowers since these loans may not be renewed while macroeconomic fluctuations and changes in borrowers’ situation may alter their credit condition. The association between collateral and lending relationship has been investigated in a number of papers and remains puzzling. Whereas a series of papers by Degryse and Van Cayseele (2000), Jime´nez et al. (2006), and Menkhoff, Neuberger, and Suwanaporn (2006) finds a mixed or an insignificant association between collateral and lending relationship; information asymmetric models by Berger and Udell (1995), Harhoff and Korting (1998), and Chakraborty and Hu (2006) predict a negative association between ex-ante information and close ties with lenders. Other studies report a positive correlation between collateral and prior credit experience with lenders (Elsas & Krahnen, 2000; Machauer & Weber, 1998; Ono & Uesegi, 2009). A study by Godlewski and Weill (2011) tests whether the extent of information asymmetry affects the relationship between collateral and loan spread. In practice, credit quality is private information accumulated and bettered through a long relationship between lenders and borrowers. Borrowers’ credit history is known to the lender through this relationship, thereby mitigating asymmetric information problems. When this relationship matures, if the borrower is required to over-collateralize, then it may be that the lender is more concerned with ex-post moral hazard problems. Using a sample of 300 Mexican debtors, La Porta, Lopez-de-Silanes, and Zamarripa (2003) report that firms with related ties to lenders pledge less collateral. Boot et al. (1991) argue that under moral hazard, the need to request collateral grows with collateral requirements as well as increased borrowing costs, reduced size, and maturity.

236

IKE MATHUR AND ISAAC MARCELIN

Transaction costs generally involve the costs of defining property rights and those of enforcing contracts
including costs of acquiring and processing relevant information (Aggarwal & Goodell, 2009); the costs of monitoring the assets; and any enforcement/disposal expenses in the event of repossession (Leeth & Scott, 1989). Booth and Booth (2006) show that the presence of multiple lenders, and the associated risk and/or information spillovers may play a role in whether collateral is pledged in bank loans; and as a result firms may be motivated to pay a higher cost to borrow unsecured loans to reap the benefit of generalized bank monitoring.

Collateral and Capital Market Development Where capital markets are underdeveloped, collecting information on borrowers and transforming this information may be costly due to a lack of technology. Resolving this information asymmetric problem may compel lenders to ask for steep collateral coverage, which at the same time, restricts access to finance by imposing a binding constraint on financing. This constraint binds even harder in undeveloped financial markets where a slew of capital stocks cannot be pledged. Despite advances in information technologies and the availability of enhanced mechanisms for information gathering, project screening and monitoring, overreliance on collateral to assess creditworthiness of finance seekers is predominant. Liberti and Mian (2010) look at how financial development affects collateral requirements for SMEs from 15 developing countries, and provide some evidence that institutions promoting financial development ease borrowing constraints by lowering the collateral spread and shifting the composition of acceptable collateral toward firm-specific assets. The authors also report the cost of collateral in terms of value of collateral and the specificity of assets pledged as collateral as negatively related with the level of financial development. Given the evidence on financial development and country-level institutions (governance), their results suggest a relationship between collateral and financial development that may be channeled through countries’ institutions.

NATIONAL ATTRIBUTES AND COLLATERAL It is difficult to draw general judgments on national attributes such as trust, culture and other national traits and the use of collateral in financing

237

Unlocking Credit

contracts. Every commercial transaction has within itself an element of trust and much of the economic non-development in the world can be explained by lack of mutual confidence (Algan & Cahuc, 2010). In spite of rising interest and national differences and access to finance, there is little prior research on why lenders withhold their trust to rely extensively on collateral and how this overreliance is related to national cultures. For instance, Feder, Tongroj, and Tejaswi (1988) underscore the difficulty in enforcing financing contracts in developing countries especially when land is pledged as collateral due to political, legal, and social considerations. Provision of credit on a secured (unsecured) basis may vary not only by a borrower’s business economics, but also by the legal, social and cultural environments, with varying impacts on access to credit. Despite the importance of collateral in financing contracts, there is virtually little work relating national attributes such as trust, cultural traits, political stability, regulatory quality, inter alia, to the use of collateral. Financial frictions including uncertainties around financial contracts, weak enforcement machinery, lack of trusts between parties, may compel agents to confront a basic trade-off between sub-optimal supply of credit and insurance against default. A recent study by Menkhoff et al. (2012) shows that lenders, in Northeast Thailand, enforce collateral-free loans using third-party guarantees and through relationship lending. Aggarwal and Goodell (2009) assume that the efficacy and efficiency of overcoming contracting costs depends not only on the legal environment, but also on ethical and other informal conventions, and social and cultural values. Whether these national traits influence the use or overuse of collateral in financing contracts remains, thus far, an open question of empirical implications. Lenders’ propensity to trust reflects their expectation with respect to the borrower’s credibility and commitment to the project success. Trust may play an important role for lenders in setting collateral to loan value. In trusting countries, collateral to loan value is expected to be relatively low.

Institutions, Court Efficiency, and Collateral While collateral is used across all income groups and legal settings, countries’ institutions may influence lenders’ expected payoffs from providing capital. Whether lenders suffer a loss due to a decline in assets’ value may depend on the ability of the court to swiftly resolve litigations arising among claimants, and enforce financing contracts with efficacy. Fleisig, Safavian, and de la Pen˜a (2006) highlight that the faster and more cheaply property can be seized and sold, the more value it has as collateral. The

238

IKE MATHUR AND ISAAC MARCELIN

300

authors note that even in advanced industrial countries with a modern judiciary, completing a civil suit can take up to three years
far too long to allow perishable or rapidly depreciating property to serve as collateral. Differences in the character of legal systems and the effectiveness of the courts in protecting creditors’ rights are reflected in financing contracts in terms of higher cost of credit, shorter maturity, smaller loans, and higher collateral coverage. We conjecture that better legal protection conveys stronger rights to creditors allowing them to lower collateral requirements. Fig. 3 illustrates that higher collateral coverage is associated with lower recovery rate subsequent to bankruptcy. While it takes on average 1,420 days7 (India) or 1,356 days (Columbia) to enforce a judgment, it takes on average 264 days (Hong-Kong) or 395 days (Germany) to enforce financing contracts. In fact, it may take up to 5.7 years8 (Philippines), 4.5 years (Indonesia) and 4 years (Brazil), compared to 0.4 years (Ireland), 0.6 years (Japan), and 0.8 years (Canada) for creditors to recover part or all of their credit in case of bankruptcy. As a matter of policy reform, there has been a tendency to bypass the court in a number of countries to allow out-of-court (settlement) enforcement of financing contracts (Alvarez de la Campa, 2010). In practice, these reforms may signal a failure of formal enforcement channels, substituted by informal enforcement mechanisms, which may operate through threats and abuses.

Lao PDR

Benin

200

Burundi Zimbabwe Nepal Yemen, Rep. Philippines Dominican Republic Togo Central African Republic Paraguay Namibia Vietnam Rwanda

100

Jamaica Ethiopia Uruguay Congo, Dem. Rep. Mauritania Sri Lanka Mali Bosnia and Herzegovina Nicaragua Macedonia, FYR Bolivia Ecuador Costa Georgia Rica Cambodia Algeria Uganda Cameroon Morocco Mexico Niger Peru Argentina UkraineKyrgyz Panama Malawi Republic El Salvador Hungary Chile Lebanon Russian Federation Honduras Albania Latvia Bulgaria Poland Colombia Croatia Guatemala Burkina Faso Zambia Romania Nigeria Armenia Thailand Chad Greece Botswana Slovak Republic Ghana Jordan Slovenia Senegal Uzbekistan India Azerbaijan Belarus Tanzania Syrian Arab Republic Kenya Kazakhstan Turkey Lithuania Czech Republic Egypt, Arab Rep. Madagascar South Africa Angola Mozambique

Ireland Korea, Rep. Spain Germany Portugal

Brazil Pakistan Lesotho Malaysia Sierra Leone Côte d'Ivoire Congo, Rep. Mongolia

0

Indonesia

0

20

40 60 Debt recovery rate in bankruptcy

80

Value of collateral as percent of a loan

Fig. 3.

Collateral Requirements and Recovery Rate in Bankruptcy.

Unlocking Credit

239

Without improvements in the judiciary, those efforts, aiming at protecting creditors’ rights to increase the flow of credit, may be detrimental to borrowers. Although out-of-court enforcement may speed up the recovery process, in countries with weak institutions and culture of corruption, however, extra-judicial settlements may wreak havoc if performed with the help of police officers or government officials even when the law is couched in languages protecting the interests of the debtor. While lending volume may respond positively to out-of-court enforcement, Haselmann et al. (2010) note that not all legal change is equally effective. Private enforcement mechanisms are necessarily less efficient than a reliable common legislation, since agents unconnected through a scheme are unable to transact (Modigliani & Perotti, 1997a, 1997b). Several studies concur that where contracts enforcement is weak, and the judiciary ineffective, collateral requirements are tougher (e.g., Modigliani & Perotti, 2000; Rajan, 1992). Poor quality of the judiciary may amplify lenders’ fear of collecting their claims on default. The efficacy of the legal framework may reduce systemic risk lowering thus collateral coverage, that is, reducing barriers to finance. In fact, Qian and Strahan (2007) warn that when lending to a company in emerging economies, banks must assess not only the credit quality of the borrower but also the risks due to weak institutions and laws. Further, the authors highlight that in the presence of better legal protection during bankruptcy and reorganization lenders are more willing to extend credit on favorable terms ex-ante. Where there are gaps between the laws and their application, property pledged as collateral might not have clear title. To facilitate the flow of credit, many countries have introduced collateral registries (Love, Perı´ a, & Singh, 2013). These structures record and provide relevant information on security interests created between parties in secured lending transactions (Alvarez de la Campa, 2010). This is especially important when there are competing claims against the same asset pledged as collateral. Lenders operating in ineffective legal frameworks may demand higher collateral coverage to extend credit. In other words, the enforcement machinery or the judiciary may play an important role in reducing systemic risk for lenders to lower collateral thresholds. Although the lines between institutions and legal machinery are at best blurry, several contributions to the literature explicitly assess the impact of institutions on access to finance. There is broad consensus that legal institutions play a determining role in shaping financial contracting. An important strand of literature, the law and finance theory, documents a positive

240

IKE MATHUR AND ISAAC MARCELIN

relation between institutions and access to finance (Demirgu¨c¸-Kunt & Huizinga, 1998; La Porta et al., 1997, 1998; Laeven & Majnoni, 2005; Levine, 1999; Marcelin & Mathur, 2014; Djankov et al., 2007). The positive correlation between institutions and collateral coverage implies that where institutions are weak, collateral coverage is higher. While it is an empirical question as to the extent to which ineffective institutions drive prohibitive collateral coverage, the institutional and legal infrastructures provide the statutory framework governing financing contracts and security interests. Fleisig et al. (2006) stress the need for reforming collateral laws to expand access to finance, while Safavian (2008) argues that many legal systems place unnecessary restrictions on creating collateral, leaving lenders unsure whether a loan agreement will be enforced by the courts. Many studies on collateral focus on reforming collateral laws to allow movable assets, machinery, and/or intangible assets to be pledged to increase access to external finance (Dahan & Simpson, 2008; Fleisig, 2008; Lopez-deSilanes, 2008; Safavian, 2008). An important body of studies emphasizes the role of creditor and property rights (Bae & Goyal, 2009; Besley & Ghatak, 2008; Esty & Megginson, 2003). The legal efficiency of secure transactions has been considered in Dahan and Simpson (2008). Reforming the systems governing movable property may transform some of a country’s stock of dead capital into productive capital. There is a need to address institutional weaknesses causing lenders to withhold their trust and set collateral coverage beyond the reach of would-be borrowers to reduce some barriers to finance. Well-functioning secured transaction systems enable businesses to use their assets as security to generate capital for expansion (Alvarez de la Campa, 2010). The inability to post collateral suggests poor creditworthiness, often associated with tougher lending requirements or steeper barriers to finance. As a consequence of weak institutions, Fleisig et al. (2006) note that collateral provides insurance against risk such as strategic default because sale of the good taken as collateral will provide funds for the lender. In a corrupt judicial system, proceeds from collaterals may differ significantly from the intrinsic value of the seized asset. Auctioning off collateral entices numerous fees including court costs, appraisal fees, legal representation and auctioneers. Where these fees are arbitrarily set, often without transparency, the economic value of collateral may be significantly reduced. The spread in collateral coverage and the relative ease of access to finance is amplified along corruption practices across countries. In the law and finance literature, institutions have been proxied by control of corruption in a number of studies (Aggarwal & Goodell, 2009; Bae & Goyal,

241

300

Unlocking Credit

Lao PDR

Benin

100

200

Burundi Zimbabwe Nepal Yemen, Philippines TogoRep. Dominican Republic Central African Republic Paraguay Vietnam Namibia Rwanda EthiopiaJamaica Uruguay Congo, Dem. Rep. Mauritania Sri Lanka Mali Bosnia and Herzegovina Nicaragua Macedonia, FYR Bolivia Ecuador Georgia Costa Rica Cambodia Uganda Cameroon Morocco Mexico Argentina Niger Algeria Peru Ukraine Panama Malawi Lebanon Kyrgyz Republic El Salvador Hungary Russian Federation Chile Honduras Albania Zambia Bulgaria Latvia Ireland Poland Colombia Burkina FasoCroatia Romania Nigeria Guatemala Armenia Chad Rep. Thailand GreeceKorea, Botswana Slovak Republic Ghana Spain Jordan Slovenia Senegal Uzbekistan Azerbaijan India Germany Belarus Syrian Arab Republic Tanzania Kazakhstan Kenya TurkeyLithuania Portugal Czech Republic Egypt, Arab Rep. MadagascarSouth Africa Angola Mozambique SierraPakistan Leone Côte d'Ivoire Congo, Rep.Mongolia

Brazil Lesotho Malaysia

0

Indonesia

-2

-1

0 Control of corruption

1

2

Value of collateral needed as a percent of a loan

Fig. 4.

Collateral Requirements and Control of Corruption.

2009; Fernandes & Kraay, 2007; La Porta, Lopez-de-Silanes, & Shleifer, 1999). Fig. 4 shows that in countries with a higher score on the control of corruption index (least corrupt), collateral coverage is lower. The more corruption is contained, the easier it is for firms to access finance. Since collateral is associated with better loan terms, that is, lower interest rates, longer maturity, and larger size, the inability of firms to post collateral due to higher collateral coverage may be a hindrance to private credit expansion, investment and growth communicated through poor institutions. The institutions literature holds that English and French legal institutions are at the opposite sides of the institutional spectrum. Whereas the common law heritage is credited with affording the most effective and least corrupt institutions, the civil law system is characterized by ineffectiveness and crippling corruption.9 While the literature has supplied abundant evidence that firms face steeper barriers to finance in civil law countries, there is less direct evidence on whether poor institutions influence lenders’ propensity to require more collateral coverage. Legal traditions and principles need not constitute a barrier to the introduction of modern secured transactions systems (Alvarez de la Campa, 2010). Nevertheless, we observe that collateral coverage is steeper in countries with weak institutions, that is, 164% (civil law), 152% (common law), 145% (German legal heritage), and 120% (socialist legal system).

242

IKE MATHUR AND ISAAC MARCELIN

English

French

German

Socialist

40

60

80

100

40

60

80

100

Financial contracting is by nature particularly sensitive to the legal framework in which it takes place (Modigliani & Perotti, 1997a, 1997b). The basic function of a secured transactions law is to allow the creation of a security right over assets, which can be seized and whose proceeds are applied toward satisfaction of a claim to the extent of the latter’s priority; if there is no right to enforcement, the law fails to achieve it basic legal function (Dahan & Simpson, 2008). When institutions are poor and ineffective, financial contracts unenforceable, court effectiveness impeded by politics, and formal enforcement mechanisms overshadowed by informalities, lenders will be more reluctant to grant credit and this should be noticeable in collateral coverage. Acemoglu and Johnson (2005) and Fernandes and Kraay (2007) delineate property rights institutions from private property rights institutions (often referred to as contracting rights). When assessing the types of institutions unlocking obstacles to external finance, such a distinction may be useful. In these foregoing investigations, corruption is used as a proxy not only for institutions but for property rights protection as well. Beck et al. (2008) find that protection of property rights increases external financing of small firms significantly more than large firms. Secured lending increases with weak property rights protection may increase with collateral coverage. Fig. 5 shows a negative correlation between collateral to loan value and the

0

20

40

60

80

0

20

40

60

80

Percent of firms with bank loans Proportion of loans requiring collateral

Fig. 5.

Proportion of Loans Requiring Collateral by Legal System.

Unlocking Credit

243

percent of firms with a bank loan. If this correlation is weak in common law, German and socialist legal systems, it appears to be stronger in civil law system corroborating and underscoring the relative ineffectiveness of the civil law system. Differences in property rights and contracting rights institutions may explain some of the variations in collateral coverage across countries. Reducing obstacles to finance needs good institutions tailored within effective investor protection frameworks. The channels through which improvements in the institutions percolate into the economy are varied. The legal system of a country (common law, civil law, or other) provides the structure within which secured transactions occur (Alvarez de la Campa, 2010). A loan agreement between a bank and its customer is a contract terms that includes the interest rate and all non-price variables (Harris, 1974). Should these non-price factors incorporate risk imparted by institutional quality, many projects with positive NPVs may remain unfunded in countries with ineffective institutions. Policies aimed at removing barriers to external finance should focus on institutions that matter for lowering collateral coverage, such as promoting effective courts to enforce financing contracts, and reducing the level of corruption. Marcelin and Mathur (2014) stress the importance for secured lenders to have priority claims in bankruptcy proceedings to commit important financial resources to firms’ long-term financing needs.

POWER INFORMATION THEORY OF CREDIT AND COLLATERAL A large body of literature focuses on the use of collateral by lenders to ably monitor their borrowers (Scott, 1986), mitigate lack of information on borrowers (Igawa & Kanatas, 1990; Inderst & Mueller, 2007), decrease bankruptcy cost and moral hazard (Scott, 1977, 1979; Smith & Warner, 1979), screen and monitor (Bester, 1985), and reduce information asymmetries and adverse selection (Chan & Thakor, 1987). Nevertheless, Alvarez de la Campa (2010) observes that even in the most advanced jurisdictions where reliable credit information and wide range of financial products are available, only the largest and best connected business can obtain unsecured loans. Recent studies on private and public information agencies and the incidence of collateral highlight the use of credit scoring technology by lenders to reduce ex-ante private information problems (Berger, Espinosa-Vega,

244

IKE MATHUR AND ISAAC MARCELIN

et al., 2011; Menkhoff et al., 2012). Collateral requirements are stringent in developing countries because lending environments involve opaque information (Hainz, 2003; Menkhoff et al., 2006). In addition, the power of information theory of credit, emphasized in a number of studies (Ang, 2013; Djankov, McLiesh, & Shleifer, 2007; Stiglitz & Weiss, 1981), underscores the importance of information for lenders to distinguish between honest and dishonest borrowers when setting lending and collateral premiums into the market interest rate. Based on the argument that financial intermediation will be higher if repayment can be more easily enforced through obtaining collaterals and gaining control over firms, this theory shows the importance of creditor protection and information sharing in reducing financial market frictions and deepening financial systems (Ang, 2013).

NEW PERSPECTIVES AND POLICY RECOMMENDATIONS The conventional view on secured lending is that collateral removes barriers to credit finance. Despite disagreements in the field, there are abundant insights from finance theory suggesting riskier borrowers face steeper collateral requirements. Without collateral offers, financing opportunities fail to materialize for a number of borrowers. In spite of new reforms to broaden the scope of assets that can be pledged, collateral coverage remains prohibitive. One reason appears to be that lenders fear that seized assets may fetch prices well below their values in best use when liquidated
while these proceeds may be insufficient to cover outstanding loans. Collateral and credit registries may help reduce information asymmetric problems in terms of the ability of the borrower to pledge collateral. However, collateral is still an imperfect and controversial ex-ante proxy for borrower’s credit quality. Under the current collateral regime when lenders cannot remove the potential for losses on liquidation, the only remaining avenue for paring their losses may be through protecting themselves with high collateral coverage. In a recent study, Marcelin and Mathur (2014) suggest a structure, a CBS, to cope with information asymmetric problems. Issues related to adverse selection and ex-post moral hazard may be mitigated by implementing a CBS. This structure would collect firms’ financial statements to build a database with participating firms opting in voluntarily. It would assign a score
reflecting the probability of

245

Unlocking Credit

bankruptcy
to those firms. When lenders face credit decisions, they would request from the CBS a score reflecting borrower’s creditworthiness upon a service fee, which should be affected toward a general guarantee fund to provide credit insurance for a lender granting credit to a borrower participating in the CBS. A predetermined collateral threshold given a credit score should guide credit decisions, and in the event of bankruptcy, the collateral may be seized and liquidated. If the liquidated asset fetches a price below the outstanding loan, the credit insurance would cover the balance. This proposed framework for secured transactions may increase a lender’s payoff from a loan since the potential loss is covered through the credit insurance, lessening thus the need for steep collateral coverage. Generally, the lender bears the majority of the risk involved in financing contracts.10 One of the main reasons creditors set high collateral coverage is the fear that the differential between the collateral’s value and the amount that the lender would recover on the loan after collateral liquidation may be too low. This gap would be practically filled by the credit insurance. In fact, the only cushion for the lender against downside risk is to have ample credit insurance from adequate collateral coverage. The most effective way to mitigate lenders’ risk aversion and inducing them into setting lower collateral coverage is to limit their potential losses when assets are grabbed and liquidated. Lenders would be less reluctant to provide credit given low collateral coverage (even without collateral) as risk of nonpayment is perceived to be insignificant. Implementation of a CBS may increase level of private credit, investment and growth; particularly, in developing countries where collateral coverage is the main inhibitor to finance. In developing countries with weak institutions, a CBS would accomplish several objectives including reducing information asymmetric and adverse selection problems while alleviating the need for excessive collateral
which could unleash private credit and enhance resource allocation.

CONCLUSION Perhaps one of the much more complex and potentially intertwined relationship between parties in secured lending is when, for liquidity reasons, the same asset used as collateral is involved in other financing transactions, that is, as the lender decides to re-hypothecate the asset as collateral to another creditor. This may have serious implications for risk propagation

246

IKE MATHUR AND ISAAC MARCELIN

and financial stability especially when the rehypothecation chain is extended over numerous lenders. Also, there are wide spreads in collateral coverage across countries and the differences are observable along the institutional settings. This study addresses the issue of high collateral to loan value in relation with some specific aspects of the law and finance and financial development literature. Institutions and the courts provide the framework within which financing contracts take place. We find that where institutions are ineffective and contract enforcement weak, lenders set collateral coverage high, erecting obstacles to access finance. Overcollateralization seems to cause inefficient credit allocation, credit rationing, and underinvestment. The immediate consequence of high collateral coverage along with poor institutions is that many firms cannot put certain assets to work using external finance to expand. Since collateral information contents reduce information asymmetric problems and ex-post frictions, to better estimate the riskiness of the borrowers, lenders have resorted to secured lending over the ages. Consequently, as a consistent feature, through different stages of financial development, most borrowers have to pledge collateral to access finance. Relaxing collateral requirements may have large effects on credit markets and firms’ ability to invest and grow. While legal institutions ought to function in a manner imparting lower systemic risk, many businesses with valuable assets are locked away from credit markets for insufficient collateral. Well-functioning legal systems with low prevalence of corruption allow firms to put up their assets as collateral to access capital for expansion. Other scholarly research explains, without consensus, why collateral is a prominent feature in financing contracts. In contrast, this study proposes a policy to make the use of collateral more beneficial to borrowers and lenders while unlocking credit for greater investment, economic growth, and development. It expects to be useful to a number of developing countries where policymakers are struggling to increase domestic finance.

NOTES 1. Alvarez de la Campa (2010). 2. Collateral to loan ratios are computed from the World Bank Enterprise Surveys database. 3. See Love et al. (2013). 4. Average computed from the World Bank Enterprise Surveys database.

247

Unlocking Credit

5. See Francis, Hasan, and Song (2012) for a detailed discussion. 6. Firms in Indonesia, Ireland, South Africa, Portugal and Croatia, and other countries falling in the low collateral and low constraints to finance. 7. Counted from the moment the plaintiff decides to file the lawsuit in court until payment (Source: World Bank Doing Business database). 8. The period of time from the company’s default until the payment of some or all of the money owed to the bank (Source: World Bank Doing Business database). 9. See Acemoglu and Johnson (2005) and Acemoglu, Johnson, and Robinson (2001) for detailed discussions on institutions. 10. When a project generates significant abnormal returns, the lender’s expected payoffs (return on investment) remains unchanged while the borrower’s return on equity is magnified. When the project fails, however, the lender’s loss may be substantially large while the borrower’s loss is limited to the borrower’s equity investment, which may be insignificant given the project capital structure.

REFERENCES Acemoglu, D., & Johnson, S. (2005). Unbundling institutions. Journal of Political Economy, 113(5), 949
995. Acemoglu, D., Johnson, S., & Robinson, A. J. (2001). The colonial origins of comparative development: An empirical investigation. American Economic Review, 91(5), 32. Acharya, V. V., Shin, H. S., & Yorulmazer, T. (2010). Crisis resolution and bank liquidity. Review of Financial Studies, 24, 2166
2205. Acharya, V. V., & Viswanathan, S. V. (2011). Leverage, moral hazard, and liquidity. Journal of Finance, 66, 99
138. Aggarwal, R., & Goodell, W. J. (2009). Markets and institutions in financial intermediation: National characteristics as determinants. Journal of Banking and Finance, 33, 1770
1780. Aghion, P., & Bolton, P. (1992). An incomplete contracts approach to financial contracting. Review of Economic Studies, 59(3), 473
494. Aghion, P., & Bolton, P. (1997). A theory of trickle-down growth and development. Review of Economic Studies, 64(2), 151
172. Algan, Y., & Cahuc, P. (2010). Inherited trust and growth. American Economic Review, 100, 2060
2092. Allen, F., Qian, J., & Qian, M. (2005). Finance, law and economic growth in China. Journal of Financial Economics, 77, 57
116. Almeida, H., & Campello, M. (2007). Financial constraints, asset tangibility, and corporate investment. Review of Financial Studies, 20, 1429
1460. Alvarez de la Campa, A. (2010). Secured transactions systems and collateral registries. International Finance Corporation, Washington DC, USA. Ang, B. J. (2013). Are modern financial systems shaped by state antiquity? Journal of Banking & Finance, 37, 4038
4058. Angbazo, A. L., Mei, J., & Saunders, A. (1998). Credit spreads in the market for highly leveraged transaction loans. Journal of Banking & Finance, 22, 1249
1282. Bae, K.-H., & Goyal, V. K. (2009). Creditor rights, enforcement, and bank loans. Journal of Finance, 64(2), 823
860.

248

IKE MATHUR AND ISAAC MARCELIN

Beck, T., & Demirgu¨c¸-Kunt, A. (2006). Small and medium-size enterprises: Access to finance as growth constraint. Journal of Banking and Finance, 30, 2931
2943. Beck, T., Demirgu¨c¸-Kunt, A., & Laeven, L. (2006). The determinants of financing obstacles. Journal of International Money and Finance, 25, 932
952. Beck, T., Demirgu¨c¸-Kunt, A., & Levine, R. (2003). Law endowments and finance. Journal of Financial Economics, 70, 137
181. Beck, T., Demirgu¨c¸-Kunt, A., & Maksimovic, V. (2008). Financing patterns around the world: Are small firms different? Journal of Financial Economics, 89, 467
487. Berger, A. N., Espinosa-Vega, M. A., Frame, W. S., & Miller, N. H. (2011). Why do borrowers pledge collateral? New empirical evidence on the role of asymmetric information. Journal of Financial Intermediation, 20, 55
70. Berger, A. N., Frame, W. S., & Ioannidou, V. (2011). Reexamining the empirical relation between loan risk and collateral: The roles of collateral characteristics and types. Working Paper No. 2011-12. Federal Reserve Bank of Atlanta. Berger, A., Imbierowicz, B., & Rauch, C. (2014). The role of corporate governance in bank failures during the recent financial crisis. Unpublished Working Paper, University of South Carolina. Berger, A. N., & Udell, G. F. (1990). Collateral, loan quality, and bank risk. Journal of Monetary Economics, 25(1), 21
42. Berger, A. N., & Udell, G. F. (1992). Some evidence on the empirical significance of credit rationing. Journal of Political Economy, 100(5), 1047
1077. Berger, A. N., & Udell, G. F. (1998). The economics of small business finance: The roles of private equity and debt markets in the financial growth cycle. Journal of Banking and Finance, 22, 613
673. Bernanke, B., & Gertler, M. (1989). Agency costs, net worth, and business fluctuations. American Economic Review, 79(1), 14
31. Bernanke, B., & Gertler, M. (1990). Financial fragility and economic performance. Quarterly Journal of Economics, 105(1), 87
114. Besanko, D., & Thakor, A. (1987). Collateral rationing: Sorting equilibria in monopolistic and competitive credit markets. International Economic Review, 28, 671
689. Besley, T., & Ghatak, M. (2008). Creating collateral: The de Soto effect and the political economy of legal reform. Unpublished Working Paper. London School of Economics. Bester, H. (1985). Screening vs. rationing in credit markets with imperfect information. American Economic Review, 75(4), 850
855. Birge, R. J., & Ju´dice, P. (2013). Long-term bank balance sheet management: Estimation and simulation of risk-factors. Journal of Banking and Finance, 37, 4711
4720. Boot, W. A., Thakor, V., & Udell, G. F. (1991). Secured lending and default risk: Equilibrium analysis, policy implications and empirical results. Economic Journal, 101, 14. Booth, J., & Booth, C. L. (2006). Loan collateral decisions and corporate borrowing costs. Journal of Money, Credit and Banking, 38(1), 67
90. Booth, J. R. (1992). Contract costs, bank loans, and the cross-monitoring hypothesis. Journal of Financial Economics, 31, 2
41. Brick, I., & Palia, D. (2007). Evidence of jointness in the terms of relationship lending. Journal of Financial Intermediation, 16(3), 452
476. Chakraborty, A., & Hu, C. X. (2006). Lending relationships in line of credit and non-line of credit loans: Evidence from collateral use in small business. Journal of Financial Intermediation, 15, 86
107.

Unlocking Credit

249

Chan, Y.-S., & Kanatas, G. (1985). Asymmetric valuations and the role of collateral in loan agreements. Journal of Money, Credit and Banking, 17(1), 84
95. Chan, Y.-S., & Thakor, A. V. (1987). Collateral and competitive equilibria with moral hazard and private information. Journal of Finance, 42(2), 345
363. Chofaras, N. D. (2004). Internal ratings, supervisory weights and collateral. In Elsevier (Ed.), Economic capital allocation with basel II: Cost, benefit and implementation procedures (1 ed., pp. 440). London: Butterworth-Heinemann. Dahan, F., & Simpson, J. (2008). Legal efficiency of secured transactions reform: Bridging the gap between economic analysis and legal reasoning. In Secured transactions law to support access to credit. London: Elgar Financial Law. Degryse, H., & Van Cayseele, P. (2000). Relationship lending within a bank-based system: Evidence from European small business data. Journal of Financial Intermediation, 9(1), 90
109. Demirgu¨c¸-Kunt, A., & Huizinga, H. (1998). Determinants of commercial bank interest margins and profitability: Some international evidence. Working Paper No. 1900, World Bank Policy Research. Djankov, S., McLiesh, C., & Shleifer, A. (2007). Private credit in 129 countries. Journal of Financial Economics, 84, 30. Elsas, R., & Krahnen, J. P. (2000). Collateral, default risk, and relationship lending: An empirical study on financial contracting. CFS Working Paper No. 1999/13. Engelhardt, T., & Regitz, B. (2008). The state of nature and lending in an unreformed environment: Experience from early transition countries. In Secured transactions law to support access to credit. London: Elgar Finance Law. Esty, B. C., & Megginson, W. L. (2003). Creditor rights, enforcement, and debt ownership structure: Evidence from the global syndicated loan market. Journal of Financial and Quantitative Analysis, 38(1), 37
59. Feder, G., Tongroj, O., & Tejaswi, R. (1988). Collateral guaranties and rural credit in developing countries: Evidence from Asia. Agricultural Economics, 2, 231
245. Fernandes, A. M., & Kraay, A. (2007). Property rights institutions, contracting institutions, and growth in South Asia: Macro and micro evidence. In S. Ahmed & E. Ghani (Eds.), South Asia: Growth and regional integration. New Delhi: Macmillan. Fisher, I. (1933). The debt
deflation theory of great depressions. Econometrica, 4, 337
357. Fleisig, H. (2008). The economics of collateral and collateral reforms. In Secured transactions law to support access to credit. London: Elgar Financial Law. Fleisig, H., Safavian, S. M., & de la Pen˜a, N. (2006). Reforming collateral laws to expand access to finance. Washington, DC: The World Bank. Francis, B., Hasan, I., & Song, L. (2012). Are firm and country specific governance substitutes? Evidence from financial contracts in emerging markets. Journal of Financial Research, 35(3), 343
374. Gai, P., Kondor, P., & Vause, N. (2006). Procyclicality, collateral values, and financial stability. Bank of England Working Paper No. 304. Gan, J. (2007). Collateral, debt capacity, and corporate investment: Evidence from a natural experiment. Journal of Financial Economics, 85, 709
734. Gatev, E., Schuermann, T., & Strahan, P. E. (2009). Managing bank liquidity risk: How depositloan synergies vary with market conditions. Review of Financial Studies, 22, 995
1020. Gnabo, J.-Y., & Moccero, D. N. (2013). Risk management, nonlinearity and aggressiveness in monetary policy: The case of the US Fed. Journal of Banking and Finance. doi:10.1016/ j.jbankfin.2013.11.016

250

IKE MATHUR AND ISAAC MARCELIN

Godlewski, C. J., & Weill, L. (2011). Does collateral help mitigate adverse selection? A cross country analysis. Journal of Financial Services Research, 40, 49
79. Gorton, G., & Metrick, A. (2011). Securitized banking and the run on repo. Journal of Financial Economics, 104, 425
451. Gottesman, A. A., & Roberts, G. S. (2007). Loan rates and collateral. Financial Review, 42(3), 401
427. Hainz, C. (2003). Bank competition and credit markets in transition economies. Journal of Comparative Economics, 31, 223
245. Hanedar, E., Broccardo, E., & Bazzana, F. (2014). Collateral requirements of SMEs: The evidence from less-developed countries. Journal of Banking and Finance, 38, 106
121. Harhoff, D., & Ko¨rting, T. (1998). Lending relationships in Germany: Empirical evidence from survey data. Journal of Banking and Finance, 22, 1317
1354. Harris, D. G. (1974). Interest rates, nonprice terms, and the allocation of bank credit. Southern Economic Journal, 40(3), 428
433. Haselmann, R., Pistor, K., & Vig, V. (2010). How law affects lending. Review of Financial Studies, 23(2), 549
580. He, Z., & Xiong, W. (2012). Rollover risk and credit risk. Journal of Finance, 67, 391
429. Holmstrom, B., & Tirole, J. (1997). Financial intermediation, loanable funds, and the real sector. Quarterly Journal of Economics, 112(3), 663
691. Houben, A., & Slingenberg, W. J. (2013). Collateral scarcity and asset encumbrance: Implications for the European financial system. Financial Stability Review, 17, 197
206. Igawa, K., & Kanatas, G. (1990). Asymmetric information, collateral, and moral hazard. Journal of Financial and Quantitative Analysis, 25(4), 469
490. Inderst, R., & Mueller, H. M. (2007). A lender-based theory of collateral. Journal of Financial Economics, 84, 33. Jackson, T. H., & Kronman, A. T. (1979). Secured financing and priorities among creditors. Yale Law Journal, 88, 1143
1182. Jime´nez, G., & Saurina, J. (2004). Collateral, type of lender and relationship banking as determinants of credit risk. Journal of Banking & Finance, 28, 2191
2212. Jime´nez, G., Salas, V., & Saurina, J. (2006). Determinants of collateral. Journal of Financial Economics, 81, 26. Kashyap, A., Scharfstein, D., & Weil, D. (1990). The high price of land and the low cost of capital: Theory and evidence from Japan. Unpublished Working Paper, MIT. Kirkwood, J. (2010). Securitisation and banks’ net interest margins. Economic Record, 86(274), 329
341. Kiyotaki, N., & Moore, J. (1997). Credit cycles. Journal of Political Economy, 105, 211
248. Kripke, H. (1985). Law and economics: Measuring the economic efficiency of commercial law in a vacuum of fact. University of Pennsylvania Law Review, 133(5), 929
985. Krishnamurth, A. (2003). Collateral constraints and the amplification mechanism. Journal of Economic Theory, 111, 277
292. Laeven, L., & Majnoni, G. (2005). Does judicial efficiency lower the cost of credit? Journal of Banking and Finance, 29(7), 1791
1812. La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (1999). Corporate ownership around the World. Journal of Finance, 54(2), 471
518. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, W. R. (1997). Legal determinants of external finance. The Journal of Finance, 52(3), 1131
1150.

Unlocking Credit

251

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, W. R. (1998). Law and finance. Journal of Political Economy, 106(6), 1113
1155. La Porta, R., Lopez-de-Silanes, F., & Zamarripa, G. (2003). Related lending. Quarterly Journal of Economics, 118(1), 231
268. Leeth, J. D., & Scott, J. A. (1989). The incident of secured debt: Evidence from the small business community. Journal of Financial and Quantitative Analysis, 24(3), 379
393. Levine, R. (1999). Law, finance, and economic growth. Journal of Financial Intermediation, 8(1
2), 8
35. Levmore, S. (1982). Securities and secrets: Insider trading and the law of contracts. Virginia Law Review, 68, 117
160. Liberti, J. M., & Mian, R. A. (2010). Collateral spread and financial development. Journal of Finance, 65(1), 147
177. Lopez-de-Silanes, F. (2008). Turning the key to credit: Credit access and credit institutions. In Secured transactions reform and access to credit. London: Elgar Financial Law. Love, I., Perı´ a, M. S. M., & Singh, S. (2013). Collateral registries for movable assets: Does their introduction spur firms’ access to bank finance? Working Paper No. 6477, World Bank Policy Research. Machauer, A., & Weber, M. (1998). Bank behavior based on internal credit ratings of borrowers. Journal of Banking and Finance, 22, 1355
1383. Mann, R. (1997). Explaining the pattern of secured credit. Harvard Law Review, 110(3), 625
683. Manove, M., Padilla, A. J., & Pagano, M. (2001). Collateral versus project screening: A model of lazy banks. Rand Journal of Economics, 32, 726
744. Marcelin, I., & Mathur, I. (2014). Financial development, institutions and banks. International Review of Financial Analysis, 31, 25
33. Martin, A. (2009). A model of collateral, investment, and adverse selection. Journal of Economic Theory, 144, 1572
1588. McAleer, M., Jimenez-Martin, J.-A., & Perez-Amaral, T. (2013). Has the Basel accord improved risk management during the global financial crisis? North American Journal of Economics and Finance, 26, 250
265. Menkhoff, L., Neuberger, D., & Rungruxsirivorn, O. (2012). Collateral and its substitutes in emerging markets’ lending. Journal of Banking and Finance, 36, 817
834. Menkhoff, L., Neuberger, D., & Suwanaporn, C. (2006). Collateral-based lending in emerging markets: Evidence from Thailand. Journal of Banking and Finance, 30, 1
21. Modigliani, F., & Perotti, E. (1997a). Protection of minority interest and the development of security markets. Managerial and Decision Economics, 18(7/8), 9. Modigliani, F., & Perotti, E. (1997b). Security markets versus bank finance. Legal enforcement and investor protection. International Review of Finance, 1, 81
96. Modigliani, F., & Perotti, E. (2000). Security markets versus bank finance: Legal enforcement and investors’ protection. International Review of Finance, 1(2), 81
96. Oehmke, M. (2014). Liquidating illiquid collateral. Journal of Economic Theory, 149, 183
210. Ono, A., Sakai, K., & Uesugi, I. (2012). The effects of collateral on firm performance. Journal of the Japanese and International Economies, 26, 84
109. Orgler, Y. (1970). A credit scoring model for commercial loans. Journal of Money, Credit and Banking, 2(4), 435
445. Qian, J., & Strahan, E. P. (2007). How laws and institutions shape financial contracts: The case of bank loans. Journal of Finance, 62(6), 2803
2834.

252

IKE MATHUR AND ISAAC MARCELIN

Rajan, G. R. (1992). Insiders and outsiders: The choice between informed and Arm’s-Length debt. Journal of Finance, 47(4), 1367
1400. Rajan, R., & Winton, A. (1995). Covenants and collateral as incentives to monitor. Journal of Finance, 50(4), 1113
1146. Safavian, S. M. (2008). Firm-level evidence on collateral and access to finance. In Secured transactions reform and access to credit. London: Elgar Financial Law. Sakurai, Y., & Uchida, Y. (2013). Rehypothecation dilemma: Impact of collateral rehypothecation on derivative prices under bilateral counterparty credit risk. Journal of Banking and Finance. doi:10.1016/j.jbankfin.2013.11.024 Schwartz, A. (1981). Security interests and bankruptcy priorities: A review of current theories. Journal of Legal Studies, 10(1), 1
37. Scott, J. H., Jr. (1977). Bankruptcy, secured debt, and optimal capital structure. Journal of Finance, 32(1), 1
19. Scott, J. H. Jr. (1979). Bankruptcy, secured debt, and optimal capital structure: Reply. Journal of Finance, 34(1), 253
260. Scott, R. E. (1986). A relational theory of secured financing. Columbia Law Review, 86(5), 901
977. Shim, I., & von Peter, G. (2007). Distress selling and asset market feedback. Financial Markets, Institutions and Instruments, 16(5), 243
291. Shin, S. H. (2009). Securitisation and financial stability. Economic Journal, 119, 309
332. Shleifer, A., & Vishny, W. R. (1992). Liquidation values and debt capacity: A market equilibrium approach. Journal of Finance, 47(4), 1343
1366. Smith, C. W., Jr., & Warner, J. B. (1979). Bankruptcy, secured debt, and optimal capital structure: Comment. The Journal of Finance, 34(1), 247
251. Steijvers, T., & Voordeckers, W. (2009). Collateral and credit rationing: A review of recent empirical studies as a guide for future research. World Bank Economic Review, 4, 271
295. Steijvers, T., Voordeckers, W., & Vanhoof, K. (2010). Collateral, relationship lending and family firms. Small Business Economics, 34, 243
259. Stiglitz, J., & Weiss, A. (1981). Credit rationing in markets with imperfect information. American Economic Review, 71(3), 393
410. Stiglitz, J., & Weiss, A. (1983). Incentive effects of terminations: Applications to the credit and labor markets. American Economic Review, 73, 912
927. Stulz, R., & Johnson, H. (1985). An analysis of secured debt. Journal of Financial Economics, 14(4), 501
521. Swary, I., & Udell, G. F. (1988). Information production and the secured line of credit. Unpublished Working Paper. Wang, Y.-L. (2010). Does collateral cause inefficient resource allocation? Journal of Economics and Business, 62, 220
233. Williamson, O. E. (1988). Corporate finance and corporate governance. Journal of Finance, 43, 567
592.

PART IV COUNTRY STUDIES ON MONETARY POLICY, CREDIT, AND FINANCIAL INTERMEDIARIES

MONETARY POLICY AND BANK LIQUIDITY IN CHINA Nan Shi, Xin Sun and Fan Zhang ABSTRACT The interbank market in China experienced remarkable squeezes in liquidity in 2013. In particular, the overnight Shanghai Interbank Offered Rate reached a historical high in June. Banks were unprepared, facing the occurrence of various liquidity demands simultaneously. Effects of the liquidity squeeze spread across markets, and concerns were expressed about the health of the banking sector in the world’s second largest economy. Yet the central bank of China maintained an unswerving view that the tightness of liquidity was only structural, and could be overcome by the commercial banks themselves. While it may be too early to judge whether the central bank was correct, or whether there is systematic liquidity risk in the banking sector, markets received a clear signal from the People’s Bank of China. The central bank stopped acting as a ‘perpetual put option’ for commercial banks and refused to take responsibility to satisfy liquidity needs in the interbank market. Its intention is clear; that is, to adjust monetary policy and support economic reform in China. The new Chinese government seems determined to steer a new course away from the previous growth episode. Its resolution has been published and actions have been taken. Among them, the central

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 255
276 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096010

255

256

NAN SHI ET AL.

bank’s changes to monetary policy have received responses from the markets, and the People’s Bank of China is now in the vanguard of a battle to squeeze liquidity. It is difficult to predict what further actions the government will take. However, it should be aware that the driving force of economic reform in China comes from structural change and productivity improvement. Without follow-up policies, complication in the financial system could undermine the central bank’s effort and international capital flows may quickly substitute the opening position of the central bank in the interbank market. More wisdom is required if China is to win the battle for deleveraging and structural reform. Keywords: Money squeeze; liquidity; interbank market; monetary policy; open market operation

INTRODUCTION In June 2013, the interbank market in China experienced a storm, represented by spikes in the Shanghai Interbank Offered Rate (SHIBOR). The rates on all short-term products reached record highs. Since the introduction of SHIBOR in 2006, the overnight rate had normally floated below 4%, and seldom had it reached 8%. Surprisingly, for the whole of June 2013 it remained at over 4%, and on June 19, 2013 it reached a peak at 13.44% (Fig. 1). The shortage of liquidity in the interbank market spread to other markets. The 7-day repo rate reached 10.77% and the auction price for 6-month government cash deposits shot to 6.5%. The Shanghai Composite Index slumped by 5.3%, with over a hundred stocks falling by 10%, reaching their daily limit. In China the term ‘Qian Huang’ (‘money squeeze’) is used to describe the tightness in liquidity in the markets. Markets were unprepared for the sudden money squeeze and many banks approached desperation, seeking all possible means to generate funds to avoid default. The rates returned to normal levels by the end of June, after the People’s Bank of China (PBOC) injected liquidity into the interbank market. Yet the record high rates and the fragility of the markets caught global attention. Debates revolved around the causes of the liquidity crunch. In particular, when the PBOC was inactive and refused to intervene, markets were initially uncertain about the intention of the central bank. In early June when there were signs of a spike in the SHIBOR, many people were

257

Monetary Policy and Bank Liquidity in China 14 12 10 8 6 4 2 0

O/N

Fig. 1.

1W

1M

Shanghai Interbank Offered Rate (Per cent). Source: Shanghai Interbank Offered Rate (SHIBOR) Data Services.

confident that the PBOC would step in and that the high rates would not last long. However, by the time the PBOC did intervene it was already late June, and much later than had been expected by many. The PBOC was given a new nickname, ‘Yang Ma’ (the central mum), which describes people’s strong feeling of reliance on its operation in the interbank market, and also their disappointment in this case. Furthermore, people were uncertain about the future direction of monetary policy. Although the Third Plenum of the Communist Party of China (CPC), held in November 2013, proposed very ambitious plans for reform, details have still to be published, and it will be difficult to examine the determination of the CPC until the government proceeds with further actions. Doubts on the determination of the PBOC, and uncertainty about future policies, still overshadow the markets. Moreover, international markets are watching the health of the Chinese economy. Asset bubbles and increasing leverage may cause a hard landing if China’s reform is not equipped with careful and prudent actions. A turbulent Chinese economy could threaten the global pace of economic recovery, which would be unwelcome to domestic and foreign markets alike. More than 6 months after the liquidity squeeze, echoes of the liquidity crunch still reverberate. Interbank rates swung significantly in the second half of 2013. In December, another squeeze of similar pattern to the one in

258

NAN SHI ET AL.

June emerged in the interbank market and recalled the recent nightmare. If we are to understand and interpret the deleverage path of the Chinese economy, a comprehensive review of the June 2013 liquidity crunch is required. Studies carried out soon after the event may not be capable of reviewing the long lasting effects. In this chapter, we contribute a detailed analysis of the evolvement, causes and implications of the liquidity crunch. The June 2013 liquidity crunch recalled to mind the history of money squeeze in China. The section ‘The Road to Liquidity Crunch’ of this chapter reviews the historical episodes in order to identify their similarities. The evolvement of the event was a progressive process, and in this section we provide a chronicle of its unfolding. The section ‘Causes of the Liquidity Crunch’ lists the causes of the liquidity crunch. Most evidence indicates that the liquidity crunch is not a black swan event. The role of the PBOC is clear and should not be underestimated. In order to have a comprehensive picture of change to monetary policy, the section ‘Monetary Policy and PBOC in the Liquidity Crunch’ first discusses the unconventional policies of the Chinese central bank in recent years. There is a strong connection between the rationale of PBOC inaction and the financial reform now being undertaken in China. The PBOC’s intention and reasons for changing its direction of policy are discussed in the section ‘Monetary Policy and PBOC in the Liquidity Crunch’. Finally, conclusions and implications are summarized in the section ‘Concluding Remarks’.

THE ROAD TO LIQUIDITY CRUNCH Difficult though it may be to imagine, recorded money squeeze can be traced back almost one thousand years in China. Even today, with the modern financial system and strict regulation, flaws in the financial system and numerous streams of capital flowing across economies could easily push markets to dangerous limits. Historically, as a regional economic centre, China has a long record of events showing similarities with the June 2013 liquidity crunch. The liquidity squeeze recalled to mind stories from the past, and this partly explains the rapid spread of negative feelings among the public. Historical Episodes The history of money squeeze can be traced back to the Song Dynasty of about one thousand years ago. Shortage of currency was always a problem

Monetary Policy and Bank Liquidity in China

259

during the dynasty’s several centuries’ long governance, but the reason was not that insufficient amounts of currency were issued. In the early Northern Song, the government issued millions of Guans of copper coins1 per year, whereas the previous Tang Dynasty had never issued more than 330,000 Guans per year. Shenzong, the sixth Song emperor, issued 5 million Guans of copper coins per year, a huge number, but the quantity in circulation in the economy was still low. During that period China had a dominant position in the regional economy: many nearby countries used Song copper coins as hard currency and traded on Song copper coins in daily life (Zheng, 2005). Demand from other countries accounted for a large proportion of the issuance, and the number of coins circulating within Song territory decreased. This early form of capital outflows challenged a Bretton Woods-like system between the Song and neighbouring countries. Flaws in the financial system worsened the problem. When the value of the metal exceeded the face value, people tended to hide copper coins and melt them to produce copper (Chen & Xiong, 2010). Governments failed to ban such arbitrage; nor did they succeed in withdrawing coins from people’s hands through taxation. Finally, the dynasty was even unable to maintain its military force, due to economic recession. More recent liquidity crunches in 2011 reflected the modern situation in China, and the symptoms and causes were closely related to the 2013 crisis. In January and June 2011, Chinese banks encountered system-wide liquidity tightness. The overnight SHIBOR shot up to 7.98% and 7.47% in January and June, respectively. Even the large banks normally considered sufficiently capitalized were not immune to liquidity problems. In January 2011, the PBOC injected 1.3 trillion renminbi into the banking system. Following the 2008 global financial crisis, Chinese banks were criticized by the central bank of China for aggressive counter-cyclical operations (Chu, Wen, & He, 2011). The credit boom fuelled the increase in real estate and infrastructure, and the erosion of liquidity in the financial sector was a worrying sign of systematic increase in non-performing loans. In addition, new deposits attracted from individuals and corporations dwindled in this period. Many corporations showed early signs of liquidity tightness and relied strongly on a shadow banking system and offshore financing channels. Banks experienced growing disparity between interim and period-end data. The fluctuation in the quantity of deposits before and after quarter end was unprecedented, and cash cushion on regular dates was even thinner than at quarter ends. While the memory of the tensions in liquidity in 2011 had not yet faded, in June 2013 a new, similar, liquidity crunch occurred. After a few

260

NAN SHI ET AL.

rehearsals in recent years, banks should have stocked enough bullets to fight against the squeeze. However, their performance in 2013 was disappointing. Meanwhile, the changing attitude of the central bank towards the same situation is interesting. In the next section, we offer a chronicle of the June 2013 crisis and the actions of related parties.

The June 2013 Crisis The liquidity crunch that occurred towards the end of June 2013 shocked the world. Unlike previous episodes in China, it is unique in terms of the central bank’s reluctance to intervene. Some raised concerns that it would lead directly to events as serious as the collapse of Lehman Brothers in the United States in the 2008 financial crisis. Although this intensive episode had passed away by the end of June, further attention has continued to be paid to the role of different players involved in it and to the health of the second largest economy in the world. To review the process of this event, we create a timeline covering the whole month of June 2013. The fuse of the liquidity crunch explosion was lit on June 6. A sale of 6-month bills issued by the Agricultural Development Bank of China failed to achieve its target of 20 billion renminbi, reaching only 11.51 billion. The SHIBOR overnight interest rate rose by 135.9 basis points to 5.98% (Fig. 1), and short-term interest rates began to exceed long-term rates. On the same day, a more influential news event triggered panic. Unable to make repayments on time to the Industrial Bank Co. Ltd., China Everbright Bank (CEB) had to pay penalties for the delay. Traders believed the amount delayed was up to 6 billion renminbi, and that the delay should be classified as default. Although on the next day, both banks officially denied the delay in repayment, rumours had already spread around the markets. In the following days, SHIBOR fluctuated like a roller coaster. On June 14, the Ministry of Finance intended to sell 15 billion renminbi worth of local government bonds, but raised only 9.5 billion. This was the first time in 23 months that such a sale had not been cleared out. In regular periods, such bonds normally offer 23 basis points higher than central government bonds. However, 60% of the bonds sold were 39 and 40 basis points more than three- and five-year central government bonds, respectively. On June 18, the Agricultural Development Bank of China again tried to issue two bonds with three years and five years maturity. The target was 26 billion renminbi, but for the second time, the full amount was not sold.

Monetary Policy and Bank Liquidity in China

261

On June 19, interbank liquidity tightness reached its climax. The SHIBOR overnight rate shot to 13.44%, the highest daily average in its history, and the 7-day repo rate also went to a record high of 12.25%. All other short-term fixed income products in various markets surged. On the same day, the interbank trading system postponed closing for half an hour. According to some interbank traders, the delay to system closing time was due to a large bank having failed to borrow enough cash to repay borrowings within trading time. On the next day, the nightmare continued. Rumours spread around the market that the Bank of China (BOC), the fourth largest commercial bank, had defaulted. The overnight repo rate once hit 30% and the 7-day repo rate hit 28%. The Ministry of Finance auctioned 6-month government cash deposits, and the interest rate shot to 6.50%, another record high in the last 15 months, so that it was now on a similar level as interbank rates. Still, there was no direct intervention by the PBOC, but it is evident that a few actions were taken on that day to stabilize the market. According to Fan, Wang, and Zhang (2013), the PBOC instructed the Industrial and Commercial Bank of China (ICBC) and the China Development Bank to release liquidity. From 2 p.m. on that day, several deals were struck at the overnight rate of 8%, a significant drop from the previous day’s high. Near closing time of the trading system, there were even deals at an overnight rate of 4.3%. On June 23, the central bank still refused to intervene directly and stated the importance of liquidity management by commercial banks themselves. The central bank bill, which would tighten liquidity, was issued normally. On June 24, the panic triggered by the June 6 repayment delay by CEB and sent spiralling by the June 19 SHIBOR outburst started to get out of control, and quickly spread to stock markets. The Shanghai Composite Index slumped by 5.3%. Over a hundred stocks hit the 10% daily slumping floor, and the average loss rate of bank stocks amounted to 6.68%. To prevent the problem in the interbank market spreading to other markets and resulting in a larger crisis, the central bank eventually stepped in on June 25. The PBOC published an announcement on its website restating its unswerving view that the liquidity tightness was structural and should be overcome by banks themselves (PBOC, 2013a). It estimated that there should be 1.5 trillion renminbi extra cash beyond the capital reserves in all financial institutions and that an amount of 1 trillion renminbi would be sufficient to support liquidity needs. Meanwhile, it also announced that liquidity had been injected into a few prudent financial institutions. With the 12 billion renminbi central bank repo due the PBOC paused extension

262

NAN SHI ET AL.

of the repo, which largely relaxed the tension in the market. The overnight repo rate lowered to 5.83%, 592 basis points down from that on June 20. Banks were seeking funds from all possible sources. Of all the wealth management products sold that day, 23 had return rates higher than 6% and 8 had investment threshold lower than 100,000 renminbi and maturity shorter than 3 months, in order to attract small personal funds. The liquidity crunch finally eased, but the liquidity tightness did not disappear. The Economist (2013) commented that the event delivered a clear warning from the central bank that it would not continue to act as ‘perpetual put’ to help commercial banks manage liquidity. Although markets had calmed down by the end of June, the battle between the PBOC and commercial banks was not over. In October and December 2013, the PBOC again squeezed liquidity, abstaining from performing reserve repos. From September 2013, overnight SHIBOR climbed steadily (Fig. 1). Banks were preparing themselves with cash, as indicated by the increase in yields of fixed income securities. Demand for cash rather than high-quality assets triggered an increase in sell-offs in fixed income markets, especially with many lower-rated corporate bonds hitting yields at all-time highs. At the end of October, one-year BBB + rated corporate bonds were offered at 11.2%. Although the crisis was not of similar magnitude to that of June 2013, the tensions in the markets were here to stay into the near future.

CAUSES OF THE LIQUIDITY CRUNCH In early 2011, commercial banks in China started to experience periodical shortage of liquidity. However, a broader market-wide crisis did not show up until June 2013, when the PBOC refused to inject money into the interbank market. The central bank’s decision to refrain from liquidity injection exposed participants in the interbank market to a tremendous gap between short-term demands and supply of funds (Feng, 2013). Periodical quarterend needs and a few policy changes created a greater than regular shortage of funds for banks. Overall, banks were estimated to require funds totalling as much as 1.2 trillion renminbi during that period. Periodic needs at quarter end create regular large demand in the interbank market. Commercial banks should meet their capital requirement by the end of the second quarter. Meanwhile, an estimated 450 billion renminbi deposit is required for companies to pay corporate tax by May 31. This is equivalent to a 0.5% rise in the required reserves rate for

Monetary Policy and Bank Liquidity in China

263

banks. However, the effects from these two requirements should be expected by banks and markets. Compared with their situation in the early 2000s, commercial banks in China today do not have as severe a lack of capital. The five largest banks in China are subject to 20% required reserve ratio, while the figure for smaller banks is 16.5%. When investigating the June liquidity squeeze, new regulatory restrictions prohibiting banks from a few unconventional ways of generating funds for liquidity needs, and the failure of the interbank market under uncertainty, should not be ignored. Since early 2013, the Chinese government has made a series of crackdowns on shadow banking and irregularities in international trades. On March 25, the China Banking Regulatory Commission (CBRC) issued File No. 8 to regulate wealth management products, which have been used as out-of-balance tools for the commercial banks to circumvent deposit interest ceiling regulations and to attract deposits. In fact, the CBRC did not intend to constrict the wealth management product market. The vice president of the CBRC pointed out that legitimate wealth management products would not be classified as shadow banking (Shen, 2014). Rather, the government aimed to regulate the market and encourage asset securitization, which is a crucial means to deleverage the economy (Wei & Song, 2013). However, the side effect of the long-term benefits is short-term contraction in the market. Under the new regulation, transition between wealth management products and banks’ on-balance sheet accounts is cut. Increasing cost and stronger competition continue to push reward rates of wealth management products. New regulation on capital flows and credit in foreign trades worsened the situation. On May 5, the State Administration of Foreign Exchange (SAFE) issued File No. 20 to forbid cross-border artificial trade and restrict shadow banking. One popular method of arbitrage is to apply a letter of credit (LC) by a company and the company import copper from foreign areas to areas bounded within China. It would later sell the copper and the funds would be used to provide liquidity or invest in interest generating projects, including wealth management products. The normal repayment period of LC is from 90 to 180 days, but companies normally need less than 60 days to sell the copper. Companies use this kind of arbitrage to generate cheap funds for more than one month. More seriously, one deal of copper could be used to generate LCs on multiple occasions. In Zhejiang Province alone, more than 400 trading companies were reduced to band B by the end of June. In that situation a company’s trade income becomes subject to more strict supervision and it is required to prove the validity of its trades. It is estimated that arbitrage using a similar approach reduced

264

NAN SHI ET AL.

by more than 90% in the May
June period. As a result, funds outstanding for foreign exchange increased by only about 67 billion renminbi in May 2013 (Fig. 2), far less than in April 2013, when they increased by almost 200 billion. The PBOC had been looking closely at this kind of arbitrage from 2012, and had decided to take action; however, this contributed greatly to the liquidity tightness during this period (Fan et al., 2013). Regulatory change is not the only cause of capital outflows. Uncertainty regarding the progress of global recovery and the pulling back of quantitative easing by the US drove investors to withdraw funds from developing economies in 2013. Following large slowdown in the increase of funds outstanding for foreign exchange, June and July saw net capital outflows, for the first time since November 2012 (Fig. 2). Nevertheless, whether the combined effects of the above factors alone could have caused a sizable crisis is questionable. If the central bank is correct, the 1.5 trillion funds available in all institutions should have met the gap due to these effects (PBOC, 2013a). If regular and irregular demands were the fuel to boost interbank rates, then the structural problem in China’s banking sector, and market sentiment, were oxidizers that caused the liquidity crunch. The difference in liquidity situations between large commercial banks and small commercial banks is evident. Smaller banks faced much more serious liquidity issues because of the widespread mismatch between shortChange

Level 29,000

800

28,500

700

28,000

600

27,500

500

27,000

400

26,500

300

26,000

200

25,500

100

25,000

0

24,500

–100

24,000

–200

Change

Fig. 2.

Level

Funds Outstanding for Foreign Exchange (Billions of Renminbi). Source: People’s Bank of China and authors’ own calculation.

265

Monetary Policy and Bank Liquidity in China

term interbank borrowings and their medium- and long-term loans. Before early 2011, when markets first showed signs of tightening liquidity, small and medium banks had a faster pace of increasing medium- and long-term loans than did large banks (Fig. 3). Small and medium banks had to increase their reliance on the interbank market for funding due to their inability to generate sufficient deposits. Around the same time, small and medium banks overtook large banks in funding in the interbank market (Fig. 3). Until July 2013, small and medium sized banks generated more than twice the amount of funds from the interbank market as large banks. Term mismatch and difference in liquidity situation between large and small banks represent major structural problems in the Chinese banking sector. Relying on large banks to supply surplus funds in the interbank market was not as successful as expected. The PBOC has stated its disappointment at the failure of large banks to stabilize the market by making available their surplus funds. By the end of May 2013 the ICBC, a leading bank globally in terms of liquidity, had 4.6 trillion renminbi of highly liquid assets. Yet it may be too hash to blame the banks for their unwillingness to Interbank funding (billion renminbi)

Loan increase % 7.5 7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 –0.5

8,000 7,500 7,000 6,500 6,000 5,500 5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 Jan Apr Jul Oct Jan Apr Jul Oct Jan Apr Jul Oct Jan Apr Jul Oct 2010 2010 2010 2010 2011 2011 2011 2011 2012 2012 2012 2012 2013 2013 2013 2013 Medium and long term loan increase % (Large banks) Medium and long term loan increase % (Small & medium banks) Funding in interbank market (Large banks) Funding in interbank market (Small & medium banks)

Fig. 3.

Term Mismatch and Reliance on Interbank Funding. Source: People’s Bank of China and authors’ own calculation.

266

NAN SHI ET AL.

lend. In an interview with Reuters on June 26, 2013, the president of the ICBC complained that it had taken several days to understand the source of fluctuation in the market and the associated risks (Bi, Xie, & Jian, 2013). The president’s statement reveals the banks’ delay in reading market information and market sentiment under huge uncertainty. Market sentiment during the liquidity crunch can be traced back to the CEB event. On June 6, rumours of a 6.5 billion renminbi default on the part of CEB spread through the market. Although CEB initially denied the default, in its IPO document submitted to the Hong Kong Stock Exchange in late 2013 it admitted that the default had happened, although claiming this was due to technical problems, not a lack of liquidity. On June 20, while the market was spreading the rumour of the default of BOC, banks quickly recalled the CEB event of earlier that month. Moreover, BOC is the fourth largest bank in China according to capital, much larger than CEB. Although BOC promptly announced the rumour to be false, this was no longer important. It is understandable that banks refused to provide liquidity when they learnt that counterparts in the market were subject to huge default risk.

MONETARY POLICY AND PBOC IN THE LIQUIDITY CRUNCH The previous phase of economic development in China was supported by a 4 trillion renminbi fiscal stimulus and credit expansion through the banking system. The massive fiscal stimulus and supporting monetary policy helped China recover quickly from negative spillover transmitted from other economies. However, this phase also boomed a growth style relying on surplus liquidity in the whole economy, including the interbank market. The PBOC may have favoured this approach in previous years, but in 2013, its attitude towards fighting to deleverage the economy became more and more apparent. Previously, such liquidity squeeze had happened periodically and the central bank had been generous in satisfying banks’ liquidity demands. In 2011, communications between bank traders and the central bank were frequent. Normally the communication happened in the morning. Before 4pm the central bank would issue reverse repo to provide liquidity (Fan et al., 2013). However, ‘Yang Ma’ refused to take similar action this time. The PBOC’s changing attitude demonstrated its decision to say farewell to the previous phase of growth. In this section, we first

Monetary Policy and Bank Liquidity in China

267

review China’s turn away from her recovery policies after the 2008 crisis. Then we explain the PBOC’s intention and possible difficulties in achieving its goals.

China’s Banking Sector and Policies Since the 2008 Crisis In order to withstand the global crisis in 2008 and maintain economic growth, China’s government delivered a surprising announcement to a world reeling from financial crisis. It pledged to spend 4 trillion renminbi on a stimulus plan, three times as much as the US government was promising at that point. Later it proved that its contribution to the Chinese economy went beyond fiscal stimulus. In fact, it comprised an enormous surge in credit from the banking system, allowing local governments to launch thousands of infrastructure and other investment projects (Zhang, Li, & Shi, 2009). The consequences of that decision still shape China’s economic prospects today. The gross debt of China’s non-financial sectors, including government, households and corporations, rose from 160% of GDP in 2008 to 200% in 2010. The substantial 40% increase is equal to the increase in the US credit-to-GDP ratio in the five years before 2007. With rough estimations, of the 200% just mentioned for 2010, 65% was in the public sector, including government proper and off-balance-sheet borrowings by infrastructurebuilding state-owned enterprises (SOEs); 103% was in corporate debt and 32% in household debt. Compared with 2008, when public debt took about 46% of the 160%, corporate debt was about 93% and household debt was about 21%, leverage rose across all sectors of the economy sharply, as a result of loose monetary policy. However, China today is very different than before 2008, and now she has a slower pace of economic growth. This creates a dilemma. How can China maintain GDP growth with slower credit growth? One way is to boost productivity across the economy through structural reform and ultimately the ability to repay debt. To achieve this goal, the authorities aim to redirect the banking system to serving the needs of real private sector businesses rather than government-sponsored projects, by pushing banks to lend more to small businesses with marketized interest rates. On the other side, higher funding cost raises the pressure on all companies to invest more effectively. This plan is backed up by the new economic governance under Premier Li Keqiang, and was formally installed in March 2013. Many market

268

NAN SHI ET AL.

researchers use the term ‘Likonomics’ to refer to Premier Li’s strong market-oriented approaches in tackling economic structural reforms. Li’s policy strategy is based on three independent fundamental pillars, of which fiscal policy is one. In terms of fiscal policy, the government is more focused on ‘activating the stock of fiscal funds’ than on another round of massive stimulation. Under this approach, the government debt would be slowly absorbed by asset securitization, with reduced emphasis on fixed assets investment. The room for further massive innovation on fiscal policy is limited. This largely depends on the volatile trend of fiscal revenues in recent years. Moreover, following the Third Plenum, which took place in November 2013, the expansion of business tax-to-VAT reform will further squeeze government tax revenues. With increasing liabilities of government-owned financial vehicles and off-budget funds, government debt takes about 45% of GDP, and its inherent risks become a burden to China’s real economy (IMF, 2013). In 2012, all forms of credit expanded faster than expected, as the government loosened monetary policy to cushion the economic turndown. Household and public sector debt level remained stable and not unusually high, but a big chunk of local government borrowing had been reclassified as normal corporate borrowing, hence the increase concentrated on corporate sectors, with the corporate debt/GDP ratio rising from 108% in 2011 to 122% in 2012. For many years, the PBOC needed to adjust interbank liquidity in only one direction: it used higher reserve requirements and sales of central bank bills to soak up the liquidity created by huge inflows of foreign exchange. By the end of 2011, with the slowdown of China’s economic growth, this pattern had started to change. Total inflows of foreign currency ground nearly to a halt as China’s trade surplus shrank and net capital inflows fell. In 2012, as capital inflows declined, rather than having to aggressively sterilize the increase in domestic money, China’s central bank needed to address a shortfall of liquidity. This resulted in some signs of distress in the interbank market. In response, in December 2011, February 2012 and May 2012 the PBOC cut the reserve requirement ratio by a total of 150 base points, which helped free up some funds for banks. By May 2012, one-week interest rates were down below 3%. The central bank also cut benchmark loan and deposit rates in June and July 2012. From that time, however, the central bank changed its strategy: it stopped cutting reserve requirements, and began to using reverse repos to add liquidity to the market. Although the transparency of PBOC monetary policy may be low, it appears that it

Monetary Policy and Bank Liquidity in China

269

continues to use open market operations to provide a stable liquidity environment, rather than implementing a massive expansionary policy. This change is an early sign of the PBOC redirecting monetary policy to focus on interest rate rather than the quantity of loans. Keeping the required reserve ratio high may help facilitate the transition to a rate-based monetary policy. Meanwhile, China’s interest rate liberalization has made progress. The objectives of the interest-rate liberalization in China are to establish market-based pricing for the cost of capital, to improve efficiency of monetary policy transmission and to optimize financial resource allocation. The original roadmap was set to liberalize the money market and bond market interest rates, and then gradually liberalize deposit and lending rates. The sequencing of deposit and lending rates liberalization reform is designed to minimize the negative repercussions for financial market stability. Ultimately, the PBOC has revealed its plan; not surprisingly, it shows itself ambitious to contribute to China’s economic and financial reform and to promote reform in the interbank market.

PBOC Battle: Goals and Difficulties In 2013, the PBOC launched a series of actions in the interbank market. In May of 2013, the central bank bill was restarted to retrieve liquidity for the first time in a year and a half. On June 17, the PBOC issued an official letter to all banks stating that ‘at the present time, the overall liquidity level in China’s banking system is at a reasonable level’, requiring the commercial banks to ‘strengthen the analysis and forecast of the factors influencing recent liquidity demand, and make reasonable arrangement for the half year liquidity stress’ (PBOC, 2013b). On June 18 and 20, central bank bills worth a total of 2 billion renminbi were issued, regardless of the rising interbank rates. Even against a background of a sharp turn in monetary policy, banks were reluctant to change their previous behaviour, or were unable to accommodate to the current situation. Under initial tension in the market, participants seemed unprepared and believed that ‘central mum’ would step in to reduce liquidity tightness as usual. The expected intervention did not happen until the market showed early signs of a more widespread crisis. Some may blame the central bank, but it is possible that the liquidity crunch is exactly the signal the PBOC wished to deliver. No wonder that, after the June 2013 crisis, many in the markets had the feeling that

270

NAN SHI ET AL.

‘the party is over’. The intention of the central bank is clear, and consistent with Prime Minister Li Keqiang’s comments on ‘activating the existing money stock’, which were published three times in May. A similar policy goal was mentioned again at the conference of the State Affairs Council on June 19. The PBOC’s intention to soak up excess liquidity is not news. Notwithstanding the interruption due to the 2008 financial crisis, the PBOC has taken a succession of steps to promote economic reform. As early as 2007, there was consensus in the country that China needed to find a new way to boost economic growth, and that the government must change long adopted policies. After a year’s rocketing growth with GDP increasing more than 10%, in late 2007 the Central Economic Work Conference set the goal of macroeconomic policy to avoid overheating of the economy. In an interview on October 18, 2007, Governor Zhou Xiaochuan stated that the economy was overheated and policies to soak up excess liquidity had not been intensive enough (Zhao & Han, 2007). At that stage, reducing excess liquidity and its potential harm to economic growth was already among the goals of the central bank. Policy instruments available to the PBOC included interest rate, exchange rate, deposit reserve rate and open market operation. However, the 2008 financial crisis delayed China’s progress in economic reform. To some extent, macro policies adopted from 2007 to 2008 were clumsy. From 2007 to June 2008, the PBOC increased the deposit reserve rate 15 times, from 9% to 17.5%, and the interest rate 6 times, from 2.52% to 4.14% (Fig. 4). Disappointingly, the response from the market seemed sluggish. Moreover, the bold steps using interest rate and deposit reserve rate brought at least two negative effects. First, the market underestimated the PBOC’s intention to pursue economic reform and the uncertainty due to rapid change in macro policies caught the eye of the market. Secondly, the PBOC’s later response to the 2008 financial crisis led to a series of de facto flip-flops of policy. By the end of 2008, the central bank had reduced the interest rate 5 times and the deposit reserve rate 4 times (Fig. 3). Increase in credit was no longer a concern, and mortgages and a wide range of lending were again encouraged. The ongoing exchange rate reform was also paused, and the renminbi was pegged to the dollar again to boost exports. Fiscal policy included injections of fiscal funds totalling 4 trillion renminbi to various sectors. In spite of the success in bringing stability to the market with these actions, the government also received increasing criticism for printing money, which could result in liquidity surplus. By 2013, it was again widely

271

Monetary Policy and Bank Liquidity in China Benchmark deposit rate

Deposit reserve rate

5.0%

25%

4.5% 4.0%

20%

3.5% 3.0%

15%

2.5% 10%

2.0% 1.5% 1.0%

5%

0.5% 0% Jan 2007 Apr 2007 Jul 2007 Oct 2007 Jan 2008 Apr 2008 Jul 2008 Oct 2008 Jan 2009 Apr 2009 Jul 2009 Oct 2009 Jan 2010 Apr 2010 Jul 2010 Oct 2010 Jan 2011 Apr 2011 Jul 2011 Oct 2011 Jan 2012 Apr 2012 Jul 2012 Oct 2012 Jan 2013 Apr 2013 Jul 2013 Oct 2013

0.0%

Benchmark deposit rate

Fig. 4.

Deposit reserve rate

Benchmark Deposit Rate and Deposit Reserve Rate. Source: People’s Bank of China.

agreed among markets that pursuit of economic growth by expansionary monetary and fiscal policy is not sustainable. Meanwhile, there were clear signs of global economic recovery, and the United States was discussing a possible pull back of quantitative easing. It was time for China to resume the long paused reform. For many years, the ‘central mum’ acted as the manager of the interbank market in China. The interbank rate was stabilized at around 3%, and short-term actions such as open market operation were used to manage interbank liquidity. The interbank market was managed in order to stabilize the liquidity in the market and reduce market fluctuation. Yet while the central bank carefully managed the liquidity in the interbank market, this seemed incompatible with its strict goal to limit overall lending. Commercial banks used the interbank market as one major source of funding to support increases in loans. In fact, the interbank borrowing provided cheap funds to maintain high increases in loans, and created a maturity mismatch between bank assets and liabilities. For many reasons, the determination of the central bank should not be read as an independent event. The fundamental economic reform includes

272

NAN SHI ET AL.

capital market liberalization, exchange rate and interest rate marketization. As early as June 29, 2010, China had restarted exchange rate reform by increasing the flexibility of the renminbi exchange rate. The country could not wait until there had been full recovery from crisis, because the shortterm reactions to that crisis were already showing their effects. On March 16, 2013, PBOC Governor Zhou Xiaochuan was re-elected, which was unexpected due to his age but in accordance with his popularity. The re-election of a governor who was at retirement age indicated the government’s approval of the past role of the PBOC and signalled the continuation of the reform. Sure enough, the central bank pushed forward a series of actions to accelerate the reform, including marketization of the interbank rate. Following the reform of the renminbi exchange rate, interest rate liberalization is now on the agenda. The central bank has stated on various occasions that liberalization does not necessarily imply a free deposit rate, and has refrained from fully freeing the rate due to fear that this would cause instability. Perhaps the deposit rate marketization will have to wait until a full deposit insurance scheme is launched. However, this will not obstruct other planned actions, including for the interbank market. On March 25, 2013, the CBRC issued File No.8 to regulate wealth management products. This is not intended to slow down the development of wealth management services in China. Instead, it aims to regulate the market and to urge the banks to adopt appropriate ways to reduce relevant risks. This officially stated intention accords with the strategy of using securitization to deleverage. At the same time, competition in the wealth management market offers competing return rates to investors and this could be viewed as another step towards interest rate marketization. The central bank action to reduce supply in the interbank market clearly reflects the policy goal of the Chinese government. The PBOC has been successful in terms of soaking up liquidity in the interbank market and pushing up short-term interest rates. However, in order to promote structural reform, more steps need to be taken. Banks should certainly receive a clear signal from the central bank, but they will not necessarily be able to reduce their leverage as expected. In fact, after the relatively peaceful second half of 2012, loans and credit increased quickly in the first half of 2013 (Fig. 4). Even worse, banks experiencing liquidity problems look for funds by offering higher interest rates, which may increase their leverage. The June liquidity crunch pushed up the interest rate for wealth management products, which represent a major source of funds for banks. Also, banks may invest in riskier projects

Monetary Policy and Bank Liquidity in China

273

expecting higher returns to repay their borrowing and to cover loss due to higher interest rate. In doing so they would increase their non-performing loans and operation risks, as well as the liquidity problem. Also, tightening liquidity and pushing up short-term interest rates attracts hot money to fill the gap between demand and supply. The first half of 2013 saw tremendous quantities of arbitrage money slipping over the border in terms of artificial trading. Eventually, SAFE acted to strengthen capital control and block hot money, and this is an essential factor that cannot be ignored in the June liquidity crunch. Simply halting the flood of money from outside China may only work in the short run. Once the liquidity crunch was over, foreign inflows started a steady recovery, and China again had a huge increase in funds outstanding for foreign exchange (Fig. 2). Analysts warned that artificial trading may constitute a large part of the abnormal increase in exports seen in November 2013. Although SAFE again issued a document to strengthen regulation (SAFE, 2013), various ways to bypass capital control and arbitrage interest margins have been invented and circulate in the markets. Flows of foreign funds to supply liquidity could undermine PBOC efforts. Also, more strict regulation contradicts the steps towards releasing capital control, which is another policy goal for China. Uncertainty about the future policy of the central bank may have aggravated the reluctance in the market. Banks still expected the PBOC to inject money to ease liquidity tightness. After a few weeks’ inaction, at the end of 2013 the PBOC did intervene in the market, using short-term liquidity operation and reverse repo. It did not wait until the market showed sign of crisis. However, in taking action it may have encouraged the reliance of the market on the intervention of the central bank. Questions on the consistency of the policy could bring more uncertainty and prevent the goal being achieved. There are several reasons why the PBOC remained reluctant to use interest rate and deposit reserve ratio in 2013. As discussed, frequent changes and incoherent strategies would lead to the PBOC being accused of policy flip-flops. Moreover, the progress of global recovery is still not clear. In her speech at the National Press Club in Washington early 2014, IMF president Christine Lagarde even warned of the risk of deflation in the recovery. However, being prudent is not equivalent to inaction. It may not be possible to achieve a structural economic reform without touching monetary instruments like interest rate and deposit reserve rate. Under an increasingly clear process of global recovery, it is perhaps worthwhile for China to take action similar to that taken by the United States. Publishing a plan of

274

NAN SHI ET AL.

moderate and gradual change to monetary policy could help the market to form appropriate expectations of future macroeconomic policies. This could increase credibility of the central bank and allow banks time to change their operation according to a visible future. There is no doubt that it is time for China to make the transition from her previous phase of growth to one that is more sustainable. Government policy, including central bank efforts to promote the structural reform of banks, has the support of people in China. However, this does not mean the policies should not be equipped with appropriate procedures to avoid risks in the reform. Counting the causes of the June liquidity crunch, perhaps the CEB event and the rumour about BOC could not have been anticipated. Simply denying the rumour was not enough, and the failure to convince the market came close to causing widespread crisis. The government needs to be bold in promoting the reform. On the other hand, equipping the banks with safeguarding practice is also necessary. Even before the 2008 crisis, there had been discussion regarding a deposit insurance scheme. Clearly the scheme is a necessary step towards the marketization of interest rates. Also, the regulator could provide funds with a penalty rate as an emergency source of funds for banks, and the penalty rate could also work as a ceiling of the rate in the interbank market.

CONCLUDING REMARKS The banking sector reform in China has seen the liberalization of regulations and an increasing role for the market. Banks are now well equipped with means such as securitization to generate higher profits. However, they are still far from able to cope with the risks on their own. The market’s fragility became visible in June 2013, with the liquidity crunch. Despite the PBOC’s frequent expressions of concern about soaring credit, and statements that the expansionary monetary policy would not continue, banks were slow to react. Some ignored the signals and even gambled on monetary easing by the PBOC. Early signs of the problem appeared with the CEB event in early June. When the PBOC refused to intervene, against the expectations of the market, the market found itself unable to adjust. Similar situations continued to recur after the June liquidity crunch. Thanks to the lessons learnt from that crisis, the later liquidity shortages were more alarming than harmful. By then, the PBOC’s intentions were clearly understood across the markets. Reforming the banking sector is one

Monetary Policy and Bank Liquidity in China

275

essential pillar in the economic reform. However, the government must exercise wisdom to demonstrate its credibility and determination. Increasing the transparency of monetary policy could help the market to form appropriate expectations of the future. On the other hand, the promoters of the reform have to be equipped with instruments and procedures to respond to emergency in the markets.

NOTE 1. 1 Guan coins = 1,000 copper coins

ACKNOWLEDGEMENT Nan Shi, Xin Sun and Fan Zhang would like to thank Dr. Zhichao Zhang in Durham University Business School for his generous comments on this chapter.

REFERENCES Bi, X., Xie, H., & Jian, S. (2013). Exclusive interview: ICBC expressed willingness to act as stabilizer for liquidity. Reuters. Retrieved from http://cn.reuters.com/article/ CNAnalysesNews/idCNCNE95P07K20130626. Accessed on January 1, 2014 (in Chinese). Chen, Y., & Xiong, Y. (2010). Copper coin devaluation and government reserves: North Song liquidity crunch re-examination. Hanshan Shifan Xueyuan Xuebao [Journal of Hanshan Normal University], 31(1), 5 (in Chinese). Chu, C., Wen, C., & He, H. (2011). Chinese banks: Cash cushions thinning as liquidity erodes and forbearance burdens rise. In Fitch Wire: FitchRatings. The Economist. (2013). What caused China’s cash crunch. The Economist. Retrieved from http://www.economist.com/blogs/economist-explains/2013/07/economist-explains-2. Accessed on Jan 1, 2014. Fan, D., Wang, X., & Zhang, M. (2013).Why there is liquidity squeeze when money is adequate: The June crisis in China banking system. Southern China Weekend. Retrieved from http://www.infzm.com/content/91820. Accessed on July 19, 2014 (in Chinese). Feng, H. (2013). The PBOC, the liquidity squeeze and market liberalisation. Central Banking, 24(1), 61
66. IMF. (2013). People’s Republic of China. IMF Country Report no. 13/211. PBOC. (2013a). Appropriate adjustment to liquidity and stability of currency markets. People’s Bank of China (PBOC). Retrieved from http://www.pbc.gov.cn/publish/goutongjiaoliu/

276

NAN SHI ET AL.

524/2013/20130625182009056961206/20130625182009056961206_.html. Accessed on January 1, 2014 (in Chinese). PBOC. (2013b). An official letter from the People’s Bank of China about related issues on liquidity management in commercial banks. Retrieved from http://www.pbc.gov.cn/publish/ goutongjiaoliu/524/2013/20130624105915701332763/20130624105915701332763_.html. Accessed on January 1, 2014 (in Chinese). SAFE. (2013). Notice about management of foreign exchange related to financing for trades in banks by SAFE. Retrieved from http://www.safe.gov.cn/wps/portal/!ut/p/c5/04_ SB8K8xLLM9MSSzPy8xBz9CP0os3gPZxdnX293QwML7zALA09P02Bnr1BvIyNvc6 B8pFm8s7ujh4m5j4GBhYm7gYGniZO_n4dzoKGBpzEB3eEg-_DrB8kb4ACOBvp-Hvm 5qfoFuREGWSaOigDuOwR_/dl3/d3/L2dJQSEvUUt3QS9ZQnZ3LzZfSENEQ01LRz EwODRJQzBJSUpRRUpKSDEySTI!/?WCM_GLOBAL_CONTEXT=/wps/wcm/ connect/safe_web_store/safe_web/zcfg/jcxmwhgl/jcksfhyhxgl/node_zcfg_jcxm_jck_store/ 8486de0042197e8581fafb3e0e4da78e+. Accessed on January 1, 2014 (in Chinese). Shen, Y. (2014). Yan Qinmin: Legitimate wealth management products shall not be classified as shadow banking. Reuters. Retrieved from http://cn.reuters.com/article/cnInvNews/ idCNCNEA1C1SI20140116. Accessed on January 16, 2014 (in Chinese). Wei, R., & Song, X. (2013). Reviving two trillion existing funds. International Financial News. Retrieved from http://paper.people.com.cn/gjjrb/html/2013-08/29/content_1290814. htm. Accessed on January 1, 2014 (in Chinese). Zheng, Z. (2005). Shortage of money (silver) in the Tang-Song Dynasty and Ming-Qing Dynasty: History and enlightenment. Heze Xueyuan Xuebao. [Journal of Heze University], 27(1), 78
82 (in Chinese). Zhang, Z., Li, W., & Shi, N. (2009). Handling the global financial crisis: Chinese strategy and policy response. Unpublished manuscript. Zhao, X., & Han, J. (2007). Zhou Xiaochuan’s response to recent hot topics about the Chinese economy. Xin Hua Net. Retrieved from http://news.xinhuanet.com/fortune/2007-10/18/ content_6902782.htm. Accessed on January 1, 2014 (in Chinese).

MONETARY POLICY AND BANK CREDIT RISK IN VIETNAM PRE AND POST GLOBAL FINANCIAL CRISIS Xuan Vinh Vo and Phuc Canh Nguyen ABSTRACT A thorough understanding of transmission mechanism is a key to a successful conduct of monetary policy. This chapter attempts to improve knowledge in this respect by examining the impacts of commercial bank risks on the transmission of monetary policy. We investigate the impact of monetary policy on bank risk in Vietnam pre and post 2008 global financial crisis employing a unique and disaggregated bank level data set from 2003 to 2012. The results of panel data estimation indicate that the bank lending channel of monetary is evidenced in Vietnam. In addition, we find that the transmission mechanism is affected by characteristics of commercial banks. Keywords: Monetary policy; bank risk; crisis; transmission channel JEL classifications: E52; E58; G28

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 277
290 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096011

277

278

XUAN VINH VO AND PHUC CANH NGUYEN

INTRODUCTION There are many factors attributing to the 2008 global financial crisis, however, it is commonly argued that monetary is one of the factors contributing to excessive risk taking by banks (Taylor, 2009). Many published papers refer to a new transmission mechanism of monetary policy, termed as “risk taking channel” existing where low interest rates lasting for too long (Adrian & Shin, 2009; Borio & Zhu, 2008). In addition, there is an increasing volume of research domiciled on the importance of bank characteristics and conditions in the transmission mechanism of monetary policy. However, there is an inconclusive evidence regarding this issue in the current literature (Altunbas, Gambacorta, & Marques-Ibanez, 2010). This chapter investigates bank lending channel by estimating the banks’ supply function in dynamic panel data framework. Theoretical and empirical work suggests that monetary policy is transmitted into the economy via many important channels including interest rate channel, exchange rate channel, asset price channel, and credit channel (Dabla-Norris & Floerkemeier, 2006, Disyatat & Vongsinsirikul, 2003; Honda, 2004; Mugume, 2011; Mukherjee & Bhattacharya, 2011; Wulandari, 2012). In addition, commercial banks are an integral part of monetary transmission mechanism especially in a country where banks are major sources of credit to the economy. The bank lending channel operates through the central bank, limiting the supply of reservable deposits to banks when tightening monetary policy. This may force them to adjust the balance sheets on the asset side by reducing the supply of loans. The bank lending channel is one mechanism that amplifies the transmission of monetary policy because of credit market frictions such as asymmetric information. In the absence of such frictions, monetary policy operates through the conventional interest rate channel (Mora, 2013). The extent to which monetary policy can influence short-term real interest rates (because prices are sticky) alters the real cost of capital for firms and households and thus investment. This conventional interest rate channel is magnified and propagated by frictions that are collectively known as the “credit channel of monetary policy transmission” (Bernanke & Gertler, 1995). Information frictions in credit markets create a gap between a borrower’s internal funds (retained earnings) and more costly sources of external finance. Monetary policy endogenously influences the borrower’s external finance premium because of two amplifying mechanisms: the “balance sheet channel” affecting the borrower’s net worth and the “bank lending channel” influencing the supply of loans by depository institutions. For example, a

Monetary Policy and Bank Credit Risk

279

monetary policy tightening will negatively impact the borrower’s net worth, raising the borrower’s external finance premium and reducing its credit demand. Monetary policy becomes more and more important as an instrument for macroeconomic policy making and understanding the mechanism of transmission is a key to successful conduct of monetary policy. Monetary policy, especially interest rates, has strong impact on banks and this is multifaceted. When the central bank increase the interest rate, spot interest rate increases and expected long-term interest increases and this leads to the decrease in investment and consumption (Cecchetti, 1995; Friedman, 1956; Taylor, 1995). This leads a decrease in income and ability to repay the loan and increase in bank risk. Different bank characteristics will be affected differently (Altunbas et al., 2010, 2012). Recent literature on the transmission mechanism relating to bank characteristics and roles of banks by focusing on banks’ incentive (Borio & Zhu, 2008; Rajan, 2006a, 2006b). In this chapter, we argue that bank risk, together with other bank characteristics, should be considered when analyzing and examining the function and mechanism of bank lending channel of monetary policy. We argue that bank risk also influences the way banks react to GDP shocks. Loan supply from low-risk banks is less affected by economic slowdowns, which probably reflects their ability to absorb temporary financial difficulties on the part of their borrowers and preserve valuable long-term lending relationships. This chapter focuses on the implications bank risk for the provision of credit supply and the monetary policy transmission mechanism. We examine the roles of financial factors in the transmission mechanism of monetary policy. The evidence that favors the link between bank risk taking and interest rate is becoming more and more widespread, the understanding of this link and the channel through which it operates is of clear interest to policy makers concerned with the likely effects of economic measures designed to operate through interest rate channels (Dell’Ariccia & Marquez, 2013). Using an ex-post measure of risk which is provision for loan loss to total assets, we analyze the effect of bank risk on loan supply in the context of Vietnam, a small partially open economy. There are many published papers investigating the importance of bank conditions in the transmission mechanism of monetary policy. The empirical results are more focused for the United States while recent research in the Euro area provides inconclusive evidence. However, there are not many published papers concerning this issue using Vietnam data and one of our contribution is on this regard. We contribute to the existing literature by

280

XUAN VINH VO AND PHUC CANH NGUYEN

extending the analysis of the relevance of monetary lending channel and bank risk in a small partially open economy by employing a unique and disaggregated data of commercial banks in Vietnam. In addition, the current situation of the commercial bank system is one of the difficult issue that needs to be addressed. The remainder of the chapter proceeds as follows. The section “Overview of Banking system in Vietnam Pre and Post 2008 Global Financial Crisis” provides an overview of the banking system in Vietnam. The section “Model and Data” suggests the model and data for the analysis. The section “Results and Discussions” shows the results and the final section “Conclusions” concludes the chapter.

OVERVIEW OF BANKING SYSTEM IN VIETNAM PRE AND POST 2008 GLOBAL FINANCIAL CRISIS As a result of the Global Financial Crisis, Vietnam experienced a dropdown in economic growth rate for the post-crisis period from 2008 to 2012 (Fig. 1) compared to the pre-crisis period. In 2009, the government conducted a stimulus program which included a government’s makeup of 2
4% in interest rate for corporate borrower in order to promote economic growth however the results is limited as growth rate increased slightly in 2010 and started to decrease since 2011. In addition to slow economic growth, Vietnam also has problem with high inflation since the financial crisis. Fig. 2 presents an overview of the Vietnam inflation rate for the period 2003
2012. In terms of monetary policy implementation, it appears that Vietnam central bank (State Bank of Vietnam) focused on dealing with inflation other than other issues. Compare to the neighboring country of Thailand, the Bank of Thailand (BOT) considers exchange rate movement, which is uncertain, in setting the policy rate. Moreover, the BOT focuses more on the contemporaneous economic condition than the lag of interest rate. Specifically, the rule illustrates that the BOT follows the Taylor principle and puts more weight on exchange rate stabilization relative to the output stabilization (Lueangwilai, 2012). The change in Vietnam’s macro-economy has strong impact on Vietnam’s banking sector. Vietnam’s banking sector is expected to have one of the highest growth rates in Asia in the next few years due to the

281

Monetary Policy and Bank Credit Risk Vietnam GDP annual growth rate 9.00 8.40 8.00

8.50

8.20

7.80 7.30

7.00

6.80 6.30

6.00

5.90 5.30

5.00

5.03

4.00 3.00 2.00 1.00 -

2003

Fig. 1.

2004

2005

2006

2007

2008

2009

2010

2011

2012

Vietnam GDP Annual Growth Rate from 2003 to 2012. Source: Vietnam Key Indicators (2013), ADB.

Vietnam inflation 25.00% 23.08% 20.00% 18.55% 15.00% 10.00%

10.00% 7.70%

8.34%

8.33% 6.57%

6.75%

5.93%

5.00% 3.25% 0.00%

Fig. 2.

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

Vietnam’s Inflation from 2003 to 2012. Source: Vietnam Key Indicators (2013), ADB.

country’s continued economic expansion, rising household incomes, and the pivotal roles of commercial banks in the economy. Bank lending channel plays an important role in Vietnam as if banks react to restrictive monetary policy shocks by reducing their supply of loan, this should effect the economy especially firms and households that are dependent on bank loans. Bank lending channel of monetary policy transmission is

282

XUAN VINH VO AND PHUC CANH NGUYEN

based on the view that bank plays an important role in the financial system as an external source for financing the firms and households. Bank is an important source of credit for firms and households in emerging market economies and Vietnam banks share the similar special roles. Most of Vietnamese firms and households fully rely on banks as the only source of finance in the credit market. Other channel to raise funds is the stock markets however it is still under developments and there is a problem of liquidity in the recent years that makes it difficult for firms to raise fund in stock markets. It is therefore reasonable to expect that the lending channel is more effective than in advanced markets with developed capital market innovation making the transmission weakened (Carpenter & Demiralp, 2012). Agency costs are greater in developing economies, hindering the ability of banks to raise external finance. Imperfect substitution between sources of funds is also more severe in emerging economies because of the more systemic nature of shocks. Any change in the monetary policy stance will affect the bank behavior on the asset and liabilities sides. Over the past two decades, the Vietnamese government has undertaken a series of reforms to strengthen and modernize the banking sector as a strategic part of the country’s move toward a more open and market oriented economy. Since the reform of banking system in 1988
1989 as part of the “Doi Moi” process, the current two-tier banking system is transformed from the monobank system to serve the need of a centrally planned economy. There are four major state-owned banks in the economy as a result of this reform. In the early years of 1990s, the financial liberalization precipitate the formation of a number of small commercial banks where shareholders of these small banks are state-owned commercial banks, state-owned enterprises, and a limited numbers of private entities. In 2006
2008 can be considered a significant development in Vietnam banking system when a numbers of small banks considerably increases their capital base and diversifies shareholders. Within a relative short period of time, Vietnam banking sector has transitioned to more competitive and stronger one with a more diversified set of market participants. There are four primary types of institutions including 6 state-owned commercial banks, 37 joint stock commercial banks, 5 joint venture banks, and 5 wholly foreign-owned banks. Prior to the 2008 global financial turmoil, more lenient credit risk management by Vietnamese commercial banks significantly contribute to the easing of credit standards applied to loans and credit lines to borrowers.

Monetary Policy and Bank Credit Risk

283

Since the 2008 global financial crisis, Vietnam’s banking system exposes significant weaknesses as a result of easy and unsustainable credit growth of 32
54% per annum and Vietnamese government responses to this crisis by announcing a roadmap for banking sector restructuring. Reported nonperforming loan across Vietnamese banking sector has been significantly increasing since 2009 and 4.67% as of April 2013. Independent rating agencies as well as other economists believe the “unreported” number to be much higher and many think that this non-performing number does not reflect the true problematic status of Vietnamese bank soundness nor the credit quality of the banking system. The problems are mainly because of gaming behaviors and imprudent lending on the part of some commercial banks. According to Vietnam central bank data, loan is only account for 57% of the bank total assets on average as of the end of 2012 and there is a large proportion of assets investing in financial stocks and real estate making risk is a central issue. The risk is accelerated with the recent problem of non-performing loan. There is a number of measures from the State Bank of Vietnam in 2013 to control and manage the bank risk situation in Vietnam, however, these measures need time to be effective. Stronger capital and enhanced corporate governance are key components of reform efforts to improve the competitiveness and stability of Vietnam’s domestic commercial banks. A strong banking sector is an essential ingredient for any emerging country and the Vietnamese government and the State Bank of Vietnam start launching the restructuring plan to reform and consolidate the banking system. One of the aims is to reduce the number of small and weak credit institutions and to restructure or establish some large banks. In addition, the increased presence of foreign banks as a condition of WTO agreements drives the need for quick and strong reform further.

MODEL AND DATA Due to the difficulty in entangling demand and supply factors in measuring the effect of bank conditions on the supply of credit using aggregate data, we assume that certain characteristics of banks including size, liquidity, owner equity that influence the supply of loan. In addition, we assume that loan demand is largely independent of bank-specific characteristics and mostly dependent on macroeconomic factors.

284

XUAN VINH VO AND PHUC CANH NGUYEN

We apply the model Altunbas et al. (2010) to consider banks’ risk and bank characteristics of monetary policy transmission mechanism of lending channel. The estimated equation is as follows. ΔLOANi;t = αΔLOANi;t − 1 þ ¥2 SIZEi;t − 1 þ ω1 LIQi;t − 1 þ λ1 CAPi;t − 1 þ ν1 LLPi;t − 1 þ ϕ1 Δrt þ ϕ2 Δrt − 1 þ φ1 Δit SIZEi;t − 1 þ φ2 Δit CAPi;t − 1 þ τ1 Δit LIQi;t − 1 þ τ2 Δrt LLPit − 1 þ ψ 1 Δrt GDPi;t − 1 þ ɛi;t

ð1Þ

where LOAN is the total loan of bank, r is the interest rate, Δr is the interest change, SIZE is the logarithm of bank to total assets, LIQ is the securities and other liquid assets over total assets, CAP is the capital asset ratio, GDP is logarithm of nominal GDP. LLP is the loan-loss provisions as a percentage of loans and this is a standard ex-post accounting measure of credit risk widely used in the literature. Variables representing bank characteristics, other control variables, macroeconomics variables and interest rate variables enter the equation in lag form of one year to allow for the policy lag (Bernanke & Gertler, 1995; Ehrmann, 2000). The estimation is performed using an approach similar to that of Altunbas et al. (2010); (Altunbas et al., 2012) to analyze the link between securitization and the bank lending channel. To tackle problems derived from the use of a dynamic panel and to control for the problem of endogeneity, we use GMM panel estimator as suggested by Arellano and Bond (1991) for our analysis. We use a unique data set of bank balance sheet items macroeconomic variables for Vietnam over the period 2003
2012. Our data are collected from different sources. Bank characteristics data are hand collected from financial reports and annual reports. Other data are from Vietnam Bureau of Statistics and Asian Development Bank. Range of data from 2003 to 2012 is the longest available one. Table 1 describes variables in our analysis. Table 1. Mean Median Maximum Minimum Std. Dev. Observations

Data Description.

ΔLOAN

SIZE

LIQ

CAP

LLP

Δr

GDP

8,690,930 3,407,309 72,798,920 −9,610,010 13,537,502 223

17.2601 17.3333 20.2418 12.1975 1.5556 223

0.3811 0.3940 0.8228 0.0360 0.1490 223

0.1304 0.0946 0.7115 0.0267 0.1032 223

1.1801 0.9949 7.4249 0.0322 1.0045 223

0.1878 0.3583 6.1425 −5.3392 3.8474 223

14.2709 14.3214 14.7948 13.4805 0.4028 223

Monetary Policy and Bank Credit Risk

285

RESULTS AND DISCUSSIONS The results of the study are summarized in Table 2. The models have been estimated using the GMM estimator suggested by Arellano and Bond (1991), Arellano and Bover (1995), and extended by Blundell and Bond (1998). This technique has an advantage for addressing the Nickell (1981) bias associated with the fixed effects in short panels (e.g., bias due to the presence of the lagged dependent variable and bias due to the endogeneity of other explanatory variables) which ensures efficiency and consistency, provided the models are not subject to serial correlation of order two and the instruments used are valid (when assessed using the Sargan test). The negative effects of bank size and bank capital on bank lending suggesting that Vietnamese banks are less affected by the adverse implication of informational frictions. This is consistent with Ehrmann (2000), Ehrmann et al. (2003), and contrary to the results for the United States in many studies. This might be due to the fact that Vietnamese banks are supported by the government from failure and the good relationship between small and large banks. In addition, the result implies the fact that large and well-capitalized banks in Vietnam are less prone to increase credit without considering risks. In addition, bank risk will increase for large bank if there is a tighten monetary policy. However, if we consider in lag term, larger banks have the ability to distribute their portfolio more efficiently then bank risk will decrease. Prior crisis, large banks and well-capitalized banks tend to increase their lending significantly without considering risk. Postcrisis, large and well-capitalized banks are better in taking into account of risk by reducing their supply of credit. Liquidity has a positive impact on bank credit to the businesses and this suggests that liquid banks are well equipped to increase the loan portfolio when there is a change in the monetary policy. This reflects the fact that during the period of analysis, Vietnamese banks are having problems with liquidity and liquid banks are less controlled by the government in providing credit. This is consistent both before and after financial crisis. Therefore, liquidity is a very important factor in supply of credit to firms and households in Vietnam market. This result is also supported by the liquidity focused policies of Vietnam government on controlling of commercial banks that favor credit growth of more liquid banks and limit the growth of illiquid banks. The riskness of the credit portfolio has a positive and significant effect on the bank’s capacity to provide lending. This result is contrary to

286

Table 2. GMM Estimation Results. Variables

Whole Period

Pre-Crisis

Post-Crisis

Coeff.

P-value

Coeff.

P-value

Coeff.

P-value

Coeff.

P-value

0.6075181*** −0.0152092 0.0305502 −0.5799197 2.134499* −879103.6*** −557349.6*** −0.0013331 −0.1202333 0.0334586*** −0.3738257

0.000 0.750 0.732 0.227 0.052 0.000 0.006 0.842 0.120 0.003 0.135

6762418***

0.000

0.6039599*** −0.02462 0.0584784 −0.4717543 1.947918* −2135083*** −572375.6*** −0.0019227 −0.1640331** 0.0318969*** −0.2106317 0.8745395*** 7036670***

0.000 0.589 0.495 0.303 0.063 0.000 0.003 0.763 0.029 0.003 0.387 0.002 0.000

0.0582799 0.1543726 0.1521193 2.897592 2.130323 −4937373 −4480502** −0.0001863 −0.9046716** 0.0769726** 0.2857198 3.609171 −5500039

0.922 0.166 0.482 0.151 0.176 0.538 0.048 0.993 0.019 0.017 0.655 0.602 0.296

0.3934875*** −0.0648755 0.1043472 −0.5769268 0.4143179 −1981539*** −625150*** −0.0002556 −0.174888** 0.0290516*** −0.2386288 0.8152411*** 13200000***

0.001 0.210 0.229 0.348 0.827 0.000 0.000 0.967 0.012 0.002 0.289 0.001 0.000

*, **, *** indicates significance level of 10%, 5%, and 1% respectively.

XUAN VINH VO AND PHUC CANH NGUYEN

Δloant−1 sizet−1 liqt−1 capt−1 llpt−1 Δi Δit−1 Δi*sizet−1 Δi*capt−1 Δi*liqt−1 Δi*llpt−1 Δi*gdpt−1 Cons

Whole Period

Monetary Policy and Bank Credit Risk

287

Altunbas et al. (2010) Theoretically, other factors being equal, higher provision for loan loss reduces bank’s profit and bank capital and therefore, bank should provide less lending supply. However, Vietnamese banks seem to cover the reduction in profit by increasing credit supply and this is a serious problem to undermine the soundness of the banking system. This result confirms the suggestion from the literature that banks’ risk condition matter for the supply of loan. In this regard, there is evidence that Vietnamese banks’ debts are sensible to bank risk. More importantly, this sensitivity seems to increase as bank becomes riskier. This finding implies a very imprudent lending practice of Vietnamese commercial banks during the analyzed time period and results in a very high non-performing loan rate at the current time. The results of regression analysis indicate that Vietnamese banks are strongly affected by monetary policy shocks in the current and previous year. This result is consistent for both prior and after financial crisis. We find that increase in interest rate significantly reduce bank lending. Therefore, we can conclude that bank lending channel is very effective monetary policy transmission chancel in Vietnam. The interaction terms between size and monetary policy, bank capital and monetary policy have negative signs. In contrary with the literature on bank lending channel, large and well-capitalized Vietnamese banks are not be able to better buffer their lending activities against shocks. This is true because since the 2006 until recently, most Vietnamese banks follow the strategy of easy credit resulting in many problems for the financial stability of banks. The interaction term between liquidity proxy and monetary policy is positive and significant at the 1% level. This is very interesting to state that Vietnamese commercial banks with liquid asset structure are better equip to cope well with shock of monetary policy. However, the interaction term between loan loss provision and monetary policy has the predicted negative sign even though not significant, indicating that low-risk banks in Vietnam are more sheltered from the effect of monetary policy shocks. The effect of bank risk on lending supply may be different over the business cycle due to diverse perception of this risk. The interaction term between GDP and monetary policy is positive and significant at the 1% level. Hence, the negative effects of an increase in risk on bank loan supply is reduced in an expansionary phase and vice versa because the market perception of risk is typically reduced in good times and increased in bad times (Borio et al., 2001). There are several possible explanations for such

288

XUAN VINH VO AND PHUC CANH NGUYEN

observable facts which are myopia and herd-like behavior (Brunnermeier, 2008; Minsky, 1975), perverse incentives in managerial remuneration schemes (Rajan, 2006a), widespread use of Value-at-Risk methodologies for economic and regulatory capital purposes (Danielsson et al., 2004), pro-cyclicality of bank leverage (Adrian & Shin, 2008).

CONCLUSIONS Similar to many emerging markets, Vietnam economy has been growing at very fast rate in recent years however the economy especially firms and households are highly dependent on banks for credit. Even though the role of bank lending channel in monetary transmission has been widely studied in advanced economies, little attention has been given to investigating this issue in Vietnam. This chapter fills the gap by analyzing how bank risk influences bank credit supply and its ability to shelter that supply from the effects of monetary policy changes. We suggest that it is very important to consider bank risk, together with other standard bank-specific characteristics when analyzing the functioning of the bank lending channel of monetary policy. Using a comprehensive and unique data set of Vietnamese banks, we find that bank risks play an important role in determining banks’ loan supply and in sheltering them from the effects of monetary changes. Low-risk banks can better shield their lending from monetary shocks as they have better prospects and an easier access to uninsured fund raising. This is consistent with the “bank lending channel” hypothesis. Interestingly, the greater exposure of high-risk bank loan portfolios to monetary policy shock is attenuated in the expansionary phase, consistently with the hypothesis of a reduction in market perception of risk in good times. This chapter focuses on the analysis of banks risks and monetary policy effects, a reverse relationship might also be holding and this is an interesting revenue open to further research. Moreover, if banks were to expect some kind of insurance from the State Bank of Vietnam against the economic downturns, this is a very serious problem of moral hazard issues as banks tend to follow excessive risk taking strategies on average over the business cycle. The current situation raises an awareness to call for a careful analysis of the evolution of a number of indicators of risk premia and credit aggregates by the central bank to anticipate excessive risk taking by commercial banks.

Monetary Policy and Bank Credit Risk

289

ACKNOWLEDGMENT The authors wish to thank the Editors, Professor Jonathan A. Batten and Professor Nicklas F. Wagner, for their many helpful comments and suggestions which greatly enhance the brevity and quality of the chapter. We are grateful for the financial support from the University of Economics, Ho Chi Minh City, in conducting this research. Any remaining errors are of course our own responsibility.

REFERENCES Adrian, T., & Shin, H. S. (2008). Financial intermediaries, financial stability, and monetary policy. FRB of New York Staff Report No. 346. Adrian, T., & Shin, H. S. (2009). Financial intermediaries and monetary economics. Federal Reserve Bank of New York Staff Report No. 398. Altunbas, Y., Gambacorta, L., & Marques-Ibanez, D. (2010). Bank risk and monetary policy. Journal of Financial Stability, 6, 121
129. Altunbas, Y., Gambacorta, L., & Marques-Ibanez, D. (2012). Do bank characteristics influence the effect of monetary policy on bank risk? Economics Letters, 117, 220
222. Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The review of Economic Studies, 58(2), 277
297. Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68(1), 29
51. Bernanke, B. S., & Gertler, M. (1995). Inside the black box: The credit channel of monetary policy transmission. Journal of Economic Perspectives, 9(4), 27
48. Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(1), 115
143. Borio, C., Furfine, C., & Lowe, P. (2001). Procyclicality of the financial system and financial stability: Issues and policy options. BIS Papers, 1, 1
57. Borio, C., & Zhu, H. (2008). Capital regulation, risk-taking and monetary policy: A missing link in the transmission mechanism? BIS Working Paper No. 268. Brunnermeier, M. K. (2008). Deciphering the liquidity and credit crunch 2007
08. Cambridge, MA: National Bureau of Economic Research. Carpenter, S., & Demiralp, S. (2012). Money, reserves, and the transmission of monetary policy: Does the money multiplier exist? Journal of macroeconomics, 34(1), 59
75. Cecchetti, S. (1995). Distinguishing theories of the monetary transmission mechanism. Federal Reserve Bank of St. Louis Review, 77, 83
97. Dabla-Norris, E., & Floerkemeier, H. (2006). Transmission mechanisms of monetary policy in Armenia: Evidence from VAR analysis. International Monetary Fund Working Paper No. 06/248. Danielsson, J., Shin, H. S., & Zigrand, J.-P. (2004). The impact of risk regulation on price dynamics. Journal of Banking & Finance, 28(5), 1069
1087.

290

XUAN VINH VO AND PHUC CANH NGUYEN

Dell’Ariccia, G., & Marquez, R. (2013). Interest rates and the bank risk-taking channel. Annual Review of Financial Economics, 5(1), 123
141. Ehrmann, M. (2000). Firm size and monetary policy transmission: Evidence from German business survey data. Working Paper Series 0021. European Central Bank. Ehrmann, M., Gambacorta, L., Martinez-Page´s, J., Sevestre, P., & Worms, A. (2003). The effects of monetary policy in the euro area. Oxford Review of Economic Policy, 19(1), 58
72. Disyatat, P., & Vongsinsirikul, P. (2003). Monetary policy and the transmission mechanism in Thailand. Journal of Asian Economics, 14, 389
418. Friedman, M. (1956). The quantity theory of money: A restatement. In M. Friedman (Ed.), Studies in the quantity theory of money. Chicago, IL: Chicago University Press. Honda, Y. (2004). Bank capital regulations and the transmission mechanism. Journal of Policy Modeling, 26, 675
688. Lueangwilai, K. (2012). Monetary policy rules and exchange rate uncertainty: A structural investigation in Thailand. Procedia Economics and Finance, 2, 325
334. Minsky, H. P. (1975). John Maynard Keynes. New York, NY: Columbia University Press. Mora, N. (2013). The bank lending channel in a partially dollarized economy. Journal of Applied Economics, 6(1), 121
151. Mugume, A. (2011). Monetary transmission mechanisms in Uganda. Bank of Uganda Working Paper. Retrieved from https://www.bou.or.ug/export/sites/default/bou/bou-downloads/ research/BouWorkingPapers/2011/All/Monetary_Transmission_Mechanisms_in_Uganda_ BOU_version.pdf Mukherjee, S., & Bhattacharya, R. (2011). Inflation targeting and monetary policy transmission mechanisms in emerging market economies. IMF Working Paper (WP/11/229). Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica: Journal of the Econometric Society, 49, 1417
1426. Rajan, R. G. (2006a). Has finance made the world riskier? European Financial Management, 12(4), 499
533. Rajan, R. G. (2006b, June 8). Monetary policy and incentives. At the Bank of Spain Conference on Central Banks in the 21st Century. Retrieved from https://www.imf.org/external/np/ speeches/2006/060806.htm Taylor, J. (1995). The monetary transmission mechanism: An empirical framework. Journal of Economic Perspectives, 9(4), 11
26. Taylor, J. B. (2009). The financial crisis and the policy responses: An empirical analysis of what went wrong. National Bureau of Economic Research Working Paper. Wulandari, R. (2012). Do credit channel and interest rate channel play important role in monetary transmission mechanism in Indonesia? A structural vector autoregression model. Procedia
Social and Behavioral Sciences, 65, 557
563.

ANALYSIS OF FACTORS INFLUENCING AND CONTROLLING EXCESS CASH AND SHORT-TERM BANK LOANS IN TAIWAN Ma-Ju Wang and Yi-Ting Chang ABSTRACT This study conducts a logistic regression analysis of the ability of excess cash and short-term bank loans to substitute for each other and a multiple regression analysis of the factors influencing excess cash and short-term bank loans holdings. In addition, a questionnaire is used to survey the views of Taiwan’s corporate financial leaders on the factors influencing these two liquidity resources. The empirical results support a certain level of substitution between the two types of holdings. The regression analysis shows that for companies that would accumulate more excess cash when interest rates are low, have strong corporate performance, have low debt ratios, and whose chairman of the board and chief executive officer (CEO) are not the same person. Companies tend to have more short-term bank loans when corporate

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 291
315 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096012

291

292

MA-JU WANG AND YI-TING CHANG

performance is poor, debt ratios are high, and the chairman of the board and CEO are the same person, as well as when the degree of the deviation of control is small. We find that factors on financial structure, operating performance, cost of capital and corporate governance have significant influence on the holdings of these two liquidity facilities in regression, whereas the influence factors exclude corporate governance in questionnaire. Keywords: Excess cash; short-term bank loan; questionnaire; financial structure; corporate governance

INTRODUCTION This subject is part of working capital management decision. Especially in Taiwan, the importance of working capital management is no less than other corporate policy. A common case of black bankruptcy may result from the poor management of this policy. The research on the topic is relatively few. This study aims to bridge the gap by investigating the factors influencing and alternative of excess cash and short-term bank loans, supplementing the literature of this field and being the reference of emerging markets. The literature has mostly explored the use of corporate cash holdings for liquidity purposes as well as the factors influencing the decision to hold cash (Dittmar & Mahrt-Smith, 2007; Lin, Lin, & Kuo, 2010; Opler, Pinkowitz, Stulz, & Williamson, 1999), but few studies have investigated the use of the corporate credit lines. The common factor between the uses of these two tools is the transaction cost, and the difference between them is that the uses of cash have few regulatory and time constraints, while a credit line is subject to contractual restrictions and supervision from financial institutions. Flexible use of cash and bank loan credit lines keep the business operating smoothly and improve optional performance, therefore, both liquidity tools are essential for firms. If they are highly alternative, that will enhance the firm’s financial flexibility. Furthermore, understanding the corporate characteristics while a firm is using the two tools is valuable and can be helpful to devising working capital management policies and realizing the relationship and effect among financial decisions.

Excess Cash and Short-Term Bank Loans in Taiwan

293

Operating performance, dividend policy, financial leverage, and corporate governance are important issues for businesses, and different issues have different effects on liquidity choices. Additionally, the separation of ownership and management tends to lead to agency issues (Jensen & Meckling, 1976), thereby affecting corporate financing decisions. Yun (2009) studies the use of liquidity-influencing tools and finds that enterprises with poor corporate governance are inclined to hold cash, which is not subject to monitoring from financial institutions. Therefore, corporate governance decisions affect the corporate choice of liquidity tools. Prior to 2004, Taiwan had been ranked among the top 10 global exporters. The ranking has fallen slightly since 2004, but Taiwan is still one of the world’s major export regions in emerging markets. We use sample data from the publicly listed companies in Taiwan from 2001 to 2010 to examine whether excess cash and short-term bank loans could be substituted for each other and to investigate the impact on operating performance, dividend policy, financial leverage, and corporate governance on excess cash holdings and the utilization of short-term bank loans. Empirically, the model of Dittmar and Mahrt-Smith (2007) is employed to estimate excess cash, and data on short-term bank loans are taken from the firm’s annual report. The correlation analysis of the excess cash and short-term bank loans show that in the majority of industries, excess cash and short-term bank loans are negatively correlated. There is a different phenomenon in the loan market in Taiwan. Lins, Servaes, and Tufano (2010) study various enterprises in 29 countries and find that the median of short-term bank loans/Total Assets (TA) and cash/ TA is 15% and 9%, respectively. They find that across countries, firms make greater use of lines of credit when external credit markets are poorly developed. The average short-term bank loans/TA of the listed companies in Taiwan is approximately 7.6%, with a median of 2.7%, and the average cash/TA is approximately 12.4%, with a median of 9.3%, suggesting that the enterprises in Taiwan are more conservative in their financing than those in other countries. Moreover, the listed companies in Taiwan have their own strengths in every important industry chain, so we think that it is necessary to include the opinion of industry people on the topic of these two liquidity sources. Hence, we survey the administration of a questionnaire for chief financial officers (CFO) to replenish the omission of empirical method promoting the comprehensiveness of the issue. Logistic regression analysis shows that when interest rates are low, the firm is small and has only been listed for a short period of time, the return on equity (ROE) is low, the debt ratio is small, the chairman of the board

294

MA-JU WANG AND YI-TING CHANG

and chief executive officer (CEO) are the same person, and the degree of deviation of the control is large, there is significant substitution between the two types of funds. The multiple regression analysis of the empirical results shows that the lower the interest rates, the more likely the enterprise is to accumulate excess cash. When the debt ratio is low, the operating performance is strong, the market-to-book ratio is high, and the chairman of the board and CEO are not the same person, then the excess cash holdings of the firm are high. When the firm is small and has been listed for a short period of time, it may have higher costs of external financing and usually holds a larger amount of excess cash as well. In terms of the impact on short-term bank loans, non-electronics industries utilize more short-term bank loans. When the operational performance is good and cash dividends are paid, then short-term bank loans are utilized less frequently. Firms with many investment opportunities do not necessarily utilize short-term bank loans. However, when the chairman of the board is also the CEO, the degree of deviation of the control is small, and the enterprise is small and has been listed for a long period of time, then short-term bank loans are utilized more often. Finally, the significant variables from the regression analysis and the most important factors for the use of funds from the questionnaire are compared, and the results show nearly consistent between the significance of the empirical research and the importance indicated by the industry in the questionnaire. The economic implications of the consistent results show that operating performance, financial structure, and cost of capital dealing with banks are critical factors that affect two tools holding. Both the regression analysis and questionnaire are used to test the hypotheses in this study, which differs from the previous literature. However, the biggest difference from the two methods is that corporate governance is also a significant influence variable in the regression, whereas the ranking of options relevant to the interests of investors is almost at the back of items in the questionnaire. Therefore, the study argues that it still needs to take corporate governance into account for managers when making decision about controlling short-term capital. The chapter is structured as follows. The section “Literature and Hypotheses” discusses the literature review and hypothesis development. The next section describes the data and variables. Section “Survey Results and Empirical Analysis” provides a survey and empirical results as well as multivariate analysis and robustness checks. The last section concludes the chapter.

Excess Cash and Short-Term Bank Loans in Taiwan

295

LITERATURE AND HYPOTHESES Cash Holdings and Short-Term Bank Loans Dittmar, Mahrt-Smith, and Servaes (2003) and Dittmar and Mahrt-Smith (2007) point out that the efficiency of corporate governance structures and cash holdings are related. Good corporate governance can double the cash value. Therefore, enterprises with agency conflicts can be expected to hold more cash (Jensen, 1986; Lin et al., 2010). According to the pecking order theory, capital structure may be the result of financing decisions. Firms without optimal cash holdings are less likely to raise funds in the capital markets, preferring to hold their cash (Acharya, Almeida, & Campello, 2007; Almeida, Campello, & Weisbach, 2004; Myers & Majluf, 1984). Kim, Mauer, and Sherman (1998) construct a trade-off model to explore the factors influencing corporate holdings of liquidity tools and to balance the optimal level of liquidity between the two types of holdings (Dittmar et al., 2003; Opler et al., 1999). Opler et al. (1999) show that when a firm has great growth opportunities, its cash accounts for a higher ratio of net assets. Large enterprises have easy access to loans in the capital markets, so the ratio is relatively low. Denis and Sibilkov (2010) also find a strong correlation of the cash holdings of firms with financing restrictions and the degree of investment, mainly for the purpose of reducing the cost of external financing. Duchin (2010), however, finds that the cash holdings of firms with diversified investments are significantly lower than those of specialized manufacturing companies. Lins et al. (2010) show that only when there is a good relationship between the company and the bank will the line of credit be widely used, but the company cannot use the line of credit to cover the investment funding gap over the long-term because the credit contracts are categorized as short-term bank loans. Sufi (2009) finds that enterprises maintain stable cash flows in order to secure bank credit contracts; concurrently, to avoid default, a bank will give an appropriate line of credit to regulate the enterprises. Sufi (2009) argues that revolving credit facilities can resolve the friction involved in facing capital markets in the future and are more operationally efficient than relying on internal cash. Yun (2009) reports that managers or controlling shareholders would prefer the company to hold cash for their personal interests; therefore, using the credit line can reduce agency issues. Therefore, the company’s optimal liquidity reflects the trade-off between the costs and benefits of the

296

MA-JU WANG AND YI-TING CHANG

two liquidity tools. Lin, Ma, Malatesta, and Xuan (2012) find that the larger the deviation between the control rights of the firm’s largest and ultimate owner and the cash flow right is, the more the underwriting syndicate for financing tends to favor creditors that are geographically close to the firm and domestic banks that understand the industry for convenient supervision. Lins et al. (2010) find that when prospects are poor, the firm holds cash (non-operating cash) and uses it to cover operating needs, and when prospects are good, the line of credit can be used to provide the funds necessary for investment opportunities in the future. Broadly speaking, the credit line would provide the needed funds for the firm’s future growth opportunities, and non-operating cash can be regarded as a risk-aversion tool.

Factors Influencing Excess Cash and Short-Term Bank Loans and Hypotheses Substitution between Excess Cash and Short-Term Bank Loans Lins et al. (2010) show that there are certain relations between externally owned funds and internally owned funds, and the two sources of funding can be substituted for each other (Sufi, 2009). We deduce that the higher the firm’s excess cash holdings are, the less bank loans are utilized, and vice versa. Therefore, we propose hypothesis 1: Hypothesis 1. Excess cash is negatively correlated with short-term bank loans.

The Impact of Operating Performance on Excess Cash and Short-Term Bank Loans Sufi (2009) shows that profitability may have a positive or negative impact on the two types of liquidity. If a firm’s past revenues are stable, then nonoperating cash and profitability are positively correlated, and if a firm has stable revenues, then that firm will use internal funds accumulated from previous profits to meet its future capital needs instead of using external funding. Under these circumstances, bank credit lines and profitability will be negatively correlated. The results show that enterprises with a good profitability treat non-operating cash and credit lines as substitutes for liquidity.

Excess Cash and Short-Term Bank Loans in Taiwan

297

Based on the study of Lins et al. (2010), this study uses ROE as a measure of profitability and predicts that the higher the expected profitability is, the more the firm accumulates non-operating cash and the lower the need for external funding. Therefore, we propose hypothesis 2: Hypothesis 2-1. Profitability has a positive impact on excess cash. Hypothesis 2-2. Profitability has a negative impact on short-term bank loans. Lins et al. (2010) argue that a company will consider the use of their line of credit as a source of funds for future acquisitions or investment opportunities and hold non-operating cash as a risk-aversion tool to cover shortages of future cash flows or current financial gaps faced by the company. Sufi (2009) finds that the line of credit cannot be used to tide over the current financial distress faced by enterprises because the risk of corporate default is high when prospects are bad. Dittmar et al. (2003) show that the higher the market-to-book ratio (MB), the bigger the firm’s growth potential is, requiring more cash for investment. Therefore, the firm will retain more internal funds. We hypothesize that more investment opportunities indicate a higher need for funds; thus, investment opportunities have a positive impact on excess cash and short-term bank loans. Thus, we propose hypothesis 3: Hypothesis 3-1. Growth or investment opportunities have a positive impact on excess cash. Hypothesis 3-2. Growth or investment opportunities have a positive impact on short-term bank loans. Impact of Dividend Payments on Excess Cash and Short-Term Bank Loans Smith and Warner (1979) and Nini, Smith, and Sufi (2009) find that if the enterprises maintain a fixed dividend policy, they would prefer to hold nonoperating cash. Lins et al. (2010) argue that if the enterprise maintains dividend payments, this indicates that the firm may have consistent profits, with plenty of internal funds. Generally, when there are capital needs, the firm might cut dividends and use cash as a potential source of funding, which will reduce the need for excess cash or lines of credit. Therefore, dividend-paying enterprises will not regard the line of credit and nonoperating cash as substitutes.

298

MA-JU WANG AND YI-TING CHANG

This study assumes that dividend-paying enterprises would retain more cash, and because bank credit contracts stipulate the purposes of credit funds, the line of credit is less likely to be used to pay dividends. Therefore, we propose hypothesis 4: Hypothesis 4-1. Dividend payment is positively correlated with excess cash. Hypothesis 4-2. Dividend payment is negatively correlated with shortterm bank loans. The Impact of Financial Leverage on Excess Cash and Short-Term Bank Loans Lins et al. (2010) show that during financial difficulties, enterprises use non-operating cash to cover shortages of cash flows. Enterprises will use credit lines for future investment opportunities with growth. For firms with fixed profits, the degree of leverage is not expected to affect whether the enterprises use non-operating cash and credit lines as alternatives. In fact, financial leverage may affect the extent of corporate cash holdings. Bates, Kahle, and Stulz (2009) find that the more a firm borrows, the more that interest payments to creditors will reduce the corporate accumulated excess cash, suggesting that non-operating cash holdings and financial leverage are negatively correlated. Both Acharya et al. (2007) and Gamba and Triantis (2008), however, find that financial leverage and cash holdings are positively correlated. Firms with large amounts of financial leverage may retain more cash or have higher credit lines due to huge capital needs to cope with future investment opportunities. Gamba and Triantis (2008) find that the optimal flexible use of financing for the enterprise involves borrowing with proper use of the funds. The use of financial leverage and liquidity policy can also increase the value of the firm. Following Gamba and Triantis (2008), this study deduces that enterprises with more financial leverage have an appetite for risk and will hold more liquidity to meet their financing or investment needs. Therefore, financial leverage is expected to have a positive impact on both excess cash and short-term bank loans. Therefore, we propose hypothesis 5: Hypothesis 5-1. Financial leverage has a positive impact on excess cash. Hypothesis 5-2. Financial leverage has a positive impact on short-term bank loans.

Excess Cash and Short-Term Bank Loans in Taiwan

299

Impact of Corporate Governance on Excess Cash and Short-Term Bank Loans Lins et al. (2010), Sufi (2009), and Yun (2009) report that firms with high agency costs would not use these two types of tools as substitutes. Nonoperating cash is not subject to bank monitoring and is positively correlated with agency issues, while credit lines are subject to contractual monitoring and management and are negatively correlated with agency issues. Sufi (2009) and Yun (2009) point out that non-operational cash differs from the line of credit in that managers or controlling shareholders prefer to hold cash for their personal interests. Dittmar et al. (2003) find that the agency issue is the most important factor affecting the company’s cash holding decisions. Enterprises with managers who disregard the interests of shareholders will tend to hold more cash. Tsui, Jaggi, and Gul (2001) show that firms for which the chairman of the board also serves as the CEO operate less efficiently and tend to have more excess cash (Lin et al., 2010). Yun (2009) shows that only when the firm implements anti-takeover policies to fight against the threat of external control will managers have higher cash holdings. In contrast, Opler et al. (1999) and Mikkelson and Partch (2003) argue that there is no evidence to show that agency cost is an important factor that affects cash holdings. Harford, Mansi, and Maxwell (2008), however, find that enterprises with poor corporate governance often have less excess cash, while the enterprises with better corporate governance have higher dividend payments and retain more cash. The quality of corporate governance and cash holdings are, therefore, positively correlated. According to the previously mentioned literature, the poorer the corporate governance; the more serious the agency issue, and the higher the excess cash holdings. Conversely, more oversight from external agencies creates better corporate governance, less serious agency issues, and then more short-term bank loans. Therefore, we propose hypothesis 6: Hypothesis 6-1. Good corporate governance has a negative impact on excess cash. Hypothesis 6-2. Good corporate governance has a positive impact on short-term bank loans. The empirical part of this study also considers control variables, such as the firm size, interest rates, and the length of the listing period, and the subsequent survey also considers competitors and relationships with banks.

300

MA-JU WANG AND YI-TING CHANG

DATA AND VARIABLES Information Sources and Sample Selection The empirical analysis is divided into a regression analysis and the administration of a questionnaire. For the regression analysis, the variable data are taken from the Taiwan Economic Journal (TEJ) database. The research sample is mainly composed of the companies in non-financial and noninsurance industries listed in Taiwan from 2001 to 2010. This study includes a total of 734 companies. Because the two study variables (excess cash and short-term bank loans) have high short-term liquidity, quarterly data for a total of 10 years are used. A total of 13,696 valid quarterly observations are collected. The total number of questionnaires sent out is 150 for paper questionnaires and 1,400 for email questionnaires. The distribution period for the questionnaires is September 2011 to January 2012, with a total of 450 recovered questionnaires, including 264 valid questionnaires and 186 questionnaires with incomplete data. Those that are not filled out correctly and those from non-accounting managers or non-executives are excluded. The recovery rate of effective questionnaires is 17.03%. The electronics industry account for approximately 50.38% of respondents, and non-electronics industries account for approximately 49.62%.

Variable Definition The subsequent empirical research uses logistic regression analysis to explore the relation between excess cash and short-term bank loans and multiple regression analysis to investigate the impact of operating performance, dividend payments, financial leverage, and corporate governance on excess cash and short-term bank loans. Finally, cash holdings are evaluated as an alternative variable for excess cash. The following provides the definitions and descriptions of the variables relevant to the study hypotheses. Dependent Variables 1. The mutual substitution between excess cash and short-term bank loans (Y) This is a dummy variable for the correlation analysis of excess cash holdings and short-term bank loans. If there is a significant negative

Excess Cash and Short-Term Bank Loans in Taiwan

301

correlation, then Y = 1. If the two are not correlated or positively correlated, then Y = 0 (Lins et al., 2010). Because not every company has excess cash every year, we first categorize the companies according to industrial sector and use Pearson correlation analyses to analyze the excess cash and short-term bank loans in a given year in for that particular industry. If the excess cash and short-term bank loans are significantly negatively correlated, then Y = 1; otherwise, Y = 0. 2. Excess cash holdings (Excess Cash/TA) Dittmar and Mahrt-Smith’s (2007) cash model (1) is used to estimate the cash holdings of individual firms required under normal operating conditions. Excess cash is defined as actual cash holdings minus the cash required under the normal operating conditions per quarter. Excess cash is not the cash portion of corporate operations and investments but is the portion of the actual cash holdings greater than the normal cash holdings; therefore, only firms with positive post-calculation excess cash are used in the study sample. The postcalculation excess cash value is then divided by the book value of TA. The greater the value is, the higher the firm’s excess cash holdings.



Cashi;t FCFi;t NWCi;t Ln þ β3 = β0 þ β1 Ln NAi;t þ β2 NAi;t NAi;t NAi;t  MVi;t RDi;t þ β4 ðIndustrySigmaÞi;t þ β5 þ ɛi;t þ β6 NAi;t NAi;t

ð1Þ

where LnCashi,t is the amount of cash for firm i in quarter t, measured by the natural logarithm of cash and cash equivalents; Ln(NAi,t) is the net assets of firm i in quarter t, measured by the natural logarithm of TA minus cash and cash equivalents; FCFi,t is the free cash flow (FCF) of firm i in quarter t, calculated as the net operating profit minus interest expenses and income taxes; NWCi,t is the net operating capital of firm i in quarter t, calculated as liquid assets minus liquid debt minus cash and cash equivalents; IndustrySigmai,t is measured by the standard deviation of the particular industry per quarter of firm i in quarter t, calculated using the net operating profit minus interest expenses and income taxes divided by TA minus cash and cash equivalents (FCF/NA); MVi,t is the market value of equity plus total liabilities of firm i in quarter t; and RDi,t is the research and development expenses of firm i in quarter t.

302

MA-JU WANG AND YI-TING CHANG

1. Utilization of short-term bank loans (line of credit/TA) The line of credit in this study includes short-term loans from financial institutions and issued commercial papers. Issuing commercial papers are part of the bank credit line, and those still in circulation are included in the credit line (Sufi, 2009; Yun, 2009). The data are collected from TEJ financial statements, which are divided by TA after the shortterm loans from non-financial institutions are subtracted. 2. Cash holdings (cash/TA) The total cash holdings used for operational purposes and nonoperating cash are highly correlated, so total cash is used as an alternative variable for excess cash, calculated as cash and cash equivalents divided by TA (Lins et al., 2010). Explanatory Variables ROE is used as a variable to measure profitability, which is the net aftertax profit divided by total shareholder equity (Lins et al., 2010). The MB ratio is the market value of equity divided by the book value of equity and measures the investment or growth opportunities of the firm (Dittmar et al., 2003). Dividend payment is set as a dummy variable to measure the firm’s dividend policy. If there is payment of cash dividends, then Dividend = 1, otherwise, it is zero (Lins et al., 2010). The debt ratio (Leverage) is calculated as total liabilities/TA to measure the level of financial leverage (Lins et al., 2010). According to agency theory, if the chairman of the board is also the CEO and has both implementation as well as supervision rights, the Board has lost the function of fair and objective oversight, which lowers operating performance, possibly resulting in serious agency problems and poor corporate governance (Tsui et al., 2001). This factor is set as a dummy variable, Duality. If the chairman of the board and the CEO are the same person, the possibility of agency problems is higher, and Duality = 1; otherwise, it is zero. The earnings share deviation rate (Deviate) is calculated as the earnings distribution rights (or cash flow rights) of the control shareholders divided by their control rights and is used as a variable to measure the quality of corporate governance mechanisms. The smaller the value is, the greater the degree of deviation, making it easier for controlling shareholders to impact company policies, which is less in the interest of the general retail shareholders (La Porta, Lopez-de-Silanes, & Shleifer, 1999). Firm size (Size) is calculated as the natural logarithm of the book value of TA. The length of the listing period is the listed time extending to the

Excess Cash and Short-Term Bank Loans in Taiwan

303

time of this study and is a dummy variable. If the period is longer than 5 years, the listed variable is set to one; otherwise, it is zero (Lins et al., 2010). The interest rates of the primary market of 91
180 day commercial papers issued in Taiwan are used as interest rates (CP). With regard to industrial sectors (Industry), a dummy variable is set to one for the electronics industry and zero otherwise.

Descriptive Statistics and Variable Analysis The electronics industry has the largest share of the entire sample. The average and median of their cash holdings are higher than those of the entire sample, suggesting that there are differences in the utilization of liquidity in the electronics industry compared to other industries. Therefore, for this study, the sample is divided into electronics and non-electronics industries. For the entire sample, cash and excess cash are significantly positively correlated (0.9108); therefore, cash can be used as an alternative variable to excess cash. Excess cash or cash and short-term bank loans are significantly negatively correlated in nearly all industries. There are significant differences between the electronics and nonelectronics industries for all explanatory variables influencing excess cash and short-term bank loans except dividend payment. Table 1 shows the descriptive statistics. From 2001 to 2010, the electronics industry does not grow as quickly as it has in the past, but the average ROE of the electronics industry is higher than that of the entire sample as well as that of the nonelectronic industries. The average investment opportunities variable is 1.906 in the electronics industry, which is higher than that of the nonelectronics industries and the entire group, suggesting that high-tech industries still has growth opportunities and a competitive edge. Taken together and on average, based on the two corporate governancerelated variables, the quality of corporate governance in the electronics industry is somewhat lower than that of the non-electronics industries. Taiwan is conservative in finance compared with the data of several countries in the study of Lins et al. (2010). The degree of leverage is also lower in the electronics industry than in the non-electronics industries. Interest rates are also much lower for the electronics industry, providing financial advantages. The correlation coefficient for listing time and industry sectors (Industry) is the highest (−0.385) but still shows a low negative correlation,

304

Table 1. Variable

Total Industry Sample (N = 13,696)

Electronics (N = 6,557)

Non-Electronics (N = 7,139)

Variation

Average

Standard deviation

Average

Standard deviation

Average

Standard deviation

t-value

15.697 0.02 0.636 1.568 0.193 0.371 0.261 0.798 0.017 0.479

1.286 0.062 0.481 1.123 0.395 0.162 0.439 0.263 0.009 0.5

15.728 0.024 0.443 1.906 0.193 0.351 0.299 0.735 0.016 1

1.381 0.061 0.497 1.195 0.394 0.147 0.458 0.293 0.008 0

15.668 0.016 0.813 1.258 0.193 0.39 0.227 0.856 0.018 0

1.191 0.063 0.39 0.952 0.395 0.173 0.419 0.216 0.009 0

2.709*** 7.718*** −48.279*** 34.896*** −0.122 −14.459*** 9.535*** −27.409*** −8.344***

***, **, and * indicate a significance level at the 1%, 5%, 10%, respectively.

MA-JU WANG AND YI-TING CHANG

Size ROE Listed MB Dividend Leverage Duality Deviate CP Industry

Descriptive Statistics for Each Variable for Various Industries.

Excess Cash and Short-Term Bank Loans in Taiwan

305

suggesting that the electronics industry is relatively young. The remaining Pearson correlation coefficients are generally smaller, so there is no colinearity.

SURVEY RESULTS AND EMPIRICAL ANALYSIS Survey Using the discussion in Lins et al. (2010) as a reference in this study, questionnaire items suitable for Taiwanese companies are selected. A Likerttype scale is employed, and the subjects answer questions according to the following five-point scale: most important (5 points), important (4 points), normal (3 points), not important (2 points), and least important (1 point). The items are divided into two major domains. Table 2 shows the comparison of the responses to the two areas of focus of the questionnaire: factors that affect company decisions on the amount of excess cash (9 questions) and factors that affect company decisions on the utilization of short-term bank loans (11 questions). Table 2 shows the most important factors that affect the selection of the two liquidity tools are future short-term operating cash needs and the interest rates of bank loans. The soundness of the company’s financial structure is also considered important for bank loans. Next, among the nine questions that relate to the factors that affect company decisions on excess cash holdings, the averages for all of the questions range from 2.70 to 4.47. Similarly, among the 11 questions that relate to the factors that affect company decisions on the utilization of short-term bank loans, the averages of all of the questions range from 2.53 to 4.37. The results suggest that these two sets of questions are all valid, and according to the subjects, they are related to factors that more or less affect decisions on excess cash holdings and the utilization of shortterm bank loans. The questions that are used in this study are based on the relevant theory and on a literature review; moreover, we use feedback solicited from corporate financial managers with empirical experience. The finalized questionnaire reflects a test-run, discussions, and modifications. The survey subjects have a certain level of understanding of the research topic, and the survey questions therefore have good validity.

Ranked and Average Responses to Survey Questions on Excess Cash and Short-Term Bank Loans.

Factors that Affected Company Decisions on the Amount of Excess Cash (9 Questions) Future short-term operating cash needs The interest rates of bank loans

Percentage Ranking Average

1

4.47

70.45

2

3.73

The difficulty of securing bank loans When capital needs arise, the company could issue securities at a fair and reasonable price Opinions of major shareholders or directors and supervisors

67.05

3

3.74

62.5

4

3.61

57.2

5

3.53

Used for good investment opportunities Using excess cash as a surplus to be shared with shareholders Opinions of the creditors of the company The excess cash holdings of competitors

52.65

6

3.47

46.97

7

3.35

23.86

8

2.77

21.21

9

2.7

Factors that Affected Company Percentage Ranking Average Decisions on the Utilization of ShortTerm Bank Loans (11 Questions) Future short-term operating cash needs The soundness of the company’s current financial structure The negotiation capability of the company with the bank The interest rates of bank loans

When capital needs arise, the company could issue securities at a fair and reasonable price The difficulty of securing excess cash internally Opinions of major shareholders or directors and supervisors Used for good investment opportunities Opinions of the creditors of the company Using bank loans as a surplus to be shared with shareholders The bank loans holdings of competitors

93.03

1

4.37

90.98

2

4.25

73.77

3

3.81

73.36

4

3.81

60.25

5

3.54

56.56

6

3.53

54.1

7

3.43

42.21

8

3.19

25

9

2.85

24.59

10

2.83

15.16

11

2.53

The percentage and ranking columns in the table reflect the added value of the percentage of the most important (5 points) and important (4 points) responses for each question among all subjects and are ranked according to the percentage value. Average refers to the mean of all five responses (5
1 points) on the questionnaire, including most important, important, normal, not important, and least important.

MA-JU WANG AND YI-TING CHANG

92.42

306

Table 2.

Excess Cash and Short-Term Bank Loans in Taiwan

307

The scale scores of the two domains of the questionnaire are then added for correlation analysis, and the results show that the criterion-related validity between the first set of 9 questions and the second set of 11 questions is 0.829; the correlation coefficient is therefore significant, suggesting the strong validity of this questionnaire. In addition, Cronbach’s α values for the sets of questions related to excess cash and short-term bank loans are 0.666 and 0.8, respectively, suggesting a certain level of reliability for this questionnaire and that the test data are consistent and stable. Logistic Regression Analysis In accordance with the aforementioned hypothesis, excess cash and shortterm bank loans could substitute for each other. The Lins et al. (2010) study constructs are referenced to establish the logistic regression used in this study. Table 3 shows the empirical results. The logistic regression is as follows: Yi;t = β0 þ β1 Sizei;t þ β2 ROEi;t þ β3 Listedi;t þ β4 MBi;t þ β5 Dividendi;t þ β6 Leveragei;t þ β7 Dualityi;t þ β8 Deviatei;t þ β9 CPt

ð2Þ

þ β10 Industry þ ɛi;t

Table 3.

Empirical Correlation Analysis for Excess Cash and Short-Term Bank Loans.

Variable

β

Wald

Constant Size ROE Listed MB Dividend Leverage Duality Deviate CP Industry

2.972 −0.122*** −0.008** −0.167** −0.016 0.099 −0.613*** 0.110* −0.214* −18.370*** 20.928

59.690 28.307 4.201 6.128 0.377 2.495 18.046 3.498 3.434 46.678 0.002

The Dependent variable Y comprises the alternative dummy variables excess cash and short-term bank loans. If there is significant negative correlation, then Y = 1; otherwise, Y = 0. ***, **, and * indicate a significance level at the 1%, 5%, 10%, respectively.

308

MA-JU WANG AND YI-TING CHANG

The interest rates (CP) have the most significant explanatory power for substitution between and excess cash and short-term bank loans. When the interest rates for the money market and cost of capital are low, the firm is likely to regard excess cash and short-term bank loans as substitutes. Firm size (Size) and listing time (Listed) negatively impact liquidity substitution. For a larger firm and longer listing time, the excess cash and short-term bank loans cannot substitute for each other. Conversely, for a smaller and younger firm, sources of funding are more limited, and excess cash and short-term bank loans can substitute for each other. For financial leverage (Leverage), the greater financial leverage in a firm lowers liquidity substitution. This result is likely because the firm’s increased financial risk produce a heavier interest burden and limited financing flexibility. The firms with less financial leverage have greater financing flexibility with more flexibility in dispatching the two funds; thus, the substitution would be more significant. ROE that indicates corporate operating performance is also negatively correlated with liquidity substitution, which suggests no substitution between the two fund types in enterprises with good operating performance. This result is likely because firms with good operating performance tend to accumulate cash and have easier access to bank loans. Firms with low ROE has low profits and secured a lower line of credit from the bank. Therefore, substitution between the two liquidity tools is significant because of limited funding. For the corporate governance mechanism, if the board chairman is also the CEO, deviation from the control rights is high, and then the firm is more likely to regard excess cash and short-term bank loans as a substitute for each other. In summary, 7 of the 10 variables are significant, which indirectly suggests a certain level of substitution between excess cash and short-term bank loans and supports hypothesis 1.

Multivariate Regression Analysis Factors that Affect Excess Cash Holdings Given the aforementioned factors that influence corporate excess cash holdings, a multivariate regression model is used to empirically analyze and verify hypotheses 2-1, 3-1, 4-1, 5-1, and 6-1 herein. The impact of operating

309

Excess Cash and Short-Term Bank Loans in Taiwan

Table 4.

Factors that Affect Excess Cash Holdings and Short-Term Bank Loans. Excess Cash/TA

Variables Constant Size ROE Listed MB Dividend Leverage Duality Deviate CP Industry Adjusted R2 F-statistic N

β

t-value ***

0.1639 −0.0041*** 0.0004*** −0.0067*** 0.0066*** −0.004 −0.1019*** −0.0073*** −0.0033 −0.747*** 0.0403*** 0.1662

17.533 −7.135 3.448 −4.174 10.61 −2.392 −24.669 −4.843 −1.269 −10.023 26.732 273.91 13,696

Credit Lines/TA β

t-value ***

0.1057 9.962 −0.011*** −17.018 −0.0012*** −9.355 0.0071*** 3.894 −0.0065*** −9.183 −0.0035* −1.877 0.3969*** 84.667 0.0058*** 3.379 0.0166*** 5.565 0.5722*** 6.762 −0.0431*** −25.159 0.4318 1,041.728 13,696

***, **, and * indicate a significance level at the 1%, 5%, 10%, respectively.

performance, dividend policy, financial leverage, and corporate governance on excess cash holdings is explored, and the empirical results are shown on the left side of Table 4. The regression is as follows: ExcessCashi;t = β0 þ β1 Sizei;t þ β2 ROEi;t þ β3 Listedi;t þ β4 MBi;t TAi;t þ β5 Dividendi;t þ β6 Leveragei;t þ β7 Dualityi;t þ β8 Deviatei;t

ð3Þ

þ β9 CPt þ β10 Industry þ ɛi;t

Most variables have a highly significant impact. The variables with the highest t-value are industry sector showing a positive coefficient and the financial leverage showing a negative coefficient. This indicates that there is high excess cash holding in the electronics industry. For a higher financial leverage, the excess cash holdings are lower, which do not support hypothesis 5-1 that suppose financial leverage positively impacts excess cash. This result is most likely because a firm with a high debt ratio must also pay high interest charges to creditors, and thus, excess cash is accumulated less.

310

MA-JU WANG AND YI-TING CHANG

In operating performance, for larger MB ratios and ROE, investment opportunities are greater and excess cash is easier to accumulate; therefore, the more profitable enterprises have higher excess cash holdings. These two results support hypotheses 2-1 and 3-1. Interest rate (CP) also negatively impact excess cash. When the interest rates are high, the firms tended to pay with excess cash. For corporate governance, when the board chairman and CEO (Duality) is not the same person, the supervision quality might be better, and the firm has more excess cash. The dividend payment (Dividend) as well as earnings shares deviation (Deviate) do not affect excess cash, which do not support hypothesis 4-1 that supposes dividend payment positively impacts excess cash and hypothesis 6-1 that supposes corporate governance negatively impacts excess cash. For smaller firms, a shorter listing time, and higher external financing costs, the firms typically have higher excess cash holdings. Factors that Affect the Utilization of Short-Term Bank Loans Given the aforementioned influential factors for short-term bank loans, a multivariate regression model is used to empirically analyze and verify hypotheses 2-2, 3-2, 4-2, 5-2, and 6-2. The impact of operating performance, dividend policy, financial leverage and corporate governance on short-term bank loans is explored, and the empirical results are shown on the right side of Table 4. The regression is as follows: CreditLinesi;t = β0 þ β1 Sizei;t þ β2 ROEi;t þ β3 Listedi;t þ β4 MBi;t TAi;t þ β5 Dividendi;t þ β6 Leveragei;t þ β7 Dualityi;t

ð4Þ

þ β8 Deviatei;t þ β9 CPt þ β10 Industry þ ɛi;t Table 4 shows that all of the independent variables significantly impact short-term bank loans. The positive coefficient for financial leverage (t = 84.667) support hypothesis 5-2, which supposes that financial leverage positively impacts short-term bank loans. With more debt, firms have more short-term bank loans. For industry sector (t = −25.159), non-electronics industries utilize more short-term bank loans perhaps because non-electronics industries include traditional industries with a longer listing time and good relationship with the bank. Therefore, such industries secure a higher credit line. In addition, the electronics

Excess Cash and Short-Term Bank Loans in Taiwan

311

industry grows fast and is more risky; thus, high financial leverage is unsuitable. For operating performance, both profitability (ROE) and investment opportunities (MB) negatively impact short-term bank loans. The empirical results support hypothesis 2-2, which supposes that enterprises with good operating performance and stable revenue have sufficient internal funds and thus utilize fewer short-term bank loans. However, hypothesis 3-2, which supposes that growth or investment opportunities positively impact short-term bank loans, is not supported probably because firms need more long-term funds for investment opportunities, not short-term bank loans. Interest rates (CP) and short-term bank loans are positively correlated. For higher interest rates in commercial papers, issuing commercial papers is less attractive, and firms may switch to lending through the signed loan agreement. For corporate governance, high earnings share deviation ratio (lower deviation for control rights) may increase policy flexibility and generate more short-term bank loans. These two variables are positively correlated, which supports hypothesis 6-2. When the board chairman is not the CEO, the firm has less short-term bank loans, which does not support hypothesis 6-2. For dividend policy, a dividend payment (Dividend) has a weak but significant negative impact on short-term bank loans. The firms that pay dividends utilize less short-term bank loans, which support hypothesis 4-2 and the financing pecking order theory. For the control variables, with a smaller firm size and a longer listing time, they use more short-term loans. The left and right sides of Table 4 show that, among the 10 regression variables, the impact from 8 variables on excess cash and short-term bank loans is opposing. Although this is not a direct test of substitution, it indirectly suggests that, under most circumstances, firms consider the trade-off between the two fund types, which indirectly supports hypothesis 1.

Robust Test According to the correlation analysis for excess cash and cash, the cash for all the sample firms and estimated excess cash are significantly positively correlated. Therefore, this study also uses cash as the alternative variable to excess cash to verify the factors that influence cash holdings. Table 5 shows the robust test. A comparison shows that the factors affecting excess cash and cash as well as their correlation are almost

312

MA-JU WANG AND YI-TING CHANG

Table 5.

Empirical Results for Cash as the Alternative Variable to Excess Cash.

Variables

β

t-value

Constant Size ROE Listed MB Dividend Leverage Duality Deviate CP Industry

0.2890*** −0.0097*** 0.0010*** −0.0088*** 0.0236*** −0.0033* −0.1464*** −0.0100*** −0.0095*** −0.8886*** 0.0736***

28.473 −15.753 8.161 −5.063 34.914 −1.815 −32.659 −6.111 −3.335 −10.980 44.934

Adjusted R2 F-statistic N

0.3851 858.813 13,696

***, **, and * indicate a significance level at the 1%, 5%, 10%, respectively.

identical. Therefore, generally the coefficients are identical with greater significance. Cash could be regarded as an alternative to excess cash, and the deduction derived from the variables can be similar. The regression is as follows: Cashi;t = β0 þ β1 Sizei;t þ β2 ROEi;t þ β3 Listedi;t þ β4 MBi;t þ β5 Dividendi;t TAi;t þ β6 Leveragei;t þ β7 Dualityi;t þ β8 Deviatei;t þ β9 CPt

ð5Þ

þ β10 Industry þ ɛi;t

Consistency between the Regression Analysis and Questionnaire Survey The explanatory variables (Size, ROE, and Listed) for the regression analysis are categorized into Item 1 in the left portion of the questionnaire in Table 2. They are relevant to operating cash needs. MB is categorized into questions on investment opportunities (Item 6, Table 2). Dividend is categorized into questions on earnings distribution and sharing among shareholders (Item 7, Table 2). Duality and Deviate are categorized into

Excess Cash and Short-Term Bank Loans in Taiwan

313

questions on major shareholders and directors and supervisors (Item 5, Table 2). CP is categorized into questions on interest rates (Item 2, Table 2). Leverage is categorized into Item 2 in the right portion of Table 1 and is relevant to financial structure. The importance of the explanatory variables with strong significance, including (Size, ROE, Listed), Leverage and CP, were ranked 1st, 2nd, and 4th, respectively, in the corresponding 9 or 11 questions on the questionnaire, including relative factors about operating performance, financial structure and relevant cost of capital, as shown in Table 3. Therefore, the significance of variables in regression and the importance recognized by managers in questionnaire are similar, and that enhanced the reliability of the results herein. Only the corporate governance variables (Duality, Deviate) significantly affecting short-term bank loans receive less attention for managers.

CONCLUSION For this study, we use the companies listed in Taiwan during 2001
2010 as samples to explore substitution between excess cash and short-term bank loans and to investigate the impact on operating performance, dividend policy, financial leverage and corporate governance on excess cash holdings and use of short-term bank loans. The Dittmar and Mahrt-Smith (2007) model is used to estimate excess cash, and survey a questionnaire for managers so that the variables those are not precisely measured through regression analyses can be included. First, the results based on an industry classification show that, excess cash and short-term bank loans in all industries are negatively correlated. Further, logistic regression analysis shows that there is a certain level of substitution between excess cash and short-term bank loans. In addition, when the interest rate in the money market is low, the firm size is small, the listing time is short, the debt ratio is low, the ROE is low, the board chairman and the CEO are the same person, and the deviation for the control rights as well as cash flow rights is high, then the substitution between the two fund types is significant. Next, empirical results from the multivariate analysis show that when the interest rate in the money market is low, the operating performance is good, MB ratio is high, the debt ratio is low, and the board chairman and CEO is not the same person, the firms tended to accumulate excess cash.

314

MA-JU WANG AND YI-TING CHANG

For the impact on short-term bank loans, when the operational performance is good, a cash dividend is paid, and the debt ratio is low, then the firms use less short-term bank loans. More investment opportunities do not cause use of more short-term bank loans. When the board chairman is also the CEO, the deviation for control rights is lower; the enterprises use more short-term bank loans. In the questionnaire survey results, the results show that the operational needs, bank dealings and the company’s financial structure are considered regarding short-term funds, regardless of internal or external funds. Finally, the significant variables from the regression analysis and the top few important influential factors in fund use of the questionnaire survey are compared, the results show near consistency between significance in the empirical study and the importance recognized in the industry through the questionnaire, which enhances the reliability of the results herein. Only corporate governance variables, which are strongly significant in regression analysis, are more unnoticed for managers, reminding the managers should promote the importance of corporate governance no matter long-term or short-term corporate important decision.

REFERENCES Acharya, V., Almeida, H., & Campello, M. (2007). Is cash negative debt? A hedging perspective on corporate financial policies. Journal of Financial Intermediation, 16(4), 515
554. Almeida, H., Campello, M., & Weisbach, M. (2004). The cash flow sensitivity of cash. Journal of Finance, 59(4), 1777
1804. Bates, T., Kahle, K., & Stulz, R. (2009). Why do U.S. firms hold so much more cash than they used to? Journal of Finance, 64(5), 1985
2021. Denis, D. J., & Sibilkov, V. (2010). Financial constraints, investment, and the value of cash holdings. Review of Financial Studies, 23(1), 247
269. Dittmar, A., & Mahrt-Smith, J. (2007). Corporate governance and the value of cash holdings. Journal of Financial Economics, 83(3), 599
634. Dittmar, A., Mahrt-Smith, J., & Servaes, H. (2003). International corporate governance and corporate cash holdings. Journal of Financial and Quantitative Analysis, 38(1), 111
133. Duchin, R. (2010). Cash holdings and corporate diversification. Journal of Finance, 65(3), 955
992. Gamba, A., & Triantis, A. (2008). The value of financial flexibility. Journal of Finance, 63(5), 2263
2296. Harford, J., Mansi, S. A., & Maxwell, W. F. (2008). Corporate governance and firm cash holdings in the US. Journal of Financial Economics, 87(3), 535
555. Jensen, M. C. (1986). Agency costs of free-cash-flow, corporate finance, and takeovers. American Economic Review, 76(2), 323
329.

Excess Cash and Short-Term Bank Loans in Taiwan

315

Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305
360. Kim, C., Mauer, D., & Sherman, A. (1998). The determinants of corporate liquidity: Theory and evidence. Journal of Financial and Quantitative Analysis, 33(3), 335
359. La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (1999). Corporate ownership around the world. Journal of Finance, 54(2), 471
517. Lin, C., Ma, Y., Malatesta, P., & Xuan, Y. (2012). Corporate ownership structure and bank loan syndicate structure. Journal of Financial Economics, 104(1), 1
22. Lin, J. H., Lin, M. L., & Kuo, W. H. (2010). Corporate governance and the payment decisions of excess cash holdings. Sun Yat-Sen Management Review, 18(4), 1051
1088. Lins, K. V., Servaes, H., & Tufano, P. (2010). What drives corporate liquidity? An international survey of cash holdings and lines of credit. Journal of Financial Economics, 98(1), 160
176. Mikkelson, W. H., & Partch, M. M. (2003). Do persistent large cash reserves hinder performance? Journal of Financial and Quantitative Analysis, 38(2), 275
294. Myers, S., & Majluf, N. (1984). Corporate financing and investment decisions when firms have information the investors do not have. Journal of Financial Economics, 13(2), 187
221. Nini, G., Smith, D., & Sufi, A. (2009). Creditor control rights and firm investment policy. Journal of Financial Economics, 92(3), 400
420. Opler, T., Pinkowitz, L., Stulz, R., & Williamson, R. (1999). The determinants and implications of corporate cash holdings. Journal of Financial Economics, 52(1), 3
46. Smith, C., & Warner, J. (1979). On financial contracting: An analysis of bond covenants. Journal of Financial Economics, 7(2), 117
161. Sufi, A. (2009). Bank lines of credit in corporate finance: An empirical analysis. Review of Financial Studies, 22(3), 1057
1088. Tsui, J. S. L., Jaggi, B., & Gul, F. A. (2001). CEO domination, growth opportunities, and their impact on audit fees. Journal of Accounting, Auditing and Finance, 16(3), 189
208. Yun, H. (2009). The choice of corporate liquidity and corporate governance. Review of Financial Studies, 22(4), 1447
1475.

BANK COMPETITION, MANAGERIAL EFFICIENCY AND THE INTEREST RATE PASS-THROUGH IN INDIA Jugnu Ansari and Ashima Goyal ABSTRACT If banks solve an inter-temporal problem under adverse selection and moral hazard, then bank specific factors, regulatory and supervisory features, market structure, and macroeconomic factors can be expected to affect banks’ loan interest rates and their spread over deposit interest rates. To examine interest rate pass-through for Indian banks in a period following extensive financial reform, after controlling for all these factors, we estimate the determinants of commercial banks’ loan pricing decisions, using the dynamic panel data methodology with annual data for a sample of 33 banks over the period 1996
2012. Results show commercial banks consider several factors apart from the policy rate. This limits policy pass-through. More competition reduces policy pass-through by decreasing the loan rate as well as spreads. If managerial efficiency is high then an increase in competition increases the policy pass-through and the vice-versa. Reform has had mixed effects, while managerial

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 317
339 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096013

317

318

JUGNU ANSARI AND ASHIMA GOYAL

inefficiency raised rates and spreads, product diversification reduced both. Costs of deposits are passed on to loan rates. Regulatory requirements raise loan rates and spreads. Keywords: Banks; panel data; interest rates; net interest income; operating cost

INTRODUCTION In the wake of a balance of payment crisis, India adopted banking reform as part of a revamp of the financial system. Although the reforms focused on removing financial repression through reductions in statutory preemptions, while stepping up prudential regulations, they encompassed various dimensions. First, the level of competition was gradually increased within the banking system by allowing greater participation of domestic private and foreign banks while giving banks greater freedom in pricing and allocation of credit. Second, measures were taken to develop various segments of financial markets such as money, bond, credit, foreign exchange and equity. Newer instruments were introduced to allow financial institutions, savers and investors opportunities for diversification, optimization of return and risk on their portfolios and effective management of liquidity and other risks. Third, in order to ensure stability of the financial system, banks were subjected to international best practices in prudential regulation and supervision tailored to Indian requirements. Fourth, measures were taken to improve the institutional arrangements including the legal framework and technology platform for effective, cost efficient and sound payment and settlement system. Finally, consistent with the new institutional architecture for the financial system in general and the banking sector in particular, the monetary policy framework made a phased shift from direct instruments of monetary management such as cash reserve and statutory liquidity requirements to an increasing reliance on indirect instruments such as short-term policy interest rate including repo and reverse repo rates. The shift from a traditional quantum of money to an interest rate channel of monetary transmission mechanism envisaged that banks would be guided by market conditions and balance sheet pressures along with regulatory and prudential requirements, while pricing their assets and liability components. Commercial banks’ loan pricing decisions are important for policy. First, competitiveness and efficiency of banks in financial intermediation affect

Bank Competition, Managerial Efficiency and the Interest Rate

319

price discovery in the loan market. This is measured by loan interest rates and their spreads over deposit interest rate and risk free yield on government securities. Loan interest rates in turn affect economic growth and macroeconomic stability (Levine, 1997). Second, loan interest rates affect banks’ loan asset quality and credit risks which have implications for the stability of a bank based financial system. Third, for successful conduct of monetary policy through the interest rate channel commercial banks should adjust loan interest rates in tandem with policy actions. However, the policy interest rate can constitute only one of the several factors considered by banks in the determination of loan interest rates. Numerous studies have examined the rigidity in banks’ lending decisions in response to a policy shock. Illustratively, in the wake of recent global crisis in 2008
2009, the Reserve Bank of India (RBI) pursued a softer interest rate policy stance to stimulate the economy by way of slashing the policy rate by 475 basis points. However, banks’ lending rates declined only by 100
250 basis points. Subsequently, RBI raised the policy rate 13 times in response to hardening of inflation. However, banks did not adequately revise deposit and lending rates. Transmission had still not improved. So analysis of factors influencing banks’ loan pricing decisions would be a useful input for policy. Moreover, studies on the subject are non-existent in the Indian context. We estimate the determinants of commercial banks’ loan pricing decisions, using the dynamic panel data methodology and annual data for a sample of 33 banks including public, private and foreign banks over the period 1996
2012. Results show commercial banks consideration of several factors apart from the policy rate limits policy pass-through. The structure of the chapter is as follows: the section ‘The Literature’ presents review of literature followed by methodology and data in the section ‘Methodology’, summary statistics and empirical findings in the section ‘Empirical Analysis’ and conclusion in the section ‘Conclusion’.

THE LITERATURE The seminal works of Klein (1971), Monti (1972) and Ho and Saunders (1981) have inspired numerous studies to analyse commercial banks’ loan pricing decisions. Klein (1971) and Monti (1972) postulated a theory of a banking firm and demonstrated how in a static setting demands and supplies of deposits and loans simultaneously clear both markets. Mcshane and Sharpe (1985), and Allen (1988) have extended and modified the dealership model to a greater extent. Mcshane and Sharpe (1985)

320

JUGNU ANSARI AND ASHIMA GOYAL

considered interest uncertainty from loan and deposit returns to money market rates. Allen (1988) extended the model for various types of loans with interdependent demands. The dealership model has been criticized on the grounds that it failed to recognize the bank as a firm having a certain production function associated with provision of intermediation services (Lerner, 1981). Cost inefficiencies in the production process across banks can have a distortionary effect on the margin. Bank specific factors such as bank size, capitalization, liquidity, managerial inefficiency (MI), non-interest operating expenses, loan quality, deposit growth, interest rate risk, credit risk, ownership, non-interest incomes and risk aversion are identified by multiple studies as the important determinants of interest margins (Demirguc-Kunt & Huizinga, 1999; Liebeg & Schwaiger, 2007a). The market structure focuses on the competition in the banking sector (market power), bank concentration, and financial sector liberalization. Finally, the macroeconomic view focuses on inflation rate, GDP growth, exchange rate, interest rate policies, gross national savings, and investment and capital formation as factors driving interest spreads and margins in the banking system. Estrada, Gomez, and Orozco (2006) argue that interest margin is positively affected by inefficiency. But Hamadi and Awdeh (2012), Maudos and De Guevara (2004) and Maudos and Solı´ sc (2009) postulate that efficiency/ quality of management is negatively correlated with net interest margin (NIMs). Studies on credit risk show both negative and positive impact. Liebeg and Schwaiger (2007b) and Williams (2007) found a negative impact of credit risk on the interest margin. On the contrary, Maudos and De Guevara (2004) and Maudos and Solı´ sc (2009) showed a positive sign for credit risk as well as interest rate risk. Hamadi and Awdeh (2012) found liquidity negatively correlated with NIMs for domestic banks. As regards to operating cost, risk aversion and loan quality, Liebeg and Schwaiger (2007a), Maudos and De Guevara (2004) and Maudos and Solı´ sc (2009) in their respective studies show a positive impact of either one or all of these variables on interest margin. Implicit taxes include reserve and liquidity requirements whose opportunity cost tend to be higher as they are remunerated at less than market rates. In contrast, explicit taxes translate into higher interest margins. Studies suggest that corporate tax is fully passed on to customers in poor as well as rich countries. This is aligned with the common notion that bank stock investors need to receive a net of company tax returns that is independent of the company tax (Demirguc-Kunt & Huizinga, 1999).

Bank Competition, Managerial Efficiency and the Interest Rate

321

Goyal (2014) and Ansari and Goyal (2011) provided evidence in support of monetary transmission through the banks in Indian context for the passthrough of call money rates to bank lending rates, for different sectors and by type of bank ownership. Pass-through was also affected by size and the degree of competitiveness. Since pass-through falls with competitiveness, it is higher to the extent the Indian banking sector is less competitive. Boone (2008) assumes that more efficient firms (i.e. firms with lower marginal costs) will gain higher market shares or profits, and that this effect will be stronger the heavier competition in that market is. In order to support this intuitive market characteristic, Boone develops a theoretical measure, found to be more robust than any other methods, viz. Price Cost Margin (PCM), Herfindahl
Hirschman Index (HHI), the Panzar-Rosse measure of competition (H-statistic). The studies support that macroeconomic factors are important determinants in explaining variations in interest margin. Studies have found inflation to be associated with higher interest margins as it entails higher transaction costs (Demirguc-Kunt & Huizinga, 1999). As far as impact of GDP growth on interest margin is concerned, Liebeg and Schwaiger (2007a) and Hamadi and Awdeh (2012) have contrasting views. To summarize, the above discussion suggests that determinants and impacts of bank interest margins vary considerably. Multiple factors contribute to high spreads and margins especially in a less developed financial system. Generally, interest spreads are higher in developing countries than developed countries due to a mix of the factors explained above.1 Therefore country, time and context specific studies are required.

METHODOLOGY We assume that there are some adjustment costs stemming from asymmetric information.2 This is modelled as a quadratic loss function following Nickell (1985) and Winker (1999). It generates a tractable linear decision rule. The loss function for bank k in period t is the following: Γt;k =

∞ X s=0

φsk ½Ω1;k ðrL;k;t þ s − φk rP;t þ s Þ2 þ Ω2;k ðrL;k;t þ s − rL;k;t þ s − 1 Þ2

ð1Þ

where Ω1 and Ω2 represent the weight that the bank gives to achieving the long-run target value for the lending rate (rL) and the cost of moving to

322

JUGNU ANSARI AND ASHIMA GOYAL

that target value, respectively. The variable (rP) represents the policy rate of interest. The φk is a function of the demand elasticity and the probability of repayment that bank k faces, whereas Ωj, j = 1, 2 depends on the bank’s average loan risk. If the portion of past-due loans for bank ‘k’ is higher, the adverse selection or moral hazard problem for that bank becomes more important and the bank will give more weight to changes in the interest rate, which implies a slower adjustment. On minimizing Eq. (1) with respect to rL, we obtain: rL;k;t þ s = ðΩ1;k =ðΩ1;k þ Ω2;k ÞÞφk rP;t þ s þ ðΩ2;k =ðΩ1;k þ Ω2;k ÞÞ rL;k;t þ s − 1

ð2Þ

Eq. (2) shows that the impact coefficient depends on the size of Ω1,k relative to Ω1,k + Ω2,k and the mark up φk. Therefore, the long-run coefficient is always larger than the short-term coefficient. The bank’s loan risk determines φk and Ω2,k, the lower the probability of repayment (higher risk), the higher are both φk and Ω2,k. If the debtors are too risky and the effect on Ω2,k is more important, the bank may not completely pass through a money market interest rate increase (in the short run) because of debtor moral hazard. In the long run, however, the interest rate charged will reflect the risk characteristic of the debtor. In other words, unpaid loans should have a negative effect on the impact coefficient and a positive effect on the long-term multiplier. A dynamic panel data model that accounts for risk persistence and endogeneity of the bank specific controls and for the time persistence in the loan pricing structure is adopted. This, often used in the literature for analysing commercial banks’ loan pricing decisions, identifies and measures effects that are not detectable in pure cross-section or pure time-series data; deals with the problem of heterogeneity; and investigates the dynamics of change due to external factors affecting dependent variables. The main feature of a dynamic panel data specification is the inclusion of a lagged dependent variable in the set of explanatory variables, that is, yi;t = αyi;t − 1 þ βðLÞXi;t þ ηi þ ɛi;t ;

jαj < 1;

i = 1; :::; N;

t = 1; :::; T

ð3Þ

where the subscripts i and t denote the cross sectional and time dimension of the panel sample respectively, yi,t is the lending rate, β(L) is the lag polynomial vector, Xi,t is a vector of explanatory variables other than yi, t − 1, ηi is the unobserved individual (bank specific) effects and ɛi,t are the error terms.

Bank Competition, Managerial Efficiency and the Interest Rate

323

We have measured competitiveness index using Augmented Relative Profit Difference (ARPD). Boone (2008) proposed a competition measure with robust theoretical properties. Using a bank level panel data set, we test the empirical validity of the Augmented RPD measure for competition in the Indian loan market. Theoretically, loan market competition increases in two ways. First, competition increases when the produced services of various banks become closer substitutes and when entry costs decline. So the following relationship between market share and marginal cost can be set up (Leuvensteijn, 2011): lnðsi Þ = α þ βlnðmci Þ

ð4Þ

where the loan market share of bank i, (si) = (loan)i/total loan, and parameter β is the Boone measure of competition. A more negative beta reflects stronger competition. We calculate marginal costs instead of approximating this variable with average costs. Since marginal costs are unobservable, we have calculated marginal costs from Translog Cost Function (TCF) with linear homogeneity in the input prices and cost exhaustion restrictions using individual bank observations.

lnðcit Þ = α0 þ

H −1 X

∝h d h þ

K X K X

δt dt þ

t=1

h=1

þ

T −1 X

K X

βj lnðxijt Þ

j=1

γ jk lnðxijt Þlnðxikt Þ þ νit

ð5Þ

j=1 k=1

∂cit ∂lnðcit Þ = ðcit =xilt Þ ∂lnðxilt Þ ∂xilt X mcilt = ðcit =xilt Þðβl þ 2γ l lnðxilt Þ þ γ lk lnðxikt Þdi Þ mcilt =

ð6Þ ð7Þ

Given the estimated marginal costs (Ansari & Goyal, 2011), we are able to estimate the Boone measure by using the following equation: lnðsit Þ = α þ

X

βt lnðmcit Þ þ

X

γ t dt þ uit

ð8Þ

where s stands for market share, mc for marginal costs, i refers to bank, and d to year; dt are time dummies, and uit is the error term. dh is the bank

324

JUGNU ANSARI AND ASHIMA GOYAL

type dummies (h= public sector banks, private sector banks and foreign sector banks) This provides us with the coefficient β, the Boone Competitiveness Index. In our empirical analysis we consider alternative measures of banks’ loan pricing decisions in terms of the dependent variables bank lending interest rate (LR) and the spread of loan interest rate over deposit interest rate. The models estimated, after incorporating the competitiveness index (ARPD), are as follows. LRi;t = αLRi;t − 1 þ θmpt þ γARPDt þ βARPDt mpt þ CoFit X þ αk Xi;t þ ηi þ ei;t

ð9aÞ

LRi;t = αLRi;t − 1 þ θmpt þ γARPDt þ ϕARPDt MIt þ CoFit X þ αk Xi;t þ ηi þ ei;t

ð9bÞ

RSi;t = αRSi;t − 1 þ θmpt þ γARPDt þ βARPDt mpt X þ αk Xi;t þ ηi þ ei;t

ð10aÞ

RSi;t = αRSi;t − 1 þ θmpt þ γARPDt þ ϕARPDt MIt X þ αk Xi;t þ ηi þ ei;t

ð10bÞ

where i = 1…n, k = 1…m, t = 1…T, CoF is the cost of fund and X is the vector of control variables and bank specific characteristics viz., bank size, CRAR, loan maturity, MI, product diversification (PD), return on equity (ROE), bank liquidity and asset quality. Finally, ηi is a bank specific effect.

EMPIRICAL ANALYSIS Sample and Variables The empirical analysis based on the dependent variable interest rate spread (IRS) rests on the assumption of a complete adjustment of loan 1 2 3 interest rate ðrL;t ; rL;t and rL;t Þ with respect to deposit interest rate 1 2 3 ðrD;t ; rD;t and rD;t Þ and the spread is attributable to a host of other factors. In the second instance, we relax this assumption and thus study the loan interest rate as the dependent variable as a function of various explanatory variables including the deposit interest rate.

Bank Competition, Managerial Efficiency and the Interest Rate

325

In this context, it is useful to take note of a caveat here. In the real world, commercial banks’ loan portfolio could comprise numerous borrowers with different loan interest rates, reflecting different characteristics of borrowers. A similar argument could hold for depositors. Accordingly, empirical research has to rely on a derived measure of loan and deposit interest rates based on banks’ balance sheet data. We have experimented with three measures of loan interest rates based on annual balance sheet data for total interest income RL,t generated from loans and advances and the outstanding loans ‘L’ as shown below: RL;t Lt

ð11Þ

RL;t Lt − 1

ð12Þ

RL;t þ RL; t − 1 Lt þ Lt − 1

ð13Þ

1 rL;t =

2 rL;t =

3 = rL;t

1 The first measure LR1 ðrL;t Þ could account for effective loan interest 2 rate. The second measure LR2 ðrL;t Þ recognizes that the interest income earned in the current period relates to loans extended in the beginning 3 of the year (previous year). The third measure LR3 ðrL;t Þ recognizes stock-flow (SF) concept, that is banks could not only earn interest income from loans extended in the previous period but also current 1 period. In the same manner, we derived deposit interest rates ðrD;t ; 2 3 rD;t and rD;t Þ. As regards the explanatory variables, we have used monetary policy rate and regulatory variable prudential capital to risk weighted assets ratio (CRAR) consistent with India’s monetary policy and banking sector regulation frameworks. For bank specific variables, we have indicators of bank size (SIZE) defined in terms of ratio of a bank’s total assets to the banking industry aggregate measure; liquidity ratio, that is liquid assets less liquid liabilities to total assets ratio; operating cost to assets ratio as an indicator of MI; asset quality measured by gross non-performing loans to total loans ratio; earnings and profitability in terms of ROE; PD represented by non-interest income to total asset ratio; return on

326

JUGNU ANSARI AND ASHIMA GOYAL

investment as the ratio of interest received on G-sec investment divided by total investment in G-sec;3 and loan maturity defined as the share of term loans in total loans. For macro variables, we have used real GDP growth rate and inflation rate from the wholesale price index. Our sample consists of 33 banks comprising 27 public, 3 private and 3 foreign banks, which together account for the bulk of commercial banking system in India with three-fourth share in total deposits, credit, investment and other indicators. The majority of the sample comprises the public sector banks. We are not able to control for the ownership variable here due to the low number of bank sampling units under private and foreign sector bank groups.

Descriptive Statistics Our data set is the annual accounts of 33 commercial banks over the period 1996
2012. The data source is RBI published ‘Statistical Tables Relating to Banks in India’. Tables 1 and 2 provide descriptive statistics for some bank specific variables. Table 1 shows reduction in MI of banks as the operating cost to assets ratio fall. However, the ROE variable showed greater cross-section variability than loan IRSs. The non-interest income ratio, reflecting PD, showed an increasing trend during 1996
2007 and some moderation thereafter. The size variable exhibited steady trend during the sample period, reflecting banks’ ability to maintain their competitiveness in financial intermediation. Banks, however, showed more substantial variation in net liquidity ratio than in loan and deposit interest rates. Loan maturity showed an increasing trend during the sample period. Loan interest rate and their spreads over deposit interest rates showed some moderation during 2002
2007 as compared with the late 1990s (Table 2). For the more recent period from 2008, loan spreads have shown some firming up as compared with the first half of the 2000s but they remain lower than the late 1990s. This trend also was observed in terms of cross-section variability (standard deviation) of loan interest rates and spreads. Deposit interest rates more or less showed lower variability than loan interest rates during the late 1990s, except for the year 1997.

327

Bank Competition, Managerial Efficiency and the Interest Rate

Table 1.

Bank Specific Descriptive Statistics.

Year

Loan Maturity

Product Diversification

Managerial Inefficiency

Return on Equity

Liquidity

Size

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

29.5(13.3) 33.0(15.3) 34.5(15.2) 35.4(12.4) 36.1(13.1) 36.4(11.6) 40.3(13.1) 43.5(12.7) 48.1(10.7) 53.0(11.1) 55.3(10.9) 58.0(11) 57.5(11.6) 58.2(12.4) 57.7(12.5) 56.5(12.1) 57.4(11.9)

1.4(0.6) 1.4(0.6) 1.5(0.7) 1.3(0.6) 1.4(0.5) 1.4(0.5) 1.7(0.6) 1.9(0.5) 2.0(0.5) 1.5(0.5) 1.1(0.5) 1.1(0.5) 1.3(0.6) 1.3(0.7) 1.2(0.5) 1.0(0.4) 1.1(0.3)

3.2(0.9) 2.9(0.5) 2.7(0.5) 2.7(0.5) 2.5(0.5) 2.7(0.5) 2.4(0.5) 2.4(0.4) 2.3(0.5) 2.2(0.6) 2.1(0.4) 1.9(0.4) 1.7(0.5) 1.6(0.4) 1.6(0.4) 1.7(0.4) 1.5(0.5)

19(23.9) 13.7(8.1) 14.9(7.2) 14.2(7.0) 14.6(6.7) 13.1(7.7) 15.3(7.2) 19.3(7.5) 22.2(6.1) 15.9(6.2) 13.8(5.4) 15.8(4.1) 16.2(4.8) 16.2(4.7) 16.0(4.9) 15.1(3.9) 15.9(4.0)

9.6(56.8) 0.5(7.6) 1.4(6.4) −0.2(9.4) −1.5(12.4) −3.6(18.5) −8.9(26.6) −3.5(11.8) −3.4(13.3) −15.1(49.5) −16.8(26.5) −13.8(15.8) −11.2(10.4) −12.2(10.3) −12.0(10.9) −15.5(12.7) −13.6(11.9)

3.0(4.7) 3.0(4.6) 3.0(4.6) 3.0(4.7) 3.0(4.6) 3.0(4.8) 3.0(4.4) 3.0(4.3) 3.0(4.0) 3.0(3.8) 3.0(3.6) 3.0(3.4) 3.0(3.5) 3.1(3.6) 3.1(3.5) 3.2(3.4) 3.3(3.2)

Note: Standard deviation in parenthesis.

Table 2.

Interest Rate Spreads and Loan Rates: Descriptive Statistics.

Year

RS1

RS2

RS3

LR1

LR2

LR3

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

5.5(1.7) 6.6(1.5) 5.1(1.4) 4.5(1.1) 3.8(0.9) 3.8(1.1) 3.1(1.4) 3.5(1.1) 3.4(1.0) 3.2(0.9) 3.3(1.0) 3.5(0.9) 3.5(1.2) 4.1(1.6) 3.8(1.4) 3.9(0.8) 3.8(1.0)

9.1(7.7) 4.2(15.2) 5.9(2.1) 5.0(1.5) 5.2(1.8) 5.0(1.7) 4.4(1.8) 4.1(0.9) 4.2(1.2) 5.3(3.9) 4.8(1.0) 4.9(1.2) 4.5(1.6) 4.9(1.8) 4.3(1.2) 5.0(0.9) 4.9(1.1)

5.1(1.1) 6.1(1.3) 5.8(1.3) 4.7(1.1) 4.1(0.9) 3.8(1.0) 3.3(1.3) 3.3(0.9) 3.5(1.0) 3.3(0.9) 3.2(0.9) 3.4(0.9) 3.5(1.1) 3.8(1.4) 3.9(1.5) 3.9(1.1) 3.6(0.9)

12.4(2.3) 14.0(2.0) 12.1(1.4) 11.7(1.4) 10.9(1.0) 10.7(1.0) 9.6(1.6) 9.4(0.9) 8.2(0.9) 7.3(0.8) 7.3(0.5) 8.0(0.6) 9.0(0.8) 9.8(1.3) 8.9(0.9) 8.6(0.7) 8.7(0.8)

17.3(9.1) 17.0(6.9) 14.7(2.1) 14.0(1.9) 13.7(1.8) 13.0(1.3) 12.1(1.6) 11.0(1.3) 9.8(1.1) 10.1(3.6) 9.6(0.7) 10.5(0.8) 11.2(1.1) 11.9(1.5) 10.3(0.8) 10.5(0.8) 9.9(0.9)

11.8(1.2) 13.3(1.8) 13(1.5) 11.9(1.3) 11.2(1.1) 10.8(0.9) 10.0(1.5) 9.5(0.9) 8.8(0.9) 7.7(0.9) 7.3(0.5) 7.7(0.5) 8.5(0.7) 9.4(1.0) 9.3(1.1) 8.7(0.7) 8.6(0.9)

Note: Standard deviation in parenthesis.

328

JUGNU ANSARI AND ASHIMA GOYAL

Results We have used the ARPD measure, estimating marginal costs using Translog Cost Function. The ARPD measure quantifies the impact of marginal costs on performance, measured in terms of market shares. The original Boone’s RPD is improved by calculating marginal costs instead of approximating marginal costs by average variable costs. Competition measured by using the ARPD in Table 3 is statistically significant in a consistent manner for our sample. This table also contains the non-parametric RPD measure estimated as area under the curve and two other traditional but non-robust measures of competition. The absolute value of the ARPD measure is used in our regressions since it has a straightforward interpretation. Higher the coefficient in absolute sense, higher is the competition. The competition in the banking sector increased after 2002 except a marginal decrease during 2006
2007. All the values of the coefficients lie between 0 and 1, which indicates monopolistic competition in the Indian loan market.

Table 3.

The Augmented RPD (ARPD) Measure of Competition.

Year

ARPD (Parametric Measure)

RPD (Non-Parametric Measure)

H-Stat

PCM

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

−0.315(0.086)** −0.340(0.087)** −0.356(0.085)** −0.307(0.086)** −0.384(0.091)** −0.359(0.088)** −0.453(0.091)** −0.404(0.086)** −0.384(0.080)** −0.390(0.067)** −0.341(0.061)** −0.360(0.051)** −0.388(0.061)** −0.452(0.050)** −0.448(0.059)** −0.493(0.063)** −0.501(0.057)**

0.557 0.615 0.347 0.399 0.479 0.392 0.185 0.434 0.513 0.468 0.429 0.521 0.620 0.668 0.691 0.632 0.563

0.666 0.734 0.646 0.627 0.633 0.615 0.440 0.466 0.646 0.656 0.458 0.616 0.584 0.492 0.503 0.514 0.509

0.514 0.553 0.504 0.463 0.427 0.427 0.344 0.423 0.454 0.467 0.462 0.448 0.397 0.332 0.381 0.402 0.391

Note: Figures in parenthesis are standard deviation of estimates. **indicates level of significance at 5%.

Bank Competition, Managerial Efficiency and the Interest Rate

329

The empirical results on the determinants of banks loan pricing decisions using alternative measures of loan interest rate and its spread are presented in Tables 4 and 5. Comparing the maximum likelihood value of all the three interest rate specifications shows the second specification has the highest value, then the third specification and the least value is for the first specification in the static framework. So the second specification is the better proxy for measuring the lending rate. First, the policy rate has a statistically significant positive effect on loan interest rates but the magnitude of impact, as measured by the size of the coefficient of policy rate, is quite moderate. The short-term interest rate pass-through ranges from 17 to 56 basis points for RS specification and 23 to 60 basis points for LR specification. This suggests an imperfect monetary transmission mechanism and rigidity in loan pricing decisions of banks due to various factors as explained by the other control variables. Second, interest rate pass-through depends significantly on the competitiveness index (ARPD). The impact of competition on interest rate passthrough for both the LR specification and RS specification is negative and highly significant. The long-run (dynamic)

pass-through coefficient can be t calculated using the formula θ þ βARPD It depends upon the intensity of 1−α competition and persistence in the model. The LR model shows that the cost of funds is fully recovered while pricing a loan. A positive policy shock lead to increase in the cost of funds and hence a lower spread. The spread depends on the difference between the lending rate and the cost of funds and not only between lending rate and the deposit rate. Further the policy multiplicative coefficient of the competitiveness index is higher in the RS specifications (1 and 3) as compared to LR (1 and 3). This indicates the possibility of competition in both the deposit market and in the loan market, and implies a rise in policy rates reduces RS more than LR. Additionally, the lag effect (persistence) is higher in the RS specification which could lead to lower pass-through. Results from interaction of MI with the competitiveness index are intuitive and significant. MI increases rates and spreads, but its interaction with competition decreases both. The coefficient of competition-MI interaction is double that of the interaction of competition with policy rate. This implies to the extent the banking system is efficient, an increase in competition would increase policy pass-through. Third, banks recover the cost of deposit funds from borrowers and earn a positive spread. This is captured by the intercept term in the regressions. The intercept terms are 4.7 and 5.2 under the LR2 specification in Table 4.

330

Table 4. Variables

Determinants of Bank Lending Rate (LR).

With Competition-Policy Interaction LR2

LR1 Coefficient

LR (t − 1) Policy rate (MP) Competitiveness index (ARPD) ARPD*MP/MI Cost of deposit fund Return on investment Loan maturity Managerial inefficiency (MI) Product diversification Return on equity Size Bank liquidity Asset quality CRAR GDP growth Inflation Intercept

0.460*** 6.860 0.164* 0.270*** 3.740 0.238* −0.334** −1.980 −0.098

t

Coefficient

LR3 t

Coefficient

LR1 t

Coefficient

LR2 t

Coefficient

LR3 t

Coefficient

t

1.710 0.626*** 11.440 0.454*** 6.700 0.160* 1.660 0.618*** 11.120 1.710 0.225*** 3.710 0.500*** 6.350 0.590*** 4.400 0.395*** 5.650 −0.360 −0.056 −0.390 −0.919*** −4.810 −0.996*** −3.650 −0.488*** −2.730

−0.753*** −5.550 −1.159*** −4.710 −0.532*** −4.030 −1.933*** −5.450 −2.987*** −4.620 −1.436*** −4.080 0.791*** 8.850 0.947*** 5.330 0.504*** 4.880 0.807*** 9.070 0.962*** 5.450 0.522*** 5.080 −0.199** −2.130 −0.187** −2.050 −0.141* −1.740 −0.199** −2.200 −0.182** −2.040 −0.145* −1.850 0.001 0.010 0.005 0.580 −0.001 −0.120 −0.001 −0.100 0.005 0.680 −0.000 −0.060 0.015 0.150 0.267* 1.720 0.094 1.120 0.010 0.100 0.270* 1.740 0.090 1.100 −0.220

−1.130 −0.396** −2.010 −0.346** −2.360 −0.215

−1.120 −0.402*** −3.311 −0.341** −2.420

0.003 0.350 0.018* 1.570 0.004 0.550 0.003** 0.390 0.018* 1.600 0.005 0.610 −0.027*** −2.690 −0.057*** −5.110 −0.022** −2.160 −0.027* −2.690 −0.057*** −5.100 −0.022** −2.170 −0.002 −0.780 −0.040** −2.450 −0.001 −0.370 −0.002** −0.780 −0.040** −2.450 0.001 0.360 0.017 1.150 0.052*** 2.600 0.000 0.000 0.016* 1.120 0.052*** 2.580 0.001 0.070 0.005 0.140 0.009 0.270 0.013 0.480 0.002*** 0.060 0.005 0.150 0.012 0.430 0.191*** 4.390 0.076* 1.870 0.171*** 5.570 0.196* 4.500 0.085** 2.090 0.175*** 5.710 0.081*** 2.950 0.071* 1.770 −0.020 −0.860 0.090*** 3.300 0.084** 2.060 −0.014 −0.610 1.226 1.200 5.011*** 3.940 0.500 0.630 1.124 1.100 4.774*** 3.790 0.436 0.550

Note: ***, ** and * indicate level of significance at 1%, 5% and 10%, respectively.

JUGNU ANSARI AND ASHIMA GOYAL

Bank lending rate

With Competition-Efficiency Interaction

Determinants of Interest Rate Spread (IRS).

With Competition-Policy Interaction

Variables

RS1

RS2

Coefficient

IRS (t − 1) Policy rate (MP) ARPD ARPD*MP/MI Return on investment Loan maturity Managerial inefficiency Product diversification Return on equity Size Bank liquidity Asset quality CRAR GDP growth Inflation Intercept

0.538*** 7.750 0.171** 1.890 0.766*** 0.178** 2.190 0.242** 6.051 0.030 −0.220 −1.260 −0.097 −0.360 0.024 −0.665*** −4.920 −1.060*** −4.420 −0.441*** −0.124* −1.470 −0.147* −1.530 −0.077 0.003 0.460 0.003 0.370 −0.004 0.029 0.340 0.252* 1.800 0.069 −0.170

Coefficient

RS3

Interest rate spread

t

With Competition-Efficiency Interaction

t

Coefficient

RS1 t

Coefficient

RS2 t

Coefficient

RS3 t

Coefficient

t

17.670 0.534*** 7.570 0.169** 1.870 0.762*** 17.200 0.490 0.379*** 4.020 0.569*** 4.580 0.169*** 2.620 0.180 −0.722*** −3.470 −0.915*** −3.340 −0.322*** −2.050 −3.940 −1.639*** −4.540 −2.702*** −4.270 −1.142*** −3.800 −1.310 −0.118* −1.680 −0.141* −1.500 −0.076* −1.650 −0.850 0.002 0.380 0.004 0.460 −0.004 −0.790 1.140 0.028 0.330 0.251** 1.790 0.068 1.150

−0.930 −0.378** −2.130 −0.266** −2.180 −0.162

−0.900 −0.395*** −3.20 −0.269** −2.310

0.007 1.060 0.017 1.470 0.007 1.320 0.007 1.060 0.018 1.500 0.008 1.370 −0.026*** −2.930 −0.059*** −5.150 −0.015** −2.120 −0.026*** −2.950 −0.059*** −5.150 −0.016** −2.130 −0.001 −0.330 −0.039*** −2.460 0.000 0.150 −0.001 −0.310 −0.039*** −2.440 0.001 0.570 0.000 0.030 0.049*** 2.420 0.010 0.980 −0.002 −0.130 0.048*** 2.360 −0.010 −1.010 0.023 0.680 0.015 0.480 0.013 0.660 0.021 0.610 0.012 0.370 0.012 0.600 0.105*** 3.040 0.031 0.790 0.097*** 4.450 0.109*** 3.160 0.038 0.960 0.100*** 4.570 0.074*** 2.650 0.062** 1.840 0.015 0.730 0.081*** 2.980 0.071*** 2.150 0.019 0.960 0.341 0.470 2.597** 2.340 0.039 0.080 0.250 0.340 2.416*** 2.170 −0.100 −0.190

Bank Competition, Managerial Efficiency and the Interest Rate

Table 5.

Note: ***, ** and * indicate level of significance at 1%, 5% and 10%, respectively.

331

332

JUGNU ANSARI AND ASHIMA GOYAL

Alternatively, the pass-through of cost of funds is reflected in the coefficient of deposit interest rate in the loan interest rate equations, which varies from 0.94 to 0.96 under the different scenarios. The second specification of the interest spread is also highly significant for the intercept terms, that is the mean spreads are 2.41 and 2.59 under the two interaction specifications given in Table 5. Fourth, the capital to risk adjusted assets ratio (CRAR) has a positive effect on loan pricing but is statistically significant in only one of the regressions. Many studies find a positive impact of CRAR on loan pricing. Banks hold capital to insulate themselves against both expected and unexpected credit risk, and therefore, reflecting banks’ risk aversion (Saunders & Schumacher, 2000). Banks often endogenously choose to hold more capital than minimal regulatory requirements, against unexpected credit losses or market discipline which may induce them to hold more capital (Flannery & Rangan, 2006). However, equity capital is a more expensive funding source than debt (because of tax and dilution of control reasons). Thus, banks that have a relatively high capital ratio for regulatory or credit reasons can be expected to seek to cover some of the increase in the average cost of capital by operating with higher loan interest rate and its spread over deposit interest rate.4 Fifth, a positive relationship, a priori, is expected between asset quality variable and bank loan interest rate, reflecting the notion that banks tend to push the cost of non-performing loans to customers. Moreover, a neoclassical finance theory perspective entails that higher credit risk expected to be associated with higher return in terms of loan interest rate. A contrarian perspective entails that banks are likely to follow softer loan interest rate policy in order to avoid more loan defaults. But our results show that the effect is not consistent in loan pricing or in the determination of spread. Asset quality of loans and advances as reflected in gross nonperforming loans ratio is statistically significant and positive in LR1, RS2, LR2. In other specifications it is sometimes negative but insignificant. The positive impact of asset quality on interest rates dominates in the Indian context. Sixth, managerial efficiency which is measured by non-interest operating expenses to average assets ratio, captures expenses in processing loans and the servicing of deposits. In addition, some portion of operating cost may arise on account of non-funded activities with regard to a variety of banking transaction services. Thus, two scenarios are possible. One, banks may recoup some or all of such costs by factoring them into loan pricing. Two, banks may recover a portion of such costs from non-funded activities by

Bank Competition, Managerial Efficiency and the Interest Rate

333

way of other non-interest income, thereby, charging only a fraction of operating cost to loan interest rate to borrowers. We found that a positive effect of MI, that is higher operating cost ratio on loan interest rates and their spread over deposit interest rates. From Tables 4 and 5, we can see that the operating cost put on average 10
27 percentage point weights on the loan pricing, which is positive and highly significant. This is a critical finding because such effects persist in the presence of non-interest income variable, characterising PD. Seventh, a stable and sustainable banking system entails that banks should earn sufficient profit to satisfy shareholders while keeping credit and liquidity risks under tolerable levels. The ROE measures the rate of return on the money invested by common stock owners and retained earnings by the bank. It demonstrates a bank’s ability to generate profits for shareholders’ equity (also known as net assets or assets minus liabilities). In other words, ROE shows how well a bank uses investment funds to generate growth. Interest income is clearly a function of the yield curve and credit spreads posited under the stress scenario, but what the net impact of rising or falling rates are on bank profitability remains ambiguous.5 As expected it is positive in all the specifications but is significant only under LR1, and LR2. From the Tables 4 and 5, we see that the coefficient varied from 0.3% to 1.8% under different scenarios viz. current loan interest rate, lagged loan IRS and SF measure of loan interest rate. Eighth, liquidity affects loan pricing behaviour of banks. As the liquidity ratio increases, liquidity risks increase, implying a higher margin set by banks. However, banks with more liquid assets are expected to find it easier to fund loans on the margin, so there may be a negative sign for this variable. Our results show a negative and significant differential impact of banks’ liquidity with regard to differential measures of loan interest rates. Under the second specification, we have a negative and highly significant impact of liquidity on loan pricing. PD measured by the non-interest income variable has a significant negative coefficient in all our panel data estimations suggesting possible crosssubsidization of traditional lending activities.6 The results show that the coefficient of non-interest income (the income share of commission and fee income) are negative and significant. Our results are consistent with the hypothesis that banks decrease their lending rate when they are more reliant on fee generating products. The coefficient ranges from 22% to 40% depending on the lending rate structure chosen for the analysis. For IRS, the coefficient ranges from 16% to 39%, which is significant under second and third specifications.

334

JUGNU ANSARI AND ASHIMA GOYAL

The role of loan maturity in loan pricing derives from the terms of lending and management of asset-liability mismatches (Ranjan & Dhal, 2003). In the Indian context, the introduction of maturity-based pricing reflects bank’s continuous commitment to safeguard its financial strength based on sound banking principles, while striving to provide resources for development lending at the lowest and most stable funding costs and on the most reasonable terms.7 The brokerage function and term transformation functions of banks are blurred in the NIMs and Average Spreads, since all interest income and expenses are aggregated to create implicit returns on assets and liabilities. Nevertheless, the NIM and the Average Spread are important because aggregation highlights the overall profitability of bank management across different loan and deposit activities, as well as the role of non-interest income activities. According to Segura and Suarez (2012) banks’ do not have an incentive to set debt maturities as short as savers might ceteris paribus prefer. Their incentive for longer debt comes from the fact that there are events (called systemic liquidity crises) in which their normal financing channels fail and they have to turn to more expensive sources of funds. In this context, we find that the coefficients are positive but not significant. The coefficient of the maturity ranges from 0.1% to 0.5%. In the Indian banking system, there is no evidence of discount to the customers to keep a long-term relationship and hence, price rises with loan maturity. Lastly, on the bank specific variables, bank size is normally important in the loan price decision of banks. Larger banks are expected to have greater market power and better access to government safety net subsidies relative to smaller banks. Relatively smaller banks may be at a competitive disadvantage in attracting the business of larger loan customers. So bank size is expected to influence bank’s lending activities differentially. The theoretical model predicts a positive relationship between the size of operations and margins, since for a given value of credit and market risk, larger operations are expected to be connected to a higher potential loss. On the other hand, economies of scale suggest that banks that provide more loans should benefit from their size and have lower margins. Therefore, we do not have a prior on the expected sign of this coefficient. Our results show negative effects of bank size on loan interest rate and its spread. The coefficients of size range from −11% to −22% under the bank lending rate whereas it ranges from 15% to 19% under the interest spread. In the Indian context only the State Bank of India has a bigger size (22%) and rest are within the range of 1
5%. So loan pricing power may not be effective.

Bank Competition, Managerial Efficiency and the Interest Rate

335

Macroeconomic factors such as growth and inflation are expected to influence the loan market from the demand as well as supply side. From a theoretical standpoint, there is a positive relationship between economic activity and banks’ spreads. As the economy expands, the demand for loans increases and this in turn can lead to higher lending rates, which can serve to widen spreads.8 Economic activity is proxied by the growth rate of real gross domestic product. The coefficient ranges from 9% to 19% depending on various measures of spreads and lending rates, and is mostly positive and significant as expected in the Indian context. Inflation is included because if inflation shocks are not passed on equally in terms of magnitude as well as speed to deposit and lending rate, then the spread would change. As expected the impact of inflation on interest spread is positive when it is significant.

CONCLUSION We investigated commercial banks’ loan pricing decisions, which could be influenced by a host of factors, using dynamic panel data methodology and annual accounts data of 33 Indian commercial banks over the period 1996
2012. The determinants of loan interest rate and spreads were classified into (i) regulatory and policy variables such as capital adequacy, and the repo rate, (ii) bank specific variables pertaining to asset quality, managerial efficiency, earnings, liquidity, bank size, loan maturity, cost of funds, (iii) competition as a market structure variable and (iv) macro variables including the rate of growth of GDP and inflation rate. Our main finding is that loan interest rates and spreads are positively impacted by policy variables. At the same time they are influenced by various market structure, bank specific and macro factors. More competition reduces transmission by reducing the loan rate but a positive policy shock increases the cost of funds and reduces the spread. The interaction between policy rate and competition in the banking sector had a negative and highly significant coefficient, which is the impact of competition on interest rate pass-through. Under the ‘competition-efficiency’ hypothesis (Demsetz, 1973), increases in competition increase profit efficiency. An exogenous shock (e.g. deregulation under the Indian banking reform) forced banks to minimize costs, offer services at lower prices, and at the same time increase profits, for example through shifts in outputs. Efficient banks (i.e. those with superior

336

JUGNU ANSARI AND ASHIMA GOYAL

management and production technologies, that translate into higher profits) will increase in size and market share at the expense of less efficient banks. We found the competition-MI interaction puts significant downward pressure on loan pricing which leads to increased market share in a competitive loan market, which in turn increases profits and hence bank soundness by reducing the default rate. The competition-inefficiency interaction coefficients range from 1.4 to 2.9, which are negative in sign and highly significant. With competition-MI interaction, the interest rate passthrough was almost twice that compared to the interaction of competition with policy rate. So increase in competition increases policy pass-through for an efficient banking system and vice-versa. Regarding the bank specific variables, loan interest rates and their spreads showed statistically significant relationship with operating cost, profitability and capital adequacy, loan maturity, asset quality, bank size and liquidity indicators. Macro variables such as GDP growth and inflation rate also showed positive impact on loan interest rates. MI raises rates and spreads and PD reduces both. Reform has had mixed effects to the extent MI fell but is still high, and PD improved but reduced again after 2004. Competition increased but with a dip in the middle. Regulatory requirements raised loan rates and spreads. Costs of deposits were passed on to loan rates. The results highlight the roles of operating efficiency, risk aversion, asset-liability management, and credit risk management in commercial banks loan pricing decisions.

NOTES 1. See Barajas, Steiner, and Salazar (1999), Brock and Rojas-Suarez (2000), Chirwa and Mlachila (2004), Beck and Hesse (2009). 2. Lago-Gonza´lez and Salas-Fuma´s (2005) found that loan price adjustment speed first decreases and later increases with market concentration, which was consistent with predictions from models that assumed quantity adjustment costs. 3. Corresponding to the three regressions this variable is defined as: Interest received on investments/total investments in G-sec; Interest received on investments/total investments in G-sec (−1); Interest received on investments (average)/ total investments in G-sec (average). 4. Berger (1995) finds that there is no relationship between ROE and capital during normal times, which may reflect the fact that the smaller competitive advantage of capital during normal times may be offset entirely by the negative mechanical effect of higher capital on ROE. Gambacorta and Mistrulli (2004) suggested that bank capital is a potentially critical factor affecting banks’ behaviour, particularly in times of financial stress and showed that bank capital affects lending even when

Bank Competition, Managerial Efficiency and the Interest Rate

337

regulatory constraints are not binding and that shocks to bank profits, such as loan defaults, can have a persistent impact on lending. Another viewpoint is that since capital is considered to be the most expensive form of liability, holding capital above the regulatory minimum is a credible signal of creditworthiness on the part of the bank (Claeys & Vennet, 2008) and thus, it is expected to have positive influence on banks’ loan interest rates. 5. English (2002) found that the co-movements of long- and short-term interest rates were sufficiently close to make the effects hard to identify because of multicollinearity if both variables were included in the regression. Multicollinearity is not a general problem, however, since neither the short-term nor the long-term rate entered alone was significant. 6. However, Stiroh and Rumble (2006) have shown that diversification gains are frequently offset by the costs of increased exposure to volatile activities. 7. Brock and Franken (2002) found matched maturity spreads are conceptually similar to bid-ask spreads in securities markets, an idea that was originally put forward by Ho and Saunders (1981). In contrast, the long spread captures the premium that banks charge for bearing duration risk. 8. Bikker and Hu (2002) found that profit appears to move up and down with the business cycle, allowing for accumulation of capital in boom periods. Provisioning for credit losses rise when the cycle falls, but less so when net income of banks is relatively high, which reduces procyclicality.

REFERENCES Allen, L. (1988). The determinants of bank interest margins: A note. Journal of Financial and Quantitative Analysis, 23(2), 231
235. Ansari, J., & Goyal, A. (2011). Competition in the banking sector and monetary transmission mechanism: An empirical analysis of interest rate pass-through in India. Presented at the 5th BIS-Asian Research Network conference, Mimeo, Bank Negara Malaysia. Barajas, A., Steiner, R., & Salazar, N. (1999). Interest spreads in banking in Colombia 1974
96. IMF Staff Papers, 46, 196
224. Beck, T., & Hesse, H. (2009). Why are interest spreads so high in Uganda? Journal of Development Economics, 88(2), 192
204. Berger, A. N. (1995). The profit-structure relationship in banking
Tests of market-power and efficiency-structure hypotheses. Journal of Money, Credit and Banking, 27, 404
431. Bikker, J., & Hu, H. (2002). Cyclical patterns in profits, provisioning and lending of banks and pro-cyclicality of the new basel capital requirements. Banca Nazionale Del Lavoro Quarterly Review, 55, 143
175. Boone, J. (2008). A new way to measure competition. The Economic Journal, 188, 1245
1261. Brock, P., & Franken, H. (2002). Bank interest margins meet interest rate spreads: How good is balance sheet data for analyzing the cost of financial intermediation? Mimeo, Central Bank of Chile. Brock, P. L., & Rojas-Suarez, L. (2000). Understanding the behavior of bank spreads in Latin America. Journal of Development Economics, 63, 113
134.

338

JUGNU ANSARI AND ASHIMA GOYAL

Chirwa, E. W., & Mlachila, M. (2004). Financial reforms and interest rate spreads in the commercial banking system in Malawi. IMF Staff Papers, 51(1), 96
122. Claeys, S., & Vennet, V. R. (2008). Determinants of bank interest margins in central and Eastern Europe: A comparison with the west. Economic Systems, 32(2), 197
216. Demirguc-Kunt, A., & Huizinga, H. (1999). Determinants of commercial bank interest margins and profitability: Some international evidence. World Bank Economic Review, 13, 379
408. Demsetz, H. (1973). Industry structure, market rivalry, and public policy. Journal of Law and Economics, 16, 1
9. English, W. B. (2002). Interest rate risk and bank net interest margins. BIS Quarterly Review, (December), 67
82. Estrada, D., Gomez, E., & Orozco, I. (2006). Determinants of interest margins in Colombia. Borradores de Economia, Banco de la Republica de Colombia No. 393. Flannery, M. J., & Rangan, K. P. (2006). Partial adjustment toward target capital structures. Journal of Financial Economics, 79(3), 469
506. Gambacorta, L., & Mistrulli, P. E. (2004). Does bank capital affect lending behavior? Journal of Financial Intermediation, 13(4), 436
457. Goyal, A. (2014). Banks, policy and risks. International Journal of Public Policy, 10(1), 4
26. Hamadi, H., & Awdeh, A. (2012). The determinants of bank net interest margin: Evidence from the Lebanese banking sector. Journal of Money, Investment and Banking, 23, 85
98. Ho, T. S. Y., & Saunders, A. (1981). The determinants of bank interest margins: Theory and empirical evidence. Journal of Financial and Quantitative analysis, 16, 581
600. Klein, M. (1971). A theory of the banking firm. Journal of Money, Credit and Banking, 3(2), 205
218. Lago-Gonza´lez, R., & Salas-Fuma´s, V. (2005). Market power and bank interest rate adjustments. Banco De Espan˜a, Documentos De Trabajo No. 539. Lerner, E. M. (1981). Discussion: The determinants of banks interest margins: Theory and empirical evidence. Journal of Financial and Quantitative Analysis, 16(4), 601
602. van Leuvensteijn, M., Bikker, J., van Rixtel, A., & Kok-Sørensen, C. (2011). A new approach to measuring competition in the loan markets of the euro area. Applied Economics, 43(23), 3155
3167. Levine, R. (1997). Financial development and economic growth: Views and agenda. Journal of Economic Literature, 35, 688
726. Liebeg, D., & Schwaiger, M. S. (2007a). Determinants of bank interest margins in central and Eastern Europe. Financial Stability Report No. 14, Oesterreichische Nationalbank (Central Bank of Austria). Liebeg, D., & Schwaiger, M. S. (2007b). What drives the interest rate margin decline in EU banking
The case of small local banks. Kredit and Kapital. Maudos, J., & De Guevara, F. J. (2004). Factors explaining the interest margin in the banking sectors of the European union. Journal of Banking and Finance, 28(9), 2259
2281. Maudos, J., & Solı´ sc, L. (2009). The determinants of net interest income in the Mexican banking system: An integrated model. Journal of Banking and Finance, 33(10), 1920
1931. Mcshane, R. W., & Sharpe, I. G. (1985). A time series/cross section analysis of the determinants of Australian trading bank loan/deposit interest margins: 1962
1981. Journal of Banking and Finance, 9, 115
136.

Bank Competition, Managerial Efficiency and the Interest Rate

339

Monti, M. (1972). Deposit, credit and interest rate determination under alternative bank objective functions. In K. Shell & G. P. Szego (Eds.), Mathematical methods in investment and finance (pp. 431
454). Amsterdam: North-Holland. Nickell, S. (1985). Error correction, partial adjustment, and all that: An expository note. Oxford Bulletin of Economics and Statistics, 47(2), 119
129. Ranjan, R., & Dhal, S. (2003). Non-performing loans and terms of credit of public sector banks in India: An empirical assessment. Reserve Bank of India Occasional Papers, 24(3), 81
121. Saunders, A., & Schumacher, L. (2000). The determinants of bank interest rate margins: An international study. Journal of international Money and Finance, 19, 813
832. Segura, A., & Suarez, J. (2012). Dynamic maturity transformation. Working Paper, No. 1105, CEMFI. Stiroh, K., & Rumble, A. (2006). The dark side of diversification: The case of US financial holding companies. Journal of Banking and Finance, 30, 2131
2161. Williams, B. (2007). Factors determining net interest margins in Australia: Domestic and foreign banks. Financial Markets, Institutions and Instruments, 16(3), 145
165. Winker, P. (1999). Sluggish adjustment of interest rates and credit rationing: An application of unit root testing and error correction modeling. Applied Economics, 31(3), 267
277.

FINANCIAL ARCHITECTURE AND MONETARY POLICY TRANSMISSION MECHANISM IN KENYA Roseline Nyakerario Misati, Alfred Shem Ouma and Kethi Ngoka-Kisinguh ABSTRACT All over the world, the role of central banks is being redefined following the outbreak of the global financial crisis and subsequent breakdown of the “great moderation” consensus. Consequently, most advanced economies adopted non-conventional approaches of monetary policy which resulted in spill-overs to emerging markets and developing countries with implications on their financial system and monetary policy transmission. This, coupled with, internal developments in the financial systems of developing countries necessitated modifications of not only monetary policy frameworks but also responsibilities of most central banks. This chapter acknowledges possible evolutions of the financial structure variables in developing countries and uses data from Kenya to analyze the dynamic linkages between financial sector variables and monetary policy transmission in the light of the financial crisis. The study used structural

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 341
364 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096014

341

342

ROSELINE NYAKERARIO MISATI ET AL.

vector autoregression to examine the relationship between financial structure variables and monetary policy as well as assess the relative importance of various monetary transmission channels in Kenya. The results show that the changing financial structure represented by credit to the private sector and stock market indicators in Kenya only slightly altered relative importance of monetary policy transmission. The insignificance of credit to the private sector suggests that the importance attached to the bank lending channel in previous studies is waning while the marginal significance of the stock market indicator signals the potential for asset price channel. The results also indicate that the interest rate and exchange rate channels are relatively more important in Kenya while the asset prices is only marginally significant and bank lending channel is the weakest in the intermediate stage of monetary policy transmission. However, transmission of monetary policy to the ultimate objectives is somewhat slow and weak to inflation and almost absent to output. The result implies a limited role of monetary policy on growth and questions the wisdom of pursuing multiple objectives. Keywords: Non-conventional monetary policy; financial structure; transmission channels

INTRODUCTION Interest rate policy should continue to be a central focus of monetary policy deliberations despite the existence of time-varying credit frictions that complicate the relationship between central bank’s policy and financial conditions. (Curdia & Woodford, 2010)

The role of central banks in maintaining macroeconomic stability and supporting stable development in the real economy using conventional methods was severely tested following the global financial crisis. Economic growth remained elusive with the advent of the credit crunch yet interest rates in most advanced economies were nearly zero percent and would therefore not be effective in propping up growth. The period after 2007 therefore saw advanced economies adopt unconventional monetary easing methods popularly referred to as credit-easing or quantitative easing.1 The main thrust of adopting non-conventional means of conducting monetary policy is founded on the belief that central banks can influence prices and output even when short term interest rates are near their zero floor by influencing liquidity through the central bank balance sheet (Bowdler & Radia,

Financial Architecture and Monetary Policy Transmission Mechanism

343

2012; Chen, Filardo, Dong, & Feng, 2012; Curdia & Woodford, 2010, 2011; Fawley & Ezeoha, 2013; Joyce et al., 2012; Shirakawa, 2013; Trichet, 2013). These monetary easing practices in advanced economies are believed to have affected trade, growth, and financial markets of emerging and developing countries, necessitating review of monetary policy frameworks and realignment of monetary policy through economic stimulus in some countries. One critical aspect on spill-overs of monetary easing measures in advanced countries to emerging markets and developing countries that is dominantly attracting research debate is the effect capital inflows in search of relatively higher returns and asset price increases are having on these economies. It is argued that these spill-overs may have affected the evolution of financial structures of developing and emerging economies (Chinn, 2013; Sugimoto, Takashi, & Yushi, 2013). Thus, it is possible that the rapid development of capital markets that has been experienced in emerging markets and developing countries in the recent past partly resulted from the spill-overs of the global financial crisis. Such dynamism in the financial structure necessitates re-examination of how it interacts with monetary policy. Existing literature on financial structure-monetary policy transmission nexus argue that the effectiveness of monetary policy depends on a set of structural parameters not directly under the control of the central bank. These structural parameters including elasticities of the demand and supply of financial and real assets to money market interest rates are affected by the structure of the financial system. The demand for loans for each bank for instance is less elastic in markets that have fewer competitors, higher barriers to entry, and no alternative finance sources. In such markets, lending rates may show a limited response to changes in money market rates in the short run (Cecchetti, 1999; Cottarelli & Kourelis, 1994; Elbourne & Jakob, 2006; Gigineishvili, 2011; Wolfgang, 2007). Although most African countries have shifted from financially repressed systems to more market based economies, their financial sectors are still relatively underdeveloped and relatively bank dominated, compared to some emerging markets and developed countries. In Kenya, for instance, the financial sector has rapidly evolved since the initiation of financial liberalization in the 1990s (Misati et al., 2011). Many models such as mobile financial services and agency banking have rapidly revolutionized the Kenyan financial landscape. The structure of the financial system is rapidly developing with many new products entering the financial markets, increasing role of other sources of financing such as the capital market and

344

ROSELINE NYAKERARIO MISATI ET AL.

increasing capital flows following capital account liberalization. It is therefore possible that, although bank financing still dominates, there may have been a shift in the form of credit financing from bank loans to securities. Hence, the interaction between financial variables and monetary policy variables may have changed with possible implications on the monetary transmission mechanism and the relative importance of various transmission channels. Previous work in Africa on this subject has mainly focused on individual monetary transmission without capturing the dynamic interactions between financial structure variables and monetary policy transmission (Al-Mashat & Billmeier, 2007; Buigut, 2010; Cheng, 2006; Fadiran & Ezeoha, 2012; Kovanen, 2011; Mangwengwende, Chizara, & Nel, 2011; Misati et al., 2011; Misati & Nyamongo, 2012; Mishra & Montiel, 2013; Montiel, Adam, Mbowe, & O’Connell, 2012; Mugume, 2011; Neaime, 2008; Ngalawa & Viegi, 2011; Sichei & Njenga, 2012; Walker, 2012). The objective of this study therefore is to assess monetary policy transmission channels in an evolving domestic financial structure and a transformed global financial system reflective of the global financial crisis. The chapter used trend analysis of financial variables to obtain some assessment of the effects of the financial crisis on financial variables and by implication on monetary policy transmission. The study also used structural vector autoregression (SVAR) to assess the interaction between financial sector variables and monetary policy in the light of the changed financial structure. In the next section, a brief analysis of selected financial variables before and after the global financial crisis and the monetary policy regime in Kenya is provided while the methodology is presented in the section “Methodology.” The section “Discussion of Results” discusses the results and the last section concludes the chapter.

STYLIZED FACTS Monetary Policy Regime The conduct, strategy, and operation of monetary policy in Kenya has transformed markedly over the years to reflect rapid domestic and global financial sector developments in the recent past. The traditional monetary aggregate targeting that dominated monetary policy decisions in the 1990s and part of 2000s is now becoming increasingly unreliable in the

Financial Architecture and Monetary Policy Transmission Mechanism

345

face of rapid financial innovations and developments in Kenya that have rendered the relationship between monetary aggregates and the real sector unstable and unreliable. The central bank of Kenya therefore considers a variety of indicators to inform its monetary policy decisions and it uses central bank rate (CBR) as the policy signaling rate to other market interest rates.

Trends of Selected Financial Variables In this sub-section, we review trends of some of the financial sector variables in the light of the changing domestic and global financial environment. In Fig. 1, we present trends of three major financial structure variables that may have been affected by the internal and external financial system developments. It can be observed that in the period before 2007, credit growth to the private sector was generally lower than the period after 2007. Except for the period 2003
2005 when higher levels of credit growth are observable on account of a major change of political regime during this period, the rest of the period before 2007 registers low credit growth figures. After 2007, credit growth to the private sector generally picked up with averages of above 20 percent except in 2009 and 2012 partly because of a restrictive monetary policy adopted in the latter period. While trends of the housing index exhibit similar trends as the credit to the private sector growth with relatively higher levels after the global financial crisis, trends in NSE (Nairobi 20 Share Index) show that the index levels were at the peak during the global financial crisis period. After the crisis, the NSE index slightly dropped but remained at a relatively higher level than the period before the global crisis. The upward trend of the house index in the period after the crisis may also be explained by the possibility of preferring collateralized lending in the property markets with increased risk aversion resulting from the crisis. Fig. 2 shows that at the beginning of the crisis short term capital flows reduced marginally but recovered in 2010. Remittances also proved resilient and have maintained an upward trend even during the crisis while official, medium, and long-term flows that were negative before the crisis turned positive after the crisis. While acknowledging that there were many internal events that may have influenced the trends of the reviewed financial variables, the possibility of the effects of spill-overs from the crisis cannot be entirely ruled out.

346

400

6,800 40

350

5,800 35

Credit growth to the private sector 300 4,800

30

250 25

3,800 200 2,800 150

15 1,800

10

100

5 50

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Fig. 1.

June

Jun

Dec

Jun

Dec

Jun

Dec

June

December

June

December

June

December

June

December

June

December

June

December

June

December

June

Jun

December

June

Apr

Nov

Sep

Jul

Feb

May

December

March

October

August

June

January

April

November

September

July

February

December

Oct

May

Mar

Aug

Jan –5 –10

Dec

Dec

800 0

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 20122013 Housing Index, RHS

NSE Index, LHS

Credit Growth to the Private Sector, Housing Index and NSE Index. Source: The housing index is proxied from the Hass Consult property index.

ROSELINE NYAKERARIO MISATI ET AL.

In percent

20

347

Financial Architecture and Monetary Policy Transmission Mechanism 3,000

US$, Million

2,500 2,000 1,500 1,000 500 – (500) 2004

2005

2006

2007

2008

2009

2010

2011

2012

Official, Medium and Long-Term

Private, medium and Long-Term

Short Term Capital Flows

Remittances

Fig. 2.

Financial Flows.

METHODOLOGY This study uses a SVAR specification, which unlike reduced and recursive Vector Autoregressives (VARs) uses economic theory to cater for the contemporaneous links among variables. SVAR allows imposition of additional apriori restrictions termed “identifying restrictions” based on economic arguments (Renee & Adrian, 2011). The identified assumptions produce instrumental variables that permit the contemporaneous links to be estimated using instrumental variables regression (Enders, 2010; Stock & Watson, 2001; Watson, 1994). The SVAR recovers the structural parameters from reduced form VAR model and dynamic features of the model give a more structural interpretation (Sharifi-Renani, 2010). SVAR models are the most suitable tools for the analysis of monetary transmission mechanism and it is a superior methodology over reduced form VAR and recursive VAR (Chuku et al., 2011; Gottschalk, 2001; Plaff & Taunus, 2008). The structural VAR model is expressed as follows (Paolo & Antonio, 2012; Cevik & Teksoz, 2012; Goyal & Pujari, 2005; Lack & Lenz, 2000; Mariano, 2005; Mihira & Sugihara, 2000; Ngalawa & Viegi, 2011; Singh & Pattanaik, 2010; Verheyen, 2010): BðLÞXt = ɛ t

ð1Þ

where Xt is a vector of economic variables and ɛt is a (n × 1) vector of structural error terms with

348

ROSELINE NYAKERARIO MISATI ET AL.

0

Eðɛt ɛt Þ = I

ð2Þ

Eq. (3) provides the Wold representation of Eq. (1): Xt = DðLÞɛ t

ð3Þ

where DðLÞ = BðLÞ − 1

ð4Þ

Estimation of reduced form VAR model provided in Eq. (5) facilitates identification of this structural model: AðLÞYt = et

ð5Þ

where yt is an nx1 vector of endogenous variables and A(L) =A(0)+A(1) L + A(2) L2 + … + A(P) Lp and L is a lag operator; A(0) = I and et is an nx1 vector of reduced form VAR residuals with covariance matrix E (etet’) = Ω which is, in general, non-diagonal. The moving average form of Eq. (5) is expressed as follows: Yt = CðLÞet

ð6Þ

where C(L) = A(L)−1 is an infinite order lag polynomial. From Eqs. (3) and (6), D(L)ɛt = C(L)et. The reduced form model residuals are a linear combination of structural residuals implying that the parameters in the reduced form model and those in the structural form are related by the following representation: et = Dð0Þɛt

ð7Þ

such that the E(etet) = D(0)E(ɛtɛt)D(0)’ and variance-covariance matrix: Ω = Dð0ÞDð0Þ0

ð8Þ

The structural moving average representation in Eq. (3) is also called the final form of an economic model since the endogenous variables Xt are

Financial Architecture and Monetary Policy Transmission Mechanism

349

expressed as distributed lags of the exogenous variables, given by elements of ɛt. The elements of ɛt are indirectly observed through their effects on the elements of Xt. The basic model is formulated based on seven variables to reflect interactions between output, lending interest rates, credit to the private sector, price level, nominal exchange rate, policy rate, and stock market indicators. Although previous studies using Kenyan data reveal weak linkages between short term interest rates and ultimate objectives of monetary policy, we still include a short term interest rates to capture monetary policy stance since the Central Bank of Kenya uses CBR as a signaling mechanism to commercial banks (Misati et al., 2011; Misati et al., 2012; Misati & Nyamongo, 2012). We therefore use a price variable rather than a monetary aggregate variable, in any case, in the literature, it is considered that each level of interest rate has a corresponding money supply and money demand equates to money supply to clear the market (Tahir, 2012). Implicitly, the interest rate, asset prices, and exchange rate channels of monetary transmission are captured in this framework.2 In all the models, we also include international oil prices and world food prices as exogenous variables to reflect the dominant role often played by supply side factors in influencing inflation. The theoretically plausible restrictions imposed on the structure of the model to identify various structural shocks are thus specified as follows:   yt = GDPIt ; CPICOREt ; CredPt ; Lendt ; StCapt ; ERt ; Interbankt 2

1 6 0 6 6 b31 6 6 b41 6 6 b51 6 4 b61 0

0 1 0 b42 b52 b62 0

0 0 1 b43 0 b63 0

0 0 b34 1 0 b64 0

0 b25 0 b45 1 b65 b75

0 0 0 b46 b56 1 0

ð9Þ

32 GDPI 3 2 GDPI 3 0 e ɛ 6 ɛCPI 7 6 eCPI 7 0 7 7 6 Credpriv 7 76 Credpriv 7 6e 7 6 0 7 7 6 7 76 ɛ 6 ɛ Lending 7 = 6 eLending 7 b47 7 7 6 7 76 ER 7 6 eER 7 6 b57 7 7 6 7 76 ɛ b67 54 ɛNSE 5 4 eNSE 5 1 ɛ Interbank eInterbank

where ɛGDPI, ɛCPI, ɛCredpriv, ɛLending, ɛER, ɛNSE, ɛInterbank are the structural disturbances, while eGDPI, eCPI, eCredpriv, eLending, eER, eNSE, eInterbank are the residuals in the reduced form equations. Eq. (1) represents output, which adjusts slowly to shocks of other variables in the system.3 GDP responds to all variables with a lag. Although it is assumed that credit is spent as soon as the funds are obtained,

350

ROSELINE NYAKERARIO MISATI ET AL.

immediately adding to aggregate demand, this does not apply in this case due to the high frequency data used (Berkelmans, 2005; Verheyen, 2010).4 All coefficients in this matrix are therefore restricted to zero. However, it is important to note that output might respond contemporaneously to inflation because nominal incomes, and therefore spending, may be motivated by the Lucas Phelps imperfect information model in which producers face a signal extraction problem. Contemporaneously, producers only observe their own price and so are unaware whether an increase in price reflects inflationary pressures or an increase in demand. Consequently, they prefer to increase production, even if the price increase is purely inflationary (Berkelmans, 2005; Naceur, Boughrara, & Ghazouani, 2009). But since we are using monthly data, we assume that this increase in production may not occur rapidly. The exchange rate (imported inflation) has a contemporaneous effect on inflation. Since the study is using monthly data, it is logical to assume that CPI responds to output with a lag, besides the well acknowledged assumption of nominal rigidities in price adjustments. We impose restrictions on all variables in this equation except the exchange rate. The first two variables are included as ultimate target variables or policy goals while the stock market and credit to the private sector capture the financial structure of the economy. Bank lending rates, exchange rate, stock market indicators, bank loans are intermediate variables, representing interest rate channel, exchange rate channels, asset prices channel, and credit channel, respectively. Credit to the private sector in Eq. (3) responds to bank lending rate reflecting the fact that a certain percent of loans is on flexible interest rate terms. Credit demand may also be affected by GDP growth, which affects the demand for credit. Bank lending rates in Eq. (4) is assumed to contemporaneously respond to all variables since it is argued that expectations of future activity form an important determinant of credit demand. Thus it can be assumed that current output, price level, exchange rate, and interest rate provide some indication of what is expected in the future (see Ngalawa & Viegi, 2011). Eq. (5) captures the foreign exchange market. It captures the effects of depreciation of the shilling on inflation and controls for the components of interest rate movements that are systematic responses to a depreciation of the domestic currency (Elbourne & Salomons, 2004). Asset prices represented by a stock market indicator is allowed to react instantaneously to all other types of shocks reflecting the efficient market hypothesis theory that financial markets reflect all the information in the system (Singh & Pattanaik, 2010). Thus no restrictions are imposed in

Financial Architecture and Monetary Policy Transmission Mechanism

351

Eq. (6). We assume that monetary policy does not respond contemporaneously to output and prices since the data isn’t available contemporaneously. So the central bank responds contemporaneously to the variables that are available contemporaneously, in this case the exchange rate (Elbourne & Salomons, 2004). The exchange rate responds immediately to all variables except credit to the private sector and lending interest rate. In our study, we used the unrestricted VAR representation since the goal of a VAR analysis is to determine the interrelationships among variables, not to determine the parameter estimates (Khan & Ahmed, 2011; Sims, Stock, & Watson, 1990). Moreover, differencing of variables may lose important information concerning the co-movements in the data. Estimation of SVAR in levels is appropriate if the error terms of each SVAR equations are stationary and serially uncorrelated (Berkelmans, 2005). In addition, various studies have found that usage of unrestricted VAR is superior in terms of forecast variance to a restricted VAR at short horizons (Clements & Hendry, 1995; Engle & Yoo, 1987; Hoffman & Rasche, 1996). Other studies have also established that the performance of unrestricted VARs and restricted VARs for impulse response analysis is similar over the short run (Naka & Tufte, 1997).5 See summarized details of the advantages and disadvantages of different VAR specifications in Farzanegan and Markwardt (2009).

DISCUSSION OF RESULTS In this section, we briefly report the results of impulse response functions (IRFs) based on Montiel et al. (2012) three-variable and six-variable models for comparison purposes with similar results in Tanzania, the only other country that has conducted a similar study in the East African Region. More importantly, we present results that incorporate financial structure variables in the SVAR framework, which also allows determination of the relative importance of various monetary policy transmission mechanisms in Kenya.

Three-Variable SVAR and the Six-Variable SVAR In this sub-section, we present IRFs of the three-variable model in Fig. 3 and IRFs of the six-variable model in Fig. 4.

352

ROSELINE NYAKERARIO MISATI ET AL. Response of LCPICORE to LRM

Response of LER to LRM

0.03

0.04 0.03

0.02

0.02 0.01 0.01 0.00

0.00

–0.01

–0.01 5

10

15

20

25

Fig. 3.

30

35

5

10

15

20

25

30

35

Three-Variable SVAR.

Response of LGDPI to LRM

Response of LCPICORE to LRM 0.03

0.06

0.02

0.04

0.01 0.02 0.00 0.00

–0.01 –0.02

–0.02 5

10

15

20

25

30

5

35

Response of LER to LRM

10

15

20

25

30

35

30

35

Response of LEND to LRM 1.0

0.03 0.02

0.5

0.01 0.0 0.00 –0.5

–0.01

–1.0

–0.02 5

10

15

20

25

Fig. 4.

30

35

5

Six-Variable SVAR.

10

15

20

25

Financial Architecture and Monetary Policy Transmission Mechanism

353

In Fig. 3, we include reserve money (RM), non-fuel non-food consumer price index (CPICORE) representing a core measure of inflation and exchange rate (ER). All the variables are in logs, represented by L before the abbreviation of the variable. In Fig. 3, the results show that an increase in RM increases core inflation and first appreciates and then depreciates the exchange rate but the effect is not significant in both cases. In Fig. 4, we add output (GDPI), broad money (M3) and lending interest rates (Lend) to the three-variable model. In this case, an increase in RM has no significant effect on all the variables of interest. Changes in RM do not change output except a one off insignificant increase in the first three months and it has no effect on lending interest rates. However, a monetary policy change first reduces core inflation before increasing it and it also first appreciates the exchange rate before depreciating it. The insignificance of all the variables considered in this six-variable VAR would be a reflection of the waning importance of monetary aggregates in monetary policy transmission, in contrast to Tanzania where monetary aggregates are still more important.

IRFs for VAR with Financial Structure Variables In this sub-section, we empirically focus on the main objective of this chapter as specified in Eq. (9). We therefore replace monetary aggregates with interest rate and add financial structure indicators. Credit to the private sector (credP) represents the level of banking sector development while stock market capitalization (STCAP) captures the level of development of the capital market besides being indicative of the availability of alternative sources of financing to bank finance. Unlike in the three and six VAR cases, we consider both overall inflation (CPIH) and core inflations measures and how they interact with financial and policy variables. We also report results that indicate the relative importance of monetary policy transmission channels, where the bank lending channel is captured by credit to the private sector; lending interest rate represents the interest rate channel; the stock market indicator capture the asset price channel and the exchange rate captures the exchange rate channel. In Fig. 5, we present IRFs when overall inflation is used in the SVAR. We analyze the effects of the variables in the model on overall inflation as well as the effect of monetary policy on indicators of the channels of

354

ROSELINE NYAKERARIO MISATI ET AL. Response of LCPIH to LCREDP

Response of LCPIH to LEND

0.02

0.02

0.01

0.01

0.00

0.00

–0.01

–0.01

–0.02

–0.02 5

10

15

20

25

30

35

5

10

15

20

25

30

35

Response of LCPIH to LSTCAP

Response of LCPIH to LER 0.02

0.02

0.01

0.01

0.00

0.00

–0.01

–0.01

–0.02

–0.02 5

10

15

20

25

30

35

5

Response of LCPIH to INTERBANK

10

15

20

25

30

35

Response of LCREDP to INTERBANK

0.02

0.02

0.01

0.01

0.00

0.00

–0.01

–0.01

–0.02

–0.02 5

Fig. 5.

10

15

20

25

30

35

5

10

15

20

25

30

SVAR with Financial Structure Variables and Overall Inflation.

35

355

Financial Architecture and Monetary Policy Transmission Mechanism Response of LEND to INTERBANK

Response of LER to INTERBANK 0.04

0.8

0.03 0.4 0.02 0.01

0.0

0.00 –0.4 –0.01 –0.02

–0.8 5

10

15

20

25

30

35

5

10

15

20

25

30

35

Response of LSTCAP to INTERBANK 0.10

0.05

0.00

–0.05

–0.10 5

10

15

20

25

30

Fig. 5.

35

(Continued )

monetary policy transmission. The results show that a shock to lending interest rate has a negative significant effect on overall inflation with a lag of seven months. Innovations to the interbank capturing the monetary policy stance also has a negative and significant effect on overall inflation after a lag of nine months but the significant effect lasts for over eighteen months. However, the exchange rate, stock market indicators and credit to the private sector have no significant effects on overall inflation. A positive shock to the interbank interest rate increases the lending interest rate immediately and the effect remains significant for fifteen months. Innovations to the interbank interest rate cause an appreciation of the exchange rate after a lag of three months and the significant effect lasts for ten months. However, a shock to the interbank interest rate has no effect on credit to the private sector and very marginal effects on the stock market indicators. Thus, in terms of relative importance of monetary policy transmission channels, these results imply that, in Kenya, the interest rate

356

ROSELINE NYAKERARIO MISATI ET AL.

and the exchange rate channels are relatively more important while the bank lending and the asset price channels are very weak. These results further imply that the changing financial structure in Kenya only slightly altered the relative importance of monetary policy transmission. The insignificance of credit to the private sector suggests that the importance attached to the bank lending channel in previous studies is waning while the marginal significance of the stock market indicator signals the potential for asset price channel. It should also be noted that, although the exchange rate channel seems to exist, it only ends at this stage of transmission. The passthrough does not seem to reach the ultimate objective since we have already observed no effect of exchange rate on overall inflation. The importance of the exchange rate channel and relative weakness of the asset prices channel is consistent with some existing studies (Cheng, 2006; Misati & Nyamongo, 2012). However, the strength of the interest rate channel contrasts previous studies that indicate that it is weak both in terms of magnitude and speed of pass-through (Misati et al., 2011). It is possible that these conflicting results would be explained by usage of different methodologies and different data frequencies in previous studies. In our view, the more plausible explanation of these contrasting results would be, first, usage of different indicators of monetary policy stance. Second, this particular study uses most recent data which incorporates the gains achieved over the recent past on effective monetary policy communication that has sensitized the market on the role of the policy rate as a signaling mechanism to commercial banks. Third, this study uses data incorporating different monetary policy regimes with both extended periods of monetary policy tightening and loosening. The lack of demarcation of these periods distorts the asymmetric nature prevalent in responses of commercial banks to different monetary policy regimes. In this case, it seems that the monetary policy tightening periods dominate the monetary policy loosening periods. The results of this study therefore do not necessarily imply that problems of stickiness and rigidity of lending interest rates that are observed in practice, particularly following monetary policy loosening, have disappeared. Further work on interest rate asymmetries with demarcated monetary policy regimes can possibly have different implication on the importance of the interest rate channel. In Fig. 6, we present IRFs when the overall inflation is replaced with non-food non-fuel inflation, a measure of core inflation. The results are largely the same as in Fig. 5. The only difference is the lags and the length of the period of persistence. For example, a shock to lending interest rate has a negative and significant effect on core inflation after six months but the

357

Financial Architecture and Monetary Policy Transmission Mechanism Response of LCPICORE to LCREDP

Response of LCPICORE to LEND

0.04

0.04

0.03

0.03

0.02

0.02

0.01

0.01

0.00

0.00

–0.01

–0.01 –0.02

–0.02 5

10

15

20

25

30

35

5

Response of LCPICORE to LER

10

15

20

25

30

35

Response of LCPICORE to LSTCAP

0.04

0.04

0.03

0.03

0.02

0.02

0.01

0.01

0.00

0.00

–0.01

–0.01 –0.02

–0.02 5

10

15

20

25

30

5

35

Response of LCPICORE to INTERBANK

10

15

20

25

30

35

Response of LCREDP to INTERBANK 0.04

0.04 0.03

0.02 0.02 0.00

0.01 0.00

–0.02 –0.01 –0.04

–0.02 5

Fig. 6.

10

15

20

25

30

35

5

10

15

20

25

30

SVAR with Financial Structure Variables and Core Inflation.

35

358

ROSELINE NYAKERARIO MISATI ET AL. Response of LEND to INTERBANK

Response of LER to INTERBANK 0.04

0.8

0.03 0.4 0.02 0.01

0.0

0.00 –0.4 –0.01 –0.02

–0.8 5

10

15

20

25

30

5

35

10

15

20

25

30

35

Response of LSTCAP to INTERBANK 0.10

0.05

0.00

–0.05

–0.10 5

10

15

20

25

30

Fig. 6.

35

(Continued )

significant effect only lasts for six months, shorter than it does in the case of overall inflation. A shock to the interbank interest rate also has a negative effect on core inflation after a period of fifteen months but with less pronounced significance compared to its effect on overall inflation. As was the case in overall inflation, credit to the private sector, stock market indicator and exchange rate have no significant effects on core inflation. The relative importance of various transmission channels is similar to the analysis on overall inflation.

Variance Decomposition: Transmission Channels In Table 1, it is clear that the interest rate channel dominates the other monetary policy transmission channels. Within 12 months, the interbank rate explain 35 percent of variations in lending interest rate while it explains about 16 percent of variations in the exchange rate over a similar period.

359

Financial Architecture and Monetary Policy Transmission Mechanism

Table 1. Monetary Policy Variable (Interbank) with Overall Inflation. Horizon (Months) 6 12 18 24

Credit to the Private Sector

Lending Interest Rate

Stock Market Indicator

Exchange Rate

0.08 1.66 5.53 10.10

25.17 35.09 32.20 28.82

2.86 3.33 2.90 2.69

6.02 16.43 22.81 25.25

Overall GDP CPI 3.13 14.47 26.59 32.89

0.57 1.59 3.76 5.56

Table 2. Monetary Policy Variable (Interbank) with Core Inflation. Horizon (Months) 6 12 18 24

Credit to the Private Sector

Lending Interest Rate

Stock Market Indicator

Exchange Rate

Core CPI

GDP

0.10 1.51 4.47 7.32

18.14 24.94 21.90 18.63

1.73 1.69 1.34 1.13

5.61 15.27 20.90 22.49

0.41 3.40 8.22 10.81

0.59 1.08 2.51 3.72

However, the innovations to the interbank interest rate only explain about 3 percent and 1.6 percent of the variations in the stock market indicator and credit to the private sector, respectively within 12 months. In terms of relative importance therefore, the interest rate and the exchange rate channels are the most important in that order while the bank lending and asset price channels of monetary transmission are the weakest in that order. Transmission of monetary policy to the ultimate objectives, mainly inflation in this case, is somewhat slow and weak, with about 14 percent of the variations in overall consumer price indices being explained within 12 months but it is significantly weak to output, with only 1.6 percent of the variations in GDP explained by interbank changes. Whether this is an indication of the limited role of monetary policy on growth and if by implication it would be less beneficial for central banks to pursue multiple objectives are still open questions. When a core inflation measure is used (Table 2), the results in terms of relative importance of the various monetary policy transmission channels is not significantly altered, it is only muted in this case. However, a significant difference is noted in terms of the impact of monetary policy on the ultimate price stability objective. Shocks to monetary policy explain relatively higher magnitudes of variations, 14 percent within one year and 33 percent within two years in overall inflation compared to 3 percent and 10 percent of variations in core inflation, respectively, over a similar period. Whether this implies that countries where food and fuel occupy huge weights in the

360

ROSELINE NYAKERARIO MISATI ET AL.

consumer price index and where prices of such commodities are highly volatile should focus their monetary policy on overall inflation rather than core inflation is open to debate.

CONCLUSIONS AND POLICY OBSERVATIONS The financial crisis of 2007
2012 necessitated deviations from the traditional method of conducting monetary policy, particularly in advanced economies. Spillover effects to emerging and developing markets such as capital inflows and fears of later reversals, increased asset prices and realignments of interest rates through economic stimuli dominated the period after 2007 with possible effects on the financial structure and implications for monetary policy transmission. The changing global financial environment coincided with diverse financial reforms in Kenya and modifications to monetary policy framework. While the monetary aggregate targeting framework was not formally abandoned in Kenya, now, a variety of indicators to inform monetary policy decisions and a policy signaling rate have been adopted. The financial architecture and with it monetary policy transmission is thus different in Kenya than the period before 2007. This chapter used SVAR methods to examine monetary policy transmission to ultimate objectives within this changing financial environment. Within the same framework, the study also analyzed the relative importance of the standard monetary policy transmission mechanisms. The results show that the changing financial structure represented by credit to the private sector and stock market indicators in Kenya only slightly altered the relative importance of monetary policy transmission. The insignificance of credit to the private sector suggests that the importance attached to the bank lending channel in previous studies is waning while the marginal significance of the stock market indicator signals the potential for asset price channel. The study also shows that the interest rate and exchange rate channels of monetary policy transmission are relatively more important while the bank lending and asset price channels of monetary policy transmission are relatively weak in Kenya. The variance decomposition results show that within 12 months, a monetary policy shock explain 35 percent of variations in lending interest rate, 16 percent of variations in the exchange rate, 3 percent in the stock market indicators, and 1.6 percent of the variations in credit to the private sector. However, transmission of monetary policy to the ultimate objectives, mainly inflation in this case, is somewhat slow, with about 14 percent of the variations in overall

Financial Architecture and Monetary Policy Transmission Mechanism

361

consumer price indices being explained within 12 months but it is significantly slow to output, with only 1.6 percent of the variations in GDP explained by interbank shocks. These results imply that, first, the central bank can somewhat rely on the interest rate and exchange rate channels in its pursuit of price stability objective. This observation does not however imply that the retail interest rate rigidities observed in practice, especially following a monetary policy loosening have disappeared. Further work on interest rate asymmetries, demarcating different monetary policy regimes, if possible, is critical to authoritatively validate the role of interest rate channel of monetary transmission. Second, the significantly slow and weak transmission of monetary policy actions to economic growth suggests a limited role of monetary policy on growth and thus questions the wisdom of pursuing multiple objectives. Third, the results showed some marginal effects of the interbank on stock market indicators, implying that the asset channel is likely to be important in future thus prompting a recommendation for further measures to develop the capital market. In the light of the global financial crisis, this last point requires further empirical analysis to quantitatively estimate the potential spill offer effects from advanced monetary policies to the Kenyan capital market.

NOTES 1. Non-conventional monetary policy approaches differ across central banks with some providing additional direct monetary stimulus to the economy while others aim at enhancing monetary policy transmission of standard interest rate policy (Cour-Thimann & Winkler, 2012). 2. However, we consider the three-variable and six-variable VARs, based on Montiel et al. (2012) only at the result level. Montiel et al. (2012) mainly used monetary aggregates to reflect the current monetary policy practice in Tanzania. However, in Kenya although formally, Kenya still uses a monetary targeting framework, the main signaling mechanism to commercial banks that is relevant for monetary policy transmission is interest rate which is our main focus in this study and which informs our final model of analysis. 3. The non-zero coefficients of bij in the equations indicate that variable j affects i instantaneously. 4. GDP monthly data was interpolated based on the method proposed in Chow and Lin (1971) and Fernandez (1981). 5. Nevertheless, we conducted unit root tests which revealed some nonstationarity in some variables. We further estimated SVAR in differences and the results were similar to the unrestricted SVAR.

362

ROSELINE NYAKERARIO MISATI ET AL.

REFERENCES Al-Mashat, R., & Billmeier, A. (2007). The monetary transmission mechanism in Egypt. IMF Working Paper No. WP/07/285. International Monetary Fund, Washington, DC. Berkelmans, L. (2005). Credit and monetary policy: An Australian SVAR. Research Discussion Paper No. RDP 2005
06, Reserve Bank of Australia. Bowdler, C., & Radia, A. (2012). Unconventional monetary policy: The assessment. Oxford Review of Economic Policy, 28(4), 603
621. Buigut, S. (2010). Is there a bank lending channel of monetary policy in Kenya. International Research Journal of Finance and Economics, 45, 183
192. Cecchetti, S. (1999). Legal structure, financial structure, and the monetary policy transmission mechanism. FRBNY Economic Policy Review, 5(2), 9
28. Cevik, S., & Teksoz, K. (2012). Lost in transmission? The effectiveness of monetary policy transmission channels in the GCC countries. IMF Working Paper No. WP/12/191. International Monetary Fund, Washington, DC. Chen, Q., Filardo, A., Dong, H., & Feng, Z. (2012). International spillover of central bank balance sheet policies. BIS Papers, (66), 230
272. Cheng, K. (2006). A VAR analysis of Kenya and monetary policy transmission mechanism: How does the central bank’s REPO rate affect the economy? IMF Working Paper No. WP/06/300. International Monetary Fund, Washington, DC. Chinn, M. (2013). Global spillovers and domestic monetary policy: The effects of conventional and unconventional measures. 12th BIS Annual conference WP 436 “Navigating the great recession: What role for monetary policy”. Chow, G., & Lin, A. (1971). Best linear unbiased interpolation, distribution, and extrapolation of time series by related series. Review of Economics and Statistics, 53(4), 372
375. Chuku, C. A., Usenobong, F., Ndifreke, R., & Ekpeno, L. (2011). Oil price shocks and the dynamics of the current account balances in Nigeria. OPEC Energy Review, 35(2), 119
139. Clements, M., & Hendry, D. (1995). Forecasting in cointegrated systems. Journal of Applied Econometrics, 10(2), 127
146. Cottarelli, C., & Kourelis, A. (1994). Financial structure, bank lending rates, and transmission mechanism of monetary policy. IMF Staff Papers, 41(4), 587
623. Cour-Thimann, P., & Winkler, B. (2012). The ECB’s non-standard monetary policy measures: The role of institutional factors and financial structure. Oxford Review of Economic Policy, 28(4), 765
803. Curdia, V., & Woodford, M. (2010). Conventional and unconventional monetary policy. Federal Reserve Bank of St. Louis Review, 92(4), 229
264. Curdia, V., & Woodford, M. (2011). The central-bank balance sheet as an instrument of monetary policy. Journal of Monetary Economics, 58(1), 54
79. Elbourne, A., & Jakob, D. (2006). Financial structure and monetary policy transmission in transition countries. Journal of Comparative Economics, 34(1), 1
23. Elbourne, A., & Salomons, R. (2004). Monetary transmission and equity markets in the EU. Retrieved from http//som.eldoc.ub.rug.nl/files/reports/theme/2004/04E15/04E15 Enders, W. (2010). Applied econometric time series (3rd ed.). NY, USA: Wiley. Engle, R., & Yoo, B. (1987). Forecasting and testing co-integrated systems. Journal of Econometrics, 35(1), 143
159.

Financial Architecture and Monetary Policy Transmission Mechanism

363

Fadiran, G., & Ezeoha, A. (2012). South African market volatility, asymmetry and retail interest rate pass-through. South African Journal of Economics, 80(2), 157
180. Farzanegan, M., & Markwardt, G. (2009). The effects of oil price shocks on the Iranian economy. Energy Economics, 31(1), 134
151. Fawley, G., & Ezeoha, A. (2013). Four stories of quantitative easing. Federal Reserve Bank of St. Louis Review, 95(1), 51
88. Fernandez, R. (1981). A methodological note on the estimation of time series. The Review of Economic and Statistics, 63(3), 471
476. Gigineishvili, N. (2011). Determinants of interest rate pass-through: Do macroeconomic conditions and financial market structure matter? IMF Working Paper No. WP/11/176. International Monetary Fund, Washington, DC. Gottschalk, J. (2001). An introduction into the SVAR methodology: Identification, interpretation and limitations of SVAR models. Kiel Working Paper No. 1072, Germany. Goyal, A., & Pujari, A. (2005). Identifying long run supply curve of India. Journal of Quantitative Economics, 3(2), 1
15. Hoffman, L., & Rasche, H. (1996). Assessing forecast performance in a cointegrated system. Journal of Applied Econometrics, 11(5), 495
517. Joyce, M., Miles, D., Scott, A., & Vayanos, D. (2012). Quantitative easing and unconventional monetary policy
An introduction. The Economic Journal, 122(564), F271
F288. Khan, M., & Ahmed, A. (2011). Macroeconomic effects of global food and oil price shocks to the Pakistan economy: A structural vector autoregressive (SVAR) analysis. Retrieved from http://www.PIDE.org.Pk Kovanen, A. (2011). Monetary transmission mechanism in Ghana: Does the interest rate channel work. IMF Working Paper No. WP/11/275. International Monetary Fund, Washington, DC. Lack, C., & Lenz, C. (2000). A program for the identification of structural VAR-models. University of Basel. Retrieved from http://www.unibas.ch/fileadmin/wwz/makro/ papers/clc2000a.pdf Mangwengwende, T., Chizara, Z., & Nel, H. (2011). Bank concentration and interest rate passthrough in Sub-Saharan African countries. ERSA Working Paper No. 233. Cape Town, South Africa. Mariano, M. (2005). A SVAR model for estimating core inflation in the Euro zone. Applied Economic Letters, 12(3), 149
154. Mihira, T., & Sugihara, S. (2000). A structural VAR analysis of the monetary policy of Japan Economic Planning Agency. Discussion Paper No. 94. Tokyo, Japan. Misati, R., Nyamongo, E., & Kamau, A. (2011). Interest rate pass-through in Kenya. International Journal of Development Issues, 10(2), 170
182. Misati, R., Nyamongo, E., Njoroge, L., & Kaminchia, S. (2012). Feasibility of inflation targeting in an emerging market: Evidence from Kenya. Journal of Financial Economic policy, 4(2), 146
159. Misati, R., & Nyamongo, E. (2012). Asset prices and monetary policy in Kenya. Journal of Economic Studies, 39(4), 451
468. Mishra, P., & Montiel, P. (2013). How effective is monetary transmission in lowincome countries? A survey of the empirical literature. Economic Systems, 37(2), 187
216.

364

ROSELINE NYAKERARIO MISATI ET AL.

Montiel, P., Adam, C., Mbowe, W., & O’Connell, S. (2012). Financial architecture and monetary transmission mechanism in Tanzania. Oxford. CSAE WPS/2012-03. The University of Oxford, UK. Mugume, A. (2011). Monetary transmission mechanism in Uganda. The Bank of Uganda, Uganda. Retrieved from http:www.bou.or.ug Naceur, S., Boughrara, A., & Ghazouani, S. (2009). Monetary policy and asset pricing. Retrieved from http://ssrn.com/abstract=1554520 Naka, A., & Tufte, D. (1997). Examining impulse response functions in cointegrated systems. Applied Economics, 29(12), 1593
1603. Neaime, S. (2008). Monetary policy transmission and targeting mechanism in the MENA region. ERF Working Paper Series No. 395. Cairo, Egypt. Ngalawa, H., & Viegi, N. (2011). Dynamic effects of monetary policy shocks in Malawi. South African Journal of Economics, 79(3), 224
250. Paolo, C., & Antonio, P. (2012). Private sector balance, financial markets and U.S. cycle: A SVAR analysis. Journal of Economic Studies, 39(6), 709
723. Plaff, B., & Taunus, K. (2008). VAR, SVAR and SVEC: Implementation with R packages, Journal of Statistical Software, 27(4), 1
32. Renee, F., & Adrian, P. (2011). Sign restrictions in structural vector autoregressions: A critical review. Journal of Economic Literature, 49(4), 938
960. Sharifi-Renani, H. (2010). A structural VAR approach of monetary policy in Iran. ICOAE, 631
640. Shirakawa, M. (2013). Central banking: Before, during and after the crisis. IJCB, 9(S1), 373
387. Sichei, M., & Njenga, M. (2012). Does bank-lending channel exist in Kenya: Bank level panel data analysis. International Journal of Finance and Economics, 84, 25
40. Sims, C., Stock, J., & Watson, M. (1990). Inferences in linear time series models with some unit roots. Econometrica, 58(1), 113
144. Singh, B., & Pattanaik, S. (2010). Should monetary policy in India respond to movements in asset prices. Reserve Bank of India Occasional Papers, 31(3), 1
34. Stock, J., & Watson, M. (2001). Vector autoregressions. Journal of Economic Perspectives, 15(4), 101
115. Sugimoto, K., Takashi, M., & Yushi, Y. (2013). The global financial crisis: An analysis of the spillover effects on African stock markets. MPRA Paper No. 50473. Munich, Germany. Tahir, M. (2012). Relative importance of monetary policy transmission channels: A structural investigation; case of Brazil. Retrieved from Ecomod.net Trichet, J. (2013). Unconventional monetary policy measures: Principles-conditions. IJCB, 9(S1), 229
250. Verheyen, F. (2010). Monetary policy, commodity prices and inflation: Empirical evidence from the US. Ruhr Economic Papers No. 216. Walker, S. (2012). BLC in the EAC: The bank-lending channel of monetary policy transmission in the countries of the East African community. CSAE conference, March. Watson, M. (1994). Vector autoregressions and cointegration. In R. F. Engle & D. L. McFadden (Eds.), The handbook of econometrics (Vol. 4, pp. 2844
2910). North Holland: Elsevier. Wolfgang, S. (2007). Financial structure and its impact on the convergence of interest rate passthrough in Europe: A time-varying interest rate pass-through model. ENEPRI Working Paper No. 51. Brussels, Belgium.

PART V COUNTRY STUDIES ON MARKETS AND RISK

THE DIM SUM BOND MARKET IN HONG KONG Ike Mathur and Soumen De ABSTRACT The Dim Sum bond market in Hong Kong, which allows China to regulate the amount of offshore yuans that flow back into the mainland, has grown steadily since its inception in 2007 and is expected to surpass in 2013 the threshold level that would attract insurers and long-term issuers to the market. Yet, the market has not matured sufficiently relative to the yuan deposit market in Hong Kong that has grown at a much faster pace on account of trade liberalization and the use of yuans in China’s international trade settlements. Even though Hong Kong has fulfilled its role as an offshore currency center for the yuan, it is being challenged by Taiwan, Singapore, and London in terms of being the premier location for the issuance of yuan-denominated bonds outside of Mainland China. Keywords: Renminbi (RMB) Internationalization; Off-Shore RMB Deposit Market; Dim Sum Bond Market; Cross-Border Trade Settlement JEL classifications: F02; F31; F33; G15; G28

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 367
388 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096015

367

368

IKE MATHUR AND SOUMEN DE

INTRODUCTION Whether a country with supernormal economic growth prospects but with underdeveloped financial markets can deflect the rising demand for its currency and for unrestricted entry into its domestic capital markets by creating a highly regulated offshore center for its currency is currently being put to the test in Hong Kong. The yuan-denominated bond market in Hong Kong, also known as the Dim Sum bond market or the Renminbi (RMB)1 bond market, enables nonresidents of China
who are usually barred2 from participating in China’s onshore bond markets, to invest in yuandenominated bonds, issued by a select group of pre-screened entities and financial institutions, outside of China. This bond market, which allows accumulated offshore yuans to flow back into the Mainland under strict regulatory controls, was initiated in 2007 as part of China’s multi-faceted plan to begin internationalizing the yuan, launch a platform that would enable China to leverage Hong Kong’s superior financial infrastructure (Chen & Cheung, 2011; Fung & Yau, 2012) and gradually liberalize its capital account (Eichengreen, 2014). China is relying on Hong Kong’s relatively superior financial infrastructure to develop its offshore capital market which will eventually be fully integrated with its domestic financial markets in the Mainland. In recent years, large amounts of Chinese yuans have moved offshore as China implemented a sequence of measures to liberalize its current account and encourage settlements with its trading partners in yuans instead of the U.S. dollar (Sekine, 2011). Specifically, as (a) proportionately more bilateral trades between China and its Asian and Latin American trading partners have been transacted in yuans instead of U.S. dollars, (b) China invested in infrastructure projects around Asia, Latin America, and Africa through its sovereign wealth funds, and (c) Mainland Chinese tourists spent record amounts of yuans abroad, mainly in Hong Kong, the amount of offshore yuans seeking investable outlets have increased steeply over the years. China has the unique advantage of being a nation with two systems: the relatively restrictive, government-controlled regime in the Mainland and the free, market-based economic environment in Hong Kong. Even though it lacks at the moment the financial sophistication in the Mainland to permit full and unrestricted internationalization of its currency, China can utilize the platforms in Hong Kong to phase in its financial strategies gradually, thereby affording for itself time to upgrade the financial systems in the Mainland (Eichengreen, 2000).

The Dim Sum Bond Market in Hong Kong

369

China has allowed increasing numbers of entities to issue bonds in the Dim Sum bond market and officially move the offshore RMBs onshore via enhanced channels. Prior to October 2011, only companies outside of Mainland China, multinational agencies and financial institutions in China could issue Dim Sum bonds in Hong Kong without prior approval from regulatory agencies in China; beginning October 2011, companies domiciled in China can also issue yuan-denominated bonds in Hong Kong and repatriate the yuans into the Mainland if they satisfy the preconditions for such transfers. In March 2013, the China Securities Regulatory Commission (CSRC) expanded the scale and scope of the Renminbi Qualified Financial Institutional Investor (RFQII) program, which was launched in December 2011, to allow foreign financial institutions, Hong Kong subsidiaries of Mainland commercial banks and insurance firms and Hong Kong domiciled financial institutions to participate in the onshore RMB market. Clearly, China appears to be supplying the world with increasing amounts of offshore yuans on the one hand and creating on the other enhanced mechanisms of entry, albeit regulated, for these offshore yuans into the Mainland. Yet, the Dim Sum market remains small in its capacity to absorb all the yuans that have moved offshore due to enhanced trade settlements in yuans. It is small not because holders of RMBs are using other channels to move money onshore. It is small because the number of institutions issuing bonds in this market is not expanding as expected. The prevalence of strict oversight as to how the offshore RMBs can flow back into the mainland is discouraging major institutions and corporations from issuing longer term bonds on a large scale in the Dim Sum bond market. The object of our study is to evaluate the performance of the Dim Sum bond market since its inception, both in absolute terms and in relation to the measures that have been taken to promote the market and generate international interest in RMB bonds via the platform in Hong Kong. Despite being a smaller market for bonds than in the Mainland, the Hong Kong bond market for RMB bonds was projected to be the preeminent offshore center for the Chinese currency and RMB bonds. Whether Hong Kong has fulfilled its intended role as the center for RMB bonds to date will be explored in our study. We find that even though the Dim Sum bond market has grown appreciably since its inception in 2007 and is expected to surpass the critical threshold of $25 billion in 2013, the market is not absorbing an increasing proportion of the RMB deposits which have accumulated in Hong Kong on account of enhanced levels of trade settlements in RMBs and RMB

370

IKE MATHUR AND SOUMEN DE

spending in Hong Kong. The ratio of RMB bonds to RMB deposits in Hong Kong was the same in 2012 as it was in 2009. On the other hand, RMB holders have shown a preference for time deposits in Hong Kong, which have been growing at a much faster pace since 2007. Also, the majority of issues in the Dim Sum bond market have maturities less than five years. Even though investors were willing in the past to accept lower yields in anticipation of currency-related gains when the yuan appreciated, they have demanded since December 2012 higher yields and rated bonds on the Dim Sum market as the expectation that the yuan can only move up has weakened substantially. The expectation that the yuan can fluctuate is eventually good for the Dim Sum bond market since nobody would like to borrow in a currency that will only appreciate with time. We review in the section “Origins of the Dim Sum bond market” the economic and environmental circumstances that prompted Chinese authorities to inaugurate the Dim Sum bond market in Hong Kong. We provide a detailed review of the RMB bond market, emphasizing (a) the origins of the RMB bond market in Hong Kong, and (b) the drivers of current and future growth in the RMB bond market. The actions of the regulatory authorities in terms of expanding the scale and scope of trade settlements in yuans, permitting an enlarged set of institutions and entities to issue bonds in the Dim Sum bond market and opening up at the same time additional channels by which offshore yuans can flow back into the Mainland are documented and their effects on the Dim Sum bond market are explored. In the next section, we analyze the performance of the Dim Sum bond market since its inception, both in absolute terms and relative to the bond markets in the Mainland and in Hong Kong. In the section “Hong Kong’s Role in the Development of the RMB Bond Market”, the pivotal role Hong Kong has played so far in promoting the RMB bond market is explored and we examine the question whether Hong Kong can continue to serve as the nerve center for the RMB bond market. The last section of the chapter concludes.

ORIGINS OF THE DIM SUM BOND MARKET The Dim Sum bond market was inaugurated in January 2007, well before the global financial crisis struck in August of 2008.3 As a direct consequence of the post-1997 Asian crisis initiatives to increase the resiliency of nations to foreign currency crises, bond issuances in Asian nations were on

The Dim Sum Bond Market in Hong Kong

371

the rise since 2000. Domestic and foreign entities were being encouraged to participate in both the Asian local currency and Asian foreign currency bond markets (Goswami & Sharma, 2011; Hesse & Dalla, 2009). The Asian bond market has grown steadily since the 1997
1998 crisis until 2009 when it dipped due to the global financial crisis, but steeper growth rates commenced in 2010 and record levels of bonds have been issued by Asian nations since then (Shim, 2012). In particular, rising global demand for Asian bonds, increasing bond issuance in ASEAN countries, enhanced foreign participation in Asian local currency bond markets amidst declining preference for G34 bonds, and greater share of bonds vis-a`-vis stocks in global portfolios were the principal forces driving the growth in Asian bond market (Eichengreen, 2006). China’s liberal policies with regard to its current account that led to a much-enhanced supply of yuans in offshore markets and the concurrent events in the Asian bond market convinced Chinese authorities to begin an offshore market for yuan-denominated bonds. Lacking the necessary financial infrastructure in the Mainland to support both unrestricted capital inflows and unregulated participation by foreigners in its local currency bond markets, China opted instead to implement a controlled experiment in Hong Kong, hoping that the Dim Sum bond market will permit the Chinese authorities to regulate the amount of yuans that will move back into the Mainland from the pool of yuans which have flown, and will continue to flow in increasing amounts, out of China as a result of a step up in its current account liberalization policies since 2001. To be on par with trends in other Asian nations, restrictions on the issuance of yuan-denominated bonds by the Chinese subsidiaries of foreign banks in Hong Kong were formally lifted by the State Council in January 2007. Financial entities in Mainland China were allowed by the People’s Bank of China and the National Development and Reform Commission to issue RMB-denominated bonds in Hong Kong. The decision by the State Council of the People’s Republic of China in January 2007 to allow banks headquartered in the Mainland to issue in Hong Kong bonds denominated in RMB was the first official measure implemented by China to permit nonresidents to hold RMB-denominated bonds issued outside the Mainland. Named as “Dim-Sum” bonds, these RMB bonds issued in Hong Kong have allowed holders of RMB outside of China to invest in RMBdenominated bonds issued in Hong Kong by banks, corporations, and government institutions. This marked the beginning of a new, emerging market for RMB bonds, miniscule at that moment compared to the local and foreign currency bond markets both in Hong Kong and Mainland China, but

372

IKE MATHUR AND SOUMEN DE

with a potential to become even larger than the Euro yen bond market in Asia in the years ahead. First issue of RMB bonds by a financial institution in the Mainland were undertaken in June, 2007 when China Development Bank issued the first yuan currency bond outside Mainland. The chronology of policy measures implemented by Chinese authorities detailed in appendix would indicate that the Dim Sum bond market was not launched to take advantage of the crisis in developed economies. Rather, the Dim Sum bond market was started to herald China’s commitment to its “one nation two systems” goals and to begin an offshore bond market in Hong Kong in light of what was then happening in Asia. However, as our analysis in the section on The RMB Market in Hong Kong would demonstrate, the Dim Sum market grew rapidly only after the Chinese authorities extended the scale and scope of use of yuans in trade settlements in 2011 and the supply of offshore yuans increased appreciably to fuel the growth of the Dim Sum bond market.

Current Account Liberalization, Increasing Supply of Offshore Yuans and the Growth of the Dim Sum Bond Market China announced a 4 trillion RMB ($586 Billion) fiscal stimulus package in November 2008 and in December 2008, eligible Mainland and Hong Kong enterprises were permitted to use Renminbi to settle trade transactions. Restrictions on the issuance of yuan-denominated bonds by the Chinese subsidiaries of foreign banks in Mainland China were formally lifted by the State Council in December 2008. In April 2009, Sate Council decided to introduce a pilot scheme for using Renminbi in cross-border trade settlements and allowed Hong Kong banks to issue RMB bonds, and in May 2009, Shanghai City government confirmed the lifting of the restrictions on the issuance of yuan-denominated bonds by the Chinese subsidiaries of foreign banks in Mainland China. By June 2009, scope of trade settlements in RMB was greatly enhanced. China’s State Council decided to use RMB in cross-border trade settlement as part of its long-term plan to globalize its currency and reduce the domination of the U.S. dollar. The scheme only applied to dealings with 450 designated Chinese enterprises in five cities
Shanghai, Shenzhen, Guangzhou, Dongguan, and Zhuhai. Outside the Mainland, it was open only to 10 ASEAN nations, Hong Kong and Macau. By the end of 2010, the number of approved Mainland Designated Enterprises (MDE)
companies which could settle both exports and imports in RMB
was

The Dim Sum Bond Market in Hong Kong

373

expanded to 67,359 from 365; a non-MDE could only settle imports in RMB. Within a year, China would sign a momentous agreement with Japan. By December 2011, PBoC announced policies to promote the use of RMB in trade and financial settlements between China and Japan; Japan became the first developed economy to hold RMB bonds as reserve assets.5 By the end of 2011, China had committed itself irrevocably to internationalizing its currency and had taken the all-important step of convincing a major developing nation, Malaysia, to hold RMBs as a reserve currency. Liberalization of current account transactions requires China to open up bilateral swap lines for RMBs with its major trading partners. China should be prepared to provide RMBs should there ever be a scarcity of RMBs in global financial markets to settle terms of trade. But China’s swap lines with its trading partners are heavily underutilized to say the least. Of the 21 swap lines China has opened as of July 2013, only a few nations have drawn on their balances since China runs a deficit with most of its trading partners in the developing economies. China, on average, buys more from its Asian trading partners than it sells to them. As such, with each instance of enhanced trade liberalization only more RMBs accumulate in offshore markets. In the near future, with enhanced liquidity in offshore yuan markets, one would expect the Dim Sum bond market to play even a greater role in allowing the offshore yuans to flow back into the Mainland. For all practical purposes, the capital account of China remains closed. More importantly, the access to the onshore debt market is closed to nonresidents. Foreign institutions, with the exception of banks, are barred from participating in the onshore bond markets. RMB market in Hong Kong provides a way out of this problem. With regulated policies, holders of RMB-denominated assets in Hong Kong will be able to get access to onshore RMB accounts. Corporations in Hong Kong are now permitted to issue bonds in onshore markets of China and companies in Mainland China are now allowed to issue Dim Sum bonds in Hong Kong and repatriate the funds onshore. Nevertheless, the ultimate growth of the Dim Sum bond market will depend on the ability of issuers to invest the RMB proceeds directly and freely in Mainland China since investment opportunities are growing fastest there. Presently, RMBs raised in Hong Kong can be invested in China if the investments are in form of foreign direct investments and equity, approved by the authorities. Clearly, the RMB bond market cannot grow to its potential if such restrictions remain in force. With more RMBs leaving China on account of current account liberalization and China’s sovereign investments in other nations, pressures to

374

IKE MATHUR AND SOUMEN DE

liberalize the flow of RMBs into the Mainland will only intensify in the years ahead.

The Sources of Future Growth in the Dim Sum Bond Market Standard Chartered Bank (Special Report, 2011) has projected that annual settlement in RMB will total $1 Trillion by 2015. Imports account for 80% of RMB trade settlement and they represent an increasingly important source of RMB migration from onshore to offshore. Additionally, China is pushing to replace dollar invoicing by RMB invoicing for its trading partners in Asian, African and Latin American markets in the years ahead. Emerging markets which often use U.S. dollars to settle terms of trade are expected to go along with these measures as their trade with China increases in the future. The China ASEAN Free Trade Area (CAFTA), signed in November 2004 to develop a free trade area between China and ASEAN nations, went into effect January 1, 2010, for six nations
Brunei, Indonesia, Malaysia, the Philippines, Singapore, and Thailand. The remaining four nations, Myanmar, Cambodia, Laos, and Vietnam, will join CAFTA in 2015.6 China is slowly but steadily displacing the United States as a major trading partner of ASEAN nations. In addition to providing trading privileges to the member nations, CAFTA is committed to imparting a fair investment mechanism for investors from both sides. Chinese companies investing in ASEAN countries can also enjoy most favored nation (MFN) treatment. In addition, China is investing in infrastructure projects in Africa and Latin America and providing infrastructure assistance to developing and emerging economies using RMBs as the transaction medium. Such practices will allow greater migration of RMBs out of Mainland China and these financial innovations will have benevolent consequences for the RMB bond market in Hong Kong as more offshore yuans migrate to Hong Kong. RMB deposits in Hong Kong are expected to grow to a level of $800 billion by the end of 2015, according to Standard Chartered Bank
one of the three principal banks responsible for developing the RMB bond market in Hong Kong. Another important determinant of the growth of RMB-denominated markets in Hong Kong is the pressure on Chinese firms to go global and develop centralized treasury centers in selected regions around the world. Chinese companies have lagged behind some of their peers in Korea and Japan but there has been a sea change in the international strategies of

The Dim Sum Bond Market in Hong Kong

375

Chinese firms beginning 2010. Chinese companies in the energy, resources and telecommunications sectors have been more open to going abroad and they have selected Hong Kong, Singapore and Dubai as international treasury center locations. The strong demand for Chinese imports in the United States will always make it a prime location for expansion but Chinese companies have chosen these three locations for treasury management functions for the moment and it is not unreasonable to expect that Hong Kong will continue to be the central location for offshore RMBs. Therefore, the influx of RMBs into Hong Kong may usher in new paths of intermediation that will only benefit the Dim Sum Bond market in the future.

Flows of Offshore RMBs into the Mainland and the Role of the Quota System Current rules allow money to enter and leave China freely as long as it is backed by trade documents. As noted earlier, China has been fairly liberal with regard to its current account since the mid-1990s. Further, it had to agree to several relaxations with regard to banking transactions in order to get admission to the World Trade organization in 2001. But the situation is quite different when currencies have to be exchanged and converted. Offshore participating banks are allowed to purchase RMB from their onshore counterparts for purposes beyond trade settlement subject to quota and other regulatory constraints. China maintains strict administrative controls on this channel of currency flows in and out of Mainland China. A quota system is in force, but in February 2010, the Hong Kong Monetary Authority (HKMA) permitted banks in Hong Kong to develop new RMB-based business models as long as the money did not flow back to the Mainland. This has led to freer circulation of RMB in Hong Kong and has resulted in the growth of several RMB instruments and markets, especially the Dim Sum bond market. On the other hand, there are other forces which fuel the accumulation of RMB deposits in Hong Kong. Exporters to China who received settlement in RMB and merchants in Hong Kong who are recipients of tourismrelated RMBs tend to keep the RMB offshore and not return to the onshore banks. Such tendencies resulted in offshore banks overshooting their quotas for RMB transactions unrelated to trade very quickly and the deposits in Hong Kong rising very rapidly. While HKMA responded to this trend by raising the quota for RMB transactions for offshore banks in Hong Kong, Mainland Chinese

376

IKE MATHUR AND SOUMEN DE

authorities allowed more companies in China to settle exports and imports in RMB, adding only to the buildup of RMBs offshore which were gradually being reflected in the burgeoning RMB deposits in Hong Kong. While the Dim Sum bonds provided an outlet for these deposits to a degree, the major destination of these flows were time deposits with banks in Hong Kong. China has limited the number of offshore access points to onshore markets with active deliberation and has entrusted Hong Kong to develop, promote, and extend the RMB business globally. In effect, China is conducting a controlled experiment through Hong Kong, which is a reputed international financial center capable of ushering in financial innovation and tapping international capital markets for the ultimate benefit of China. The HKMA has world-class clearing and settlement systems already in place and is in a position to deal with an expanded set of RMB products. For the foreseeable future, it would seem that all RMB flows into and out of China will be governed by a quota system or a case-by-case regulatory approval. The quotas may be periodically raised to match demand to a degree but it is readily evident that China would maintain the quota system as it liberalizes its capital account gradually over the years. Authorities in China have signaled that full convertibility of the RMB will be phased in gradually and expansion of the RMB market will be attempted with controls in place. Full convertibility of the RMB is not a necessary condition at the moment for the RMB market to grow but there will come a time the absence of full convertibility will hinder the growth of the RMB market. The Dim Sum bond market is case in point. If this market is to attract foreign issuers intent on raising RMBs that can be used to finance their expansions in Mainland China, the offshore RMB will have to be fully convertible to onshore RMBs. In addition to comparing liquidity, spread and pricing, foreign issuers will also consider the ease with which RMBs raised offshore can be repatriated to onshore. Despite recent relaxations, several restrictions still remain in force.7

THE RMB MARKET IN HONG KONG VIS-A-VIS LOCAL AND FOREIGN CURRENCY BOND MARKETS IN MAINLAND CHINA AND HONG KONG: A COMPARATIVE ANALYSIS The size of the RMB bond market has grown from $1.34 billion in 2007 to $23.85 billion at the end of 2012 and up to $18.48 billion for the eight

The Dim Sum Bond Market in Hong Kong

377

months ending August 2013. On the other hand, total RMB deposits in Hong Kong have ballooned from $1.5 billion in 2004 to $95.6 billion in 2012 and to $112.8 billion by the end of June 2013. The number of banks engaged in RMB-related business has grown from 39 in 2008 to 140 in 2013. The deposit base was only $9.2 billion at the end of 2009. The phenomenal increase in the deposit base since 2009 is attributable to China’s decision in December 2008 to allow eligible Mainland and Hong Kong enterprises to use Renminbi in trade settlements. The RMB bond market which was only $6.15 billion in 2010 rose almost fourfold to $23.60 billion in 2011. This increase can be attributed to the increased supply of offshore RMBs on account of trade liberalization policy measures implemented by China in 2010. The RMB bond market capitalization to RMB deposit ratio, which was 25.47% at the end of 2009, receded to 13% at the end of 2010, has risen back to 24.96% by the end of 2012. Evidently, the RMB deposit base in Hong Kong has grown at a rapid rate and the RMB bond market has been able to absorb only a small fraction of the RMBs circulating in Hong Kong. Holders of RMBs in Hong Kong have shown a preference for RMB-denominated deposits, especially for time deposits, which have increased from $0.8 billion in 2004 to $112.76 billion by June 2013. As a proportion of total RMB deposits in Hong Kong, time deposits have risen from 55.35% in 2004 to 81.73% at the end of June 2013. The amount of RMB bond issuance in 2013 is expected to exceed $25 billion
a size deemed by experts to be large enough to draw pension and insurance funds to the market. The market for the bonds is still small but the interest it is generating among major issuers holds the key to its future potential growth. The range of issuers has expanded over time, from Mainland financial institutions at the beginning to the Chinese Ministry of Finance and the Mainland subsidiaries of Hong Kong banks in 2009, and eventually to Hong Kong-based and multinational corporations and financial institutions in 2010.8 With the entry of Chinese companies to the Dim Sum bond market in October 2011, the growth in this market is expected surpass prior records. The maturity of issues has lengthened from 2 to 3 years to 10 years in 2010 but the maturity for majority of bond issues in the Dim Sum market is less than three years (Fung & Yau, 2012). According to data compiled from Thomson One, investors are demanding more rated bonds than ever before and the yields on Dim Sum bonds since December 2012 have risen since the expectation that the yuan will only rise in the future has weakened substantially since the beginning of 2013. Table 1 lists some important financial variables pertaining to China and Hong Kong for the years 2007 and months in 2013 for which the latest

378

Table 1.

IKE MATHUR AND SOUMEN DE

Important Financial Indicators in 2007 and 2013 in Comparison to the RMB Deposits and Bond Issuance in Hong Kong. Unit

Foreign Currency Demand Deposits in Hong Kong Foreign Currency Savings Deposits in Hong Kong Foreign Currency Time Deposits in Hong Kong Local Currency Bonds in China Local Currency Bonds in Hong Kong Foreign Currency Bonds in China Foreign Currency Bonds in Hong Kong Total Remittance of the Renminbi Cross-Border Trade Settlement in Hong Kong Number of Authorized Banking Institutions in Hong Kong Number of Institutions Doing Business in RMB in Hong Kong Total RMB Deposits in Hong Kong RMB Bonds Issued in Hong Kong (1) As of September 2010 (2) As of June 2013 (3) As of August 2013

2007

2013

% Increase

U.S.$ million

20,816

60,532 (2)

190.8

U.S.$ million

66,604

182,952 (2)

174.7

U.S.$ million

269,253

307,962 (2)

U.S.$ million

1,688,000

4,044,744 (2)

136.5

U.S.$ million

97,980

177,490 (2)

81.1

U.S.$ million

34,970

U.S.$ million

44,550

129,020 (2)

189.6

U.S.$ million

4,508 (1)

43,750 (2)

870.5

Unit

200

201

Unit

37

140

U.S.$ million

4,531

112,756

2388.5

U.S.$ million

1,340

18,480 (3)

1279.1

159.520 (2)

14.37

356.2

data are available for the variable(s). From Table 1, we have noted that holders of RMB in Hong Kong prefer to hold time deposits. For the three classes of foreign currency deposits in Hong Kong for the time period under review, foreign currency time deposits comprise the largest proportion of the foreign currency deposit base even though they have shown the slowest rate of growth. Thus, the preference among RMB holders to hold time deposits is in accordance with the typical characteristic of Hong Kong’s foreign currency deposit holders.

The Dim Sum Bond Market in Hong Kong

379

As of June 2013, the foreign currency bond markets in China and Hong Kong were of comparable sizes but the local currency bond market in China is approximately 23 times the size of the corresponding market in Hong Kong. Despite Hong Kong’s financial sophistication, the size of its local currency bond market is indeed very small compared to that of China’s. Also, since the time China floated the Dim Sum bond market in Hong Kong, China’s foreign currency bond market has expanded by more than 356 percent and Hong Kong’s by 190 percent. It would seem that even though China is very conservative about offshore yuans moving in onshore, it has encouraged foreign currency bond issuance in the Mainland (Peiris, 2010) and the size of the market in China has surpassed that of Hong Kong’s as of June 2013. Foreign currency bond issuers have preferred Mainland China over Hong Kong even though Hong Kong is deemed to be financially more open and sophisticated than the Mainland. Cross-border trade in Renminbi was initiated in Hong Kong in September 2010. As of June 2013, total remittance of the Renminbi for cross-border trade purposes has expanded by 870 percent and amounted to $43.75 billion, which is far less than the total RMB deposits in Hong Kong ($112.76 billion). Thus, it is likely that offshore RMBs from places other than Hong Kong are moving to Hong Kong. Offshore RMBs can either be held as deposits in Hong Kong, invested in Dim Sum bonds, repatriated onshore through regular channels, extended as loans to entities onshore or swapped into other currencies. The size of the RMB deposit base in Hong Kong, in relation to the cross-border trade settlement amount, points to the possibility that Hong Kong residents are converting a substantive amount of Hong Kong dollars into RMBs and holding them as RMB deposits in Hong Kong. Residents of Hong Kong are permitted to convert 20,000 Hong Kong dollars per day to RMBs in Hong Kong. However, a recent report from the Bank of International Settlements9 notes that the RMB has overtaken the Hong Kong dollar for the first time in terms of daily turnover in foreign exchange transactions. The HKMA also noted that trading between U.S. dollar and Renminbi rose by 3.6 times to account for 17.7% of total average daily turnover, up from 4.5% in 2010, surpassing Hong Kong dollar against U.S. dollar as the most heavily traded currency pair in the Hong Kong market. This indicates that the RMB is playing a more dominant role in trade and investment activities and that as far currency trading is concerned, Hong Kong has lived up to China’s expectations. Foreign exchange trades involving offshore Renminbi (CNH) were around 63% of all Renminbi

380

IKE MATHUR AND SOUMEN DE

trades. Even though the Dim Sum bond market has not developed as expected, Hong Kong as an offshore center for the RMB has.

HONG KONG’S ROLE IN THE DEVELOPMENT OF THE RMB BOND MARKET Hong Kong’s privileged position with China is consolidated through seven different Close Economic Partnership Arrangements (CEPAs), which allow banks in Hong Kong to open branches in China with less capital and within a shorter time frame. Given the fact that Hong Kong has one of the highest concentration of large banking institutions in the world (GarciaHerrero, 2011), it is reasonable to expect that more banks will move into Hong Kong in order to have easier access to Mainland China in the future. As a matter of policy, China will first promote polices which will allow the RMB to be first regionalized before it can be truly internationalized. Hong Kong will play a pivotal role in the regionalization process and there is much to be gained in being in Hong Kong before investing opportunities are opened up in the Mainland. The Dim Sum bond market has been compared with the Japanese Euro Yen bond market that evolved to allow foreigners to finance Japan’s corporate growth in the 1980s amidst rising pressures on Japan to revalue its currency (Koo, 2003). China today is in a similar state Japan was in the 1980s. Foreigners would like to invest in China’s growth but China would like to maintain strict administrative controls over the direction of that growth. There are mounting pressures on China to revalue the RMB and internationalize its currency and permit free movement of capital across its borders. But, China lacks the financial sophistication that will permit it to open up its capital account without concurrent concerns about what free convertibility might entail. It is useful to recall that Japan delayed the opening of its markets fearing that hot money would flow in and exacerbate the problem of a rising yen. But when the markets were effectively opened up, capital actually flowed out of Japan on a net basis. Japan lacked suitable investment options at home and residents were keen to invest overseas. Such an outcome for China would have serious implications for its banking system that fuels the state-controlled domestic sector. China is determined not to open up its domestic markets too soon and markets in Hong Kong will be relied upon to moderate the flow of offshore yuans into the Mainland.

The Dim Sum Bond Market in Hong Kong

381

Of course, allowing multinational financial institutions and corporations to issue onshore in Shanghai would be viewed more favorably than allowing such entities to issue in Hong Kong. It is not that bond markets in Mainland China lack the desirable characteristics which Hong Kong is perceived to have. In fact, the bond market in Hong Kong, contrary to general beliefs, is presently not so active a market that it would be expected to stimulate the demand for Dim Sum bonds by virtue of its own inherent momentum. Not only is the market small compared to that of Mainland China’s, the need for long-term financing in Hong Kong barely goes beyond five years (Latter, 2008). The residents of Hong Kong have a distinct, dichotomous preference function: they prefer either the banking market or the stock market. The majority of the demand for long-term bonds comes from property developers, whose projects are usually completed within the five year time frame. Overall interest in bonds in Hong Kong, both from issuers and investors, is moderate at best (Latter, 2008). Multinational corporations and financial institutions will prefer to have the privilege of issuing securities in the Mainland rather than in Hong Kong and there indeed might be a demand for such securities in the Mainland. The Chinese authorities understand this motivation all too well and yet they continue to liberalize gradually the markets in Hong Kong and delay the opening the markets in the Mainland to foreign competition. The internationalization of a currency is a gradual process and a currency is internationalized only after it has been regionalized to a certain degree. In that respect, the structure and dynamics of the Dim Sum bond market will be carefully watched by global investors and the ultimate test of its success will be determined by the market’s ability to attract foreign investors and issuers alike. If foreign financial institutions and banks were keen on positioning themselves in Hong Kong in anticipation of China opening up its channels of repatriation of offshore RMBs through Hong Kong, we would expect an increase in the number of authorized institutions in Hong Kong since the Dim Sum bond was initiated in 2007. This has not occurred. In fact, the number of authorized institutions has remained stable around 200 since 2007. The reason might be that foreign institutions actually prefer to access the Chinese market directly as shown by the rapid growth of foreign financial institutions in Shanghai (Garcia-Herrero, 2011). Chinese authorities may have already apprehended this unexpected course of events.10 China has initiated offshore centers in Taiwan (2012), London (2012) and Singapore (2013) and has indicated that it would expand the scope of the offshore center in London soon. Such enhancements will not only

382

IKE MATHUR AND SOUMEN DE

provide alternative venues to Hong Kong but also reduce the pressures on itself to open up Shanghai to foreign currency and offshore RMB inflows soon. If the RMB is to become an international currency in the future, Hong Kong will assuredly play an important role but it might not be the dominant one. Hong Kong specializes in institutional banking relationships whereas Singapore emphasizes corporate banking relationships (Garcia-Herrero, 2011). The Dim Sum bond market will get the needed boost when corporations begin to issue bonds in this market on a larger scale. In May 2013, both London and Singapore recorded their first RMB-denominated bond sales whereas Taiwan sold its first RMB bond in February 2013. London and Singapore’s superior corporate linkages could lead to bigger RMB bond markets in their respective regions. Corporate issuers might prefer to issue in London and Singapore than in Hong Kong. The London market will carry the RMB bond market beyond the Asian time zone and this can only add to the global liquidity of RMB bond markets. Also, Taiwan’s trade with Mainland China has greater promise than Hong Kong’s. Therefore, Singapore and Taiwan will compete with Hong Kong to divert offshore balances to their respective hubs and it is evident that both nations have scalable advantages over Hong Kong. Hong Kong will face formidable challenges in the years ahead.

CONCLUSION We review the origins of the Dim Sum bond market in Hong Kong and analyze its growth since its inception in 2007. Even though the market was not initiated to absorb the increased levels of offshore RMBs circulating on account of China’s policies to foster trade settlements in RMBs, trade liberalization policies enacted by China in 2010 did in fact lead to a fourfold increase in the size of the RMB bond market in 2011. The trend continues even today and the RMB bond market is expected to surpass the critical size of $25 billion in 2013. This, according to the conventional wisdom about emerging bond markets, is likely to encourage foreign institutions to issue in the Dim Sum bond market in the near future. Hong Kong has played a pivotal role in promoting the Dim Sum bond market. But China’s recent moves to begin offshore centers in Taiwan, Singapore and London and the issuances of RMB-denominated bonds in each of these three regions in 2013 point to a motive to diversify the

The Dim Sum Bond Market in Hong Kong

383

strategy with regard to promoting the RMB as an international currency. Hong Kong no longer enjoys the unique status of being the only place where RMB bonds are issued. In the ultimate analysis, the offshore market in Hong Kong will be integrated with the Mainland markets; the markets in Taiwan, Singapore and London will not be. Thus, there are natural limits to the extent Hong Kong can grow as an independent offshore center for Dim Sum bonds and eventually the Chinese yuan.

NOTES 1. The Renminbi, known in its abbreviated form as RMB, is the official currency of the People’s Republic of China but not of Hong Kong or Macau. The supply of RMB is determined entirely by monetary authorities in China. Like the Federal Reserve in the United States, which determines the supply of U.S. dollars to the global economy, it is the Peoples Bank of China that determines and controls the supply of RMBs; yuan is the unit of account. 2. For the most part, nonresidents are not permitted to participate in domestic bond markets in Mainland China; only a handful of financial institutions are allowed to do so. 3. The active phase of the crisis, which manifested as a liquidity crisis, began on August 9, 2008, when BNP Paribas terminated withdrawals from three hedge funds citing “a complete evaporation of liquidity.” 4. G3 bonds are bonds denominated in U.S. dollars, Euro and the Japanese yen. 5. As of December 2011, China was Japan’s biggest trading partner ($340 billion by value) 6. According to the ASEAN Secretariat, trade between ASEAN nations and China grew at a rate of 20% annually between 2003 and 2008, and China’s trade deficit with ASEAN nations soared to $7.5 billion in the first seven months of 2010. For the same period, trilateral trade between China and ASEAN members rose to $161 billion
a 50% increase year to year. 7. The reforms of October 2011 have made repatriation of offshore funds easier but authorities in China still insist on retaining both regulatory and supervisory powers over fund flows into the Mainland over a statutory limit. China announced its policies with regard to RMB FDI effective October 13, 2011, expanding on its policies for U.S. dollar FDI flows already in place. Any application for RMB FDI valued at or above RMB 300 million ($47 million) must be submitted to the Ministry of Commerce for prior approval. Amounts below RMB 300 million need not be. FDI investments in securities and derivatives investments are still prohibited but foreign investors are allowed to participate in private placements of publicly listed companies subject to approval by the Ministry of Commerce. RMB funds for M&A activities will have to be approved by the Ministry of Commerce. These actions open up new channels of flow for offshore RMB into onshore in many ways and bode well for the offshore RMB market.

384

IKE MATHUR AND SOUMEN DE

8. To date, apart from the Ministry of Finance, the issuers that have issued or launched RMB bonds in Hong Kong included China Development Bank, China Export and Import Bank, Bank of China, Bank of Communications, China Construction Bank, HSBC (China), Bank of East Asia (China), Honeywell Highway Infrastructure Limited, McDonald’s, Asian Development Bank and International Finance Corporation. 9. Triennial Central Bank Survey of Foreign Exchange and OTC Derivatives Markets Activity, BIS Reports. 2013. 10. On September 29, 2013, Shanghai was designated as a free trading zone. However, specifics about policy changes have not been announced as of November.

REFERENCES Chen, X., & Cheung, Y. (2011). Renminbi going global. China and World Economy, 19, (2), 1
18. Eichengreen, B. (2000). Taming capital flows. World Development, 28(6), 1105
1116. Eichengreen, B. (2006). The development of Asian bond markets. In Asian bond markets: Issues and prospects, BIS Papers 30. Eichengreen, B., & Kawai, M. (2014). Issues for Renminbi internationalization: An overview. ADBI Working Paper No. 454. Asian Development Bank. Fung, H., & Yau, J. (2012). The Chinese offshore Renminbi currency and bond markets: The role of Hong Kong. China and World Economy, 20(3), 107
122. Garcia-Herrero, A. (2011). Hong Kong as an international banking center: Present and future. Journal of the Asia Pacific Economy, 16(3), 361
371. Goswami, M., & Sharma, S. (2011). The development of local debt markets in Asia. Working paper WP/11/132. International Monetary Fund, June. Hesse, H., & Dalla, I. (2009). Rapidly growing local-currency bond markets offer a viable alternative funding source for emerging-market issuers. Vox. Retrieved from http:// www.voxeu.org/article/growing-local-currency-bond-markets. Koo, R. (2003). Several observations on capital flows into Japan. BIS paper 15. Latter, T. (2008, October). What future for the Hong Kong dollar corporate bond market? Working paper 19/2008. Hong Kong Institute of Monetary Research. Peiris, S. J. (2010). Foreign participation in emerging markets’ local currency bond markets. IMF Working Paper WP/10/88. International Monetary Fund, Washington, DC. Sekine, E. (2011). Renminbi trade settlement as a catalyst to Hong Kong’s development as an offshore Renminbi center. Nomura Journal of Capital Markets, 3(1), 1–14. Shim, I. (2012). Development of Asia-Pacific corporate bond and securitization markets. BIS Paper 63. Standard Chartered Bank. (Special report, 2011, January). CNY supports CNY reserveification.

The Dim Sum Bond Market in Hong Kong

385

APPENDIX: CHRONOLOGY OF POLICY MEASURES ENACTED TO DEVELOP THE DIM SUM BOND MARKET Date

Introduction of/Changes in Policy Measures

January 2004

Ban against Hong Kong banks accepting yuan deposits (including money exchanges and remittances) is lifted

January 11, 2007

Restrictions on the issuance of yuan-denominated bonds by the Chinese subsidiaries of foreign banks in Hong Kong were formally lifted. The People’s Bank of China (PBoC), China’s central bank, announced that mainland financial institutions can issue Renminbi (RMB) financial bonds in Hong Kong with clearance from the central bank

June 26, 2007

China Development Bank issued the first yuan currency bond outside Mainland

August 9, 2007 The active phase of the global financial crisis, which manifested as a liquidity crisis, began on this date when BNP Paribas terminated withdrawals from three hedge funds citing “a complete evaporation of liquidity” November 9, 2008

China announced 4 trillion RMB fiscal ($586 Billion) stimulus package

December 2008

Eligible Mainland and Hong Kong enterprises to use Renminbi to settle trade transactions

December 8, 2008

Restrictions on the issuance of yuan-denominated bonds by the Chinese subsidiaries of foreign banks in Mainland China was formally lifted by the State Council

April 8, 2009

State Council decided to introduce a pilot scheme for using Renminbi in cross-border trade settlements and allowed Hong Kong banks to issue RMB bonds

May 8, 2009

Shanghai City government confirmed the lifting of the restrictions on the issuance of yuan-denominated bonds by the Chinese subsidiaries of foreign banks in mainland China

June, 2009

Scope of trade settlements in RMB greatly enhanced. China’s State Council decided to use RMB in cross-border trade settlement as part of its long-term plan to globalize its currency and reduce the domination of the U.S. dollar. The scheme only applies to dealings with 450 designated Chinese enterprises in five cities
Shanghai, Shenzhen, Guangzhou, Dongguan, and Zhuhai. Outside the mainland, it is open only to 10 ASEAN nations, Hong Kong, and Macau. The scheme is expected to cover more areas and countries when it wins widespread appeal.

386

IKE MATHUR AND SOUMEN DE

(Continued ) Date July 6, 2009

Introduction of/Changes in Policy Measures China and Hong Kong begin yuan-denominated trade settlement.

February 2010 HKMA streamlines operational arrangements of offshore RMB business in Hong Kong June 22, 2010

China authorities announced a major expansion of the RMB trade settlement scheme with the rest of the world

July, 2010

Offshore RMB trading begin in Hong Kong

July 19, 2010

Regulatory Changes that freed the yuan to be used for private investment setting off the issuance of bonds in Hong Kong. China expanded the trade settlement program that lets Chinese companies pay for imports with yuan instead of dollars. As a result foreign companies are building up yuan deposits. Only a limited amount of yuan can be brought from Hong Kong to China and vice versa without permission.

July 19, 2010

Clearing Agreement on RMB business amended; restrictions on account opening of corporate and interbank transfers removed. RMBs in Hong Kong can move around freely. Several financial products denominated in RMBs initiated in Hong Kong.

August 17, 2010

China authorities launched a pilot scheme for eligible institutions outside the Mainland to invest in the Mainland’s interbank bond market

September 2010

China issued its first yuan-denominated bond issue in Hong Kong

December 6, 2010

The number of approved Mainland Designated Enterprises (MDE)
companies which can settle both exports and imports in RMB
was expanded to 67,359 from 365. A non-MDE can only settle imports in RMB.

December 2010

Measures introduced to refine Hong Kong’s offshore RMB business

January 13, 2011

Domestic companies can move Renminbi offshore for investment purposes. Chinese companies can use yuan to buy foreign assets

January 13, 2011

Shanghai will make it easier for foreigners to invest in the private equity sector

January 13, 2011

Overseas sovereign wealth funds, pension funds, insurance companies and funds of funds will be able to directly invest in Chinese companies

January 2011

Pilot scheme for settlement of Mainland’s overseas direct investment in Renminbi launched. Chinese companies can now take funds out of China for M&A.

The Dim Sum Bond Market in Hong Kong

387

(Continued ) Date August 17, 2011

Introduction of/Changes in Policy Measures China endorsed the long-awaited mini-QFII and indicated that an initial quota of $3.1 billion would be shared among Hong Kong subsidiaries of Chinese securities companies and fund houses to allow them to invest directly into the mainland’s securities market. Chinese firms will also be allowed to buy into an ETF linked to Hong Kong stocks. Mainland Chinese companies were given the official go-ahead with regard to issuing offshore Renminbi bonds in Hong Kong. Earlier, only mainland financial institutions and overseas companies were able to issue

October 13, 2011

China finalized and published its rules allowing RMB FDI effective October 13, expanding on its policies for U.S. dollar FDI flows already in place. Any application for RMB FDI valued at or above RMB 300 million ($47 million) must be submitted to the Ministry of Commerce for prior approval. Amounts below RMB 300 million need not be. FDI investments in securities and derivatives investments are still prohibited but foreign investors are allowed to participate in private placements of publicly listed companies subject to approval by the Ministry of Commerce. RMB funds for M&A activities will have to be approved by the Ministry of Commerce. These actions open up new channels of flow for offshore RMB into onshore in many ways and bode well for the offshore RMB market.

June 14, 2012

The Hong Kong Monetary Authority (HKMA) announced a facility for providing Renminbi (RMB) liquidity to Authorized Institutions participating in RMB business (Participating AIs) in Hong Kong. The facility will make use of the currency swap arrangement between the People’s Bank of China and the HKMA. With effect from June 15, 2012, the HKMA will, in response to requests from individual Participating AIs, provide RMB term funds to the Participating AIs against eligible collateral acceptable to the HKMA. The introduction of the facility is to support the continuous deepening of the RMB capital market in Hong Kong and to reinforce Hong Kong’s role as the global hub for offshore RMB business.

October 11, 2012

PBoC announced that an RMB clearing bank will be set up in Taiwan. This will consolidate Taiwan’s position to be the second key offshore RMB center behind Hong Kong but ahead of London and Singapore. Bank of China (Hong King) became Hong Kong’s RMB clearing bank in 2003. It could lead to the second offshore Chinese currency CNT
as opposed to CNH in Hong Kong.

January 28, 2013

The Central bank of Taiwan (CBC) approved Bank of China (Taipei) as the offshore Renminbi clearing bank for Taiwan

January 30, 2013

13 companies and 6 banks are allowed to participate in China’s groundbreaking cash management pilots designed to better funnel Renminbi and foreign currencies across the border.

388

IKE MATHUR AND SOUMEN DE

(Continued ) Date

Introduction of/Changes in Policy Measures

January 30, 2013

PBoC appointed ICBC (Singapore) to be the RMB clearing bank in Singapore

January 30, 2013

PBoC doubled Singapore’s swap line from 150 million RMB to 300 million RMB

February 6, 2013

China continues to award QFII quotas more swiftly. The State Administration of Foreign Exchange (SAFE) granted Abu Dhabi Investment Authority $500 million and Kuwait Investment Authority $700 million in additional quota, taking both to $1 billion each. They join five others (all State backed) which have $1 billion quota as well: Government of Singapore Investment Corporation, Hong Kong Monetary Authority, Norway’s Norges Bank, Qatar Investment Authority, and Singapore’s Temasek Fullerton Alpha Investments.

June 4, 2013

The Singapore Branch of the Industrial and Commercial Bank of China (ICBC) kicked off its RMB clearing service in Singapore, marking an important step in Singapore’s development as an offshore RMB center. ICBC’s Singapore branch is the first RMB clearing bank designated by China in another country; with Hong Kong and Taiwan previously the only places outside of Mainland China with designated RMB clearing banks.

June 22, 2013

Bank of England and the People’s Bank of China signed a 3-year 20 billion yuan ($32.6 billion) swap deal

July 15, 2013

Renminbi Qualified Foreign Institutional Investor (RQFII) program extended to include institutions in Singapore and London. The program was extended to Taiwan in June and to Hong Kong in March. Total quota raised from $80 billion to $150 billion in an effort to increase foreign investment into the mainland

July 15, 2013

PBoC lifts a host of restrictions designed to make RMB cross-border trade settlement easier to conduct and make inter and intra company loans denominated in RMB. RMB trade settlement to China’s total trade increased from 3% in 2010 to 9% in 2011 and 12% in 2012.

July 25, 2013

The use of RMB in international payments hit a record market share of 0.87% in June, growing threefold from 0.24% two years ago. The RMB now ranks 11th in the rankings of world payments currencies.

August 31, 2013

In 2012, China’s total trade amounted to $3867 billion, surpassing the United States by a billion U.S. dollars

Source: People’s Bank of China, Asiamoney, Euromoney, FinanceAsia press releases, Xin Hua press releases and news items.

SENTIMENT AND BETA HERDING IN THE BORSA ISTANBUL (BIST) Nazmi Demir, Syed F. Mahmud and M. Nihat Solakoglu ABSTRACT This study searches for sentimental herding in Borsa Istanbul (BIST) during the last decade using a state-space model employing cross-section standard deviations of systematic risk (Beta). It has been found that herding toward the market in the BIST-100 is both statistically significant and persistent independently from market fundamentals such as the volatility of returns and the levels of market returns. Herding trends over the sample period indicate that the financial crisis in 2000
2001 appeared to bring about sentimental herding in BIST which was followed by a calm period during which investors turned to fundamentals. Thereafter, we observe a volatile adverse herding pattern till the end of 2011 due to the confusing environment caused by the internal and external events. Keywords: Beta herding; state-space model; market fundamentals; cross-section volatility JEL classifications: C13; C31; G14; G12

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 389
400 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096016

389

390

NAZMI DEMIR ET AL.

INTRODUCTION Herding in stock markets continues to be one of the popular issues in financial literature, as it may defy the efficient market hypothesis (EMH). Investors are normally expected to make informed decisions based on their expectations about returns, using models such as the Capital Asset Pricing Model (CAPM). Under the presence of herding, however, investors may follow the actions of the crowd, disregarding market fundamentals, thus leading to a mispricing of securities. A review of the existing literature, both empirical and theoretical, reveals two different views on herding: institutional herding and market-wide herding. “Institutional herding” is described as the tendency of a group to invest in (sell) certain stocks at the same time as opposed to a pattern expected when investors act independently of one another (Bikhchandani & Sharma, 2001). “Market wide herding,” on the other hand, focuses on the collective behavior of all participants toward the market as a whole. In this chapter, we examine the market herding behavior in the BIST from 2000 to 2011, a time which includes not only the crisis years of 2000
2001 and 2007
2008 but also a period of successful structural reforms in the country. Different from earlier herding studies on BIST, this study utilizes cross-sectional standard deviation (CSSD) of systematic risk, rather than CSSD of returns, in a state-space framework. ˙ BIST is the new name for the Istanbul Stock Exchange (ISE), which was founded in 1986 and brings together all exchanges operating in Turkish capital markets. From its inception, with only about 657 million USD and 80 listed companies, by July 2013, the market capitalization had increased to 261 billion USD with 229 listed companies in the national market. The years considered in this study are particularly interesting as they include one period of political and economic instability combined with high and volatile inflation and a second period of political and economic stability caused by a set of structural reforms and credible macroeconomic policies with the support of the IMF, the prospect of EU accession, and a singleparty majority government. A strong recovery was witnessed during 2002
2007, when real GDP grew at an average of 6.8% annually. According to a recent European Commission Report, it can be concluded that the economic reforms after 2001 have both corrected the previous problems and given Turkish markets a clear break with the past (EC Report, 2009). The report also concluded that Turkey was able to weather recent global stormy conditions and to avoid the currency and financial crises in 2007
2008. These policies also

Sentiment and Beta Herding in the Borsa Istanbul (BIST)

391

helped Turkey in the “catching-up” process and facilitated convergence with member countries of the EU. For example, the GDP per capita increased from a low level of 3,250 USD in 2002 to 9,000 USD by 2007 (EC Report, 2009). With the ongoing negotiations and prospect of Turkey’s accession to the EU, the successful implementation of domestic reforms, and the recent history of significant inflows and outflows of foreign capital from and into BIST have all further motivated the authors to undertake this study. In this chapter, the literature is briefly reviewed in the section “Review of the Literature.” The section “Methodology” outlines the empirical model and the section “Data and Results” provides details of the data employed. Finally, discussions of the results as well as conclusions are presented in the section “Conclusions.”

REVIEW OF THE LITERATURE Herding in financial markets describes a situation in which investors follow the action of others. Theoretical and empirical research in explaining and testing such behavior, however, does not seem to converge to a single approach. On the one hand, studies have focused in explaining the behavior of investors, either institutional or private, to follow the actions of others. Such behavior is further classified as rational when following the majority or a group of investors who may be perceived to have access to better information (see Banerjee, 1992; Frey, Herbst, & Walter, 2007; Gompers & Metrick, 2001; Kim & Sias, 2005; Lakonishok, Shleifer, & Vishny, 1992; Oehler & Chao, 2000; Puckett & Yan, 2007; Voronkova & Bohl, 2005, among several others). The second group of studies employs a “market wide” approach, which focuses on the cross-sectional dispersion of returns or betas. One of the pioneer studies with this approach is Christie and Huang (1995), who utilized the CSSD of returns as a measure of the average proximity of individual asset returns to the realized market returns. In their approach, the CSSD of individual security returns is being regressed using a constant and two dummy variables to capture extreme positive and negative market returns. In their model, negative values of the two coefficients, under market stress, would imply herding in the market. When they applied this model to US stock markets, they could not find any evidence of significant herding. As a different approach, Chang, Cheng, and Khorana (2000) extended the work of Christie and Huang by establishing a nonlinear relationship

392

NAZMI DEMIR ET AL.

between the level of cross-section absolute deviations (CSAD) of returns and the overall market return. While they found no evidence of herding in developed economies, herding was found in emerging economies such as South Korea and Taiwan. In recent years, several papers have studied herding behavior in the Turkish market using BIST data and following either Christie and Huang (1995) or Chang et al. (2000). For example, Kapusuzoglu (2011) and Kayalıdere (2012) found that herding behavior was detected during the days of the rising market represented by the BIST100. Coban (2009) and Dog˘ ukanlı and Ergu¨n (2011) also using the Turkish BIST data, did not find any supporting evidence for herding. Hwang and Salmon (2004) have argued that the Christie and Huang approach, with its use of dummy variables to capture extreme positive and negative market returns, is subjective at best in defining extreme situations. Furthermore, the presence of a strong correlation between the CSSD of individual stock returns and dummy variables in the Christie and Huang approach may lead to an inability of identifying whether changes are a result of volatility or of herding. Believing that herding is an outcome of investors’ unobserved sentiments, Hwang and Salmon (2004) developed a state-space model to reveal empirically herd behavior toward the market. In this chapter, we follow the Hwang and Salmon model to measure herding behavior in BIST.

METHODOLOGY The hypothesis of Hwang and Salmon (2004) emphasizes sentiment behavior of investors which is unobservable and moves in association with the systematic risk indicator beta. They also use the cross-sectional behavior of assets similar to that of Christie and Huang (1995). Their model is different, however, in that it is based on the notion of market-wide herding using betas rather than returns. In market-wide herding, investors may decide to follow market trends, and this may cause the individual asset returns to move in tandem with the market returns. As the sentiments of the investors vary, the beta values of the stocks will also change from their constant initial values, confining themselves to the market beta of unity. In short, the theoretical and empirical modeling of herd behavior used by Hwang and Salmon differs fundamentally from Christie and Huang and Chang et al. (2000) in that they make a clear distinction between unobserved sentimental herding and herding associated with market fundamentals, and they provide a framework in which to estimate it. The other models rely on

Sentiment and Beta Herding in the Borsa Istanbul (BIST)

393

observed dispersions of returns brought about by a combination of sentimental and fundamental movements, in addition to the other limitations discussed earlier. The model is based on a relationship between observed biased beta (βbimt) and unobserved true beta (βimt) as follows: Etb ðrit Þ = βbimt = βimt − hmt ðβimt − 1Þ Et ðrmt Þ

ð1Þ

where Etb(rit) is the biased short-run conditional expectation on the excess return of asset i at time t, and Et(rmt) is the conditional expectation of the market excess return at time t. The unobserved herd behavior indicator hmt is the parameter assumed proportional to the deviation of the individual true beta from market beta unity. The cross-sectional variation of βbimt becomes: Stdc ðβbimt Þ = Stdc ðβimt Þð1 − hmt Þ

ð2Þ

Taking logarithms of both sides of Eq. (2), we get, ln½Stdc ðβbimt Þ = ln½Stdc ðβimt Þ þ lnð1 − hmt Þ

ð3Þ

We may now re-write Eq. (3) as: ln½Stdc ðβbimt Þ = μm þ Hmt

ð4Þ

where μm = ln [Stdc(βimt)] is an assumed constant in the short time and Hmt = ln(1 − hmt). Hwang and Salmon (2004) now allow herding, Hmt, to follow a dynamic process AR(1), such that the system becomes:1 ln½Stdc ðβbimt Þ = μt þ Hmt þ vmt

ð5Þ

Hmt = φm Hmt − 1 þ ηmt where the two error terms are vmt ∼iid(0,σ2mv) and ηmt∼iid(0,σ2mη), respectively.

394

NAZMI DEMIR ET AL.

The two equations in Eq. (5) constitute the standard state-space model. In our estimations, Eq. (5) is referred as Model 1. One of the key parameters of interest in Eq. (5) is the variance of the error term of the state equation σ2mη. When σ2mη is zero, it would imply that there is no herding, since Hmt = 0 for all t. A statistically significant value of σ2mη however, would indicate the presence of herding in the market. Furthermore, a significant φm, provided that |φm|≤1, would support the autoregressive process. Hwang and Salmon (2004) further tested the robustness of their model by including both market volatility and market returns in the first equation of the model. They argued that if Hmt becomes insignificant after the inclusion of these market fundamentals in the model, then changes in Stdc(βbimt) can be explained by market fundamentals rather than by herding. Model 1 can therefore be modified to include these fundamentals as control variables to test for robustness as follows: ln½Stdc ðβbimt Þ = μm þ Hmt þ θc1 ln σ mt þ θc2 rmt þ vmt

ð6Þ

Hmt = φm Hmt − 1 þ ηmt where lnσmt and rmt represent market volatility and log return in time period t. The two equations in Eq. (6) constitute our Model 2.

DATA AND RESULTS The daily share prices for the firms in the BIST-100 index as well as their index levels were obtained from the data base of Borsa Istanbul.2 The data cover the dates between January 4, 2000 and October 4, 2011, providing us with 2,896 usable observations. In calculating the monthly beta of a stock, we utilized the market model and regress log returns of firm i on the log returns of the market index, using daily data for each month. Hence, we have one beta estimate for each firm for each month, giving us 100 beta estimates per month (unless there are missing data in the sample). Once we have the individual betas, we calculate the CSSD of betas for each month based on the following formula: StdðbetaÞt =

ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn 2 ðbeta − beta Þ it t i=1 n−1

where t represents the month, i represents the firm i, and betat represents the cross-sectional average of all betas in month t.

395

Sentiment and Beta Herding in the Borsa Istanbul (BIST)

Table 1 reports some of the statistical properties of the estimated CSSD of the betas covered by the BIST-100 market portfolio. The results indicate that the Stdc(βbimt) is significantly different from zero. The Jarque
Bera statistics for normality clearly suggest that the distribution of the stdc(βbimt) is not Gaussian. However, estimated log-CSSD does not seem to deviate significantly from Gaussianity.3 The maximum likelihood estimates of the parameters for Model 1 and Model 2 are reported in Tables 2 and 3. Table 1. Sample Descriptive Statistics of Cross-Sectional Std. (βb). Std. Dev. of Betas

Log-Std. Dev. of Betas

0.379321 0.097064 1.429002 10.44427 373.5636

−1.000187 0.250089 −0.180645 4.278039 10.363

Mean Std. Dev. Skewness Kurtosis Jarque
Bera

Table 2.

Kalman Filter Results of the Main Herding Model (Model 1).

Kalman Filter State-Space Model Maximum likelihood method Number of observations: 141 Coefficient Std. Error z-Statisitcs

Prob.

Variable

C(1)

−1.035022

0.110726

−9.347618 0.0000 µm

−1.035022

C(2)

0.971149

0.029436

32.99142

0.0000 φm

0.971149

C(3)

−3.326986

0.112373

−29.60673

0.0000 σmη

0.0391

C(4)

−6.485165

0.771385

−8.407174 0.0000 σmv Signal to Noise Ratio

Log likelihood

21.2588

Akaike criterion

−0.2448

Schwarz criterion

−0.1611

0.1895

σmη/log Std. β

a

0.1564

a log-Std. β represents the time series standard deviation of log-cross-sectional standard deviation of betas.

396

Table 3.

NAZMI DEMIR ET AL.

Kalman Filter Results for Herding with Market Fundamentals (Model 2).

Kalman Filter State-Space Model Maximum likelihood method Number of observations: 141 Coefficient C(1)

−0.928155

C(2)

Std. Error

z-Statisitcs

Prob.

Variable

0.063347 −14.65190

0.0000 µm

−0.928155

0.972594

0.035059

27.74201

0.0000 φm

0.972594

C(3)

−7.738176

1.247388

−6.203501 0.0000 σmη

C(4)

−3.578688

0.100245 −35.69937

C(5) C(6)

−793.4523 0.607242

80.61130 0.350114

0.0000 σmv

0.1671

−9.842941 0.0000 θc1

−793.4523

1.734412 0.0828 θc2 Signal to Noise Ratio

Log likelihood

44.7768

Akaike criterion

−0.5500

Schwarz criterion

−0.4245

0.0209

σmη/log Stdβa

0.0607242 0.0836

log-Std. β represents the time series standard deviation of log-cross-sectional standard deviation of betas.

a

The estimated value of φm and its statistical significance suggest that the Hmt is highly persistent with a signal to noise ratio of about 16%. Similarly, the estimated σmη, which is the standard deviation of the error term ηmt in the state equation, is also highly significant. We may therefore conclude that there is strong evidence of herding toward the market portfolio for the BIST-100. However, a low value of “signal to noise ratio” implies that the process of herding has been smooth. The results of Hmt in Model 2, which also takes into account the market volatility of returns and the levels of returns in order to test for the robustness of the model, clearly indicate the presence of herding toward the market portfolio for the Turkish stock market. The signal to noise ratio drops to 8%, however,

397

Sentiment and Beta Herding in the Borsa Istanbul (BIST)

indicating that both market return and volatility are responsible for some of the herding patterns. Nevertheless, the overall results of Models 1 and 2 are fairly comparable (see Tables 2 and 3). The significance of σmη and φm in Model 2 shows that the evidence of herding is robust even after the inclusion of the two market fundamentals into the model. This enforces the hypothesis that it is the investors’ sentiment rather than market fundamentals that steers herd behavior (Hwang & Salmon, 2004). The herding measure hmt in the Turkish market has also been plotted with the BIST-100 index as well as with return volatility over the sample period (see Fig. 1). Furthermore, BIST monthly foreign shares have also been plotted as percentages in the same figure for the time period for which data was available (from May 2004 onward). It shows that trends in foreign participation may also help in explaining some of the patterns in herding behavior. Trends and fluctuations in herding behavior reveal interesting points when linked with market trends and some significant events (see Fig. 1). (a)

(b) 70,000

0.4

60,000 0.2

50,000 40,000

0.0 30,000 20,000

–0.2

10,000 –0.4

0 2000

2002

2004

2006

2008

2010

2012

(c)

2000

2002

2004

2006

2008

2010

2012

2000

2002

2004

2006

2008

2010

2012

(d)

0.0014

0.75

0.0012

0.70

0.0010

0.65

0.0008 0.60 0.0006 0.55

0.0004

0.50

0.0002

0.45

0.0000 2000

2002

2004

2006

2008

2010

2012

Fig. 1. Evolution of Herding with Market Volatility, BIST-100 Index and Foreign ˙ ITY ˙ Investment in BIST. (a) ht: The Herd Indicator. (b) VOLATIL (Std. Deviation of BIST-100 Returns). (c) BIST-100 INDEX. (d) Shares of Foreign Investment in BIST-100.

398

NAZMI DEMIR ET AL.

Firstly, we found evidence of significant herding during the domestic financial crisis when the market followed a downward trend from 2000 to 2001. Unlike Hwang and Salmon (2004), we found either adverse herding or no herding during the non-crisis years, when the market followed a bullish trend. Secondly, the overall range of hmt is found to be between −0.3 and +0.3, indicating that herding and adverse herding were never very extreme in the Turkish market during the sample period. This conclusion was also substantiated by the low signal to noise ratio in Models 1 and 2 (see above).4 A careful inspection of Fig. 1 shows that there seem to be three distinct phases of herd behavior in the market over the time of the sample (from 2000 to the end of 2011). The first phase, from the beginning of 2000 to the beginning of 2003, relates to one of the worst periods of financial crisis in the country. The herding at this time, a period which also included a declining market, appears to be intense (ht > 0). It is also noticeable that the herding line makes a sharp decline toward the ht = 0 line by the third quarter of 2001. This is when it became evident to investors that the AK Party (the single majority ruling party from 2002 to present) would be in power with a strong majority, a party that had pledged to pursue austerity programs began by previous government, and backed by the IMF and the EU. This downtrend of herd behavior continues until May 2003. Thereafter, it enters the second phase, when the market was transparent and easy to predict, with wide-spread optimism (Hwang & Salmon, 2004). Investors were confident that the government was sincere in its implementation of structural reforms. This calm trend in the second phase with no herding (with ht = 0), however, comes to an end by the beginning of 2005. Thereafter, the market seems to enter a third phase, with volatile patterns of adverse herding behavior (ht < 0), a stage which continues in the BIST until the end of 2011, possibly suggesting that internal and external events caused financial investors to be in a state of confusion and to lose confidence in the market.5 When it becomes difficult to predict the market, as Hwang and Salmon (2004) mention, investors rely more on the fundamental values of firms and less on sentiment. This in turn causes asset prices to adjust from mispricing back to the long-term equilibrium.

CONCLUSIONS In this study, the presence of herding behavior in the Borsa Istanbul is investigated, employing monthly data between 2000 and 2011. The State-Space

Sentiment and Beta Herding in the Borsa Istanbul (BIST)

399

Model, as proposed by Hwang and Salmon (2004) is estimated and the Kalman filter is used to make herding observable. It has been argued that the measure of herding under this approach is preferable than other popular measures of herding, such as Christie and Huang (1995) and Chang et al. (2000). It is found that herding toward the market in the BIST-100 is both significant and persistent independently from market fundamentals such as the volatility of returns and the levels of market returns. It is concluded that it is the investors’ sentiments rather than market fundamentals that cause herding in the market. The path of herding during the sample period also reveals several interesting trends. First, significant evidence of herding has been found during the domestic financial crisis of 2000
2001. Second, herding in the Turkish market has never been extreme, as it has fluctuated between −0.3 and +0.3, with a low signal to noise ratio. Third, evidence of prolonged adverse herding has been found during the 2005
2011 period, with both domestic and international events lowering the predictability of the market by creating a state of confusion and lack of confidence. This situation may have caused investors to rely more on fundamentals about the firms and less on sentiments in decision-making via adverse herding, which in turn leads asset prices to revert back to the long-term equilibrium. Therefore, the evidence of prolonged adverse herding may imply efforts by investors of reversing to long-term equilibrium risk-return relationship in BIST during 2005
2011. The findings of this study are important for the emerging markets because herding leads to mispricing of assets and hence brings about inefficiencies in the market. As for the future work, given the significant increase in the share of foreign investments in emerging markets, it would be valuable to differentiate firms with and without a high share of foreign institutional investments in order to test the existence of sentimental herding in future work.

NOTES 1. Higher order autoregressive processes were not statistically significant. 2. The data is not available at the web site www.borsaistanbul.com in a ready format, and needs to be requested. 3. There is, however, a significant drop in the value of Jarque
Bera statistics from 373.56 to 10.36. 4. The question of whether the path of hmt has been statistically significant at each of the dates of the sample has also been investigated based on confidence

400

NAZMI DEMIR ET AL.

intervals. Herding was found to be statistically significant during April and August of 2001, while adverse herding was statistically significant during February of 2010. 5. The two events, both of which happened in the first quarter of 2008, were: (1) constitutional court action against the government; and (2) the impact of the global financial crisis.

REFERENCES Banerjee, A. (1992). A simple model of herd behavior. Quarterly Journal of Economics, 107(3), 797
817. Bikhchandani, S., & Sharma, S. (2001). Herd behaviour and financial markets. IMF Staff Papers, 47(3), 279
310. Chang, E. C., Cheng, J. W., & Khorana, A. (2000). An examination of herd behavior in equity markets: An international perspective. Journal of Banking and Finance, 24(10), 1651
1679. Christie, W. G., & Huang, R. D. (1995). Following the pied piper: Do individual returns herd around the Market. Financial Analysts Journal, 51(4), 31
37. ˙ Coban, A. T. (2009). IMKB de su¨ru¨ davranıs¸ının test edilmesi, yu¨ksek lisans tezi. [A test of herd ˙ behavior in Istanbul Stock Exchange.] Adana: C¸ukurova Universitesi Sosyal Bilimler Enstitu¨su¨. ˙ Dog˘ ukanlı, H., & Ergu¨n, B. (2011). IMKB’de Su¨ru¨ Davranıs¸ı: Yatay Kesit De˘gis¸kenlik ˙ ˙ Temelinde Bir Aras¸tırma C¸ukurova U¨niversitesi I˙ IBF Is¸letme Bo¨lu¨mu¨1. [Herd behavior ˙ in Istanbul Stock Exchange: A research based on cross-section variation.] Osmaniye ˙ Korkut Ata U¨niversitesi I˙IBF I˙ ¸sletme Bo¨lu¨mu¨. EC Report. (2009). Growth and economic crises in Turkey: leaving behind a turbulent past? Economic Papers 386. Economic and Financial Affairs, European Commission. Frey, S. Herbst, P., & Walter, A. (2007). Measuring mutual fund herding: A structural approach. Mimeo. Gompers, P., & Metrick, A. (2001). Institutional investors and equity prices. Quarterly Journal of Economics, 116, 229
260. Hwang, S., & Salmon, M. (2004). Market stress and herding. Journal of Empirical Finance, 11(4), 585
616. Kapusuzoglu, A. (2011). Herding in the Istanbul Stock Exchange (ISE): A case of behavioral finance. African Journal of Business Management, 5(27), 11210
11218. ˙ Kayalıdere, K. (2012). Hisse Senedi Piyasasında Su¨ru¨ Davranıs¸ı: IMKB’de Ampirik Bir ˙ ˙ Inceleme Is¸letme Aras¸tırmaları Dergisi. [Herd behavior in stock market: An empirical investigation in the ISE]. Kim, K., & Sias, J. (2005). Institutional herding, business groups and economic regimes: Evidence from Japan. Journal of Business, 78, 213
242. Lakonishok, J., Shleifer, A., & Vishny, R. (1992). The impact of institutional and individual trading on stock prices. Journal of Financial Economics, 32, 23
43. Oehler, A., & Chao, G. (2000). Institutional herding in bond markets. Working Paper. Banberg University. Puckett, A., & Yan, X. (2007). The determinants and impact of short-term institutional herding. Mimeo. Retrieved from http://rssrn.com/abstract=972254 Voronkova, S., & Bohl, M. (2005). Institutional traders behavior in an emerging stock market: Empirical evidence on polish pension investors. Journal of Business Finance and Accounting, 32, 1537
1650. Retrieved from www.borsa Istanbul.com

TESTING FOR RATIONAL SPECULATIVE BUBBLES IN THE BRAZILIAN RESIDENTIAL REAL-ESTATE MARKET Marcelo M. de Oliveira and Alexandre C. L. Almeida ABSTRACT Speculative bubbles have been occurring periodically in local or global real-estate markets and are considered a potential cause of economic crises. In this context, the detection of explosive behaviors in the financial market and the implementation of early warning diagnosis tests are of critical importance. The recent increase in Brazilian housing prices has risen concerns that the Brazilian economy may have a speculative housing bubble. In the present chapter, we employ a recently proposed recursive unit root test in order to identify possible speculative bubbles in data from the Brazilian residential real-estate market. The empirical results show evidence for speculative price bubbles both in Rio de Janeiro and Sa˜o Paulo, the two main Brazilian cities. Keywords: Global financial crisis; value at risk; GARCH models; extreme value theory; back-testing JEL classifications: C15; C22; G01; R31

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 401
416 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096017

401

402

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

INTRODUCTION The efficient market hypothesis (EMH) assumes that all information is reflected instantaneously in asset prices (Fama, 1970). Hence, market prices should always be consistent with the “fundamentals.” On the other hand, sudden dramatic price changes over short periods of time, in which prices diverge from fundamentals, have been observed in several markets around the world (Roehner, 2002). Famous examples include the Dutch “Tulipmania,” the 1929 Crash in the United States and the recent American subprime crises (Garber, 1990; Kindleberger, 2000). Such behavior has raised many questions regarding market efficiency and has stimulated research on rational speculative bubbles (Cajueiro & Tabak, 2006; Evans, 1991). Rational speculative bubbles occur when there is an excessive public expectation of future price increasing, which produce rapid increases in valuations of an asset (Abreu & Brunnermeier, 2003). For example, during a real-estate bubble, investors (homebuyers) stay in the market despite the deviations of prices from fundamentals because they expect significant further price increases. Additionally, when the bubble is emerging, people think that real-estate prices are very unlikely to fall, letting to a little perceived risk associated with investing in the real-estate market. But eventually the prices may reach unsustainable levels, and crash: the bubble bursts (Abreu & Brunnermeier, 2003; Sornette, 2003a, 2003b). Bubbles, rational or not, have been occurring periodically in local or global real-estate markets and are considered of critical importance and a fundamental cause of financial crises and ensuing economic crises (Mayer, 2011; Richmond, 2007; Roehner, 2002, 2006). Hence, the study of realestate bubbles is an important contribution to the literature on future economic development. Moreover, a bubble is hazardous for financial and macro stability, since it can amplify a credit boom by inflating collateral values and causing misallocation of economic resources. In order to minimize the negative macroeconomic effects of a bubble, one needs to detect it early as soon as prices begin to rise. The usual definition of a bubble is a deviation of the asset price from its fundamental value. However, it can be very difficult to evaluate what is the “fundamental” price. Alternatively, constantly diminishing dividend-price ratio can serve as a reference for identifying a rational bubble in a stock market. If price expectations are rising, but higher dividends fail to materialize, the price rise is probably not based on fundamentals. A large number of studies have been proposed to empirically detect asset bubbles, mostly

Testing for Rational Speculative Bubbles in Real-Estate Market

403

focusing in stock markets. Some of these methods have been applied to the study of real-estate bubbles. Econometrics tests, like unit root tests or cointegration tests provide direct tests for the no-bubble hypothesis. If property price to income (or rent) ratio is stationary or the property price is cointegrated with the fundamental price, the no-bubble hypothesis cannot be rejected. The first econometric tests for rational bubble detection were volatility tests such as Schiller’s (Shiller, 1981) variance-bound and West’s two-step test (Newey & West, 1987). The underlying idea is to compare the volatility between the assumed fundamental asset price and the actual asset price. An asset bubble is detected indirectly if the volatility of the actual asset price is significantly larger than that of the assumed fundamental asset price. However, Marsh and Merton (Marsh & Merton, 1986) pointed evidence that variance bounds test fails when dividends and stock prices are nonstationary. Campbell and Shiller (Campbell & Shiller, 1987) and Diba and Grossman (Diba & Grossman, 1988) introduced the most commonly used methods for detecting asset price bubbles in the literature, namely the righttailed unit root test and the cointegration test. The aim of a cointegration test is to examine whether or not variables are trending together. If two series are cointegrated, that means the movements of these two variables are highly correlated. In other words, there is an equilibrium relationship between these two series. Related tests using data sets that combine time series and cross-sections (Panel-based tests) have been applied recently in order to investigate the existence of real-estate bubbles in Taiwan (Tsai & Peng, 2011), China (Hui & Chen, 2006), as well as the United States and Europe (Taipalus, 2006). These methods, however, suffer from a serious limitation, first pointed-out by Evans (Evans, 1991), who showed that cointegration tests of asset prices and dividends are not able to detect explosive bubbles when the sample data includes periodically collapsing bubbles. A novel approach to identifying and dating bubbles in real time has recently been introduced by Phillips and Yu (2011). Considering that the explosive property of bubbles is very different from random walk behavior, they developed a recursive econometric methodology interpreting mildly explosive unit roots as a hint for bubbles. The method provides a supremum Dickey
Fuller (DF) test (Dickey & Fuller, 1979) that overcomes the problem identified with unit root and cointegration tests. The supremum DF test improves power significantly with respect to the conventional unit root and cointegration tests, and has the advantage of allowing estimation of the origination date and final date of a bubble. The idea is to detect

404

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

speculative bubbles as they emerge, not just after their burst. This methodology was generalized in Phillips, Shi, and Yu (2012) for detecting multiple bubbles. Yiu, Yu, and Jin (2013) applied the latter for detecting bubbles in the Hong Kong residential property market, Chen and Funke (2013) for the Chinese housing market and Gomez-Gonza´lez, Ojeda-Joya, Guerra, and Sicard (2013) for the Colombian housing market. A related approach was employed by Kivedal (2013) for detecting rational bubbles in the US housing market. It is noteworthy to mention a few tests for bubbles not based on econometric approaches. Sornette and Johansen (1997) intend to predict the end of the bubbles assuming that a crash follows after rapid growth of economic indicators faster than an exponential function. Zhou and Sornette (2006, 2008) examined whether the price bubble burst in the United States, and predicted that the turning point of the bubble would occur around mid-2006. Watanabe et al. (Hui, Zheng, & Wang, 2010; Watanabe, Takayasu, & Takayasu, 2007) propose to identify bubbles and crashes by exponential behaviors detected in the systematic data analysis. More recently, Ohnishi et al. (2012) observed that the land price distribution in Tokyo had a power law tail during the bubble period in the late 1980s, while it was very close to a lognormal before and after the bubble period. This led them to argue that a characteristic of real-estate bubbles is not the rapid price hike itself but a rise in cross-sectional dispersion of prices. Over the past few decades, the real-estate market has been a target of government fiscal and monetary policies aimed at achieving balanced economic growth, low unemployment and low inflation. When the housing market or the overall economy is on a downturn, the governments tend to encourage banks to adopt a lenient mortgage policy, increasing the available credit as well to implement policies favorable to construction companies. Another usual government policy consists on subsiding disadvantaged groups through preferred loan rates. As a consequence, the macroeconomic market is filled with speculative capital demand and supply, resulting in inefficiency in the operation of various markets. This situation creates the conditions to the onset of a bubble (Tsai & Peng, 2011). Such situation is exemplified by the Brazilian real-estate policies. At the end of 2007, the world was affected by the subprime mortgage crisis in the United States. In response, Brazilian government established various market stimulus policies, which includes: (i) Approval of a new law of fiduciary alienation that minimizes the risk for the Brazilian banks in case buyers default on their loans. This makes the banks more willing to lend to

Testing for Rational Speculative Bubbles in Real-Estate Market

405

potentially riskier buyers. (ii) Brazil Growth Acceleration Program (referred to as PAC1) which included loans provided by government-owned banks, as well long-term credit for infrastructure, sanitation improvements, and more under a new investment fund that began with 657.4 billion Brazilian Reals (BRL)2 in the period 2007
2010 and more 955.1 billion BRL for the period 2011
2014. (iii) Programs administered by government-owned banks with the aim of development of 1,000,000 houses and apartments with subside for poor families, resulting in a credit of 96 billion BRL in 2011, 18 times the amount of credit available in 2003. In fact, real-estate prices in Brazil have raised rapidly in the last few years, and some analysts have been arguing the possibility that a bubble has been inflated and could potentially burst (D’Agostini, 2010; Gamble, 2011; Gaulard, 2012; Marshall, 2011; Mendonc¸a & Sachsida, 2012). In contrast, others argue that the pricing growth is sustainable and based on fundamentals, since Brazil was one of the fastest-growing major economies in the world in recent years with an average annual Gross Domestic Product (GDP) growth rate of over 5 percent, which made Brazilian economy the world’s seventh largest by nominal GDP by the end of 2012. As property is a sizable component of household and corporate balance sheets, a sudden collapse in property prices may have negative spillover effects on the overall macroeconomic situation and may pose macroeconomic and financial stability risks. Therefore, the aim of this chapter is to investigate the bubble-like behavior of the recent real-estate market prices in Brazil. Identifying speculative bubbles is not an easy task even in mature markets with long time series. Since in Brazil the time series for house prices are short, the recent developed test by Phillips et al. (2012), aimed at identifying explosive bubbles in real time, provides an adequate tool for such analysis (Homm & Breitung, 2012).

METHODOLOGY Identifying speculative bubbles is a hard task even in data sets with long time series. Recently, based on previous works of Shiller (2000), Mikhed and Zemcik (2009) regarded a house as an investment asset and used a standard present-value formula to derive implications for the relationship between house prices and the cash flow associated with owning a property (rent).

406

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

The fundamental price is derived from the standard no arbitrage condition: Pt =

Et ½ Rt þ 1 þ Pt þ 1  1þr

ð1Þ

where Pt is the property price index at period t, Et[.] denotes the mathematical expectation conditional on information at time t, Rt is the rent, and r is a constant risk-free discount rate. Since the formula above holds for all t, the property price index for the time t + 1 is given by: Pt þ 1 =

Et ½ Rt þ 2 þ Pt þ 2  1þr

ð2Þ

Therefore, Eq. (1), by repeated forward iteration, can be written as:

Pt = Et

Rt þ 1 Rt þ 2 Rt þ k Pt þ k þ þ ::: þ þ 2 k 1 þ r ð1 þ rÞ ð1 þ rÞ ð1 þ rÞk

 ð3Þ

Solving Eq. (3) yields the fundamental price: Pft =

∞ X


 1 j Et Rt þ j j = 1 ð1 þ rÞ

ð4Þ

also called price reflecting fundamentals. This equation means that the fundamental price contains all expected future rents. In absence of a bubble, one has the no-bubble condition (Mikhed & Zemcik, 2009): Et ½ Pt þ k  =0 k → ∞ ð1 þ rÞk lim

ð5Þ

This yields that the unique solution of Eq. (1) is Pt = Pft under the hypothesis of nonexistence of a bubble. The spread St between the house price and rent can be defined (Campbell, Lo, & MacKinlay, 1997; Mikhed & Zemcik, 2009) as: S t ≡ Pt −

1 Rt r

ð6Þ

Testing for Rational Speculative Bubbles in Real-Estate Market

407

which, based in the no-bubbles condition (Eq. (5)), can be rewritten as: " # ∞ X 1 ΔRt þ j þ 1 1 S t = Et ð7Þ = Et ½ΔPt þ 1  j r r j = 1 ð1 þ rÞ where ΔRt + j + 1 = Rt + j + 1 − Rt + j and ΔPt + 1 = Pt + 1 − Pt. Note that the stationarity of St implies the series {Pt/Rt} is stationary in the absence of a speculative bubble, since Pt/Rt = 1/r if St = 0. In order to estimate the fundamental value of property prices, we will employ the price-rent ratio, PR. The PR ratio is defined as the average cost of ownership divided by the estimated rent that would be paid if renting: PR =

House price Annual rent

ð8Þ

The PR ratio follows the concept of stocks’ price-earnings ratio (PER), which is defined as the ratio of the price of a share to the annual earnings of a company in the current year. The PER ratio contains information about if a given stock is over (or under) valuated (Campbell et al., 1997; Shiller, 2000). Analogously, rents, as well as corporate and personal incomes, are usually connected very close to supply and demand fundamentals. This is the cause one rarely sees an unsustainable “rent bubble” or an unsustainable “income bubble.” Therefore, a rapid increase of housing prices combined with a flat or slow-increasing renting market can be a signal of the beginning of a bubble (Gallin, 2008). As mentioned previously, in the absence of bubble the series PRt is stationary. The way to ascertain whether or not bubbles exist is testing the stationarity of the house price-to-rent ratio. This leads us to seek methods to determine whether series are stationary or not. Unit root tests are commonly used to determine whether a time series is stationary by using an autoregressive model. In its simplest form, the DF test (Dickey & Fuller, 1979) estimates the following first order autoregressive AR(1) regression equation: Δpt = α þ ðβÞpt − 1 þ ɛt ; ɛ t ∼ iidð0;σ 2 Þ

ð9Þ

where pt is the real price of the asset, α is the drift and β is the coefficient of the model. The error term ɛt is an uncorrelated white noise process. The DF test compares the t-statistics of residuals with DF critical values. The null hypothesis of the test is H0:β = 0 which represents unit root versus

408

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

the left-tailed alternative hypothesis H1:β < 0 stable root. (If the residuals in first order autoregressive model are still correlated the test can be augmented by ΔPt − i for higher level autoregressive processes.) In contrast, the supremum augmented DF (SADF) test proposed by Phillips and Yu (2011) is a right-sided test. The basic fundamental of this test is using recursive regression techniques to test the unit root. In the context of DF test, the test is based on the following regression: Δpt = α þ ðβ − 1Þpt − 1 þ ɛt ; ɛ t ∼ iidð0;σ 2 Þ

ð10Þ

The null hypothesis still is unit root behavior, H0:β = 1, but the alternative hypothesis is explosive behavior, H1:β < 1. The right-tailed ADF statistics is computed in multiple recursive regressions for each sub-sample which start with the initial observation but the last point varies. Let r1 and r2 be, respectively, the fractional starting and ending point of each sample. The sample window rw = r2 − r1 therefore varies from the initial size window r0 to the total sample. The SADF statistic is based on the supremum value of the ADF statistics obtained, SADFðr0 Þ =

sup ADFr02 |{z}

ð11Þ

r0 ≤ r2 ≤ 1

The explosiveness of the process is tested by comparing with the righttailed critical values of its limit distribution which is given by: R r2 WdW sup ADFr02 → sup R r02 2 ð12Þ |{z} |{z} 0 W dW r0 ≤ r2 ≤ 1

r0 ≤ r2 ≤ 1

where W is a standard Wiener process and → denotes convergence in distribution. Instead of fixing the starting point of the sample, the generalized SADF (GSADF) test extends the sample sequence by changing both the starting and the ending point of the sample, implementing the right-tailed unit root test repeatedly on a forward expanding sample sequence: GSADFðr0 Þ =

sup |{z} 0 ≤ r1 ≤ r2 − r0

ADFrr21 r0 ≤ r2 ≤ 1

ð13Þ

Testing for Rational Speculative Bubbles in Real-Estate Market

409

The GSADF statistics can be defined as the largest ADF statistic over the feasible ranges of r1 and r2. It is then used to detect the presence of at least one bubble in the whole sample. Phillips et al. (2012) demonstrate that the moving sample GSADF diagnostic outperforms the SADF test in detecting explosive behavior in multiple bubble episodes, and works well even in modest sample sizes. In order to estimate the beginning and collapse dates of every bubble, Phillips et al. (2012) suggest using backward expanding sample sequences. Let the fractional ending point fixed at r2 with the starting point r1 moving in the range 0 ≤ r1 ≤ r2 − r0. These ADF statistic sequences are denoted by: n o BADFrr21 ð14Þ 0 ≤ r1 ≤ r2 − r0

Hence the backward SADF statistic is defined as the sup value of the ADF statistic sequence: n o BSADFr2 ðr0 Þ = sup BADFrr21 ð15Þ |{z} 0 ≤ r1 ≤ r2 − r0

The beginning (end) date of a bubble corresponds to the first date whose the BSADF statistic becomes greater (smaller) than the critical values, estimated by Monte Carlo simulation. (These critical values are calculated from respective empirical distributions of each statistic, generated under the null hypothesis (Eq. (10) with β = 1).)

BRAZILIAN RESIDENTIAL REAL-ESTATE MARKET Prior to the econometric analysis, let us briefly describe the data set used. Until recently, Brazil had no reliable indicators following the price behavior of residential properties. For this purpose, since 2008, Fipe (Brazilian Institute of Economic Research) developed an index, named Fipe-Zap, which use real-estate ads as a source of information. The major drawback with this source is a possible gap between the advertised price and the realized price. Nonetheless, if one assumes that both prices have a similar trend, at least in the medium and long run, an index for the advertised prices could be regarded as reliable real-estate market information.3 Fig. 1a shows the price index for the seven biggest Brazilian cities (for each index it

410

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

(a)

(b)

200

24 São Paulo Rio de Janeiro Belo Horizonte Recife Fortaleza Brasília Salvador

São Paulo

20 PR ratio

150

Rio de Janeiro

22

18 16

100 14 12

Fig. 1.

Nov/13

Jan/13

Jan/12

Jan/11

Jan/10

Jan/09

Jan/08

60 2008

2009

2010

2011

2012

2013

2014

(a) Fipe-Zap Index for the Seven Biggest Brazilian Cities. (b) Price-Rent Ratio for Sa˜o Paulo and Rio de Janeiro.

was set an arbitrary value of 100 on August 2010). Almost all the indices show high increases during the period, revealing that Brazilian properties prices rose rapidly in the last few years. Unfortunately, only the data for the two major cities, Sa˜o Paulo (SP) and Rio de Janeiro (RJ), is available since December 2007. The sample data is also larger for these both cities, therefore for a more precise statistical analysis, from now on we will concentrate our analysis on the SP and RJ real-estate market. Since Fipe also calculates the monthly rental yield from its rent and sale ads database, it is straightforward to obtain the PR ratio as the reciprocal of the annually rental yield. The resulting time series with the PR ratio for SP and RJ is shown in Fig. 1b. In SP, we observe that the PR ratio grows from around 11 to 18, and in RJ from around 15 to 22. This is equivalent to an increase of approximately 64% and 47%, respectively. It should be noted that a rising PR ratio is only a necessary but not a sufficient condition for speculative misalignment from fundamentals.

EMPIRICAL RESULTS In order to test whether the movement of house prices in the Brazilian realestate market reflects deviation from levels supported by fundamentals we

411

Testing for Rational Speculative Bubbles in Real-Estate Market

performed more detailed analysis of the data shown in Fig. 1b. Thus, we applied the recursive SADF and GSADF tests to the PR data set. In our analysis, we choose the minimal window size rw = 12 which ensures that there are enough observations for initial estimation. The finite sample critical values were obtained via Monte Carlo simulations with 10,000 iterations. The resulting SADF and GSADF statistics for the Rio and Sa˜o Paulo PR ratio are shown in Table 1. The SADF statistics indicate the existence of at least one speculative real-estate bubble in Rio de Janeiro during the period with a 95% confidence interval and in Sa˜o Paulo with a 99% confidence interval. Similar conclusions can be drawn from the GSADF test statistics, with bubble in Rio de Janeiro (at 90% level) and Sa˜o Paulo (at 99% level). Figs. 2a and 2b show the results of the recursive SADF and GSADF tests for Rio de Janeiro, respectively. The associated critical values are shown as doted (95% level) and dashed (90% level) curves. The SADF test shows a statistically significant explosive period from mid-2010 to the end of 2012. The GSADF test identifies a bubble beginning in mid-2010 from mid-2012. Figs. 3a and 3b show the results of the recursive SADF and GSADF tests for Sa˜o Paulo, respectively. As before, doted (95% level) and dashed (90% level) curves represent the associated critical values. In this case, the SADF test shows a unique and long explosive behavior starting in mid2009. On other hand the GSADF test is successful in identifying multiple bubble periods, with the last one beginning in mid-2011. The overall picture of Brazilian house price valuation provided by Figs. 2 and 3 corroborate the existence of rational speculative bubbles in its two major cities. It is also noticeable that this confirms the preliminary results from glancing at Fig. 1. On the other hand, it is needed to emphasize that price-to-rent indices have obvious disadvantages and shortcomings. Although the indices provide information about the dynamics of the Table 1. PR ratio Rio de Janeiro Sa˜o Paulo Rio de Janeiro Sa˜o Paulo

SADF and GASDF Statistics for PR Ratio Series and Respective 90%, 95%, and 99% Critical Values. SADF t-statistics 1.3203 2.6529 GSADF t-statistics 1.9180 4.1353

90 c.v. 1.0106 1.0106 90 c.v. 1.8312 1.8312

95 c.v. 1.3144 1.3144 95 c.v. 2.1804 2.1804

99 c.v. 1.9812 1.9812 99 c.v. 2.9606 2.9606

412

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

(a)

(b)

1.5

2

1

1.5

0.5

1

0

0.5

−0.5

0

−1

−0.5

−1.5 −2 2009

−1

Backward ADF sequence 90% critical value sequence 95% critical value sequence 2010

2011

2012

2013

Backward SADF sequence 90% critical value sequence 95% critical value sequence

−1.5

2014

2009

2010

2011

2012

2013

2014

Fig. 2. (a) Recursive Calculation of the SADF Test and (b) Recursive Calculation of the GSADF for Rio de Janeiro. The Doted Lines Represent the 95% and the Dashed 90% Critical Value Sequences. (a)

(b)

3

5 4

2 3 1

2 1

0

0 −1

2009

−1

Backward ADF sequence 90% critical value sequence 95% critical value sequence

−2

2010

2011

2012

2013

Backward SADF sequence 90% critical value sequence 95% critical value sequence

−2 2014

2009

2010

2011

2012

2013

2014

Fig. 3. (a) Recursive Calculation of the SADF test and (b) Recursive Calculation of the GSADF for Sa˜o Paulo. The Doted Lines Represent the 95% and the Dashed 90% Critical Value Sequences.

price-to-rent ratio over time, it cannot provide information about the actual level of the price-to-rent ratio. Analysis of other indicators such as a household income index and a land price index (Pelez, 2012), that would be useful to check the robustness of our results, were not performed due to the lack of monthly data on these indicators.

413

Testing for Rational Speculative Bubbles in Real-Estate Market

More insight from the explosive behavior can be achieved from Fig. 4 that show the price indices segmented by the number of apartments bedrooms. Recently, based on works of Landvoigt, Piazzesi, and Schneider (2011) and Hott (2011), Escobari, Damianov, and Bello (2013) exploit the idea that low-tier home prices increase at a faster pace during the boom than the high-tier home prices if cheap credit is available to consumers predominantly at the low end of the distribution of houses (Gupta & Miller, 2012), as is the case in Brazil. From Fig. 4a, we see that until 2010 the segmented price indices grew roughly at the same pace. From 2010 to 2012, the low prices (1-bedroom) apartments index increases at a faster rate than the higher price ones. Fig. 4b reveals that in Sa˜o Paulo the low price apartments price increases at a considerable faster rate than high price ones, during almost all the period under consideration. These results highlight two findings: (i) In Sa˜o Paulo, the credit for the cheap apartments’ consumers is an important element leading to the explosive behavior of real-estate prices. On the other hand, in Rio de Janeiro, other factors may be related, as an overvaluate anticipation due to the upcoming 2014 FIFA World Cup and 2016 Olympics. (ii) In agreement with our previous findings of the PR ratio, the explosive behavior occurred in a smaller period of time in Rio than in Sa˜o Paulo, which suggests that prices in Rio de Janeiro have already reached its peak. It is important to notice that this approach does not require information on market fundamentals.

(a)

(b)

400

350 1 bedroom 2 bedroom 3 bedroom 4+bedroom

350 300

1 bedroom 2 bedroom 3 bedroom 4+bedroom

300

250 250 200 200

2008

150

Rio de Janeiro

150

2009

Fig. 4.

2010

2011

2012

2013

São Paulo

2014

2008

2009

2010

2011

2012

2013

2014

Segmented Price Indices for Rio de Janeiro (a) and for Sa˜o Paulo (b).

414

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

CONCLUSIONS In this chapter, we analyzed recent data from the Brazilian real-estate market by means of a recently proposed recursive unit root test, aimed at identifying explosive bubbles in real time. The test is able to identify growing bubbles and can have an important impact on the construction of early warning systems. House prices rose dramatically in Brazil in the last few years, and our results in fact reveal the existence of speculative bubbles in the residential real-estate market for the two main Brazilian cities, Sa˜o Paulo and Rio de Janeiro during the recent years.

NOTES 1. http://www.brasil.gov.br/pac/ 2. During the period 2007
2012, the value of US dollar (USD) to BRL rate was in the range 1.55
2.46, with a mean of about 2.00. Currently 1.00 BR ∼2.30 USD. 3. The index is based on a database of approximate 200 thousands real estate ads per month, and its methodology is detailed in http://www.fipe.org.br

REFERENCES Abreu, D., & Brunnermeier, M. (2003). Bubbles and crashes. Econometrica, 71(1), 173
204. Cajueiro, D. O., & Tabak, B. M. (2006). Testing for rational bubbles in banking indices. Physica A: Statistical Mechanics and its Applications, 366, 365
376. Campbell, J., Lo, A., & MacKinlay, A. (1997). The econometrics of financial markets. Princeton, NJ: Princeton University Press. Campbell, J., & Shiller, R. (1987). Co-integration and tests of present value models. Journal of Political Economy, 95, 1062. Chen, X., & Funke, M. (2013). Real-time warning signs of emerging and collapsing Chinese house price bubbles. National Institute Economic Review, 223, R39. D’Agostini, L. M. (2010). A bolha imobilia´ria no Brasil esta´ formada. Economia e Tecnologia, 23, 41. Diba, B., & Grossman, H. (1988). Explosive rational bubbles in stock prices. American Economic Review, 78, 520. Dickey, D., & Fuller, W. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 74, 427. Escobari, D., Damianov, D., & Bello, A. (2013). A time series test to identify housing bubbles. Journal of Economics and Finance, (March), 1–17. doi:10.1007/s12197-013-9251-5 Evans, G. W. (1991). Pitfalls in testing for explosive bubbles in asset prices. The American Economic Review, 81(4), 922
930. ISSN: 00028282. Retrieved from http://www.jstor. org/stable/2006651

Testing for Rational Speculative Bubbles in Real-Estate Market

415

Fama, E. F. (1970). Efficient capital markets: A Review of theory and empirical work. The Journal of Finance, 25, 383. Gallin, J. (2008). The long-run relationship between house prices and rents. Real Estate Economics, 36, 635. Gamble, W. (2011). Brazil: Warning signs of a credit bubble being ignored. Moneylife, February 28. Garber, P. (1990). Famous first bubbles. Journal of Economic Perspectives, 4, 35. Gaulard, M. (2012). The “hot money” phenomenon in Brazil. Brazilian Journal of Political Economy, 32(3), 367
388. Gomez-Gonza´lez, J. E., Ojeda-Joya, J., Guerra, C. R., & Sicard, N. (2013). Testing for bubbles in housing markets: New results using a new method. Borradores de Economı´a, 753, 1
11. Gupta, R., & Miller, S. M. (2012). The time-series properties of house prices: A case study of the Southern California market. Journal of Real Estate Finance and Economics, 44(3), 339
361. Homm, U., & Breitung, J. (2012). Testing for speculative bubbles in stock markets: A comparison of alternative methods. Journal of Financial Econometrics, 10(1), 198
231. Hott, C. (2011). Lending behavior and real estate prices. Journal of Banking and Finance, 35(9), 2429
2442. Hui, E. C. M., & Chen, Y. (2006). Housing price bubbles in Hong Kong, Beijing and Shanghai: A comparative study. Journal of Real Estate Finance and Economics, 33, 209. Hui, E. C. M., Zheng, X., & Wang, H. (2010). A dynamic mathematical test of international property securities bubbles and crashes. Physica A, 389(7), 1445
1454. Kindleberger, C. P. (2000). Manias, panics and crashes a history of financial crisis. New York, NY: Wiley. Kivedal, B. (2013). Testing for rational bubbles in the US housing market. Journal of Macroeconomics, 38, 369. Landvoigt, T., Piazzesi, M., & Schneider, M. (2011). The housing market(s) of San Diego. Discussion paper, NBER Working Paper 17723. Marsh, T. A., & Merton, R. C. (1986). Dividend variability and the variance bounds tests for the rationality of stock market prices. American Economic Review, 76, 483. Marshall, P. (2011). Brazil may be heading for a subprime crisis. Financial Times, February 21. Mayer, C. (2011). Housing bubbles: A survey. Annual Review of Economics, 3, 559. Mendonc¸a, M., & Sachsida, A. (2012). Existe Bolha no Mercado Imobilia´rio Brasileiro? Discussion paper, IPEA, TD 1762. Braslia. Mikhed, V., & Zemcik, P. (2009). Testing for bubbles in housing markets: A panel data approach. Journal of Real Estate Finance and Economics, 38(4), 366
386. ISSN: 08955638. doi:10.1007/s11146-007-9090-2 Newey, W., & West, K. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55, 703. Ohnishi, T., Mizuno, T., Shimizu, C., Watanabe, T. (2012). Power laws in real estate prices during bubble periods. International Journal of Modern Physics: Conference Series, 16, 61–81. Pelez, R. (2012). The housing bubble in real-time: The end of innocence. Journal of Economics and Finance, 36, 211
225. Phillips, P. C. B., Shi, S., & Yu, J. (2012). Testing for multiple bubbles. Cowles Foundation Discussion Paper 1843.

416

MARCELO M. DE OLIVEIRA AND ALEXANDRE C. L. ALMEIDA

Phillips, P. C. B., & Yu, J. (2011). Dating the timeline of financial bubbles during the subprime crisis. Quantitative Economics, 2, 455. Richmond, P. (2007). A roof over your head; house price peaks in the UK and Ireland. Physica A, 375, 281. Roehner, B. M. (2002). Patterns of speculation: A study in observational econophysics. Cambridge: Cambridge University Press. Roehner, B. M. (2006). Real estate price peaks: A comparative overview. Evolutionary and Institutional Economics Review, 2(2), 167
182. Shiller, R. J. (1981). Do stock prices move too much to be justified by subsequent movements in dividends? American Economic Review, 71, 421. Shiller, R. J. (2000). Irrational exuberance. Princeton, NJ: Princeton University Press. Sornette, D. (2003a). Critical market crashes. Physics Reports, 378(1), 1
98. ISSN: 0370-1573. doi:10.1016/S0370-1573(02)00634-8 Sornette, D. (2003b). Why stock markets crash, critical events in complex financial system. Princeton, NJ: Princeton University Press. Sornette, D., & Johansen, A. (1997). Large financial crashes. Physica A, 245, 411. Taipalus, K. (2006). A global house price bubble? Evaluation based on a new rent-price approach. Bank of Finland Research Discussion Papers 26. Tsai, I.-C., & Peng, C.-W. (2011). Bubbles in the Taiwan housing market: The determinants and effects. Habitat International, 35(2), 379
390. ISSN: 0197-3975. doi:10.1016/j. habitatint.2010.11.010 Watanabe, K., Takayasu, H., & Takayasu, M. (2007). A mathematical definition of the financial bubbles and crashes. Physica A, 383, 120. Yiu, M., Yu, J., & Jin, L. (2013). Detecting bubbles in Hong Kong residential property market. Journal of Asian Economics, 28, 115. Zhou, W.-X., & Sornette, D. (2006). Is there a real-estate bubble in the US? Physica A, 361, 297. Zhou, W.-X., & Sornette, D. (2008). Analysis of the real estate market in Las Vegas: Bubble, seasonal patterns, and prediction of the CSW indices. Physica A, 387(1), 243
260. ISSN: 0378-4371. doi:10.1016/j.physa.2007.08.059

CHALLENGES IN THE APPLICATION OF EXTREME VALUE THEORY IN EMERGING MARKETS: A CASE STUDY OF PAKISTAN Jamshed Y. Uppal and Syeda Rabab Mudakkar ABSTRACT Application of financial risk models in the emerging markets poses special challenges. A fundamental challenge is to accurately model the return distributions which are particularly fat tailed and skewed. Valueat-Risk (VaR) measures based on the Extreme Value Theory (EVT) have been suggested, but typically data histories are limited, making it hard to test and apply EVT. The chapter addresses issues in (i) modeling the VaR measure in the presence of structural breaks in an economy, (ii) the choice of stable innovation distribution with volatility clustering effects, (iii) modeling the tails of the empirical distribution, and (iv) fixing the cut-off point for isolating extreme observations. Pakistan offers an instructive case since its equity market exhibits high volatility and incidence of extreme returns. The recent Global Financial Crisis has

Risk Management Post Financial Crisis: A Period of Monetary Easing Contemporary Studies in Economic and Financial Analysis, Volume 96, 417
437 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3759/doi:10.1108/S1569-375920140000096018

417

418

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

been another source of extreme returns. The confluence of the two sources of volatility provides us with a rich data set to test the VaR/EVT model rigorously and examine practical challenges in its application in an emerging market. Keywords: Value-at-risk; GARCH models; extreme value theory; back-testing; global financial crisis

INTRODUCTION Application of financial risk models in the emerging markets poses special challenges for a number of reasons. First, the economic, financial, and regulatory environment in the emerging markets continues to evolve. The economies are characterized by frequent structural shifts and the asset return distributions may not remain stable over time. Usage of risk models, however, assumes that the parameters of probability distribution of returns extracted from the historical data are stable. Moreover, many emerging markets offer only limited data histories. Second, although most of the developing countries have embraced free market principals, there is a tendency to frequently interject government in the economic sphere. An extreme example of such interference is when an entire market can be shut down by a government decree, as was the case of the stock market closure in Pakistan in 2008. It introduces another element of uncertainty that the risk managers need to contend with. Third, there could be substantial and systematic differences in the properties of the return distributions of financial assets between the developed and the developing countries. More importantly, the properties of the tails of the distributions could be quite different for the two groups of countries which could render the risk models developed in the context of the advanced economies invalid for the developing world. Fourth, a number of emerging markets are also characterized by a lack of depth and liquidity, which is reflected in low trading volumes. Thus, the markets exhibit low normal volatility but experience frequent systemic shocks giving rise to high extreme volatility.1 These shocks may originate internally from unexpected shifts in the economic fundamentals, or externally which are then exacerbated due to low resilience and high vulnerability of many developing economies. Fifth, the financial markets of the developing countries may also be more prone to episodes of speculative bubbles, which affect the dynamic properties of the return processes. Finally, for all markets, the empirical return distributions of

Challenges in the Application of EVT in Emerging Markets

419

financial assets are found to be fat tailed and skewed in contrast to the normal distribution as assumed in the theoretical models. An extensive literature in finance, e.g., the Black Swan, Taleb (2010), underscores the importance of rare events in risk management. These rare events may materialize in the shape of a large positive or negative investment returns, a stock market crash, major defaults, or the collapse of risky asset prices. Hence, on account of the above mentioned factors, there seems to be a substantial model risk associated with applying the quantitative models of modern finance in the emerging markets. One such widely used model is the Value-at-Risk (VaR) which is defined as the maximum possible loss to the value of financial assets with a given probability over a certain time horizon. A major challenge faced in implementing the VaR approach arises from the fact that the empirical return distributions are fat tailed and skewed in contrast to the normal distribution as assumed in the theoretical models. In order to address the problems of heavy tails, VaR risk measures based on Extreme Value Theory (EVT) have been developed to model tailrisk. EVT allows us to model the tails of distributions, and to estimate the probabilities of the extreme movements that can be expected in financial markets. The basic idea behind EVT is that in applications where one is concerned about risk of extreme loss, it is more appropriate to separately model the tails of the return distribution. This chapter addresses the challenges in the application of the extreme loss risk estimates to the emerging markets by considering the case of the Karachi Stock Exchange (KSE), the main equity market of Pakistan. The country offers an instructive case study since due to its turbulent political and economic environment its equity market has experienced very high volatility and incidence of extreme returns, thereby, providing a richer data set. Yet the country has one of the oldest stock markets among the developing countries with well-established institutions and regulatory structure. The backdrop of the global financial crisis (GFC) of 2007
2009 also provides us with an historical experiment to examine the tails of stock return distributions. During the GFC period stock market volatility increased many folds and large swings in the stock prices were observed with an unprecedented frequency, thus, providing us with a rich data set for applying EVT. The study first considers the suitability of various theoretical return distributions among the family of stable distributions to model risk by examining the Goodness-of-Fit (GOF) for the left tail (representing losses) of the empirical distributions. Next we address the issue of optimally modeling the dynamics of innovation distribution in the presence of the structural

420

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

breaks. We then take up the issue of determining the appropriate cut-off point for isolating the extreme observations and implementing the VaR model based on the extreme theory. Various techniques of measuring market risk based on the VaR approach are evaluated for the tail of the conditional distribution of KSE index return series using back-testing procedure over the period January 2001
June 2012.

RELATED LITERATURE Up till now only a few recent studies have examined the impact of GFC on the stock market behavior. Among these, Uppal and Mangla (2013) compare the tail distributions of stock returns for the pre- and post-Global Financial Crises periods. Uppal (2013) tests the EVT-VaR model and reports that the model does not describe the tail-risk in the United States and the United Kingdom market well during the crisis period, though it performed better in case of emerging markets. There have been, however, a number of studies using EVT following previous stock markets crashes and periods of high volatility in the developed as well as the emerging markets. For example, Gencay and Selcuk (2004) employ VaR models using EVT to a sample of emerging markets after the Asian financial crisis of 1998. Onour (2010) presents estimation of extreme risk in three stock markets in the Gulf Cooperation Council (GCC) countries, Saudi Arabia, Kuwait, and United Arab Emirates, in addition to the S&P 500 stock index, using the Generalized Pareto Distribution (GPD). Djakovic, Andjelic, and Borocki (2011) investigates the performance of EVT with the daily stock index returns for four different emerging markets, the Serbian, Croatian, Slovenian, and Hungarian stock indices. Bhattacharyya and Ritolia (2008) suggest a Value-at-Risk (VaR) measure for the Indian stock market based on the EVT for determining margin requirements for traders. Iqbal, Azher, and Ijza (2010) compute the VaR by considering 17 years data of KSE using four different parametric methods and two non-parametric methods. Qayyum and Nawaz (2011) use EVT to compute VaR of return series for KSE 100 during 1993
2009. Nawaz and Afzal (2011) compute the VaR using Historical Simulation, Pro and Risk Metrics method. They conclude that for the Pakistan’s KSE, VaR system is more effective than SLAB system for determining margin requirements. Our study differs from the previous work in so far as it specifically evaluates the empirical return distribution for GOF to the theoretical

Challenges in the Application of EVT in Emerging Markets

421

distribution, incorporates structural shifts in dynamic modeling, and backtests the competitive models. Our study addresses, firstly, the issue whether EVT can help in measuring and managing tail-risk in the emerging markets. Secondly, it addresses the validity of the models in the presence of structural breaks because even if the EVT does adequately describe the extreme return distribution, its applicability would be much restricted if the parameters of the distribution were not stable. Finally it discusses the GOF for the innovation distribution of returns and proposes the optimal distribution for extreme loss risk estimates.

DATA AND METHODOLOGY Value-at-Risk (VaR) is a high quantile (typically the 95th or 99th percentile) of the distribution of returns and provides an upper bound on tails of returns distribution with a specified probability. However, classical VaR measures based on the assumption of normal distribution of the financial asset underestimate risk as the actual return distributions exhibit heavier tails. One alternative to deal with the non-normality of the financial asset distributions has been to employ historical simulation methodology which does not make any distributional assumptions, and the risk measures are calculated directly from the past observed returns. However, the historical approach sill assumes that the distribution of past returns will be stable in the future. Another approach is to use the probability distribution models which account for such thick tails as are empirically observed. The assumption of constant volatility is also contradicted by the well documented phenomenon of volatility clustering, that is, large changes in assets values are followed by large changes in either direction. Hence, a VaR calculated in a period of relative calm may seriously underestimate risk in a period of higher volatility.2 The time varying volatility was first modeled as an ARCH (q) process (Bollerslev, Chou, & Kroner, 1992) which relates time t volatility to past squared returns up to q lags. The ARCH (q) model was expanded to include dependencies up to p lags of the past volatility. The expanded models, GARCH (p, q) have become the standard methodology to incorporate dynamic volatility in financial time series (Poon & Granger, 2003). Similarly the auto-correlation of returns is significant in many situations and there is a need to incorporate the ARMA (m, n) structure in the model too.

422

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

The current study evaluates the performance of the extreme loss risk estimates for the equity market of Pakistan, the KSE. Various techniques of measuring market risk based on the VaR approach are evaluated for the tail of the conditional distribution of KSE index return series over the period January 2001
June 2012. The sample period consists of 2972 daily observations for over 11 years. During the period the market experienced a number of political and economic shocks, including the 9/11 terrorist attack and the GFC. The study period also includes a time interval, from August 2008 to December 2008, when the KSE was effectively shut down by placing a floor on stock prices in the aftermath of the GFC. The returns are measured as the first log differences of the stock price series. Converting the daily closing prices into the geometric returns yields the stationary series which is confirmed by the Dickey
Fuller test (Results of Dickey
Fuller tests are available from the authors).

Dynamics of Volatility and Conditional Mean Let (Xt, t ∈ Z) be a stationary time series representing the daily observations of a log-return of financial asset price. We assume that dynamics of X are given by: Xt = μt þ σ t Zt

ð1Þ

where μt and σt measure the mean return and volatility of the process respectively, Zt are the innovations which follows a strict white noise process with zero mean, unit variance and marginal distribution function FZ(z). We assume that μt and σt are measurable with respect to the information set Gt − 1 . Let FX(x) denote the marginal distribution of stationary time series (Xt) and let FðXt þ 1 jGt Þ ðxÞ denote the 1-step predictive distribution of the returns over the next day, given knowledge of returns up to and including day t. The mean returns and the volatility of the GARCH (1, 1) model with normal innovations has the following specification: μt = φ Xt − 1 and σ2t = w + α(Xt − 1
μt − 1)2 + β σ2t − 1 with w, α, β > 0, and α + β :

− ðx − μÞ σ

 exp −

− 1=ξ !

− ðx − μÞ σ

x≤μ x>μ

1=ξ !

1

x≤μ x>μ

for parameters μ > 0, σ and ξ > 0. This is an extreme value analogue to the conventional central limit theory. These three distributions can be encompassed by a single one
the Generalized Extreme Value (GEV) family of distributions. Namely,  − 1=ξ ! 

   ðx − μÞ ðx − μ Þ GðxÞ = exp − 1 þ ξ where x : 1 þ ξ >0 σ σ 

− ∞ < μ; ξ < ∞ and σ > 0 Building on Fisher and Tippet work, Pickands (1975), Balkema and de Haan (1974) stated the following theorem regarding the conditional excess distribution function. Theorem. For a large class of underlying distribution functions the conditional excess distribution function Fu(y), for a large value of μ, is well approximated by: Fμ ðyÞ ≈ Gβ;ξ ðyÞ; μ → ∞

− 1=ξ ; ξ≠0 Gβ;ξ ðyÞ = 1
1 þ ξy=β

426

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

= 1
е − y=β ; ξ = 0 for y ∈[0, xF
μ] if ξ > 0, and y∈[0,- β/ξ] if ξ < 0. y = (x
μ) and μ is the threshold; xF ≤ ∞ is the right endpoint of F. Gβ, ξ(y) is known as the GPD. Fμ(y) can also be reformulated in terms of F(x) describing the entire time series Xt to construct a tail estimator for the underlying distribution. In modeling financial time series the preferred approach, the peak-overthreshold (POT) approach, takes large values of the sample which exceed a certain threshold u. The distribution function of these exceedances is then obtained employing fat-tailed distributions models such as the GPD.

Choice of Threshold Selecting an appropriate threshold to identify the tail region is a critical problem with the POT methods. It involves a trade-off: a very high threshold level may provide too few points for estimation, while a low threshold level may render a poor approximation (Weismann, 1978). The idea is to pick as low a threshold as possible subject to the limit to provide a reasonable approximation. Following an approach suggested by Kluppelberg (2001), the mean excess function is used to determine the threshold, as follows. Consider the ordered observations x(1), x(2), …, x(k) as a subset of the independent identically distributed random variables x1, x2, … , xn which exceeds a particular threshold μ. Define the threshold excesses by yj = xj − μ for j = 1, 2, …, k. The method requires plotting the points, Nu 1 X μ; ðxðiÞ − μÞ Nu i = 1

! : μ < xmax

and choosing the threshold where the function becomes increasingly linear. However, several researchers, (e.g., McNeil, 1997, 1999) suggested employing a high enough percentile (like 90 or 95) as the suitable threshold.

Estimation of Quantiles The next step is to estimate the quantiles. Using as an estimator of F(u) the ratio (n
Nu)/n, where n is the total number of observations and Nu is the

427

Challenges in the Application of EVT in Emerging Markets

number of observations above the threshold, the tail estimator is defined as:

− 1=ξ F ðxÞ = 1
Nu =n 1 þ ξðx − μÞ=β for x > u. For a given probability, q > F(u), the VaR estimate (qth qunatile) is obtained by inverting the tail estimation formula above to get (see Embrechts, Kluppelberg, & Mikosch, 1997). zq = VaRq = μ þ β=ξ



n=Nu ð1
qÞ

−ξ


1



The estimation of the GPD parameters, ξ and β is made using the method of maximum likelihood. Alternatively, Hill’s tail index estimator is calculated; Hill (1975). The estimated dynamic or conditional VaR using Eq. (1) is: x^t q = μd tþ1 t þ 1 þ zbq σd

ð3Þ

Back-Testing We further back-test the method on historical series of negative log-returns {x1, x2, …, xn} from January 2001 to June 2012. We calculate x^tq on day t in the set T = {m, m + 1, …, n
1} using a time window of m days each time for three different cases. In the first case, AR(1)-GARCH(1,1) model is fitted without inclusion of dummy variable. In the second case a dummy variable is included that captures the onset of the GFC. The third case uses a reduced sample, that is, observations regarding the period when KSE is closed (i.e., from August 1, 2008 to January 15, 2009) are removed from the data set. For each case we consider 50 extreme observations from the upper tail of the innovation distribution, that is, we fix k = 50 each time. On each day t∈T, we fit a new AR(1)-GARCH(1,1) model and determine a new GPD tail estimate. We compare x^t q with xt + 1 for q∈{0.95, 0.975, 0.99, 0.995}. A violation is noted whenever xt þ 1 > x^tq . We then apply a one-sided binomial test based on the number of violations for the model adequacy. We also compare the method with four other well-known parametric methods of estimation: (i) the Static Normal method in which returns are

428

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

assumed to be normally distributed and the VaR is calculated as the qth upper quantile from the normal distribution, (ii) the Dynamic or Conditional Normal in which AR(1)-GARCH(1,1) model with normal innovations is fitted by the method of maximum likelihood to the return data and x^tq is estimated, (iii) the Conditional t, in which innovations are assumed to have a leptokurtic distribution and the AR(1)-GARCH(1,1) model with t-innovations is fitted to the return data, and (iv) the Static EVT method in which returns are assumed to have fat-tailed distribution and EVT is applied to the upper tail of the returns.

EMPIRICAL RESULTS AND DISCUSSION A preliminary examination of the descriptive statistics of the negative logreturns (Rt) showed that the return distributions have heavier tails than of a normal distribution. The Jarque-Bera statistic is significant at very low levels. High values of the Kurtosis statistics indicate that the distributions have fat tails. Hence, there is a case for applying fat-tailed distributions rather than the normal distribution. A plot of the daily stock returns (see Fig. 1, top panel) indicates that large returns tend to be followed by large returns of either sign while small returns tend to be followed by small returns. It implies that returns are not independent and identically distributed, and the volatility clustering phenomenon is present in the data. The observation suggests employing a GARCH model to incorporate dynamic volatility. As noted above, the stock market was closed down during 2008; the Chow’s breakpoint test is, therefore, conducted to test whether there are significant differences in the estimated equations. The results of the test support the presence of a structural break in the data. It indicates that a dummy variable should be incorporated in the model. Correlogram of the returns, in Fig. 1 bottom panel, indicated that the auto-correlation of returns is significant up to lag 1 and, therefore, we incorporate an AR(1) component in the model to capture the effect of conditional mean. The preliminary tests results are not reported here for brevity. The next step is to estimate the parameters of AR(1)-GARCH(1,1) model. We fit two different models to make comparison, that is, one with a Dummy variable and the other without it. The models are fitted using the pseudo-likelihood method to obtain the parameter estimates θ^ =   ^ ω; ^ α; ^ β^ for the GARCH(1,1) model with normal innovations; the φ;

Results from AR-GARCH Estimation.

Without Dummy Coeff.

With Dummy

Std. error

z-Stat.

Prob.

Coeff.

3.3650

0.0008

0.0631 0.0008

13.5508 14.5711 77.4652

0.0000 0.0000 0.0000

Mean Equation AR(1) DUM1

0.0641

0.0190

Variance Equation C RESID(-1)2 GARCH(-1) DUM1

9.33E-06 0.1486 0.8073

6.89E-07 0.0102 0.0104

Durbin
Watson

1.9263

1.12 E-05 0.1469 0.8059 −3.83E-06 1.9228

Std. error 0.01933 0.0004 8.67 E-07 0.0105 0.0104 6.45E-07

z-Stat.

Prob.

3.2635 2.2125

0.0011 0.0269

12.8661 13.9996 77.5187 −5.9339

0.0000 0.0000 0.0000 0.0000

Challenges in the Application of EVT in Emerging Markets

Table 1.

429

430

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

results are reported in Table 1. All the coefficients of the volatility and mean equations are significant. The model with the Dummy variable indicates that although the coefficient DUM1 is significant in the mean and the variance equations, but the values of conditional mean and variance coefficients are not different from the other model. The relatively small size of the DUM1 coefficients also indicates that the Dummy variable may not have a large impact, implying that the proposed model without Dummy variable may be satisfactory and the structural break does not appreciably impact the validity of the model. The Durbin
Watson Statistics are within the acceptable range implying that the model’s specification is tenable. We also ran the ARCH-LM residual test for the standardized residuals extracted from both models and found no evidence against the independent identically distributed (iid) hypothesis for the residuals. The descriptive statistics and QQ-plot of standardized residuals also indicate a departure from normality and a fat right tail. We conclude that the fitted model is tenable. We next consider the choice of innovation distribution from among the closed form family of stable distributions. Table 2 reports the GOF test results for the standardized residuals. The distributions are ranked according to the statistic value for different GOF tests, and the distribution with lowest GOF test statistic value is considered as a better fit. The results indicate that the Cauchy distribution which is known to be heavy tailed is best for modeling the distribution of residuals. It is supported by all three different GOF tests. However, our interest lies in determining the distribution of the tails of the innovations or more precisely the distribution of right tail which represents losses. Therefore, we next consider different extreme value probability distributions and make a comparison by goodness-of-fit test measures. Table 3 provides the Goodness-of-Fit results for the upper tail of distributions of innovations (loss side) for six types of distributions. It also Table 2. Distribution

Cauchy Normal Levy

Goodness-of-Fit Test Results for the Standardized Residuals. Kolmogorov Smirnov

Anderson Darling

Chi-Squared

Statistic

Rank

Statistic

Rank

Statistic

Rank

0.0409 0.0940 0.5624

1 2 3

12.76 52.23 1254.10

1 2 3

161.09 550.35 20109.00

1 2 3

431

Challenges in the Application of EVT in Emerging Markets

Table 3.

Goodness-of-Fit Test Results for Right Tail of the Residuals.

Models of Tail Distribution

General Pareto Weibull General Extreme Value Gumbel Maximum Frechet Pareto

Chi-Squared

Kolmogorov Smirnov

Anderson Darling

Statistic

Rank

Statistic

Rank

Statistic

Rank

16.992 22.515** 40.127*** 65.705*** 423.380*** 3589.900***

1 2 3 4 5 6

0.0273 0.0219 0.0652*** 0.1447*** 0.1638*** 0.4294***

2 1 3 5 4 6

1.2208 1.1735 8.5026*** 27.2110*** 75.3720*** 402.5300***

2 1 3 5 4 6

Note: *, **, *** indicate rejection of the null at 5%, 2% and 1%, respectively. Critical values for the tests at 5%, 2% and 1% level of significance, respectively are χ2 = {18.307, 21.161, 23.209}; KS = {0.0375, 0.0419, 0.0450}; AD = {2.5018, 3.2892, 3.9074}.

shows the ranking of distributions on the basis of three Goodness-of-Fit test (Chi-Squared, Kolmogorov Smirnov and Anderson Darling) statistics. For four distributions, GEV, Gumbel Maximum, Frechet and Pareto, the null is rejected at 1% level of significance. It indicates that the empirical distribution does not conform to any of these four distributions. For the General Pareto and Weilbul distribution the null is not rejected at 5% as per Kolmogorov Smirnov and Anderson Darling tests. The test statistics for the two distributions are quite close which implies that both distributions may provide satisfactory models for the empirically observed extreme values. However, the Chi-squared test rejects the null for the Weibull distribution at 2% level of significance. Therefore, we would prefer GPD over the Weibull distribution. In order to estimate the right tail index of the General Pareto Distribution we first need to select an appropriate threshold to identify the tail region. The mean excess function is examined for this purpose. It is done by examining the mean excess plot for the upper tail of losses. The plot became linear for the parameter values higher than 2.00 which indicated the appropriate threshold to be around 2.00. We also used three different threshold values to judge the stability of the estimated parameter and it seems to be stable for the three values. Following this exercise, we fix the threshold value at 2.0826 corresponding to 90th percentile to estimate the shape parameter of the GPD. The parameter estimate is positive, which implies that the distribution of the residuals above the threshold value is Pareto. We also verify this by running GOF test for the residuals above the threshold value, comparing the Pareto Distribution with two light-tailed

432

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

distributions (i.e., Exponential and the Normal). Since the GOF test statistic value for Pareto distribution is smaller than for the other distributions in three different tests, the Pareto distribution is an appropriate fit for to the upper tail of the distribution of losses.3 Next, we calculate the dynamic or conditional VaR using Eq. (2). The point estimates of the conditional quantiles, zq, using the estimated parameters of the Pareto Distribution are estimated as: zq = {1.8107, 2.1842, 2.6391, 3.3959, 4.1145} at confidence levels for {0.90, 0.95, 0.975, 0.99, 0.995}, respectively. We finally proceed to conduct back-tests using methodology explained in the section above employing five different models: (i) Dynamic EVT or GARCH-EVT model, (ii) Static EVT model, (iii) GARCH model with Student-t innovations, (iv) GARCH model with normally distributed innovations, and (v) Static Normal model in which returns are assumed to be normally distributed. Panel A of Table 4 reports the back-testing results without inclusion of the Dummy variable, Panel B indicates the backtesting results of the dynamic model with dummy variable and Panel C indicates the back-testing results for the reduced sample which excludes a period for which market remained closed. The table reports the number of observed violations and compares these with theoretically expected number of violations, and the achieved significance levels, p-values. Panel A of Table 4 indicates that Dynamic EVT correctly estimates all the conditional quantiles, since the achieved p-value is greater than the 5%, significance level for all quantiles. Static EVT method fails at 95%, 99%, and 99.5%. Dynamic-t fails at 95%, but performs well at higher levels. Also, the numbers of violations are closest to the expected at 99% and 99.5%. These results suggest that the loss distribution is better modeled under the Dynamic framework. There is also some support for using the GARCH model with t-distributed innovations which also provides a good fit. Dynamic Normal and Static Normal fails at all levels except at 95% implying that the distribution of the tails of innovations is modeled better using EVT or t-distribution instead of the Normal Distribution when a higher level of confidence is sought. Panel B of Table 4 shows that Dynamic EVT correctly estimates all the conditional quantiles, since the p-value (at 5% level of significance) is insignificant for all quantile levels. The Dynamic-t method fails at 95% and 99.5% which indicates that Dynamic EVT method is more reliable than Dynamic-t method in the presence of structural breaks. Overall, the numbers of observed violations in Panel B of Table 4 are not different from the observed violations in Panel A of Table 4 for the three different dynamic

433

Challenges in the Application of EVT in Emerging Markets

Table 4.

Back-Testing Results.

Quantile

95%

97.50%

99%

99.50%

Panel A: Without Dummy Variable (Test Length = 1973 observations) Model DYNAMIC EVT STATIC EVT DYNAMIC-t DYNAMIC-NORMAL STATIC NORMAL

Expected # violations Observed # violations p-value Observed # violations p-value Observed # violations p-value Observed # violations p-value Observed # violations p-value

98 112 −0.0939 78 −0.0162 65 −0.0001 108 −0.1796 111 −0.1118

49 61 −0.0571 46 −0.3490 38 −0.0549 64 −0.0238 79 0

20 23 −0.2581 6 −0.0003 18 −0.4038 34 −0.0021 51 0

10 12 −0.2877 1 −0.0005 10 −0.8730 22 −0.0006 30 0

Panel B: With Dummy Variable (Test Length = 1973 observations) Model DYNAMIC EVT STATIC EVT DYNAMIC-t DYNAMIC-NORMAL STATIC NORMAL

Expected # violations Observed # violations p-value Observed # violations p-value Observed # violations p-value Observed # violations p-value Observed # violations p-value

98 115 −0.0533 78 −0.0162 67 −0.0003 111 −0.1118 111 −0.1118

49 60 −0.0744 46 −0.3490 38 −0.0549 65 −0.0173 79 0

20 25 −0.1410 6 −0.0003 22 −0.3331 37 −0.0003 51 0

10 12 −0.2877 1 −0.0001 20 −0.0029 24 −0.0001 30 0

Panel C: For Reduced Sample (Test Length = 1852 observations) Model DYNAMIC EVT STATIC EVT DYNAMIC-t DYNAMIC-NORMAL STATIC NORMAL

Expected # violations Observed # violations p-value Observed # violations p-value Observed # violations p-value Observed # violations p-value Observed # violations p-value

93 74 −0.0239 87 −0.2972 109 −0.0477 74 −0.0239 93 −0.9575

46 26 −0.0008 39 −0.1554 58 −0.05166 42 −0.2914 63 −0.0104

19 6 −0.0007 1 0 22 −0.2370 21 −0.3113 41 0

9 3 −0.0175 0 −0.0001 9 −0.8730 19 −0.0032 27 0

434

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

models. It implies that the inclusion of the dummy variable capturing the effect of structural break in the AR-GARCH model does not seem to impact the validity of the results. A reason underlying the results could be that the dynamic adjustments in the VaR measure may also be incorporating the structural changes. We next run the back-testing for the reduced sample, that is, we remove the observations for the period August 1, 2008 to January 15, 2009, to exclude the period when KSE remains closed. It resulted in reduction of sample to 2,852 observations. The back-testing results given in Panel C of Table 4 indicate that the observed numbers of violations are much lower than expected for the Dynamic EVT model. Similar is the case with other methods of estimation. An explanation of this result could be the fact that the market was very volatile in 2007
2008, which produced a high Valueat-Risk measure for the following period. On the other hand the market was relatively calm after it reopened in December 2008 which produced a lower number of violations. This may justify considering the whole sample for determining VaR.

SUMMARY AND CONCLUSIONS A major shortcoming of the various VaR measures has been that the actual return distributions exhibit fatter tails than the normal distribution would specify. As a remedy the EVT has been employed to explicitly incorporate extreme values, and modifying VaR accordingly. Typically, there would be a limited number of extreme observations during a given period, which makes it harder to test and apply EVT as parameters are estimated with low levels of confidence. The equity market in Pakistan provides an excellent case to study the applicability of EVT in a developing country. The stock market has exhibited a high degree of volatility reflecting a risky political and economic environment. The recent GFC has been another source of extreme returns. The confluence of the two sources of volatility provides us with an historic experiment to test the EVT more rigorously. We apply the EVT to the KSE 100 index over the 11 years period, 2001
2012. We find that the returns distributions are fat tailed and the General Pareto Distribution (GPD) model fits the observed distribution of extreme values well. Our back-testing exercise shows that the VaR measures with dynamic adjustment for volatility clustering perform better than measures which are based on normal distribution assumption, or do not

Challenges in the Application of EVT in Emerging Markets

435

take the dynamics of volatility into account. However, we find that the estimated tail-indices of the GPD distribution vary significantly over time. The implication is that the static extreme loss estimates based on one period may not be a reliable guide to the risk of actual losses during subsequent periods, and need to be updated using a dynamic framework. Our study adds to the existing empirical work in a significant way, as we specifically incorporate many of the characteristics of the emerging economies, such as structural shifts and market closures. We test for the applicability of the theoretical return distributions to model the empirical distributions. We find that, within the closed form family of stable distributions, the distribution of innovation is better modeled using the Cauchy than the Normal or the Levy distribution. However, to model the tails of the loss distribution the GPD is found to be a better fit in comparison to other extreme value distributions, such as the GEV distribution. Considering the impact of market shut-down, we find that including the closure period for back-testing procedure tends to underestimate the market volatility due to a large gap in the data, and consequently overstates the number of violation to the estimated VaR in the following period. When we remove the period for which the market remained closed from the back-testing procedure, we find that the observed number of violations is much reduced, and the dynamic EVT method does not perform as well. Since, the impact of the market closure seem to reduce the estimated volatility for the following period, it follows that by including the market closure period in the estimation period improves the predictability of the risk models. In this case the dynamic EVT provides more reliable results than the dynamic-t method in the presence of structural gaps. The main implication of this work is to validate EVT for a small frontier market like Pakistan, and by extension, it may also hold promise with respect to other emerging markets. In practical applications it implies that risk management systems based on the Dynamic VaR with tail estimation by EVT may be more helpful than simpler VaR models. For example, a Dynamic VaR EVT based system that increased the securities traders’ margin requirements with changes in the market volatility may help in containing market risk, curbing speculation and stabilizing the markets. However, we find that the estimated tail-indices of the GPD distribution vary significantly over time. The performance of various risk models also seem to depend on the choice of return distribution and the dynamics of the return process. These findings confirm results from earlier studies such as Wagner and Marsh (2005) and Wagner (2005) which document substantial differences in measured tail thickness due to small sample bias, and note that

436

JAMSHED Y. UPPAL AND SYEDA RABAB MUDAKKAR

“measuring tail thickness relies on proper assumptions about the underlying return model and its time series properties. The implication is that there may be a residual model risk associated with misspecification of the quantitative model. Therefore, extreme loss estimates based on one period may not be a reliable guide to the risk of actual losses during the subsequent periods. We may need to update not only the model parameters but also its structure since the dynamics of the market may change from one period to another. Our findings underscore the need to update the risk models in a timely fashion while considering possible structural shifts.

NOTES 1. Engel and Hakkio(1993) draw a distinction between the two kinds of volatility. 2. See Hull and White (1998). Acknowledging the need to incorporate time varying volatility VaR models employ various dynamic risk measures such as the Random Walk model, the GARCH, and the Exponentially Weighted Moving Average (EWMA). The Risk metrics model uses the EWMA. 3. The results of these tests are not reported here but are available from authors.

REFERENCES Balkema, A., & de Haan, L. (1974). Residual life time at great age. Annals of Probability, 2(5), 792
804. Bhattacharyya, M., & Ritolia, G. (2008). Conditional VaR using EVT
Towards a planned margin scheme. International Review of Financial Analysis, 17, 382
395. Bollerslev, T., Chou, R., & Kroner, K. (1992). ARCH modeling in finance. Journal of Econometrics, 52, 5
59. Chou, R. Y. (2005). Forecasting financial volatilities with extreme values: The conditional autoregressive range (CARR) model. Journal of Money, Credit and Banking, 37(3), 561
582. Diebold, F. X., Schuermann, T., & Stroughair, J. D. (2000). Pitfalls and opportunities in the use of extreme value theory in risk management. Journal of Risk Finance, 1(Winter), 30
36. Djakovic, V., Andjelic, G., & Borocki, J. (2011). Performance of extreme value theory in emerging markets: An empirical treatment. African Journal of Business Management, 5(2), 340
369. Embrechts, P., Kluppelberg, C., & Mikosch, T. (1997). Modeling extreme events for insurance and finance. Berlin: Springer. Engel, C., & Hakkio, C. S. (1993). Exchange rate regimes and volatility. Economic Review Federal Reserve Bank of Kansas City, (QIII), 78(3), 43
58.

Challenges in the Application of EVT in Emerging Markets

437

Fisher, R., & Tippet, L. (1928). Limiting forms of the frequency distribution of the largest or smallest member of a sample. Proceedings of the Cambridge Philosophical Society, 24, 180
190. Gencay, R., & Selcuk, F. (2004). Extreme value theory and value-at-risk: Relative performance in emerging markets. International Journal of Forecasting, 20, 287
303. Hill, B. M. (1975). A simple general approach to inference about the tail of a distribution. Annals of Statistics, 3(5), 1163
1174. Hull, J., & White, A. (1998). Incorporating volatility updating into the historical simulation method for value at risk. Journal of Risk, 1(1), 5
19. Iqbal, J., Azher, S., & Ijza, A. (2010). Predictive ability of value-at-risk methods: Evidence from the Karachi stock exchange-100 index. MPRA Paper 01/2010, University Library of Munich, Germany. Kluppelberg, C. (2001). Development in insurance mathematics. In B. Engquist & W. Schmid (Eds.), Mathematics unlimited
2001 and beyond (pp. 703
722). Berlin: Springer. Longin, F. M. (1996). The asymptotic distribution of extreme stock market returns. The Journal of Business, 69(3), 383
408. McNeil, A. (1997). Estimating the tails of loss severity distributions using extreme value theory. ASTIN Bulletin, 27, 117
137. McNeil, A. J. (1999). Extreme value theory for risk managers. ETH Zentru. McNeil, A. J., & Frey, R. (2000). Estimation of tail-related risk measures for heteroscedastic financial time series: An extreme value approach. Journal of Empirical Finance, 7, 271
300. Nawaz, F., & Afzal, M. (2011). Value at risk: Evidence from Pakistan stock exchange. African Journal of Business Management, 5(17), 7474
7480. Onour, I. A. (2010). Extreme risk and fat-tails distribution model: Empirical analysis journal of money. Investment and Banking, 13, 27
34. Pickands, J. (1975). Statistical inference using extreme order statistics. The Annals of Statistics, 3, 119
131. Poon, S., & Granger, C. (2003). Forecasting volatility in financial markets. Journal of Economic Literature, 41, 478
539. Qayyum, A., & Nawaz, F. (2011). Measuring financial risk using extreme value theory: Evidence from Pakistan. MPRA Working Paper 29288. University Library of Munich, Germany. Taleb, N. N. (2010). The Black Swan: The impact of the highly improbable (2nd ed.). New York, NY: Random House Trade. Uppal, J. Y. (2013). Measures of extreme loss risk
An Assessment of performance during the global financial crisis. Journal of Accounting and Finance, 13(3), 105
117. Uppal, J. Y., & Mangla, I. U. (2013). Extreme loss risk in financial turbulence
Evidence from global financial crisis. Managerial of Finance, 39(7), 653
666. Wagner, N. (2005). Autoregressive conditional tail behavior and results on government bond yield spreads. International Review of Financial Analysis, 14, 247
261. Wagner, N., & Marsh, T. A. (2005). Measuring tail thickness under GARCH and an application to extreme exchange rate changes. Journal of Empirical Finance, 12, 165
185. Weismann, I. (1978). Estimation of parameters and quantiles based on the k largest observations. Journal of American Statistical Association, 73, 812
815.

INDEX Abenomics, 182 Asia, 7, 97
100, 102
103, 147, 158
159, 165
166, 174, 216, 280, 368, 372, 384

224, 263, 272, 317
321, 323, 325, 327
331, 333, 335
336, 381 Complexity, 4
5, 15
27, 29, 31
35, 37 Corporate governance, 9, 283, 292
295, 299
300, 302
303, 308
311, 313
314 Credit, 5, 7
9, 12, 18, 20, 23, 25
26, 32, 34, 36, 61, 72, 181
182, 184
186, 188
190, 221
227, 229
235, 237
239, 241
246, 259, 263, 266
268, 270, 272, 274, 277
279, 281
285, 287
288, 292
293, 295
299, 302, 308
310, 318
320, 326, 332
334, 336
337, 342, 344
346, 349
351, 353, 355
356, 358
360, 402, 404
405, 413 Cross-Border Trade Settlement, 367, 372, 378
379 Cross-section volatility, 389

Back-testing, 36, 401, 418, 420, 427, 432
435 Bank risk, 9, 277, 279
280, 283, 285, 287
288 Banks, 5
10, 16, 22, 59
64, 68, 70, 72, 74, 77
79, 97
109, 115
118, 127, 135, 144, 148
149, 158, 181
184, 186
189, 194, 196, 208
209, 214, 223, 230
231, 234, 239, 255
256, 259
269, 271
274, 277
285, 287
288, 294, 296, 299, 317
326, 329, 332
337, 341
342, 349, 356, 359, 361, 369, 371
377, 380
381, 404
405 Basel III, 6, 59
80 Beta herding, 389, 391, 393, 395, 397, 399 BIST-100, 389, 394
397, 399 Bond markets integration, 41
42 Brazilian real estate market, 401
414

Dim Sum Bond Market, 10, 367
377, 379
383 Domestic banking sector, 97
99, 101
103, 115
116, 145, 147
150

Capital risk, 6, 59
61, 64
65, 74, 77
79 Collateral spread, 8, 221, 225, 236 Competition, 8, 10, 30, 102, 184, 211
215, 217
218,

Excess cash, 9, 291
303, 305
313 Expected shortfall, 6, 37, 83
84 Explosive behavior, 408, 409, 411, 413 439

440

INDEX

External influences, 98 Extreme value theory, 11, 401, 417
419

361, 370
371, 400
401, 417
419 Gradient tree boosting, 6, 83

Finance, 8, 15
21, 23, 25, 27, 29
33, 35, 37, 103, 110, 157
158, 161, 168, 195, 221, 224
234, 236
237, 239
247, 260
261, 278
279, 282, 303, 332, 343, 353, 376
377, 380, 384, 419 Financial architecture, 341, 343, 345, 347, 349, 351, 353, 355, 357, 359
361 Financial crisis, 3
5, 7, 9
11, 15
16, 22, 35, 37, 41
43, 46
48, 54, 59
61, 83, 97
100, 116, 119
120, 136, 147
149, 153, 160, 181, 193, 211, 221, 223, 232, 255, 259
260, 267, 270, 277
278, 280, 283, 285, 287, 291, 317, 341
345, 360
361, 367, 370
371, 389, 398
401, 417
420 Financial integration, 7, 42, 54, 153
157, 159
161, 163, 165, 167
174 Financial structure, 10, 27
28, 292, 294, 305
306, 313
314, 341
345, 350
351, 353
354, 356
357, 360 Foreign stock market, 7, 97
98, 101, 118
119, 145
150

Industries, 7, 9, 153, 156
159, 165, 172
173, 175, 293
294, 300, 303
304, 310, 313 Inflation, 8, 99, 158, 194, 196, 209, 211
218, 280
281, 319
321, 326, 330
331, 335
336, 342, 349
350, 353
360, 390, 404 Inflation persistence, 8, 211
215, 217
218 Institutions, 10, 16
17, 20, 22
23, 32
37, 62, 103, 189, 212, 221, 223
226, 232, 234, 236
237, 239
243, 245
247, 261, 264, 278, 282
283, 292
293, 302, 318, 368
371, 373, 377
378, 380
383, 419 Integration index, 153, 155, 165, 167
169, 171
172, 175 Interbank market, 22, 99, 149, 255
258, 261
263, 265
266, 268
269, 271
272, 274 Interest rates, 42, 53
54, 59
60, 64, 72
74, 78, 100, 116, 118, 184, 189, 194, 197, 234, 241, 260, 267
269, 272
274, 278
279, 291, 293
294, 299, 303, 305
306, 308, 310
311, 313, 317
319, 325
326, 329, 332
333, 335
337, 342
343, 345, 349, 353, 356, 360

GARCH models, 47, 401, 418 Global Financial Crisis, 4, 41
42, 97
98, 120, 136, 223, 259, 277
278, 280, 283, 341
345,

Kenya, 10, 229
230, 341
345, 349, 351, 355
356, 360
361 Law, 31
32, 35
37, 154, 221, 225, 239
243, 246, 404

441

Index

Linear and non-linear causality tests, 193
209 Liquidity, 4, 7
9, 11
12, 16, 22, 25
26, 34, 42, 59, 64, 99, 149, 171, 181
182, 186, 208, 223, 233, 245, 255
274, 282
283, 285, 287, 291
293, 295
296, 298, 300, 303, 305, 308, 318, 320, 324
327, 330
331, 333
336, 342, 373, 376, 382
383, 418 Liquidity trap, 182, 186

Open market operation, 256, 269
271 Operating cost, 187, 318, 320, 325
326, 332
333, 336

Macroeconomic news, 6, 41
45, 47
54 Market fundamentals, 389
390, 392, 394, 396
397, 399, 413 Market reference rate, 8, 181
188, 190
191 Monetary easing, 3
5, 15, 41, 59, 83, 97, 153, 181, 193, 211, 221, 255, 274, 277, 291, 317, 341
343, 367, 389, 401, 417 Monetary integration, 41 Monetary policy, 3
5, 7, 9
10, 42, 54, 60, 64, 78, 99, 117, 150, 182, 184, 194, 208
209, 213, 223, 255
259, 261, 263, 265
269, 271, 273
275, 277
285, 287
288, 318
319, 325, 341
345, 347, 349, 351, 353, 355
361 Money squeeze, 256, 258

Quantitative easing, 4, 8, 60, 64, 78, 181
183, 185, 187, 189, 191, 264, 271, 342 Questionnaire, 291
294, 300, 305
307, 312
314

Panel data, 212
213, 277
278, 317
319, 322
323, 333, 335 PCA, 153, 155, 161
164, 168 Price-rent ratio, 407, 410 Pro-cyclicality, 6, 59
62, 64
65, 68, 70, 72, 74, 77, 288

Net interest income, 318 Nonparametric expectile regression, 83, 85, 87
89, 91, 93

Real demand for money, 193 Recursive unit root test, 11, 414 Regulatory capital, 6, 59
69, 71
75, 77
79, 288 Renminbi (RMB) Internationalization, 10, 367
380, 382
384 Risk funds, 182, 191 Risk management, 3
5, 7, 9, 11, 15
21, 23, 25
27, 29, 31
37, 41, 59
61, 64
65, 74, 83
85, 90, 94, 97, 101, 153, 181, 193, 211, 221, 223, 228, 255, 277, 282, 291, 317, 336, 341, 367, 389, 401, 417, 419, 435 Risk measurement, 4
5, 84 R-square, 153, 155
156, 161
166, 168, 171
172, 174
175

Off-Shore RMB Deposit Market, 367

Short-term bank loan, 9, 291
303, 305
311, 313
314

442

Speculative bubbles, 11, 401
405, 407, 409, 411, 413
414, 418 Sovereign Debt Crisis, 4, 12, 42, 160 Stock prices, 7, 97
98, 100
101, 109
113, 115
116, 118, 403, 419, 422

INDEX

Subprime Crisis, 4, 11
12, 149, 160 Transmission channel, 10, 277, 342, 344, 353, 355, 358
359 Value at Risk, 6, 84
85, 401