Economic Analysis of Terrorism (Routledge Studies in Defence and Peace Economics)

The Economic Analysis of Terrorism In the aftermath of 9/11 The Economic Analysis of Terrorism examines two crucial qu...

Author: Tilman Bruck

12 downloads 703 Views 4MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

The Economic Analysis of Terrorism

In the aftermath of 9/11 The Economic Analysis of Terrorism examines two crucial questions. First, how does the new global terrorism affect the economy? And second, what contribution can economics make to the analysis of terrorism? The answers to both questions are comprehensive and intriguing. Economic analysis is used throughout the volume to explain the deadly but rational calculus of terrorists and to outline the choices available to national and global policy makers in the ﬁght against terrorism. The emphasis is on combing the main economic theories on terrorism research with cutting edge empirical evidence. The Economic Analysis of Terrorism contains 17 thought-provoking articles and essays by leading economists from Europe and North America. The authors are experts in all areas of economic analysis, ranging from ﬁnancial and insurance economics, industrial organisation and business economics, microeconomics, international macroeconomics and trade, public choice and public ﬁnance to development economics. This book represents the most thorough and systematic study of the economics of terrorism published since 9/11. It will be of interest to students, researchers and policy makers with an interest in the most challenging global policy issue facing the world today. Tilman Brück is the Head of the Department of International Economics at the German Institute for Economic Research (DIW Berlin), Germany.

Routledge studies in defence and peace economics Series Editors: Keith Hartley University of York, UK and Jurgen Brauer, Augusta State University, USA.

1 European Armaments Collaboration Policy, problems and prospects R. Matthews 2 Military Production and Innovation in Spain J. Molas-Gallart 3 Defence Science & Technology Adjusting to change R. Coopey, M. Uttley and G. Spiniardi 4 The Economics of Offsets Defence procurement and countertrade S. Martin 5 The Arms Trade, Security and Conﬂict Edited by P. Levine and R. Smith 6 Economic Theories of Peace and War F. Coulomb 7 From Defense to Development? International perspectives on realizing the peace dividend A. Markusen, S. DiGiovanna and M. Leary 8 Arms Trade and Economic Development Theory, policy, and cases in arms trade offsets Edited by Jurgen Brauer and J. Paul Dunne 9 The Economic Analysis of Terrorism Edited by Tilman Brück

The Economic Analysis of Terrorism

Edited by Tilman Brück

First published 2007 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 270 Madison Avenue, New York, NY 10016 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2007 Selection and editorial matter, Tilman Brück; individual chapters, the contributors This edition published in the Taylor & Francis e-Library, 2007. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.”

All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A catalogue record for this book has been requested ISBN 0–203–01663–7 Master e-book ISBN

ISBN10: 0-415-36523-6 (hbk) ISBN10: 0-203-01663-7 (ebk) ISBN13: 978-0-415-36523-9 (hbk) ISBN13: 978-0-203-01663-3 (ebk)

To Marta, with all my love

Contents

Figures Tables Contributors Acknowledgements

x xii xiv xvi

PART I

Introduction

1

1

3

A brief survey of the economic analysis of terrorism TILMAN BRÜCK

PART II

Features of terrorism

11

2

13

An economic perspective on transnational terrorism TODD SANDLER AND WALTER ENDERS

3

Economic conditions and terrorism

29

S. BROCK BLOMBERG, GREGORY D. HESS AND AKILA WEERAPANA

4

The economics of high-visibility terrorism

47

SANJAY JAIN AND SHARUN W. MUKAND

5

Decomposing violence: political murder in Colombia, 1946–99 JURGEN BRAUER, ALEJANDRO GÓMEZ-SORZANO AND SANKAR SETHURAMAN

64

viii

Contents

PART III

Economic effects of terrorism

81

6

83

The effects of terrorism on global capital markets ANDREW H. CHEN AND THOMAS F. SIEMS

7

Terrorism-induced structural shifts in ﬁnancial risk: the case of airline stocks in the aftermath of 11 September terror attacks

107

KONSTANTINOS DRAKOS

8

Financial markets and terrorism

121

RAFI ELDOR AND RAFI MELNICK

9

Global terrorism and the insurance industry: new challenges and policy responses

146

MICHAEL WOLGAST

10 Terrorism and international trade: an empirical investigation

173

VOLKER NITSCH AND DIETER SCHUMACHER

11 Fiscal consequences of armed conﬂict and terrorism in low- and middle-income countries

185

SANJEEV GUPTA, BENEDICT CLEMENTS, RINA BHATTACHARYA AND SHAMIT CHAKRAVARTI

PART IV

Anti-terrorist policies

207

12 Rights and citizenship in a world of global terrorism

209

DENNIS C. MUELLER

13 Decentralization as a response to terror BRUNO S. FREY AND SIMON LUECHINGER

224

Contents 14 Global threats and the domestic struggle for power

ix 231

MICHELLE R. GARFINKEL

15 Global ﬁnancial information, compliance incentives and terrorist funding

246

VALPY FITZGERALD

16 An economic analysis of security policies

262

TILMAN BRÜCK

PART V

Epilogue

283

17 Terror: the ‘ISM’ versus the ‘ISTS’

285

GEOFFREY BRENNAN

Figures

2.1 2.2 4.1 4.2 5.1 5.2 5.3 6.1 6.2 6.3 6.4 6.5 6.6

6.7 8.1 8.2 8.3 8.4 8.5

All incidents and bombings. Ordinal game matrix for retaliation. Utility from a successful attack. Expected utility of N and S attacks. Total homicides, Colombia, 1946–1999. Original data (total homicides per 100,000 people) and permanent component according to HP and BN methods. Cyclical component according to HP and BN methods. Dow Jones industrial average 1915–2002. US capital markets’ 11-day cumulative abnormal returns (Dow Jones Industrial Average Index). Global capital markets’ event-day abnormal returns following the 11 September terrorist attacks. Global capital markets’ 11-day cumulative average abnormal returns following the 11 September terrorist attacks. Global capital markets’ 11-day cumulative average abnormal returns following Iraq’s invasion of Kuwait. Global capital markets’ and banking/ﬁnancial sector 11-day cumulative average abnormal returns following the 11 September terrorist attacks. Ten-year US treasury note rates 1982–2002. Terror attacks, victims injured, victims killed and suicide attacks. The Tel Aviv 100 stock market index [January 1990–June 2003 (log)]. The Tel Aviv 100 and the S&P500 indices [January 1990–June 2003 (log)]. The basket of currencies exchange rate and the exchange rate band [January 1990–July 2003]. The Dollar and basket of currencies exchange rate [January 1990–June 2003 (log)].

17 22 51 53 65 67 68 90 92 96 97 100

103 104 123 126 126 127 128

Figures 8.6 The Tel Aviv 100 index – actual and simulated with the pre-September 27, 2000 fundamental equation, September 2000–June 2003. 11.1 Real GDP growth in conﬂict countries (average annual per cent changes). 11.2 Consumer price inﬂation in conﬂict countries (average annual per cent change). 11.3 Capital formation in conﬂict countries in per cent of GDP. 11.4 Fiscal aggregates in conﬂict countries in per cent of GDP. 11.5 Composition of government spending in conﬂict countries in per cent of GDP. 12.1 Possible optimal majorities.

xi

131 190 191 192 193 194 211

Tables

2.1 Transnational terrorism: events 1968–2001 2.2 Trend and other statistical properties of transnational terrorist incidents 3.1 Terrorist incidence around the world average annual incidence 1968–91 3.2 Estimates of 2 × 2 Markov processes for the economy and terrorism 3.3 Estimates of 4 × 4 Markov processes for the economy and terrorism 4.1 Payoff matrices with only passengers and hijackers 4.2 Payoff matrices with the possibility of sky marshals 6.1 Average abnormal returns on the Dow Jones Industrial Average Stock Index following terrorist attacks 6.2 Average abnormal returns on global capital markets following the 11 September terrorist attacks 6.3 Average abnormal returns on global capital markets following Iraq’s invasion into Kuwait 6.4 Average abnormal returns on global capital markets’ banking/ﬁnancial sectors following the 11 September terrorist attacks 7.1 IATA members’ average (international scheduled services) 7.2 Quarterly ﬁnancial review for US passenger airlines (majors); 2001 7.3 The impact of 9/11 attacks on market value 7.4 Market model for individual airline stocks: estimation results 7.5 Tests of equality of volatility 7.6 Decomposition of total risk for individual airline stocks 8.1 Summary of terror attacks by categories (January 1990–June 2003) 8.2 The damage rate of attacks by categories (January 1990–June 2003) 8.3 Proportion of people killed, injured and days of a terror attack by categories (January 1990–June 2003)

15 18 34 38 42 56 57 88 94 99

102 108 109 111 116 117 118 124 125 125

Tables 8.4 8.5 8.6

Testing for unit roots (January 1990–June 2003) Granger causality tests (6 lags) (January 1990–June 2003) The fundamental regressions [January 1990–June 2003 (3515 daily observations)] 8.7 Interpreting the news of a terror attack 8.8 Testing the impact of terror on the stock exchange (January 1990–June 2003) 8.8a Testing the impact of terror on the stock exchange (January 1990–June 2003) 8.9 Testing the impact of terror on the dollar exchange rate (February 1996–June 2003) 8.10 Testing the impact of terror on the basket exchange rate (February 1996–June 2003) 8App Dependent variable Dlog (TA100) (included observations: 3515 after adjusting endpoints) 9.1 The worst terrorist acts in terms of casualties and insured property losses 9.2 The most costly insurance losses 1970–2001 9.3 Development of major indexes on the stock market in 2001 9.4 Global insurance markets/inﬂow of new capital (USD million) (as of 12 April 2002) 9.5 Synopsis of international model solutions for the insurance of terror risks 10.1 Description of data on terrorist activity. Five countries that suffered most strongly from terrorism; period: 1968–79 10.2 The impact of terrorism on trade 10.3 The impact of internal instability on trade 10.4 The impact of military personnel and defence expenditures on trade 10.5 The impact of war on trade 11.1 Selected social indicators in countries experiencing armed conﬂicts (average annual rates of change, per cent) 11.2 Regression results 11.3 Regression results: robustness test

xiii 129 129 130 133 134 136 139 141 144 148 149 150 151 160 177 178 180 181 182 195 196 199

Contributors

Rina Bhattacharya, OECD, Paris, France S. Brock Blomberg, Department of Economics, Claremont McKenna College, USA Jurgen Brauer, James M. Hull College of Business, Augusta State University, USA Geoffrey Brennan, Social and Political Theory, RSSS, ANU, Political Science Department, Duke University Philosophy Department, UNC-Chapel Hill Tilman Brück, Department of International Economics, Economic Research (DIW Berlin), Berlin, Germany Shamit Chakravarti, Asian Development Bank, New Delhi, India Andrew H. Chen, Distinguished Professor of Finance, Cox School of Business, Southern Methodist University, USA Benedict Clements, International Monetary Fund, Washington, USA Konstantinos Drakos, Assistant Professor, Department of Economics, University of Patras, Greece Raﬁ Eldor, Arison Business School, The Interdisciplinary Center, Israel Walter Enders, Department of Economics, Finance, and Legal Studies, University of Alabama, Tuscaloosa, USA Valpy FitzGerald, Finance and Trade Policy Research Centre, University of Oxford, UK Bruno S. Frey, Institute for Empirical Research in Economics, University of Zurich, Switzerland Michelle R. Garﬁnkel, Professor of Economics, Department of Economics, University of California-Irvine, USA

Contributors

xv

Alejandro Gómez-Sorzano, Consultant, Philadelphia, USA Sanjeev Gupta, International Monetary Fund, Washington, USA Gregory D. Hess, Claremont McKenna College, USA Sanjay Jain, Department of Economics, University of Virginia, USA Simon Luechinger, Institute for Empirical Research in Economics, University of Zurich, Switzerland Raﬁ Melnick, Arison Business School, The Interdisciplinary Center, Israel Dennis C. Mueller, Department of Economics, University of Vienna, Vienna Sharun W. Mukand, Department of Economics, Tufts University, USA Volker Nitsch, Department of Economics, Free University, Berlin, Germany Todd Sandler, University of Texas at Dallas, School of Economics, Political and Policy Sciences, USA Dieter Schumacher, DIW Berlin, International Economics, Germany Sankar Sethuraman, Department of Mathematics and Computer Science, Augusta State University, USA Thomas F. Siems, Senior Economist and Policy Advisor, Federal Reserve Bank of Dallas, USA Akila Weerapana, Department of Economics, Wellesley College, USA Michael Wolgast, Chief Economist and Head of Economics Department, German Insurance Association (GDV), Germany

Acknowledgements

This book emerged from the international conference on ‘The Economic Consequences of the New Global Terrorism’ held at the German Institute for Economic Research (DIW Berlin) in 2002. The workshop was made possible by the very generous ﬁnancial support from the German Insurance Association (GDV) and the German Foreign Ofﬁce. Some of the papers presented at the conference were published in a special issue of the European Journal of Political Economy on ‘The Economic Consequences of Terror’ (Vol. 20, No. 2, June 2004), which was edited by Tilman Brück and Bengt-Arne Wickström. I am grateful to Elsevier for permission to reproduce some of the articles from that special issue in this volume. I am also indebted to Stefanie Erdrich, Wolfgang Härle and Gisela Tietke for their excellent research assistance, to the many authors of this volume for their support in preparing this text, and to the extremely helpful staff at Taylor & Francis. Tilman Brück

Part I

Introduction

1

A brief survey of the economic analysis of terrorism Tilman Brück

This introduction notes the importance of the economic analysis of the causes, workings and consequences of global terrorism and discusses the costs of antiterrorist policy. It does so by summarizing the main contributions of this volume.

1.

Introduction

9/11 has focused attention on global terrorism in a way that no previous attacks had ever done. Within the ﬁeld of economics, it seemed initially as if there was a lack of understanding about how to analyse terrorism. This early incomprehension of the attacks had two dimensions. First, most academic economists felt they had little to contribute to the debate about the nature and the workings of global terrorism. Second, most economic analysts could not predict the likely economic consequences of the attacks. Both dimensions are important as an understanding of the nature of terrorism and the magnitudes of its effects is a prerequisite for designing successful policies to prevent terror, to alleviate the costs of terrorism, and to reduce an economy’s vulnerability to attacks. Within a few months of the attacks, economists in many different sub-disciplines realised that their toolkits had, after all, prepared them to tackle the analysis of terrorism. For example, the study of shocks has long been a subject of international economics, the effects of violent conﬂict have been analysed in development economics for at least ten years and the economics of insurance naturally addresses issues of risk. To bring this young and diverse community of economic terror experts together, the German Institute for Economic Research (DIW Berlin) organised an international conference in Berlin in June 2002. Most of the chapters of this book are derived from the presentations at that conference. They hence provide one of the ﬁrst and most systematic analyses of terrorism from an economic point of view to date. The topics of this volume include on the one hand the economic analysis of the functions and mechanisms of terrorism and on the other hand the economic effects of the attacks in the short- and the long-term. Furthermore, the book offers some

4

Tilman Brück

insights from economics on the analysis of anti-terror policies. Methodologically, the chapters in this book include analytical approaches, empirical studies and policy analyses, thus providing a uniquely broad overview of the newly emerging ﬁeld of the economic analysis of terrorism. In the remainder of this chapter, the key ﬁndings of this volume will be summarized and discussed.

2.

Features of terrorism

The ﬁrst part of the book surveys some key economic features of terrorism. First, Todd Sandler and Walter Enders provide an overview of data sources on terrorism. Data that they present describe patterns of terror before the 9/11 attack and show that prior terror attacks took place in cycles. Sandler and Enders also emphasize that terrorists respond to changed incentives and that anti-terrorist policies may induce substitution effects, with terrorists moving from hard to soft targets or shifting their activities over time. Anti-terrorist policy is therefore more successful if conducted across the entire spectrum of potential terrorist activities so as to reduce terrorists’ substitution possibilities. Otherwise a domestic anti-terrorist policy directed at stopping terror against the national population would shift the focus of terrorist activities to other countries. Sandler and Enders point out that, for example, guarding American embassies more heavily abroad after the dual attacks on US embassies in Kenya and Tanzania in the 1990s reduced embassy bombings but led to an increase in shootings and abductions of embassy personnel away from the embassies themselves. S. Brock Blomberg, Gregory Hess and Akila Weerapana study the relationship between growth cycles and terrorism or civil war. They ﬁnd that in richer democratic countries terrorism is more likely during economic downturns. Their chapter emphasizes the need for further work about the relationship between conﬂict and growth in poorer countries, the role of transnational terrorism versus domestic terrorism, and the joint determination of both conﬂict and growth by omitted variables such as weak governance. With hindsight, it appears that the immediate or shorter-run economic effects of the 9/11 attacks were contained by insightful policy making by the monetary authorities in the US and Europe and by the robustness of the world economy itself. It is thus ironic that an event that some interpret as an anti-globalization attack failed to induce an international worst-case economic scenario due to the stabilizing forces of globalization. Incidentally, this result is conﬁrmed by Eldor and Melnick in their chapter on the response of ﬁnancial markets in Israel to Palestinian terror (see below), suggesting that market liberalization enhanced the capability of the Israeli economy to cope with terror. Sanjay Jain and Sharun Mukand demonstrate in a theoretical model how terrorist attacks change expectations, including the expectations of terrorists themselves. Terror leads to anti-terror policies that affect the nature of future terror. By making anti-terror policy less predictable, policy makers can increase the uncertainty facing the terrorists, which is beneﬁcial for society. Jain and Mukand show the

A brief survey of the economic analysis of terrorism

5

importance of policy makers responding to terror both verbally and through action, and that anti-terrorist policy can also contain policy elements beyond standard security or economic policies. The importance of detailed data for the analysis of terrorism is also demonstrated in the chapter by Jurgen Brauer, Alejandro Gómez-Sorzano and Sankara Sethuraman on Colombia. They demonstrate the role of cyclical political variables and permanent non-political variables using a novel source of data on politically motivated and other types of murders.

3.

Economic effects of terrorism

This section surveys contributions in this volume and elsewhere on the costs of terror, the economic effects of terror on ﬁnancial markets and international trade, and the post-terror ﬁscal and growth effects in developing and developed countries. Estimates of the costs of terrorism or violent conﬂict more generally confront methodological problems of different types, including the deﬁnition of damage, the measurement of losses, aggregation issues, avoidance of double counting of damages in different sectors or statistics, and the causality of second round and indirect effects. Against the background of such measurement problems, the OECD estimated costs resulting from the terror attacks of 9/11 of 14 billion USD for the private sector, 1.5 billion USD for state and local government enterprises, 0.7 billion USD for the US federal government, and 11 billion USD for the rescue and clean-up operations, where the latter is shared between the private and the public sectors (Lenain et al. 2002). The direct economic costs of terrorism are most pronounced in the aftermath of attacks. It appears that consumer conﬁdence in the United States had started to recover before 9/11 but then was held back by the psychological impact of the attacks. In the medium term, the loss of conﬁdence had an adverse self-reinforcing effect on growth in the United States and Europe. The heightened uncertainty reduced spending, slowed down ﬁrm investment, led to layoffs and increased unemployment (Baily, 2001). The intertemporal distribution of the indirect consequences depends on the nature of the attacks, the multiplier effects of the direct effects, and the type of policies adopted in response to the attacks. The US economy, for instance, recovered quickly from the 9/11 attacks because the shocks were transitory and caused proportionately little damage to the American capital stock, left oil prices unaffected, and did not affect the economy’s ability to generate income and wealth. This reﬂects a key difference between the economic legacies of terrorism and war, with ongoing civil war in particular damaging a country’s capacity to grow (see for example Abadie and Gardeazabal, 2003, Addison, 2003, and Stewart and FitzGerald, 2001). The indirect costs of terrorist attacks vary in their distribution across activities, sectors, countries and time. Some activities and sectors are more vulnerable to attacks than others and consequently suffer a higher burden. Service sectors, for example, experienced a sharper drop in production if they were more closely

6

Tilman Brück

related to ﬁrms affected by the twin tower attacks of 9/11 or by border closures, as inventories of service output cannot be accumulated in times of slack demand (Strauß, 2001). Highly networked and synchronized just-in-time industrial production can be expected to have been affected by supply disruptions due to 9/11. Terror attacks can also change patterns of demand in the long-term. Households, ﬁrms and governments may exhibit different preferences after exposure to terror or may re-assess their vulnerability to attacks. Reductions in demand have in particular been exhibited by the transport and tourism sectors. With regard to tourism, Fleischer and Buccola (2002) ﬁnd that foreign demand for accommodation in Israel is price elastic and sensitive to regional terrorism. Local demand for accommodation, on the other hand, is price inelastic and does not react negatively to terrorism. Such effects through costs and changed demand are reﬂected in ﬁnancial markets where asset values respond to changes in expected proﬁtability. Increases in transaction costs are one of the main indirect effects of terror. Strictly speaking, it is not the attacks themselves but the policy responses to the attacks that cause these increases. These policies can include measures to prevent and to detect terrorism. Given the nature of the attacks of 9/11, these measures are especially enacted on borders and include closer inspections of people, vehicles and goods as well as more restrictive immigration regulations. The estimates for the scale of the increase in these international transaction costs vary between 0.5 per cent and 3 per cent ad valorem (Walkenhorst and Dihel, 2002). In addition, domestic trade may also suffer from higher transaction costs if new regulations prove costly for business – though the scale of such increases has not been estimated to date. Andrew Chen and Thomas Siems investigate the magnitudes of the effects of 9/11 on global and US share prices and compare the outcome to the consequences of other political, economic or natural shocks. The magnitude of the effects of 9/11 on global and US ﬁnancial markets was signiﬁcant but not unique when placed in historical perspective. While some sectors were particularly strongly affected by 9/11, Chen and Siems show that the impact of the attacks on ﬁnancial markets varied greatly across countries. Furthermore, the reactions of ﬁnancial markets to 9/11 in the United States were less severe than the reactions to previous negative shocks. They conclude from this that ﬁnancial markets have become more resilient to political shocks in recent years and that regulatory authorities reacted wisely to the attacks by adding sufﬁcient liquidity to the global ﬁnancial system to prevent a banking crisis. Their analysis demonstrates that consumers, ﬁrms, and governments that are not directly physically harmed by terrorists thus still lose from terror attacks. The question remains why market resilience varied so extensively internationally, given that globalization has integrated national ﬁnancial markets, which should have resulted in more uniform responses to shocks. Konstantinos Drakos ﬁnds that the already embattled airline industry was strongly and aversely affected by 9/11. Changed risk perceptions by consumers reduced demand for air travel and for complementary aircraft and hotel accommodation. Airlines also faced higher insurance rates when insurance companies

A brief survey of the economic analysis of terrorism

7

reassessed the likelihood of large-scale terror attack using airplanes. Airline shares reﬂected these changes in lower stock market valuations. Airline shares thus exhibited higher systematic and individual risks post–9/11, with the systematic risk of selected US and European airlines shares more than doubling. This led fund managers to reduce their exposure to these shares in their portfolios and hence put further pressure on the value of the airline shares. The impact of the attacks may have varied by airline (Carter and Simkins, 2002) but the pressure on the sector as a whole increased signiﬁcantly due to the attacks. A policy implication from Drakos’ analysis is that state aid as a response to terror is not useful when terror induces signiﬁcant and permanent changes in demand patterns. Raﬁ Eldor and Raﬁ Melnick use daily data to enable a close matching between terror attacks and ﬁnancial-market responses. Their study indicates that the stock and foreign exchange markets in Israel functioned efﬁciently in the face of Palestinian terror. Whereas the effects of the attack of 9/11 on ﬁnancial markets was a one-off event, the attacks directed at the population of Israel were ongoing, with 639 terror attacks in the period 1990 to 2003. The empirical results indicate that the terror did not affect the foreign exchange market but that the stock market was affected, with stock prices internalizing expectations of reduced future proﬁts. The stock market (and also the foreign exchange market) never became desensitized to terror attacks. Eldor and Melnick propose that the evidence regarding the functioning of ﬁnancial markets under continuous conditions of terror in Israel has broader implications for the western world because of Israel’s well-developed ﬁnancial markets. Michael Wolgast discusses in his contribution whether the insurance sector should receive blanket subsidies in the wake of the attacks. Public-private partnerships aimed at providing some degree of insurance for large-scale terror attacks have been instituted in several countries, for example the UK and Germany. These are effectively re-re-insurers that aim to encourage insurance ﬁrms to cover terror risks that otherwise would remain uninsurable within the private sector. From a policy perspective, the main challenge in designing such schemes is to avoid diminishing incentives for private agents to reduce their or their customers’ exposure to terror risks. Volker Nitsch and Dieter Schumacher study the empirical effects of terrorism and other forms of insecurity on international trade. They demonstrate that conﬂict, broadly deﬁned, has signiﬁcant dampening effects on bilateral trade ﬂows. A doubling of terror incidents reduces bilateral trade by four per cent. The study by Nitsch and Schumacher raises the question whether policies can counteract the negative effects of terrorism on transaction costs, for example through enhanced international technical cooperation between customs ofﬁcials and police forces. Sanjeev Gupta, Benedict Clements, Rina Bhattacharya and Shamit Chakravarti provide empirical evidence on the ﬁscal effects of terrorism in low- and middleincome countries. Their focus is on the effects of terrorism on government spending and revenue, and thereby on growth. The empirical results conﬁrm that terrorism has signiﬁcant ﬁscal effects and both direct and indirect effects on growth in a range of countries. Gupta et al. make an important contribution in identifying and

8

Tilman Brück

quantifying the ﬁscal transmission mechanism for terrorism. An important role for aid conditionality is also identiﬁed, to avoid situations where governments in poorer countries permit terrorism to exist in order to reap ﬁnancial gains from the ﬁght of terrorism.

4.

Anti-terrorist policies

The next section of the volume assesses some key policy issues in the ﬁght against terrorism from an economic point of view. Dennis Mueller, for instance, addresses the policy implications of 9/11 by asking whether global terrorism should lead to a re-evaluation of how to design constitutions, how to award citizenship, and how to protect property rights. Mueller addresses the social dilemmas (or the tradeoff between the costs and beneﬁts of constitutional rights) that western liberal societies face in seeking to adhere to liberal values while at the same time protecting citizens, residents, and future citizens from terror. The policy ﬁelds affected by these dilemmas are broad, and include domestic civil rights, immigration and education policies, the regulation of religions, and the granting of citizenship. Mueller concludes that terror signiﬁcantly inﬂuences the balance of rights deﬁned in a constitution. Mueller’s premise is that global terrorism seriously challenges the democratic constitutional state. Yet 9/11 might not represent a failure of the western system from within. That might have been more appropriate in the case of the left-wing terrorism of 1970s Europe. While death, injury and capital destruction are the most visible effects of a terrorist attack, fear and the indirect effects of terror can be more harmful to the economy in the longer term. Bruno Frey and Simon Luechinger argue that terrorists are intent on causing such fear. This is particularly true for the consequences of terror attacks in a centralized economy. Their chapter hence asks whether a concerted anti-terrorist policy by a centralized government creates more fear than a less intense anti-terror policy would. The former case would yield the paradoxical result that anti-terror measures raise fear of terrorism further. Such a scenario places governments in a dilemma where both responding and not responding to terrorism plays into the terrorists’ hands. An implication of the argument made by Frey and Luechinger is, for example, that centralizing political decision-making in the European Union could attract terror attacks. Michelle Garﬁnkel proposes a model in which a terror threat has two effects on the domestic economy. On the one hand, a lower sense of domestic security reduces the value of the gross domestic product, which in turn reduces the intensity of the ﬁght for the control of the state. On the other hand, if government policy succeeds in reducing the scale or the effects of terrorism, the value of capturing the state rises and the domestic struggle for power intensiﬁes. In practice, both effects were observed sequentially in domestic US politics after 9/11. Politicians ﬁrst rallied around the ﬂag before eventually resuming and even intensifying the domestic political struggle. Garﬁnkel’s analysis raises some important questions. First, does the scale of the two effects differ according to the type of conﬂict and the type of democracy?

A brief survey of the economic analysis of terrorism

9

One or the other of the effects dominating might explain why some democracies appear to be in a high-conﬂict equilibrium (as Colombia, described by Brauer, Gómez-Sorzano and Sethuraman in this volume), while other democracies enjoy a low level of conﬂict (such as Switzerland, perhaps for reasons described by Frey and Lüchinger in this volume). Second, are terrorists aware of their impact on the domestic struggle for power and could this awareness be manipulated by policy makers to reduce the intensity of conﬂict? The second option would then open a strategic interaction between the domestic policy makers and the terrorists, affecting income in the domestic economy and hence the intensity of the domestic power struggle, and thus the probability of political survival of the domestic policy maker. Valpy FitzGerald investigates the international ﬁnancial transactions supporting global terrorism and how policy can detect and undermine terror-related ﬁnancial ﬂows. He identiﬁes obstacles to unilateral and multilateral policy initiatives in this ﬁeld. In particular, current regulatory systems do not succeed in excluding suspected groups or individuals from undertaking transactions or the transactions are reported too late for effective intervention to take place. Rather than adapting anti-money-laundering institutions to the task of combating terrorist ﬁnance, FitzGerald suggests policies of disincentives for undertaking terrorist ﬁnancial transactions and improving systems for channelling migrant remittances. In the last chapter of this section, Tilman Brück analyses security policy from an economic perspective. He discusses the role of public goods for national and global security and identiﬁes the importance of the ﬁrst- and second-order indirect effects of insecurity on economic activity, which include the behavioural responses of agents and the government to security measures, akin to such effects in insurance economics. Furthermore, key public policy trade-offs are outlined, in particular between security and efﬁciency, globalisation, equity and freedom. The chapter identiﬁes suitable policy options for raising security in the national and international contexts and in view of these trade-offs. Brück calls for a suitable balance between market and non-market instruments in achieving security to minimize the adverse effects of aiming for higher security. In addition, he emphasises the importance of the public good nature of global security, which implies that the international coordination of security policies is important, despite that process being fraught with enforcement problems.

5.

Epilogue

Together the chapters in this volume provide an introduction to and an overview of the state of the art in the economic analysis of terrorism. The chapters indicate how terrorism functions economically, how economies are affected by terror and how policies can minimize the economic costs of terrorism and maximize the defence against terror attacks. In a critical and thought-provoking epilogue to this volume, Geoffrey Brennan cautions to give too much attention to the phenomenon of terrorism. In particular, he recommends to shift the attention away from the motivations, incentives

10

Tilman Brück

and likely behaviour of terrorists and towards the mechanisms of response that effectively mobilize the terror. He also draws attention to the limited ability of most economic agents to assess correctly the expected costs of rare but extreme events, which provokes the unanswered counter-question of how terrorists are able to assess their impact in such a limited rationality framework. In his chapter, Brennan warns that the combination of imperfectly perceived terror and volatile electoral politics may yield sub-optimal or even counter-intuitive policy responses. In doing so, Brennan identiﬁes implicitly an important ﬁeld for further research – the economic analysis of the interaction of terrorism, politics and group mobilization in both Western and Muslim societies.

References Abadie, A. and Gardeazabal, J. (2003) ‘The economic costs of conﬂict: a case study of the Basque Country’, American Economic Review, 93: 113–32. Addison, T. (ed.) (2003) From Conﬂict to Recovery in Africa, Oxford: Oxford University Press. Baily, M. N. (2001) Economic policy following the terrorist attacks, Washington, DC: Institute for International Economics. Carter, D. A. and Simkins, B. J. (2002) ‘Do markets react rationally?: The effect of the 11 September tragedy on airline stock returns’, Working Paper 26, Department of Finance, College of Business Administration, Oklahoma State University, Stillwater. Fleischer, A. and Buccola, S. (2002) ‘War, terror, and the tourism market in Israel’, Applied Economics, 34: 1335–43. Lenain, P., Bonturi, M. and Koen, V. (2002) ‘The economic consequences of terrorism’, OECD Working Paper: 334, Paris: Organization for Economic Cooperation and Development. Stewart, F. and FitzGerald, E. V. K. (eds) (2001) War and Underdevelopment, Oxford: Oxford University Press. Strauß, H. (2001) ‘Assessing the effects of the terrorist attacks on the U.S. economy’, Working Paper 17, Kiel: Institute of World Economics. Walkenhorst, P. and Dihel, N. (2002) ‘Trade impacts of the terrorist attacks of 11 September 2001: a quantitative assessment’, International Conference on the Economic Consequences of the New Global Terrorism, 14 and 15 June, Berlin: German Institute for Economic Research (DIW Berlin).

Part II

Features of terrorism

2

An economic perspective on transnational terrorism Todd Sandler and Walter Enders

This paper indicates how economic analysis can be applied for enlightened policy making with respect to transnational terrorism. Both theoretical tools (e.g. game theory and utility-maximizing models) and empirical techniques (e.g. time series and spectral analysis) are used to put modern-day terrorism into perspective and to suggest policy responses. From hostage negotiations to the installation of technological barriers (e.g., metal detectors, embassy fortiﬁcation), economic methods are shown to provide policy insights. Transnational terrorism and efforts to address it are shown to involve transnational externalities and market failures. Strategic interactions abound in the study of transnational terrorism.

1.

Introduction

Economic methods – both theoretical and empirical – have been applied by a small group of economists to understand a host of issues associated with terrorism. These issues concern the policy effectiveness of alternative responses (e.g. toughening punishments, retaliatory raids, installing technological barriers), negotiation responses in hostage incidents, the terrorists’ choice of target, the economic impacts of terrorism, and others. Terrorism is the premeditated use, or threat of use, of extranormal violence to obtain a political objective through intimidation or fear directed at a large audience. An essential aspect of this deﬁnition concerns the presence of a political objective (e.g. getting the United States out of the Persian Gulf states) that the terrorist acts or campaigns are designed to achieve. Another crucial ingredient is the use of extranormal violence or brutality to capture news headlines. As the public becomes numb to their acts of violence, terrorists respond with more ghastly actions to recapture media attention. Thus, the escalation experienced on 11 September 2001 (henceforth, 9/11) came as no surprise to those who study terrorism. In a deliberate attempt to create a general atmosphere of fear, terrorists strike at a variety of targets with alternative modes of operations, thus making it difﬁcult for the authorities to anticipate the venue of the next incident. Such actions make attacks appear to be random, so that a targeted society must expend large amounts of resources to protect a wide range of vulnerabilities. This simulated

14

Todd Sandler and Walter Enders

randomness provides terrorists with a cost advantage over the stronger authorities who must defend against the threat that they pose (Hirshleifer, 1991). Because people tend to overrespond to unlikely catastrophic events while ignoring more likely daily dangers, terrorists succeed in achieving society-wide anxiety with a minimal amount of resources. When a terrorist incident in one country involves victims, targets, institutions, governments, or citizens of another country, terrorism assumes a transnational character. In the World Trade Center tragedy, citizens from over 80 countries lost their lives at the hands of terrorists who crossed into the United States from abroad. Obviously, the four hijackings on 9/11 constitute transnational terrorist attacks. Transnational terrorist incidents are transboundary externalities, insofar as actions conducted by terrorists or authorities in one country may impose uncompensated costs or beneﬁts on people or property of another country. Economic methodology is particularly well suited to provide insights in studying terrorism. Economic analysis can account for the strategic interactions among opposing interests – e.g. the terrorists and the authorities, or between two targeted countries. Rational-choice models, based on microeconomic principles, can be applied to ascertain how terrorists are apt to respond to policy-induced changes to their constraints. The same methods can be used to analyse how governments react to terrorist-induced changes to their policy-making environment. Moreover, the theory of market failures can underscore how agents’ independent optimization may be at odds with socially efﬁcient outcomes, so that governmental failures may result from well-intentioned policies. In addition, various economic empirical methods can be applied to evaluate theoretical predictions and policy recommendations. The primary purpose of this paper is to survey some essential insights gained from applying an economic perspective to a political problem – e.g. how antiterrorist actions can create unintended consequences. A second purpose is to present some new analysis of trends and cycles, policy-induced externalities, and building a coalition against terrorists. A third purpose is to identify some future research issues.

2.

A look at the data

To provide a perspective of the nature of the transnational terrorist threat, we compile Table 2.1 based on data from the US Department of State (1988–2002). This table indicates the annual number of transnational terrorist events, the associated deaths, the number of wounded, and the number of attacks against US people and/or property. A number of essential facts can be drawn from these numbers. First, transnational terrorism on average results in relatively few deaths, especially when compared with the annual 40,000 people killed on US highways, so that the events on 9/11 are clear outliers. In fact, the deaths on this single day are approximately equal to all transnational terrorist-related deaths recorded during the entire 1988–2000 period. Second, transnational terrorism appears to follow a cyclical pattern with much of the 1990s being a relatively calm era. Something that cannot

An economic perspective on transnational terrorism

15

Table 2.1 Transnational terrorism: events 1968–2001 Year

Number of events

Deaths

Wounded

Attacks on US interests

2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 1979 1978 1977 1976 1975 1974 1973 1972 1971 1970 1969 1968

348 423 392 273 304 296 440 322 431 363 565 437 375 605 665 612 635 565 497 487 489 499 434 530 419 457 382 394 345 558 264 309 193 125

3,572 405 233 741 221 314 163 314 109 93 102 200 193 407 612 604 825 312 637 128 168 507 697 435 230 409 266 311 121 151 36 127 56 34

612a 791 706 5,952 693 2,652 6,291 663 1,393 636 233 675 397 1,131 2,272 1,717 1,217 967 1,267 755 804 1,062 542 629 404 806 516 879 199 390 225 209 190 207

219 200 169 111 123 73 90 66 88 142 308 197 193 185 149 204 170 133 199 208 159 169 157 215 158 164 139 151 152 177 190 202 110 57

a Data on the number wounded in the World Trade Center attack is not available and, thus,

is not part of this ﬁgure. Source: US Department of State, Patterns of Global Terrorism (1988–2002) and tables provided to Todd Sandler in 1988 by the US Department of State, Ofﬁce of the Ambassador at Large for Counterterrorism.

be seen from Table 2.1 is that a high proportion of total casualties for a given year is typically associated with a couple of ‘spectacular’ events – e.g. the simultaneous bombings of the US Embassies in Nairobi, Kenya and Dar es Salaam, Tanzania account for 291 deaths and almost 5,000 injuries in 1998 (US Department of State, 1999). Third, attacks against US interests represent a relatively high proportion of events. This is particularly noteworthy from an externality viewpoint, because

16

Todd Sandler and Walter Enders

relatively few incidents take place on US soil – in 1998 and 2000, there were no such events, while, in 1999, there was just one such event (US Department of State, 1999–2001). By having relatively secure borders, the United States must rely on foreign governments to protect US citizens and property while abroad. Except for some annual totals, government-collected data sets have not been made available to researchers. Mickolus (1982) developed a data set, International Terrorism: Attributes of Terrorist Events (ITERATE) for 1968–77. This incidentbased data set was extended to cover 1978–87 and 1988–91 by Mickolus et al. (1989, 1993). More recently, Mickolus and Fleming (2003) have updated the data through 2002. ITERATE uses a host of sources for its information, including the Associated Press, United Press International, Reuters tickers, the Foreign Broadcast Information Service (FBIS) Daily Reports, and major US newspapers. ITERATE poses a number of shortcomings that researchers must take into account when testing theories. By relying on newspaper accounts, ITERATE is better at chronicling the actions of terrorists than in recording those of the authorities. Because ITERATE is an events data set, researchers must rely on event counts rather than on continuous measures of intensity unless casualty counts are used (Enders and Sandler, 2000, 2002). ITERATE picks up newsworthy transnational terrorist incidents, so that there is some bias, which must be recognized. Despite these difﬁculties, ITERATE is suited to a wide range of empirical tasks. For example, it can display trends and cycles for newsworthy events for forecasting purposes. The data have even been used to investigate terrorist and government bargaining behaviour in hostage-taking events by Atkinson et al. (1987), who apply a time-to-failure model, where the length of an incident is related to choice variables of the adversaries. Based on ITERATE data, we display two quarterly time series – all transnational incidents and bombings – in Figure 2.1 for 1970–2000, where transnational terrorism displays peaks and troughs. Bombings are the favourite mode of operation of terrorists, accounting for about half of all transnational terrorist incidents on average in any given year. Additionally, the bombing time series tracks the all-incident series rather well. The latter half of the 1990s represents a downturn in transnational terrorism due, in large part, to fewer states sponsoring terrorism in the post-Cold War era (Enders and Sandler, 1999). Terrorist experts have documented a change in the makeup and motivation of the general perpetrators of terrorism since the takeover of the US Embassy in Tehran in November 1979 (Hoffman, 1998). From the late 1960s until the latter 1980s, transnational terrorism has been primarily motivated by nationalism, separatism, Marxist ideology, and nihilism (Wilkinson, 1986). In the 1990s, the motivation of terrorism changed with ‘the emergence of either obscure, idiosyncratic millennium movements’ or religious-based fundamentalist groups (Hoffman, 1997, p. 2). Since the beginning of 1980, the number of religious-based groups has increased as a proportion of the active terrorist groups: 2 of 64 groups in 1980, 11 of 48 groups in 1992, 16 of 49 groups in 1994, and 25 of 58 groups in 1995 (Hoffman, 1997, p. 3). Enders and Sandler (2000) show that a signiﬁcant rise in casualties from transnational incidents can be traced back to the takeover of the US Embassy

An economic perspective on transnational terrorism

17

Figure 2.1 All incidents and bombings.

in Tehran, as speculated by Hoffman (1998). In recent years, an incident is almost 17 percentage points more likely to result in death or injury compared with the earlier eras of leftist terrorism. Judging by the public’s and media’s reaction to 9/11, one might conclude that international terrorism is on the rise, but Figure 2.1 indicates just the opposite trend. This misperception may be due to the increasing likelihood of an incident resulting in casualties, making incidents on average more newsworthy. The standard procedure for ascertaining the form of a deterministic trend is by ﬁtting a polynomial in time (t), where additional trend terms (i.e. t, t 2 , t 3 ) are added until the associated coefﬁcient is no longer statistically signiﬁcant. For 1968–2000, we investigate trends for six quarterly time series extracted from ITERATE: hostage taking, bombings (of all types), threats and hoaxes, assassinations, incidents with casualties, and all transnational terrorist incidents. Table 2.2 indicates new polynomial trend estimates for these six quarterly series (where time = t), all of which are characterized by a nonlinear trend. The t-ratios associated with the coefﬁcient estimates are indicated in parentheses beneath the constant and the time trend terms. Five of the six series are represented by a quadratic trend with a negative coefﬁcient for the squared time term. This characterization reﬂects the fact that series tended to rise in the late 1960s and to decline in the late 1990s. Only the threats and hoaxes series is represented by a more complicated cubic trend; nevertheless, this series also displays a similar inverted U-shaped pattern.

18

Todd Sandler and Walter Enders

Table 2.2 Trend and other statistical properties of transnational terrorist incidents Incident Type Hostage taking Bombings

Constant a Time

5.901 (3.832) 34.449 (4.442) Threats & 8.595 Hoaxes (2.540) Assassinations −1.521 (−1.229) Casualties 9.726 (2.441) All Events 41.689 (4.270)

0.219 (4.093) 1.139 (4.230) −0.276 (−1.256) 0.400 (9.299) 0.579 (4.497) 2.435 (7.185)

(Time)2

(Time)3

−0.001 (−3.202) −0.010 (−5.021) 0.010 −0.000 (2.572) (−3.340) −0.003 (−8.830) −0.004 (−4.635) −0.019 (−7.743)

F-stat b Variance Percent c 13.11 32.223 [0.000] 15.87 842.470 [0.000] 11.34 87.170 [0.000] 43.49 21.472 [0.000] 10.74 119.479 [0.000] 30.79 1335.560 [0.000]

0.278 0.314 0.247 0.411 0.527 0.252

a t-ratios are in parentheses. b Prob. values are in brackets under the F-statistics. c Proportion of variance of the detrended, ﬁtted-polynomial series that is accounted for by the lowest 15 per cent of the frequencies (i.e. the longest cycles).

In Table 2.2, the third column from the right reports the F-statistics and their ‘prob’ values in brackets, representing the statistical signiﬁcance of the overall regression. These signiﬁcance levels are all zero to three digits, which are strongly supportive of the ﬁtted nonlinear trend equations. Such ﬁtted trends are not useful for very long-term forecasting, because there is little reason to believe that the number of incidents will continue to decline. Instead, the ﬁt of the nonlinear trend cautions against simple statements about a decidedly upward or downward trend to any form of international terrorism.1 Such proclamations are common in the media. The trend analysis suggests that there is persistence in each of the incident series – high and low levels of terrorism come in waves or cycles. Shocks to any incident series are not permanent, so that there is a reversion toward a long-run mean. Cycles in terrorism data have been attributable to a number of factors. Alexander and Pluchinsky (1992) explain ﬂuctuations in terrorism using demonstration and copycat effects. Heightened public sensitivity following a successful terrorist attack induces other terrorists to strike when media reaction is likely to be great. The anthrax attacks following 9/11 correspond to this pattern. Economies of scale in planning terrorist incidents by terrorist groups or networks may also lead to the bunching of attacks. Cycles may also stem from the attack-counterattack process between the terrorists and authorities. Public opinion following a spate of attacks can prompt governments’ periodic crackdowns that temporarily create a lull in transnational terrorism. These downturns are subsequently followed by countermeasures and recruitment by the terrorists as they prepare for a new offensive. Enders and Sandler (1999) and Enders et al. (1992) argue that logistically complex events such as skyjackings, large suicide car bombings, and assassinations

An economic perspective on transnational terrorism

19

will have longer cycles than less sophisticated events as the attack-counterattack interaction among adversaries takes longer. The theory of Fourier series allows a wide class of functions to be expressed in terms of sine and cosine components. To uncover the underlying cycles in a series, a researcher must regress the detrended values of a series on all frequencies in the interval [1, T /2], where T is the number of observations. The frequency of a series indicates how fast the underlying cycle is completed – a low (high) frequency implies a long (short) cycle. A graph depiction of the proportionate variation explained by each frequency (called the periodogram) has large peaks representing the crucial underlying frequencies. Some series with obvious cycles, like sunspots or average daily temperatures, will display a periodogram with a single focal frequency. Given the stochastic behavior of terrorists and the measures applied to curb terrorism, there is unlikely to be one deterministic frequency that dominates the periodicity for any of the six series. Thus, we use a different approach here than trying to identify one particular frequency. Series with long periods will have most of their variance explained by the low frequencies, whereas series with short periods will have most of their variance explained by high frequencies. In accordance with spectral analysis, we detrend each series using the ﬁtted polynomial trends in Table 2.2. The last two columns of Table 2.2 report the total variance of each series and the proportion of this variance accounted for by the lowest 15 per cent of the frequencies.2 We anticipate that the logistically complex incident types will have relatively large amounts of this proportion attributable to the low frequencies. The all-events series has a large variance of 1,335.56 with just 25.2 per cent corresponding to the relatively low frequencies. In marked contrast, the more complex events of assassinations and those involving casualties have smaller variances with more of this variance (41.1 and 52.7 per cent, respectively) attributed to low frequencies. Threats and hoaxes display the greatest evidence of short cycles with just 24.7 per cent of the variance explained by the longest cycles. The variance results for hostage taking and bombing events imply moderately short cycles. Only in the case of hostage taking are our priors not realized.

3.

Game theory and hostage taking

Despite the events of 9/11, hostage taking may still have negotiations, because most such missions involve kidnappings, where the terrorists are not suicidal. Recent skyjackings in Turkey and Cuba during March 2003 demonstrate that not all skyjackings include terrorists bent on mass destruction. Nevertheless, suicide skyjackings and the reactions of desperate passengers to ﬁght back must be analysed in the future along with a government’s decision to destroy a hijacked plane. To date, there have been six economic analyses of hostage-taking events – i.e. Atkinson et al. (1987), Lapan and Sandler (1988), Selten (1988), Islam and Shahin (1989), Sandler and Scott (1987), and Scott (1991). The ﬁrst three studies stress game-theoretic aspects, while the latter three studies do not. We focus our remarks around the Lapan and Sandler (1988) study, the most general of these three game-theoretic studies. The question posed by their investigation is whether

20

Todd Sandler and Walter Enders

or not a stated policy by which a government precommits never to negotiate with hostage takers will have the intended consequence of keeping terrorists from ever taking hostages. The conventional wisdom states that if terrorists know ahead of time that they have nothing to gain that they will never abduct hostages (Selten, 1988). This belief has become one of the four pillars of US policy with respect to addressing transnational terrorism – i.e. ‘make no concessions to terrorists and strike no deals’ (US Department of State, 2001, p. iii). The conventional wisdom for the never-to-capitulate policy hinges on at least four implicit assumptions: (i) (ii) (iii)

the government’s deterrence is sufﬁcient to stop all attacks; the government’s pledge is fully credible to all would-be hostage takers; the terrorists’ gains from hostage taking only derives from the fulﬁlment of their demands; and (iv) there is no uncertainty concerning the payoffs.

Each of these assumptions may not hold in practice. Deterrence will not stop all attacks if the terrorists perceive that there is a positive expected payoff from taking hostages. Past concessions limit the credibility of the government’s noconcession pledge (Sandler et al. 1983). Moreover, a fanatical terrorist group may gain positive beneﬁts from failure in terms of heavenly rewards or media exposure. Given that the costs to the government once hostages are secured are unknown beforehand, the government’s pledge may be time inconsistent, so that the government reneges for the ‘right’ hostages (e.g. the Israeli release of 1,150 Arab prisoners in a negotiated swap for three Israeli soldiers in May 1985) if the costs of capitulating is less than holding ﬁrm.3 Thus, the pitfalls to this policy are easily laid bare by game theory. Lapan and Sandler (1988) make their analysis more realistic by allowing multiple periods and reputation costs. Concessions result in a loss in reputation that is costly in terms of more hostage taking in the future. As reputation costs increase, the no-negotiation policy is less apt to be time inconsistent. This outcome can be fostered through rules – e.g. a constitutional amendment that imposes sufﬁciently severe punishments to eliminate any discretion of government negotiators. The game can be made still more realistic by including additional sources of uncertainty in terms of the terrorists’ payoffs. Hostage-taking incidents involve asymmetric information and uncertainty on the part of both terrorists and governments.4 The beauty of game theory is that it permits the evaluation of policies while accounting for uncertainty and strategic interactions of opposing interests, so that easy ﬁxes may not be so straightforward.

4.

Game theory and governmental responses

External costs are present when deterrence at home displaces the attack abroad, while external beneﬁts are relevant when deterrence at home either protects foreigners or reduces the level of attacks globally. Depending on the opposing

An economic perspective on transnational terrorism

21

external effects, and there may be others not listed, there may result too much or too little deterrence (Sandler and Lapan, 1988). The overdeterrence/underdeterrence problem is heightened when a terrorist network (e.g. al-Qaida) operates in upwards of 60 countries and stages their attacks worldwide (US Department of State, 2001). Underdeterrence is particularly acute in countries sympathetic to a group’s grievances when the group focuses their attack on foreigners. As the number of potential targets increase, transference efforts may be especially large. By forming a global network, terrorists limit the effectiveness of countries’ efforts to thwart terrorism as externalities are maximized through countries’ uncoordinated decisions. Terrorists will naturally seek out the weakest link – i.e. the country with the least security – for the venue for their next attack and will dispatch their best team. To address these weaknesses, prime targets, such as the United States, have instituted programs to assist such weakest-link countries in bolstering their counterterrorist capabilities. In fact, this assistance is another of the four pillars of US antiterrorism policy (US Department of State, 2001). Ironically, US efforts to induce other countries to secure their airports and public places make the United States a more attractive target, as 9/11 sadly demonstrated. If the terrorist networking advantage is to be countered, then targeted nations must learn to coordinate their own efforts at counterterrorism. This poses a special problem because nations resist sacriﬁcing their autonomy over security matters to a supranational collective. With this in mind, terrorist experts have often called for piecemeal policy where intelligence is shared but not deterrence decisions (e.g. Kupperman, 1987, p. 577). Such piecemeal responses may be inadvisable when the strategic incentives are taken into account. Suppose that a terrorist network targets three countries, each of which are engaged in overdeterrence to transfer the attack abroad. Further suppose that intelligence allows the targeted countries to better judge the marginal effectiveness of diverting attacks by revealing the terrorists’ preference for alternative targets. As these nations acquire this information, they become better adept at diverting attacks, thereby augmenting the negative transference externality. The net impact of this information sharing may be to heighten the ‘transference race’ without providing more security against a group bent on attacking, so that the added deterrence costs simply make the three countries worse off. This results in a second-best outcome where the change in one policy parameter (i.e. increased information sharing), which would, under full cooperation, improve efﬁciency, may worsen inefﬁciency when a second policy (i.e. coordination of deterrence) is not chosen optimally. A similar second-best scenario may characterize other partial responses – e.g. greater actions to apprehend terrorists without coordinating efforts to increase punishments. Thus, the application of game theory again raises policy concerns previously ignored in the terrorism literature.

5.

Building a coalition against terrorists after 9/11

Actions to coordinate retaliation against either a terrorist organization or a statesponsor of terrorism has typically been characterized as a Prisoners’ Dilemma

22

Todd Sandler and Walter Enders

Figure 2.2 Ordinal game matrix for retaliation.

(e.g. Lee, 1988) with all countries playing their dominant strategy to sit back and do nothing. The forging of an alliance to wage war on terrorism in Afghanistan after 9/11 appears to abide by a different underlying game form than the Prisoners’ Dilemma for select countries that have participated in the retaliatory response against the Taliban and Osama bin Laden. We shall focus on the two most ardent participants – the United States and the United Kingdom. In Figure 2.2, we represent an underlying retaliation game in ordinal form, where payoffs are rank ordered from highest (4) to lowest (1). The payoffs for the row player – the US – are listed ﬁrst, followed by those of the column player – the UK – in each of the four strategic combinations. The ordinal payoffs displayed indicate that the highest payoffs come from these two countries jointly retaliating, followed by the next-largest payoff for free riding when the other country retaliates. The worst payoff corresponds to retaliating on one’s own, followed by the second-worst payoff when neither country retaliates. This game differs from the standard Prisoners’ Dilemma by having the ordinal payoffs of the 3s and 4s switched. That is, the heinous nature of the 9/11 attacks and its human toll on American and British citizens at the World Trade Center increased the ordinal payoff for joint retaliation and decrease this payoff from free riding, as compared with earlier terrorist incidents, including the downing of Pan Am Flight 103 over Lockerbie, Scotland on 21 December 1988.5 For the assurance game displayed in Figure 2.2, there is no dominant strategy that gives higher payoffs no matter what the other country does, but there are two pure-strategy Nash equilibriums, where either both countries retaliate or both do nothing. In this scenario, an alliance can be forged provided that one country leads and begins to retaliate, which was the role that the United States assumed.

6.

Rational-choice representations

Beginning with the Landes (1978) study of skyjackings, economists characterize terrorists as rational actors who maximize expected utility or net payoffs subject

An economic perspective on transnational terrorism

23

to constraints. Arguments in these constraints may consist of terrorists’ resource endowments or actions taken by the authorities to thwart terrorism. Using data on US hijackings for 1961–76, Landes demonstrates that greater prison sentences and enhanced likelihood of apprehension are signiﬁcant deterrents. He also indicates that the installation of metal detectors on 5 January 1973 led to between 41 and 50 fewer hijackings in the United States during 1973–6. In a subsequent analysis, Enders and Sandler (1993) examine a wide range of policy interventions, including metal detectors, fortiﬁcation of embassies, retaliatory raids, and the Reagan ‘get-tough-on-terrorists’ laws. The theoretical model for the terrorists that underlies their study is analogous to the consumer-choice model. Terrorists maximize utility or expected utility derived from the consumption of basic commodities, produced from terrorist and nonterrorist activities. For example, al-Qaida terrorists may gain utility from a reduced political resolve on the part of the United States to remain in the Persian Gulf as Americans lose their lives in terrorist attacks. This weakening of US resolve is the basic commodity that can be produced with a number of alternative attack modes. Substitution possibilities among terrorist tactics arise when alternative modes of operations produce the same basic commodities (e.g. political instability, media attention) in varying amounts. Substitution is enhanced when attack modes possess closely related outcomes and are logistically similar. This is clearly the case for hijackings and other kinds of hostage events. Complementarity results when combinations of attack modes are required to produce one or more basic commodities. When threats follow real attacks, both actions assume a heightened effectiveness and are then complementary. To produce these basic commodities, a terrorist group must choose between nonterrorist and terrorist activities, while being constrained by resources. In the latter choice, terrorists must further choose between different modes of terrorist attacks based on the perceived ‘prices’ associated with alternative operations. Choices are many and include the intended lethality of the act, its country of location, and whom or what to target. The expenditure on any activity consists of the activity’s per-unit price times the activity’s level. Each mode of operation has a per-unit price that includes the value of time, resources, and anticipated risk to accomplish the act. The securing and maintenance of a kidnapping victim in a hidden location is logistically more complex and requires more resources than leaving a small bomb in a trash bin in a railway station, so that the former has a greater per-unit price. In choosing a venue, the price is anticipated to differ based on security measures taken by the authorities, so that a country with more porous borders will be the staging ground for attacks against targets originating from other more secure countries. The prices confronting the terrorists for each tactic are determined, in large part, by the government’s allocation of resources to thwart various acts of terrorism. Government policies aimed at a single type of terrorist event adversely change its relative price and results in a substitution into now less expensive modes of attack. Thus, Landes’ (1978) measure of the success of metal detectors, in terms of fewer skyjackings, does not go far enough, because the

24

Todd Sandler and Walter Enders

application of this technology may have induced a large number of other kinds of events. To account for these substitutions, Enders and Sandler (1993) apply vector autoregression (VAR) analysis to allow for the potential interactions among various terrorist time series in response to government policies. They ﬁnd that metal detectors decreased skyjackings and threats, but increased other kinds of hostage incidents, not protected by detectors. The trade-off between events was about one for one (also see Enders et al. of 1990; Im et al. of 1987). Both substitutions and complementarities are uncovered. Fortiﬁcation of US embassies and missions reduced attacks against such installations, but were tied to a disturbing increase in assassinations of ofﬁcials and military personnel outside of protected compounds. In addition, Enders and Sandler (1993) establish that the US retaliatory raid against Libya on April 1986 (for its suspected involvement in La Belle Discothèque in West Berlin on 4 April 1986) was associated with an immediate increase in terrorist attacks against US and UK interests. This increase was shortly followed by a temporary lull as terrorists built up depleted resources. Apparently, the raid caused terrorists to intertemporally substitute attacks planned for the future into the present to protest the retaliation. Within a relatively few quarters, terrorist attacks resumed the same mean number of events.6 There are a number of ways to institute antiterrorist policies that address these likely substitutions and complementarities. First, the government must make the terrorists substitute into less harmful events. Second, the government must go after the terrorists’ resource endowment (i.e. its ﬁnances, its leadership, its membership) if an overall decrease in terrorism is to follow. Efforts to inﬁltrate groups or to freeze terrorist ﬁnances have this consequence. Third, the government must simultaneously target a wide range of terrorist attack modes, so that the overall rise in the prices of terrorist attacks becomes analogous to a decrease in resources. Success in raising the price of all modes of terrorist attacks would induce terrorists to shift into legal protests and other nonterrorist actions to air grievances. A reliance on technological barriers merely causes a substitution into other attack modes in the short run. In the long term, terrorists will develop ingenious countermeasures (i.e. plastic guns, bottles of ﬂammable liquid) to circumvent the technology. Thus, there is a dynamic strategic interaction present, where authorities must be ever vigilant to be improving the technology by anticipating ways of circumventing such barriers. Unfortunately, authorities have been reactive in practice by only responding after a technological barrier’s weakness has been exploited, leaving the public vulnerable until a new technological ﬁx is found and installed. Substitution effects abound in the study of terrorism and involve not only actions of the terrorists, but also those of the targets. For targets, the economic literature addresses two kinds of substitutions. First, there are studies that examine the tourists’ choice of vacation spot based on the perceived threat of terrorism and other costs. An alteration in travel risks, arising from increased terrorist incidents in a country, raises the price of a holiday there in comparison to other vacation venues not confronted with terrorism. In a study of Spain, Enders and Sandler (1991) employ VAR analysis to demonstrate that a typical transnational

An economic perspective on transnational terrorism

25

terrorist incident is estimated as scaring away just over 140,000 tourists when all monthly impacts are combined. Companion studies by Enders et al. (1992) and Drakos and Kutan (2001) establish and quantify terrorism-induced substitutions in tourism for Greece, Austria, Italy, Turkey, Israel, and other terrorism-ridden countries. Second, target-based substitutions involve foreign direct investment (FDI). Investors decide where to invest based on their perceived economic risks, political risks, and monetary returns. An increase in transnational terrorism directed at FDI (e.g. attacks on Euskadi ta Askatasuna (ETA) in the Basque region of Spain) is sure to divert such investment. Enders and Sandler (1996) show that an ‘average’ year’s worth of terrorism reduced net FDI in Spain and Greece by 13.5 per cent and 11.9 per cent annually. Transnational terrorism displayed signiﬁcant economic cost, not counting the billions spent on barriers and deterrence.

7.

Toward a beneﬁt-cost analysis of terrorist-thwarting policies

As a future research project, economists should assess the beneﬁts and cost of speciﬁc policies to thwart terrorism. Such an exercise has not been adequately done and poses some real challenges. Costs are fairly straightforward since ﬁgures are available in, say, the United States as to what is paid to fortify embassies and missions, or to guard airports. Consider the cost associated with airport security. To the cost of guards and screening equipment must be added the value of lost time as travellers are screened. On the beneﬁt side, calculations are less transparent. One way to estimate a portion of this beneﬁt would be to compute the reduced loss of life attributable to airport security measures – i.e. fewer people killed in skyjackings. If the net number of such lives saved, after adjusting for substitutions into other life-threatening terrorist actions, can be measured, then the average ‘value of a statistical life’ can be applied to translate these lives into a monetary ﬁgure. To this ﬁgure, a researcher must also compute and add the reduced losses in property values (i.e. from destroyed planes) attributable to the fewer hijackings. In addition, a portion of the value of net air travel revenues must be considered as a beneﬁt arising from a heightened sense of security stemming from security upgrades. The events of 9/11 clearly underscore that there is a cost to a breach in airport security as the public loses its conﬁdence in air travel. Any of these components are fraught with measurement difﬁculties, because there may be other intervening factors at work – e.g. air travel was already in a slump prior to 9/11. Every policy to thwart terrorism would entail its own stream of beneﬁts and costs. Invariably, the beneﬁt calculations are problematic. The US-led retaliation against al-Qaida and the Taliban in Afghanistan has well-deﬁned costs in terms of deployed soldiers, ordnance, diplomacy, and side payments to ‘allies’. But the true savings or beneﬁts from fewer future acts of terrorism, in terms of lives and property saved, is so much more difﬁcult to calculate. Time-series techniques, engineered by Enders and Sandler (1991, 1996) to measure losses to tourism or to

26

Todd Sandler and Walter Enders

FDI from terrorism, can be utilized following the retaliation to roughly estimate the decline in terrorist incidents and their economic value.

8.

Concluding remarks

Although economic methods have enlightened the public on a number of issues concerning transnational terrorism, there are many other issues to analyse. For instance, there is a need for applying more dynamic game methods – i.e. differential game theory – if the waxing and waning of terrorist organizations are to be understood. The terrorists try to increase their organization’s size through enhanced resources, successful operations, and recruitment, while the government tries to limit the group’s size through raids, intelligence, group inﬁltration, and actions to create failed missions. This dynamic strategic interaction needs to be modelled and empirically tested. In addition, researchers must better assess the role of information and intelligence on behalf of the terrorists and the authorities. Given how little governments really know about the strength of the terrorists that they confront – e.g. the US government has almost no clue about the size of al-Qaida prior to 9/11,7 asymmetric information characterizes efforts to thwart terrorism. Similarly, the terrorists are ill-informed about the resolve of the government and the amount of resources that it is willing to assign to curbing terrorism. Additionally, there is a need to model terrorist campaigns – i.e. the choice of the sequence and composition of attacks used by terrorists. As researchers better understand these choices, more effective policy responses can be devised that adjust for the strategic interaction.

Acknowledgements Walter Enders is the Lee Bidgood Chair of Economics; Todd Sandler is the Vibhooti Shukla Professor of Economics and Political Economy. The authors have beneﬁted from the helpful comments of Daniel G. Arce M., Arye Hillman, and two anonymous referees.

Endnotes 1 In fact, there is reasonable evidence to support the claim that each of the incident series is stationary. Using an augmented Dickey-Fuller unit-root test, we can reject the null hypothesis of a unit-root in all series, but that of threats and hoaxes, at the .05 level. For this latter series, we can reject the null of a unit-root at the .10 level. 2 We report the proportion of the variance explained by the frequencies in the interval [1, 0.15 ×T /2]. Since we are somewhat sceptical of the ﬁtted polynomial trends, we also obtained results using only demeaned data. These results are very similar to those discussed below. 3 The Arab prisoners released included Kozo Okomato, a Japanese Red Army Faction member, who was the sole surviving terrorist in the Lod Airport massacre of 1972, which left 26 people dead and 78 injured. 4 On asymmetric information models of terrorism, see Lapan and Sandler (1993) and Overgaard (1994).

An economic perspective on transnational terrorism

27

5 Britain lost the second greatest number of citizens of any country at the World Trade Center. Despite Pan Am 103 ﬂying out of Heathrow Airport and crashing in the United Kingdom (Lockerbie, Scotland), Britain lost relatively few of its citizens in this incident. 6 Analogous results are found in Brophy-Baermann and Conybeare (1994) for retaliations by Israel against Palestinian terrorists. 7 In the Patterns of Global Terrorism prior to 9/11, al-Qaida strengths is given as ‘may have several hundred to several thousand members’ (US Department of State, 2001, p. 69).

References Alexander, Y. and Pluchinsky, D. (1992) Europe’s Red Terrorists: The Fighting Communist Organizations, London: Frank Cass. Atkinson, S.E., Sandler, T. and Tschirhart, J.T. (1987) ‘Terrorism in a bargaining framework’, Journal of Law and Economics, 30: 1–21. Brophy-Baermann, B. and Conybeare, J.A.C. (1994) ‘Retaliating against terrorism: rational expectations and the optimality of rules versus discretion’, American Journal of Political Science, 38: 196–210. Drakos, K. and Kutan, A.M. (2001) ‘Regional effects of terrorism on tourism: evidence from three Mediterranean countries’, paper presented at DIW conference on The Economic Consequences of Global Terrorism, Berlin: German Institute for Economic Research (DIW Berlin). Enders, W., Parise, G.F. and Sandler, T. (1992) ‘A time-series analysis of transnational terrorism: trends and cycles’, Defence Economics, 3: 305–20. Enders, W. and Sandler, T. (1991) ‘Causality between transnational terrorism and tourism: the case of Spain’, Terrorism, 14: 49–58. —— (1993) ‘The effectiveness of anti-terrorism policies: vector-autoregressionintervention analysis’, American Political Science Review, 87: 829–44. —— (1996) ‘Terrorism and foreign direct investment in Spain and Greece’, Kyklos, 49: 331–52. —— (1999) ‘Transnational terrorism in the post-cold war era’, International Studies Quarterly, 43: 145–67. —— (2000) ‘Is transnational terrorism becoming more threatening?’, Journal of Conﬂict Resolution, 44: 307–32. —— (2002) ‘Patterns of transnational terrorism, 1970–99: alternative time series estimates’, International Studies Quarterly, 46: 145–65. Enders, W., Sandler, T. and Cauley, J. (1990) ‘UN conventions, technology and retaliation in the ﬁght against terrorism: an econometric evaluation’, Terrorism and Political Violence, 2: 83–105. Enders, W., Sandler, T. and Parise, G.F. (1992) ‘An econometric analysis of the impact of terrorism on tourism’, Kyklos, 45: 531–54. Hirshleifer, J. (1991) ‘The paradox of power’, Economics and Politics, 3: 177–200. Hoffman, B. (1997) ‘The conﬂuence of international and domestic trends in terrorism’, Terrorism and Political Violence, 9: 1–15. —— (1998) Inside Terrorism, New York: Columbia University Press. Im, E.I., Cauley, J. and Sandler, T. (1987) ‘Cycles and substitutions in terrorist activities: a spectral approach’, Kyklos, 40: 238–55. Islam, M.Q. and Shahin, W.N. (1989) ‘Economic methodology applied to political hostagetaking in light of the Iran-Contra affair’, Southern Economic Journal, 55: 1019–24.

28

Todd Sandler and Walter Enders

Kupperman, R.N. (1987) ‘Vulnerable America’, in Wilkinson, P. and Stewart, A.M. (eds), Contemporary Research on Terrorism, pp. 570–80, Aberdeen, UK: University of Aberdeen Press. Landes, W.M. (1978) ‘An economic study of US aircraft hijackings, 1961–76’, Journal of Law and Economics, 21: 1–31. Lapan, H.E. and Sandler, T. (1988) ‘To bargain or not to bargain: that is the question’, American Economic Review, 78: 16–20. —— (1993) ‘Terrorism and signalling’, European Journal of Political Economy, 9: 383–97. Lee, D.R. (1988) ‘Free riding and paid riding in the ﬁght against terrorism’, American Economic Review, 78: 22–6. Mickolus, E.F. (1982) International Terrorism: Attributes of Terrorist Events 1968–77. (ITERATE 2), Ann Arbor, MI: Inter-University Consortium for Political and Social Research. Mickolus, E.F. and Fleming, P. (2003) International Terrorism: Attributes of Terrorist Events 1992–2002, Dunn Loring, VA: Vinyard Software. Mickolus, E.F., Sandler, T., Murdock, J.M. and Fleming, P. (1989) International Terrorism: Attributes of Terrorist Events, 1978–87. (ITERATE 3), Dunn Loring, VA: Vinyard Software. Mickolus, E.F., Sandler, T., Murdock, J.M. and Fleming, P. (1993) International Terrorism: Attributes of Terrorist Events, 1988–91. (ITERATE 4), Dunn Loring, VA: Vinyard Software. Overgaard, P.B. (1994) ‘Terrorist attacks as a signal of resources’, Journal of Conﬂict Resolution, 38: 452–78. Sandler, T. and Lapan, H.E. (1988) ‘The calculus of dissent: an analysis of terrorists’ choice of targets’, Synthèse, 76: 245–61. Sandler, T. and Scott, J.L. (1987) ‘Terrorist success in hostage-taking incidents’, Journal of Conﬂict Resolution, 31: 35–53. Sandler, T., Tschirhart, J.T. and Cauley, J. (1983) ‘A theoretical analysis of transnational terrorism’, American Political Science Review, 77: 36–54. Scott, J.L. (1991) ‘Reputation building in hostage incidents’, Defence Economics, 2: 209–218. Selten, R. (1988) ‘A simple game model of kidnappings’, in R. Selten (ed.) Models of Strategic Rationality, pp. 77–93, Boston: Kluwer Academic. United States Department of State (various years) Patterns of Global Terrorism. Washington, DC: US Department of State. Wilkinson, P. (1986) Terrorism and the Liberal State, Revised edition, London: Macmillan.

3

Economic conditions and terrorism1 S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

In this paper, we develop and explore the implications of an economic model that links the incidence of terrorism in a country to the economic circumstances facing that country. We brieﬂy sketch out a theory, in the spirit of Tornell (1998), that describes terrorist activities as being initiated by groups that are unhappy with the current economic status quo, yet unable to bring about drastic political and institutional changes that can improve their situation. Such groups with limited access to opportunity may ﬁnd it rational to engage in terrorist activities. The result is then a pattern of reduced economic activity and increased terrorism. In contrast, an alternative environment can emerge where access to economic resources is more abundant and terrorism is reduced. Our empirical results are consistent with the theory. We ﬁnd that for democratic, high income countries, economic contractions (i.e. recessions) can provide the spark for increased probabilities of terrorist activities.

1.

Introduction

In this paper we ﬁrst outline a simple theory that provides structure for thinking about the channels through which economic outcomes can inﬂuence terrorist activities and then examine empirically the presence of these channels by constructing and employing a rich panel data set of 127 countries from 1968 to 1991. Our analysis investigates the importance of standard economic variables such as GDP growth per capita and investment in determining the onset and intensity of terrorist attacks. In doing so, we provide a systematic account of how economic developments inﬂuence terrorism. The theoretical foundations of our paper are based on the model of Tornell (1998). In summary, a no-conﬂict status quo will eventually be disrupted by groups who seek to increase their appropriation and agenda-setting power in the economy. Negative shocks that diminish the growth of an economy’s resource base hasten the incidence of conﬂict. These predictions are consistent with the empirical evidence in Hess and Orphanides (1995, 2001) and Blomberg and Hess (2002), among others, that ﬁnds links between adverse economic outcomes and conﬂict. We extend this latter research by breaking conﬂict down into two types: a ‘rebellion’, in

30

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

which a group seeking to disrupt the status quo overthrows the government and takes power, and a ‘terrorist attack’, a less institutionally disruptive conﬂict type in which a dissident group seeks to undertake terrorist activities to increase their voice in the economy, yet are unable to take over power. The basic prediction of the model is that the choice between a ‘rebellion attack’ and a ‘terrorist attack’ is inﬂuenced by the country’s ability to not give in to the dissident groups. In particular, during bad economic times, economies with well-established institutions and defence capabilities are more likely to be affected by terrorism, whereas economies with weak institutions and defence capabilities are more likely to see civil wars, coups and other conﬂict types designed to overthrow the government. While our theory provides structure for the links between economic weakness and terrorism, we also examine the empirical linkages between the two. We construct a data set of economic and terrorism variables by linking the Summers and Heston (1995) data set to the ITERATE data set. The empirical work estimates and identiﬁes the separate channels by which the economy and terrorism affect one another. We ﬁnd that for democratic, high-income countries economic contractions (i.e. recessions) can provide the spark for increased probabilities of terrorist activities, which in turn raises the probability of recessions in a ‘trap-like’ environment. The structure of this paper is as follows. Section 2 discusses the literature and establishes the context for our paper. Section 3 presents the basic theory and its implications. Section 4 describes the data, and provides preliminary analysis. Section 5 sets out the results from our empirical model. We conclude with Section 6.

2.

Literature summary

We ﬁrst review the seminal research into the determinants of terrorism.2 Grossman (1991) presents a general equilibrium model that treats insurrection and the suppression of insurrection as economic activities willingly undertaken by the participants. The ruler has to trade off higher taxes not only with the lower tax revenue that comes about when people devote less time to productive activities but also with the added cost of having to hire soldiering services to suppress insurrection. Grossman ﬁnds that economies in which the soldiering technology is effective can move themselves to no-conﬂict equilibria by devoting some resources to soldiering and keeping tax rates low. Lapan and Sandler (1993) present an analysis of terrorism as a signalling game in the face of incomplete information. Terrorist attacks are devices through which the two sides learn more about each others’ offensive and defensive capabilities. Lapan and Sandler (1988) examine the extent to which governments should pre-commit themselves to a strategy of never negotiating with terrorists. They show that such a strategy is not likely to be effective when terrorists have a high probability of success or when the cost of failure is low. Effective deterrence would then require taking steps to reduce the probability of success and to raise the cost of failure in addition to adopting otherwise time-inconsistent strategies of non-negotiation.

Economic conditions and terrorism

31

On the empirical side, Enders, et al. (1990) assess the effectiveness of terroristthwarting policies on terrorism. They ﬁnd little evidence for legislative activity in preventing terrorism. Atkinson, et al. (1987) examine the impact of changes in the negotiating environment (e.g. bargaining costs, blufﬁng) on the length and severity of terrorist attacks. They ﬁnd, in general, that increases in bargaining costs lengthen the duration of a terrorist incident. O’Brien (1996) considers whether terrorism is used as a foreign policy tool by international superpowers and concludes that authoritarian regimes are more likely to sponsor terrorist attacks following setbacks in the foreign policy arena. These papers provide, broadly speaking, the groundwork for analysing terrorism within an economic framework. They do not, however, explicitly address how, or even whether, the onset of terrorist incidence is related to economic circumstances in a country. There is, however, a literature that analyses how economic circumstances inﬂuence conﬂict in general. Most of the analysis to this point has, however, considered the impact on conﬂicts such as war without considering alternative types of conﬂict such as terrorism. For example, Hess and Orphanides (1995, 2001) estimate the probability of conﬂict for the US doubles when the economy has recently been in an economic contraction and the president is running for reelection. Similarly, Stoll (1984), Ostrom and Job (1986), Russett (1990), Lian and O’Neal (1993), Gelpi (1997), and Brueck (2002) have found linkages between the incidence of war and the political cycle and/or the business cycle. Broader deﬁnitions of conﬂict have been considered in more recent research. Blomberg, et al. (2002), and Blomberg and Hess (2002) provide more speciﬁc deﬁnitions of conﬂict such as external conﬂict (e.g. wars) and internal conﬂict (e.g. coups). After doing so, however, rather than ﬁnding a systematic relationship across all countries and time, they found a conditional conﬂict-growth relationship that can only be identiﬁed once the region and initial conditions are taken into account. While the above research provides a foundation for the determinants of terrorism, it still does not address the question our paper raises – do economic circumstances inﬂuence the initiation of terrorist activities in a systematic way? And if so, what is the theoretical justiﬁcation? The theory and empirical analysis presented in the next two sections seek to address these questions.

3.

The theory

In this section, we sketch a theory of the links between economic variables and terrorist attacks. The theory combines features of the static model of Grossman (1991) and the dynamic model of Tornell (1998).3 We use the simple model for theoretical structure to link economic activity and terrorism. The model describes not just the links between the economy and conﬂict, but also the links between the economy and the type of conﬂict, that is whether a civil war or a terrorist attack. This model (and our empirical results) focus only on economic explanations for terrorist activities. Economic explanations do not underlie all terrorist activity. Ideology and beliefs of course matter (see Bernholz (2003)).

32

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

We characterize two major organized groups in the economy: a government and dissidents.4 Both groups appropriate the stock of resources in the economy (with the government appropriating resources at a rate that exceeds the rate at which dissident groups extract resources from the economy). If the dissident group is unhappy with the status quo it has two options. First, it can choose to mount a ‘rebellion attack’ against the government and seek to overthrow it. If successful in this endeavour, the dissident group obtains a share of the productive resource base in the economy and also restructures and sets the new rules for the economy. The second option for the dissident group is to mount a low-intensity attack on the government in the form of a ‘terrorist attack’. This is not as intense as an overthrowing of the government but is instead designed to signal unhappiness with the status quo. To be more speciﬁc, the difference between a terrorist attack and a rebellion attack is that, if successful, the former does not provide more control of the ﬁscal assets in the economy, but it does provide more agenda-setting discretion over the rules of the economy. As an alternative to these two options, the dissidents can choose to maintain the status quo. This, of course, will only occur if the cost of mounting either a terrorist attack or a rebellion attack is too high relative to the reward of doing so. There is no reason for an attack to be initially preferred by the dissident group to the status quo. However, even if initially the status quo yields a more favourable outcome than either a rebellion or a terrorist attack, if the stock of assets grows too slowly, or if the government starts extracting a greater share of the assets then the payoff to attack will become higher than the payoff to pursuing the status quo at some point in the (perhaps distant) future. In such a situation an attack by the dissident group becomes likely. The question then becomes, what type of attack will be undertaken by the dissident group? For the dissident group, launching a rebellion attack is more costly than launching a terrorist attack, since the former is an attempt to completely overthrow the government and take control of the economy. The potential reward for incurring this additional cost is that, if successful, the dissident group takes control of the economy, gets agenda-setting power over the economy and begins to extract resources of the asset stock for its own use. In the case of a successful terrorist attack the dissident group does not receive a greater share of the assets of the economy, it only receives a partial beneﬁt from changing some of the rules of the economy. The theory posits that the possibility of an attack by dissident groups becomes more likely when payoffs to such attacks are large and costs are small. We would expect that economies with low growth rates, high government tax rates and higher political unrest (disgruntlement of dissident groups with their extraction rate) would have higher incidences of conﬂict. So, during poor economic times and when the relative share of the pie is smaller, dissidents will attack by some means. The model also provides predictions about the mode of conﬂict that will be chosen by the dissident group. The payoff to using terrorism as a mode of conﬂict instead of rebellion will be high, when the cost differential in initiating rebellion

Economic conditions and terrorism

33

is high. Intuitively, one would expect that the difference between the resources needed to initiate a rebellion and the resources needed to initiate a terrorist attack depend on the institutional processes of the economy (which can include a variety of factors including GDP per capita, income distribution, military spending, ethnic divisions etc.). When groups have fewer channels to construct an organized uprising, they resort to terrorist acts during poor economic times. However, if they can organize, it might be more beneﬁcial for them to initiate a civil war. This would be in line with the empirical results in Blomberg and Hess (2002), which show a strong correlation between adverse economic outcomes and the prevalence of civil war. However, the present paper is more concerned with the conditions under which terrorism will be chosen over war. To summarize, our model predicts conﬂict to be more likely in bad times: when the resource base of the economy shrinks, dissident groups are less likely to be satisﬁed with claiming their low share of the smaller pie and are likely to instigate some type of conﬂict to increase their share of the pie. Furthermore, the theory predicts that the choice between a rebellion, in which the dissidents overthrow the government, and a terrorist attack, in which the dissidents seek to improve their voice in the economy, depend on the degree to which the government is responsive to the terrorists demands and on the soldiering technology of the economy. Richer countries that have better institutions, stronger economies and well-equipped armies raise the cost of rebellion to the point that dissident groups prefer to resort to terrorism. In the subsequent section, we will explore this implication of our model to see if in fact terrorism is chosen by dissident groups in those countries where organized rebellion will be costly, during bad economic times.

4.

Data and preliminary analysis

In this section, we describe the data employed in the paper and then examine and present its empirical regularities. The data were obtained from two different sources. To measure terrorist activities, we employ the ITERATE data set from Mickolus et al. (1993). The economic data are obtained from the update to the Summers and Heston (1991) data set. The ITERATE project quantiﬁes characteristics, activities and impacts of transnational terrorist groups. In order to be considered an international/ transnational terrorist event, the deﬁnition in ITERATE is as follows: ‘the use, or threat of use, of anxiety-inducing, extra-normal violence for political purposes by any individual or group, whether acting for or in opposition to established governmental authority, when such action is intended to inﬂuence the attitudes and behavior of a target group wider than the immediate victims and when, through the nationality or foreign ties of its perpetrators, its location, the nature of its institutional or human victims, or the mechanics of its resolution, its ramiﬁcations transcend national boundaries’. The data set is grouped into four categories. First, there are incident characteristics which code the timing of each event. Second, the terrorist characteristics yield

34

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

information about the number, makeup and groups involved in the incidents. Third, victim characteristics describe analogous information on the victims involved in the attacks. Finally, life and property losses attempt to quantify the damage of the attack. The empirical work below focuses on the incidence of terrorist events, though other qualitative features of the data are also discussed. The remaining data employed are given in a straightforward manner deﬁned by Summers and Heston. The main advantage of employing the Summers and Heston data set is that it is calculated in PPP adjusted exchange rates so crosscountry comparisons can be made with better adjustments due to price differences. Combining the data sets, there are 127 countries over the years 1968 to 1991 with a panel of 3014 observations. We begin by examining the incidence of terrorism. Table 3.1 provides a complete list of countries and average annual terrorist attacks. The areas of the world

Table 3.1 Terrorist incidence around the world average annual incidence 1968–91 Country

Avg. no.

Country

Avg. no.

Country

Avg. no.

Algeria Angola Argentina Australia Austria Bahamas Bahrain Bangladesh Barbados Belgium Belize

0.54 2 15.58 1.63 3.29 0 0.13 0.42 0.21 4.33 0

Guatemala Guinea Guinea-Bissau Guya. Haiti Honduras Hong Kong Hungary Iceland India Indonesia

6.42 0 0 0.08 1.17 3.38 0.5 0.33 0.08 6.21 1.42

0.38 9.88 11 0.58 3.13 1.79 0.04 0 0.29 0 0.75

4.67 1.83 3.38

Paraguay Peru Philippines Poland Portugal Puerto Rico Qatar Reunion Romania Rwanda Saudi Arabia Senegal Seychelles Sierra Leone Singapore Solomon Is. Somalia South Africa Spain Sri Lanka St. Kitts & Nevis St. Lucia St. Vincent Sudan Suri.Me Swaziland

Benin Bhutan Bolivia

0 0 3.17

Iran Iraq Ireland

Botswana Brazil Bulgaria Burki. Faso

0.29 3 0.13 0.08

Israel Italy Ivory Coast Jamaica

Burundi C.A.R. Cameroon

0.08 0.13 0.13

Japan Jordan Kenya

2.17 3.29 0.33

Canada Cape Verde Is. Chad Chile China

1.71 0 0.33 5.96 0.21

Korea, Rep. Kuwait Laos Lesotho Liberia

3.04 2 0.38 0.29 0.58

9.08 14.83 0.21 0.67

0.08 0.04 0.04 0.46 0.04 0.79 0.92 10.92 0.75 0 0 0 1.92 0.25 1.13 Continued

Economic conditions and terrorism

35

Table 3.1 (continued) Country

Avg. no.

Country

Avg. no.

Country

Colombia Comoros Congo Costa Rica Cyprus Czechoslovakia Denmark Djibouti Dominica

10.17 0 0.08 2.42 5.21 0.17 1.25 0.33 0.04

Luxembourg Madagascar Malawi Malaysia Mali Malta Mauritania Mauritius Mexico

0.21 0 0.04 2.96 0 0.38 0.08 0 3.75

Dominican Rep. Ecuador Egypt El Salvador Ethiopia Fiji Finland

1.08 1.92 3.42 7.58 2.54 0.21 0

Mongolia Morocco Mozambique Myanmar Namibia Nepal Netherlands

0 0.79 2.71 0.42 0.17 0.21 6.46

France Gabon Gambia Germany, East

23.29 0.17 0 0.33

New Zealand Nicaragua Niger Nigeria

0.21 1.29 0.25 0.25

Germany, West Ghana Grenada Greece

16 0.08 0.04 15.21

Norway Oman Pakistan Panama Papua N.Guinea

0.5 0.04 5.25 1.88 0.21

Sweden Switzerland Syria Taiwan Tanzania Thailand Togo Tonga Trinidad & Tobago Tunisia Turkey U.K. U.S.A. U.S.S.R. Uganda United Arab E. Uruguay Vanuatu Venezuela Western Samoa Yemen Yugoslavia Zaire Zambia Zimbabwe

Avg. no. 1.92 2.83 1.79 0.29 0.29 2.08 0.13 0 0.38 0.79 10.92 19.92 27.63 1.79 0.79 0.25 2 0 2.83 0 0.5 0.67 0.33 0.92 1.79

Note: All information in this table was obtained from ITERATE.

that appear to be those with the most terrorism are the Americas and Europe whereas there appears to be far less terrorism in Africa. The US, which by some measures is the richest country in the world, is clearly the country with the most terrorist incidents as well. During the period 1968–91, there was an average of 28 terrorist attacks per year. How do we interpret this in the context of our model? It is unlikely that only the wealthy engage in terrorist activities whereas the poor do not. There must be some other factor at work here. The model presented in the previous section provides an explanation: groups that do not have direct access to the elites send their message by committing terrorist activities since they do not have a voice in the political process. Since the democracies of Europe and North America are largely driven by mainstream parties, fringe groups tend to get far less representation in the legislatures. Furthermore, our model showed that if there is sufﬁcient pie to ﬁght over (high GDP per capita), then in bad times the fringe groups will resort to violence to bring about change. In North America and Europe,

36

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

the combination of a sizeable economic pie and a low probability of success in a rebellion attack leads dissident groups to resort to terrorism. In the non-democratic states of Africa, there is no direct or even indirect access to the government. In this case, uniﬁed resistance and resort to actual war with the opposing government is required for change. In fact, Blomberg and Hess (2002) showed that there is indeed much more civil war in Africa vs. North America and Europe.5 The combination of low growth rates, high extraction rates of the economy’s resource base by the government and a higher probability of success in a rebellion attack leads dissident groups to engage in rebellion instead of terrorism. The ITERATE data setalso provides a breakdown of which groups engaged in terrorist activities. In the United States one of the main perpetrators, especially in the early periods of the sample, is the FALN (Armed Front for National Liberation), a group lobbying for the separation of Puerto Rico from the United States. In France, the European country with the most terrorism, the predominant entity is the Corsican National Liberation Front. Furthermore, unknown groups remain quite active throughout the sample. This highlights one of the main points in our theoretical analysis. Terrorist attacks are predominantly instigated by fringe organizations that would not ordinarily get their agenda heard in the legislature. This points to an institutional structure – broadly consistent with the theory presented in the previous section – that facilitates terrorism by fringe groups that potentially reject the status quo, yet lack the support or resources necessary for bringing about broad-based institutional change that conforms more to their preferred world view. Interestingly, more terrorism also tends to take place during the earlier time periods rather than in the later time period. Our model would explain this phenomenon through the prediction that in bad economic times there is more impetus for terrorism. Clearly the number and severity of economic contractions were greater during the 1968–84 period than during the 1985–91 period. While this quick overview of the data provides some indirect support for our theory, it is not a formal test. To explore the model’s predictions, we describe our formal empirical model.

5.

Methodology and empirical results

In this section, we analyse the dynamics of economic activity and terrorism over several relevant sub-samples. Following Burns and Mitchell (1944), we measure short-run economic activity as discrete regimes, namely, contractions and expansions. The former are periods where economic activity, as measured by the growth of real GDP growth per capita, are negative, whereas for the latter they are nonnegative. Similarly, terrorism is deﬁned if a country has any terrorist incidents in a given year. Alternatively, peace is a period of no terrorism. In this subsection, we formalize the empirical relationship between growth and terrorism by ﬁrst examining the univariate dynamics of each using Markov processes. Markov processes are dynamic processes that capture the observed transitions from one state at time period t − 1, to either remain in that state at

Economic conditions and terrorism

37

time period t or to switch to another state at time period t. This has the natural interpretation for business cycles as the economy is in either one of two states: economic contraction or expansion. For purposes of our analysis, we deﬁne a recession to be a period of negative per-capita growth of real GDP and an expansion as a period of non-negative growth. As these are mutually exclusive, we deﬁne a contraction (expansion) in period t − 1 as Ct−1 (Et−1 ). The speciﬁcation of a Markov process is an attempt to specify the extent to which a particular state of the economy in a previous period affects the probability of an expansion or contraction in the current period.6 With this generalization as background, we deﬁne pij as the conditional probability that the economy is in state i = C, E in period t − 1 and in the state j = C, E in period t. The 2 × 2 transition probability matrix is therefore:

PR(Ct |Ct−1 ) PR(Ct |Et−1 )

PR(Et |Ct−1 ) p = CC pCE PR(Et |Et−1 )

pEC pEE

where PR will denote probability. One attractive feature of this formulation is that each row of the transition matrix sums to one. So this 2 × 2 Markov transition matrix only requires us to estimate two parameters as pCE = 1.0 − pCC and pEC = 1.0 − pEE . The log-likelihood function, ln(L). for the 2 × 2 Markov process is7 ln(L) = nCC ln(pCC ) + nCE ln(1 − pCC ) + nEC ln(1 − pEE ) + nEE ln(pEE ) where nij is the number of occurrences of state i in period t − 1 and in state j in period t. It is straightforward to show that the maximum likelihood estimators of the probabilities are simply: pˆ CC =

nEE nCC and pˆ EE = , nC nE

where nj is the number of observations in state j at time t − 1. In other words, pij is the observed fraction of times that state j was observed at time t when state i was observed at time t − 1. Table 3.2 reports the results from our estimation of these transitional probabilities. The ﬁrst column of results reports the estimates for the economy whereas the second column reports the analogous exercise for terrorism. Each panel of the table reports the results for 13 samples of countries. The samples are for the entire data set (ALL), and those based on initial income: namely, the country’s whose initial real GDP per-capita in 1967 was below the median (LOW INCOME), and those that were above the median (HIGH INCOME). Finally we separately examine fully democratic countries (DEMOCRACIES), non-fully democratic countries (NON-DEMOCRACIES), African countries (AFRICA) and non-African countries (NON-AFRICA), and Democratic and High Income Countries (HIGH INCOME & DEMOCRATIC).

38

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

Table 3.2 Estimates of 2 × 2 Markov processes for the economy and terrorism EVENT X Data

Statistic

Economic Contraction (C)

Terrorist Incident (T)

NOBS

ALL

Pr(Xt |Xt−1 ) Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) Dur(X c |Xt−1 Pr(X )

0.472 1.895 0.743 3.902 0.326

0.741 3.863 0.777 4.482 0.465

3014

LOW INCOME

Pr(Xt |Xt−1 ) Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

0.465 1.87 0.677 3.096 0.373 0.002

0.6 3.312 0.826 5.753 0.312 0.001

1511

HIGH INCOME

Pr(Xt |Xt−1 ) Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

0.482 1.93 0.804 5.575 0.279 0.002

0.811 5.291 0.696 3.286 0.612 0.001

1503

DEMOCRACIES

Pr(Xt |Xt−1 ) Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

0.403 1.677 0.811 5.295 0.241 0.001

0.818 5.508 0.703 3.365 0.613 0.001

1197

NONPr(Xt |Xt−1 ) DEMOCRACIES Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

0.505

0.658

1675

2.019 0.69 3.224 0.388 0.001

2.918 0.791 4.767 0.384 0.001

Pr(Xt |Xt−1 ) Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

0.498 1.991 0.613 2.588 0.436 0.001

0.484 1.939 0.88 8.355 0.197 0.001

AFRICA

1111

Continued

Economic conditions and terrorism

39

Table 3.2 (continued) EVENT X Data

Statistic

Economic Contraction (C)

Terrorist Incident (T)

NOBS

NON-AFRICA

Pr(Xt |Xt−1 ) Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

0.447 1.809 0.803 5.067 0.264 0.001

0.788 4.714 0.665 2.982 0.614 0.001

1903

HIGH INCOME & DEMOCRACY

Pr(Xt |Xt−1 )

0.425

0.838

910

Dur(X |Xt−1 ) c ) Pr(Xtc |Xt−1 c ) c Dur(X |Xt−1 Pr(X ) p-value

1.74 0.838 6.178 0.224 0.001

6.456 0.641 2.788 0.692 0.001

X refers to the events Economic Contraction (C) and Terrorist Incident (T). The superscript c refers to the complement of an event, e.g. the complement of Contraction is Expansion, E = C c and the complement of Terrorist Incident is Peace, P = T c . PR(.) refers to probability, and PR(Xt |Xt−1 ) is the transition probability that event X will occur in period t, given that event X occurred in period t − 1. PR is the asymptotic probability of the event and DUR refers to the conditional expected duration of an event. p-value is the p-value from a likelihood ratio test that the estimated coefﬁcients from the transition matrix are the same in the sub-samples and the full samples. The test is distributed χ 2 with 2 degrees of freedom. The sub-samples are for the entire data set (ALL), countries with below the median level of initial real GDP per-capita in 1967 (LOW INCOME), and those with incomes above the median (HIGH INCOME), fully democratic countries in 1967 (DEMOCRACIES), nonfully democratic countries (NON-DEMOCRACIES), African countries (AFRICA) and non-African countries (NON-AFRICA).

The ﬁrst row of the panel reports the transitional probability of remaining in a contraction this period, given that you were in a contraction last period: pCC = Pr(Ct |Ct−1 ) The second row of the panel reports the expected duration associated with that probability which is calculated as: DUR(C|Ct−1 ) =

1 1 − pCC

Rows three and four present the analogous transition probability and duration of an expansion. The ﬁfth row of the table reports the long-run, steady-state fraction

40

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

of time that the economy is in a contraction: PR(C) =

1 − pEE 2 − pCC − pEE

Finally, the sixth row reports the p-value for the test of the null hypothesis that the sub-sample states of nature were generated from the full-sample probabilities.8 The results are quite intriguing. In the ﬁrst column of the top panel, we show that the probability of remaining in a contraction another year (pCC ) is about 0.45 for the full sample, and hence the probability of switching to an expansion phase is 0.55 (0.55 = 1.0 − 0.45). This number may seem rather high when considering industrialized business cycles. However, given that much of the sample is taken from developing countries, the estimate is not as surprising. The probability of remaining in an expansion another year is about 0.77, which implies that the corresponding probability that the expansions will switch to a contraction next year is 0.23 (0.23 = 1.0 − 0.77). These transitional probabilities translate into an expected duration of contractions of just under 2 years and just over 4 years for expansions. The top panel of the table also reports that there is a contraction (i.e. negative growth) in about one-in-three years in the sample. The remaining panels of the ﬁrst column of Table 3.2 present the estimation results of the Markov process for several important sub-samples of the data. There are three noteworthy, as well as statistically signiﬁcant, differences in the univariate results for economic contractions and expansions when we consider sub-groups. First, the duration of expansions, conditional on being in a state of expansion, is only three years for low income countries but almost six years for high income countries. However, the duration of contractions is not affected by a country’s income status. Second, Africa has both more contractions and shorter expansions as compared to non-African countries, and these differences are statistically signiﬁcant at below the conventional .10 levels as reported in the p-value row. Indeed, the expected duration of expansions is about 2.5 years, while it is over 5 years for non-African countries and over 6 years for high income, democratic countries. African countries are in the state of economic contraction about 43 per cent of the time, as compared to only 26 per cent of the time for non-African countries. Finally, democratic countries tend to have fewer and shorter contractions and longer and more frequent expansions than non-democratic countries. Similarly, using the methodology discussed above, we can also estimate the transitional patterns between terrorism and peace. The estimation results from these Markov processes are reported in the second column of results of Table 3.2. For the full sample, as shown in the top panel, these ﬁndings bear the unfortunate news that terrorism is not a rare and unusual event, as it accounts for approximately 46 per cent of the sample. Moreover, once one enters into a period of terrorism, the conditional expected duration spell of terrorism is just under 4 years, and its conditional probability of continuing an additional year is 74 per cent. The sub-samples of the data for the terrorism data also reveal a number of important empirical features. First, countries with higher income at the beginning

Economic conditions and terrorism

41

of the sample have more persistent episodes of terrorism, that are more durable and more frequent. Indeed, high income countries have terrorist events in about sixty per cent of their sample, as compared to low income countries that have terrorist events in about 30 per cent of their sample. Also, the conditional duration of terrorism continuing given that it has started is over 0.8 for high income countries, though only 0.6 for low income countries. Second, democracies appear to be more affected by terrorism than non-democracies: terrorism is more prevalent, durable and persistent in democracies as compared to non-democracies. Third, Africa has about one-third as many years with terrorist events as non-African countries; slightly more than 60 per cent of the non-African sample has a year coded with a terrorist incident, whereas slightly less than 20 per cent of the African sample has a year coded with a terrorist incident. As well, the conditional persistence of terrorist events is much lower in African (about probability 0.5) as compared to non-African countries (about probability 0.8), while the conditional persistence of peace is much higher in Africa (about probability 0.9) as compared to non-African countries (about probability 0.65). These ﬁndings suggest that for most countries, prosperity and terrorism are the norm. Moreover, while economic contractions and periods free from terrorist events do occur, the former occurs less frequently for high income and democratic countries, while the latter occurs more frequently for richer countries. A possible explanation for why high income and democratic countries have more terrorism is that they simply have fewer press restrictions: hence, they may appear to have more terrorism simply because their news agencies may be more likely to report it.9 While the ﬁndings in this sub-section may be affected by this criticism, those in the following sub-section are not. Indeed, in the following sub-section, we demonstrate that the propensity for terrorism is affected by the business cycle for democratic and high income countries. Such a ﬁnding of a cyclical economic predecessor to terrorism cannot be a direct by-product of sample selection issues driven by countries that have a free press. In this subsection, we extend the empirical model presented above to investigate the joint determination of terrorism and contractions. The methodology employed in this section is similar to that used in the previous section and in Blomberg and Hess (2002). Consider the joint determination of terrorism, Tt and contraction Ct . To keep the accounting straight, denote state 1 as the joint occurrence of internal conﬂict and contraction, Tt & Ct , state 2 as the joint occurrence of internal conﬂict and expansion, Tt & Et , state 3 as the joint occurrence of internal peace and contraction Pt & Ct , and state 4 as the joint occurrence of internal peace and expansion Pt & Et . As before, we can then estimate the transition matrix of probabilities, but now there are 4 possibilities such that pij for i, j = 1, . . . , 4 speciﬁes the transitions from state i in period t − 1 to state j in period t. Table 3.3 provides the estimation results of this bivariate Markov process. The table is organized in a similar fashion to Table 3.2. Not only are we interested in estimating the parameters from these Markov processes, however, but more importantly we are interested in using these estimated transition probabilities from the Markov matrix to help to identify ‘causal’ timing patterns in the data. To keep

0.177 0.171 [0.831] 0.515 0.577 [0.177] 0.355 0.305 [0.102] 0.502 0.497 [0.909]

PR(Tt |Pt−1 & Et−1 ) PR(Tt |Pt−1 & Ct−1 ) p-value PR(Tt |Tt−1 & Et−1 ) PR(Tt |Tt−1 & Et−1 ) p-value PR(Ct |Pt−1 & Et−1 ) PR(Ct |Tt−1 & Et−1 ) p-value PR(Ct |Pt−1 & Ct−1 ) PR(Ct |Tt−1 & Ct−1 ) p-value

0.211 0.252 [0.206] 0.718 0.782 [0.049] 0.146 0.188 [0.026] 0.466 0.477 [0.826]

LOW INC 0.209 0.258 [0.347] 0.722 0.816 [0.050] 0.213 0.242 [0.428] 0.365 0.495 [0.067]

HIGH INC 0.195 0.201 [0.271] 0.583 0.670 [0.102] 0.272 0.236 [0.838] 0.463 0.488 [0.655]

DEMO 0.121 0.148 [0.242] 0.344 0.515 [0.242] 0.388 0.381 [0.878] 0.512 0.523 [0.827]

NON DEMO 0.235 0.272 [0.228] 0.708 0.758 [0.080] 0.160 0.198 [0.035] 0.460 0.472 [0.765]

AFRICA 0.206 0.252 [0.085] 0.770 0.881 [0.041] 0.148 0.200 [0.192] 0.354 0.462 [0.313]

NON-AFRICA

0.177 0.171 [0.831] 0.515 0.577 [0.177] 0.355 0.305 [0.102] 0.502 0.497 [0.909]

HIGHINC × DEMO

See Table 3.2. P and E refer to the states of ‘Peace’ and ‘Economic Expansion’, respectively. Note that Peace is the complement of Terrorism, while Expansion is the complement of Contraction. p-value, reported in square brackets, is the test that the two preceding probabilities are equal to one another. The test is distributed Chi-squared with 1 degree of freedom.

ALL

SAMPLE

Table 3.3 Estimates of 4 × 4 Markov processes for the economy and terrorism

Economic conditions and terrorism

43

our reporting of the estimates parsimonious, we restrict our presentation to key statistics such as testing whether PR(Tt |Pt−1 & Ct−1 ) = PR(Tt |Pt−1 & Et−1 ).10 The estimates and the restriction are reported in the top panel of the table. The restriction on the likelihood function can be implemented, and the restrictions can be tested using a χ 2 likelihood ratio test with one degree of freedom. The reported p-value is presented in Table 3.3 and is labelled in the row immediately following the estimated transition probabilities. This allows us to infer that economic contractions cause an increase in the transition from internal peace to terrorism since it is temporally prior to the incident. Of course, terrorism may be more persistent when coupled with a contraction than otherwise. This too can be explored by examining whether PR(Tt |Tt−1 & Ct−1 ) = PR(Tt |Tt−1 & Et−1 ) and testing whether the two are equal, as is done in the table’s second panel. We also test for whether the pattern of contractions is affected by terrorism. For example, we examine whether the transition probability from expansion to contraction rises if a conﬂict is present, by testing the null hypothesis that PR(Ct |Tt−1 & Et−1 ) = PR(Ct |Pt−1 & Et−1 ), against the alternative that these parameters should be freely estimated – see third panel.11 Finally, we test for whether the conditional persistence of economic contractions is affected by the presence of terrorism in the fourth panel, PR(Ct |Tt−1 & Ct−1 ) = PR(Ct |Pt−1 & Ct−1 ), Not surprisingly, the results in Table 3.3 point to a broad dependence between terrorist incidents and economic contractions, though not for all countries. Indeed, for the full data sample (column 1 of results), the estimated conditional probability of terrorism is not signiﬁcantly affected by whether a contraction occurred in the most recent year or not at the 0.10 level. Similarly, the likelihood of a contraction is not affected by terrorism. For example, for the full data sample, the conditional probability of a terrorist activity next period given that a country is currently at peace is 0.196 if the economy is also expanding, while it is 0.197 if the economy is contracting: namely, PR(Tt |Pt−1 & Et−1 ) = PR(Tt |Pt−1 & Ct−1 ) are essentially identical. However, for a number of important sub-groups listed at the top of the table, there is strong dependence between terrorism and economic activity. There are

44

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

three noteworthy points centring on how income and governance account for this dependence. First, the relationship between terrorism and the economy appears to be quite important for High Income countries, though not for Low Income countries. For example, the conditional persistence of terrorist events is signiﬁcantly higher when an economic contraction has occurred (0.782) as compared to when an economic expansion has occurred (0.718). Moreover, a contraction is more likely to start after a terrorist episode has taken place as compared to when a terrorist episode has not taken place.12 Both of these ﬁndings, however, do not hold for low income countries. Second, the relationships between terrorism and the economy are quite different for Democratic and non-Democratic countries. In particular, Democratic countries have signiﬁcantly more persistent contractions during periods of terrorism (conditional on starting in a contraction), and signiﬁcantly more persistent terrorism (conditional on starting in a terrorist episode) during economic contractions. In other words, for Democratic countries, PR(Tt |Tt−1 & Et−1 ) < PR(Tt |Tt−1 & Ct−1 ) and PR(Ct |Pt−1 & Ct−1 ) < PR(Ct |Tt−1 & Ct−1 ). Again, these important ﬁndings for democracies do not hold for non-democracies. Combining high income and democratic countries together also reveals an important pattern of statistically signiﬁcant relationships between economic activity and terrorist incidents. Strikingly, it appears that the critical ﬁnding is that current periods of economic contractions make future terrorist events more likely. This result can be gleaned from the following two observations. First, periods of peace are more likely to turn to periods of terrorism if the economy is in an economic contraction. In other words, PR(Tt |Pt−1 & Ct−1 ) > PR(Tt |Pt−1 & Et−1 ). Second, periods where terrorist events take place are more likely to remain in the state of terrorism if the economy is in an economic contraction: namely, PR(Tt |Tt−1 & Ct−1 ) > PR(Tt |Tt−1 & Et−1 ). Taken together, these two ﬁndings suggest that for countries that are both high income and democratic, economic contractions make future terrorism more likely. In summary, the results from Table 3.3 are quite strong and statistically signiﬁcant. Economic contractions and terrorist events are simply not independent events that can be considered in isolation. The strongest link between the two appears to be from economic contractions to increased frequencies of terrorism. This link, however, is not constant across all countries but rather is driven to a large extent by higher income and democratic governance. Contractions make countries more likely to transition to terror and remain there. There is some additional evidence that terrorism leads to an increase in the initiation and continuation of economic contractions.

Economic conditions and terrorism

6.

45

Conclusion

This paper has developed a theory whereby groups with limited access to opportunity rationally engage in terrorist activities while policy-making elites ﬁnd it rational not to engage in opening access to these groups. The result is a pattern of reduced economic activity and increased terrorism. To explore the model’s implications, we constructed a panel data set of 130 countries from 1968 to 1991 of terrorist and economic variables. We have found that economic activity and terrorism are not independent of one another. In particular, high income and democratic countries appear to have a higher incidence of terrorism, and a lower incidence of economic contractions. Furthermore, terrorism appears to be related to the economic business cycle: periods of economic weakness increase the likelihood of terrorist activities.

Endnotes 1 We thank Bengt-Arne Wickstrom and Joseph Joyce for helpful comments. Archana Ravichandran provided excellent research assistance. We also thank Peter Fleming and Todd Sandler for their help in unscrambling the data set. This paper was ﬁrst written for the DIW Berlin’s workshop entitled ‘The Economic Consequences of Global Terrorism’. Part of this paper was written while Gregory D. Hess was an academic consultant to the Federal Reserve Bank of Cleveland and the IMF Institute. The opinions expressed are those of the authors and do not necessarily reﬂect views of the Federal Reserve Bank of Cleveland, the Federal Reserve System, or the IMF. 2 See also Sandler (2000). 3 We only provide an overview of the theory here. A more formal model that explicitly derives the results mentioned here is available on request from the authors. 4 Just as Tornell categorizes potentially heterogeneous sub-groups under the broad rubric of ‘unions’ or ‘corporations’, the rubric of ‘dissidents’ can cover many groups with different objectives and ideologies. 5 Many of the civil wars in Africa also have to do with control over the economy by controlling natural resources. 6 This speciﬁcation of a ﬁrst order Markov process follows Blomberg and Hess (2002). Higher order Markov processes can be speciﬁed by allowing the economy’s state in period t − 1 and earlier to independently affect the economy’s state in period t. For the purposes of this study, however, we maintain a ﬁrst order structure on our analysis as higher order Markov processes dramatically increases the number of parameters to be estimated and reduce the precision of these estimates. 7 We ignore terms that are not functions of the probabilities. 8 More speciﬁcally, one evaluates the log-likelihood for the sub-sample using the maximum likelihood estimates obtained from the sub-sample and the full sample and then constructs a likelihood ratio test between the two. The p-value is obtained from a χ 2 distribution with two degrees of freedom, stemming from the two estimated transition parameters. 9 We thank Todd Sandler for bringing this point to our attention. 10 This requires estimating the parameters of the log-likelihood function subject to the additional constraint that p31 + p32 = p41 + p42 . 11 This again requires estimating the parameters of the log-likelihood function subject to the additional constraint that p22 + p24 = p42 + p44 . 12 PR(Ct |Pt−1 & Et−1 ) = 0.146 < 0.188 = PR(Ct |Tt−1 & Et−1 ) with a p-value of 0.026.

46

S. Brock Blomberg, Gregory D. Hess and Akila Weerapana

References Atkinson, S., Sandler, T. and Tschirhart, J. (1987) ‘Terrorism in a bargaining framework’, Journal of Law and Economics, 30: 1–21. Bernholz, P. (2003) Supreme values as the basis for terror, unpublished paper, University of Basel, Switzerland. Blomberg, S.B. and Hess, G.D. (2002) ‘The temporal links between conﬂict and economic activity’, Journal of Conﬂict Resolution, 46: 74–90. Blomberg, S.B., Hess, G.D. and Thacker, S. (2002) ‘Is there a conﬂict-poverty trap?’, Wellesley College Working Paper 2000–06, Wellesley, MA. Brueck, T. (2002) ‘The macroeconomic effects of the war on Mozambique’, QEH Working Paper #11, Oxford, UK. Burns, A. and Mitchell, W. (1944) Measuring business cycles, New York: NBER. Enders, W. and Sandler, T. (1993) The effectiveness of anti-terrorism policies: vector autoregression intervention analysis’, American Political Science Review, 87: 839–44. Gelpi, C. (1997) ‘Democratic diversions: governmental structure and the externalization of domestic conﬂict’, Journal of Conﬂict Resolution, 41: 255–82. Grossman, H.I. (1991) ‘A general equilibrium model of insurrections’, American Economic Review, 81: 912–21. Hess, G.D. and Orphanides, A. (1995) ‘War politics: an economic, rational-voter framework’, American Economic Review, 85: 828–46. —— (2001) ‘Economic conditions, elections and the magnitude of foreign conﬂicts’, Journal of Public Economics, 80: 121–40. Lapan, H. and Sandler, T. (1988) ‘To bargain or not to bargain: that is the question’, American Economic Review Papers and Proceedings, 78: 16–21. Lian, B. and O’Neal, J. (1993) ‘Presidents, the use of military force, and public opinion’, Journal of Conﬂict Resolution, 37: 277–300. —— (1993) ‘Terrorism and signaling’, European Journal of Political Economy, 9: 383–97. Mickolus, E., Sandler, T., Murdock, J. and Fleming, P. (1993) International terrorism: attributes of terrorist events (ITERATE), Dunn Loring, VA: Vinyard Software. O’Brien, S.P. (1996) ‘Foreign policy crises and the resort to terrorism: A time series analysis of conﬂict linkages’, Journal of Conﬂict Resolution, 41: 320–35. Ostrom, C.W. and Job, B.L. (1986) ‘The president and the political use of force’, American Political Science Review, 80: 541–66. Russett, B. (1990) ‘Economic decline, electoral pressure, and the initiation of international conﬂict’, in Gochman, C.S. and Sobrosky, A.S. (eds), Prisoners of war, Lexington, MA: Lexington Books. Sandler, T. (2000) ‘Economic analysis of conﬂict’, Journal of Conﬂict Resolution, 44: 723–9. Stoll, R. (1984) ‘The guns of November: presidential reelections and the use of force, 1947–82’, Journal of Conﬂict Resolution, 28: 231–46. Tornell, A. (1998) ‘Reform from within’, Working Paper 6497, Cambridge, MA: NBER.

4

The economics of high-visibility terrorism Sanjay Jain and Sharun W. Mukand

This paper analyses some implications of very visible, discrete, large-scale terrorist actions, such as the 11 September attacks. By becoming part of the set of common knowledge, such events of mass terror have direct implications on both the supply and composition of the ‘mix’ of terrorist attacks, as well as responses to them. First, we analyse the implications of greater anti-terror operations on the quantity as well as the mix of terrorists supplied by terrorist organizations. Second, we present a novel argument for the role of public announcements in endogenously coordinating responses in the face of a terrorist hijacking. Public announcements, by injecting strategic uncertainty, can give rise to a Pareto-superior outcome, at minimal resource cost to the government.

1.

Introduction

One of the most distinct aspects of the event of mass terror witnessed on 11 September, was its ‘visibility’. The high visibility had an immediate and dramatic effect on the information sets and beliefs of all agents – potential victims, government and law enforcement agents, as well as terrorists. Although suicide attackers had previously inﬂicted casualties in the Middle East, and had attacked US targets in Lebanon and Yemen, the high visibility of the 11 September attacks meant that, at one stroke, the presence of suicide terrorists became ‘common knowledge’ amongst all agents. This paper analyses the implications of such a visible event of mass terror on the incentives of terrorists and law enforcement as well as the potential victims. We use game-theoretic analysis to examine two related questions. First, what is the likely impact of such an event of mass terror on the mix of terrorists supplied by terrorist organizations? Second, we consider the impacts of such acts of mass terror on the likely responses of law enforcement, the government and potential victims, and use that to derive the implications of a particular policy proposal. A starting point of our analysis is that we do not see terrorists who are willing to commit suicide as irrational zealots, but rather as instruments used by terrorist organizations to achieve their goals.1 Once we treat acts of terrorism as being conducted by rational actors, anti-terror policies such as increased vigilance,

48

Sanjay Jain and Sharun W. Mukand

and the devotion of greater resources to law enforcement and anti-terror operations, can have consequences that are not obvious. It is well understood from the literature on the economics of crime and punishment that greater resources allocated to anti-terror operations will have a negative impact on the level of terrorism (see Becker, 1968, and the survey by Sandler and Enders, 2004). However, we show that although increased anti-terror operations decrease the level of terrorism, the mix or composition of terrorists may shift adversely, towards those who are more willing to use more extreme methods. Indeed this shift in the mix of terrorists arises not because of any new information about their effectiveness, but rather precisely because of the increased vigilance of law enforcement as well as potential victims (such as airline passengers). This shift in the composition of terrorists takes place for purely strategic reasons and is quite different from the reasons identiﬁed in the literature on the roots of terrorism (see, for example, Krueger and Maleckova, 2002). Our analysis is closer in spirit to Sandler et al. (1983) and Landes (1978), who focus on how an increase in penalties and law enforcement affects terrorists’ incentives. In contrast to their analysis, however, the focus of our analysis is not the tradeoff between terrorism and other activities, but rather on the tradeoff between alternative ‘means’ of terrorism to achieve the same end. The dramatic visibility of the mass terror of 11 September and the willingness of the terrorists to commit suicide has had a second important effect – on the incentives of potential victims to choose between alternative actions. In particular, the fact that the dramatic demonstration of the possibility of suicide terrorists is common knowledge gives rise to a multiplicity of equilibria. Depending on their beliefs about the possibility of coordinating resistance with their fellow passengers, a passenger may prefer to co-operate with or resist against the terrorists. As is well understood, this multiplicity of equilibria associated with self-fulﬁlling beliefs is a consequence of the ‘common knowledge’ of fundamentals by all agents. However, less well understood is the fact that the government can take advantage of this situation by introducing strategic uncertainty through public announcements – which remarkably, will result in a unique equilibrium, where passengers resist the terrorists. For instance, the public announcement of the introduction of sky marshals in ﬂights can serve as a coordinating device, such that passengers end up endogenously attacking terrorists, even though they are aware that there are no sky marshals on the ﬂights. This remarkable insight draws on recent work by Morris and Shin (1997, 1998) on the role of ‘almost common knowledge’ in generating unique equilibria in coordination games. It is useful to distinguish this result from the more standard result common to the literature on random audits or costly state veriﬁcation (Townsend, 1979) or the literature on the economics of crime. Our claim is not that introducing costly sky marshals (or any other anti-terror investments) in ﬂights, by lowering the ability of terrorists to carry out a successful hijacking, will reduce the incidence of plane hijacking. That we take to be self evident. Rather, our insight is more subtle and relies on the fact that when common knowledge about the presence (or absence) of sky marshals no longer holds, then that is enough to induce a unique equilibrium where passengers endogenously choose to resist hijackers in a coordinated

The economics of high-visibility terrorism

49

fashion. The role played by information in our analysis is quite nuanced. Even if all passengers are aware that there are no sky marshals on the ﬂight, so long as they are unsure of the beliefs of the other passengers, a unique equilibrium, in which passengers resist, may arise endogenously. More generally, our analysis highlights the importance of public announcements by the government, and suggests that even terrorist warnings can (under some conditions) be much more effective than is commonly thought. There is remarkably little related literature in this area, on the application of game-theoretic tools to model decision-making in the context of hijackings and hostage-taking.2 This is surprising given the importance of formulating policy responses to counteract and respond to terrorism, even before the 11 September attacks, and the applicability of available tools in the theoretical literature on game theory and models of asymmetric information. For example, Sandler and Enders (2004) identify a mere handful of papers that deal with the economic analysis of hostage-taking events, of which only three utilize game-theoretic tools in any depth – Atkinson et al. (1987), Selten (1988) and Lapan and Sandler (1988). We differ from these analyses in that we focus on the composition of terrorists, and not just the level of terrorism. Indeed a distinctive feature of our analysis is that we explore the incentives of terrorist organizations from alternative strategies – e.g. suicide-hijackers as against the more conventional hijackers who are willing to negotiate. By examining both the supply of terrorists as well as the endogenous response of potential victims of terror, we identify a potential multiplicity of equilibria, and suggest insights into how the equilibrium might change. The importance of governments coordinating their responses to terrorist actions is emphasized in other lines of research. For example, in choosing the levels of deterrence effort (or expenditure) to undertake, governments impose a ‘transference externality’ on one another when deterrence levels are chosen independently. This externality can be positive or negative, depending on the relative strength of two opposing effects: on the one hand, country A’s deterrence efforts make country B’s citizens safer (by making attacks on them less likely in country A). On the other hand, increased deterrence in country A also shifts the attacks abroad, to country B. The analysis of remedies again emphasizes the importance of increasing coordination among countries, and points out that a piecemeal approach is likely to be unsuccessful. Similarly, in choosing retaliatory action against the terrorists, coordination may be key. Typically, the cost of retaliation (in terms of attracting subsequent terrorist actions) is privately borne, whereas the beneﬁt accrues to all countries. In this situation, the classic prisoners’ dilemma prediction is that governments under-provide a resource that has value to all (Lee, 1988, Sandler and Enders, 2004). Again, as with most public good provision problems, a coordinated response can go some way to solving this problem. This is further explored in Kunreuther and Heal (2002), who examine the role of tipping mechanisms in ensuring coordinated investments and responses to terrorism. Our analysis of the coordination problem differs from these analyses in that we focus on implicit rather than explicit coordination. In other words, there is no overt coordination among the concerned parties, for example, among passengers

50

Sanjay Jain and Sharun W. Mukand

and ﬂight crew choosing whether to attack aircraft hijackers. Such communication may simply be impossible in the circumstances. Thus, convergence to a (new) equilibrium happens because public announcements by governments, by injecting strategic uncertainty amongst the affected parties, may affect beliefs sufﬁciently to ensure endogenous coordination at a unique equilibrium. Strikingly, this coordination is predicted to occur in the absence of any formal coordinating agency. The paper is organized as follows. Section 2 examines how the choice among alternative means of terrorism is inﬂuenced by terrorists’ expectations of the level of resistance that they might encounter, for example, from potential victims. In section 3, we examine the role of public announcements in generating strategic uncertainty and enabling endogenous coordination amongst potential victims. Section 4 concludes.

2.

Resistance and the supply of terrorists

We begin by looking at the ‘supply’ of terrorists – or, more accurately, at the composition of the mix of various means used by terrorists in the pursuit of their political or ideological aims. We model the ‘choice of technique’ by (potential) terrorists as the outcome of an optimization decision by the terrorists (or terrorist groups), given the underlying parameters regarding the level of resistance and deterrence they are likely to encounter, their own willingness to kill or be killed, etc. In this section, we relate this choice decision, and the resultant composition of terrorist actions, to the anticipated level of resistance from the potential victims and deterrence by government agencies. In the next section, we conduct the complementary analysis: for a given mix of terrorist actions, we describe the potential victims’ responses, and use the resultant equilibria to analyse the implications for a particular policy proposal, namely, the introduction of sky marshals on some airline ﬂights. We model heterogeneity among (potential) terrorists as resulting from variations in their ‘type’, denoted by θ, which we can interpret as the propensity towards violence, or the willingness to shed blood in pursuit of their objectives. If the terrorist decides on violent means to pursue his or her cause, then we consider two possible ways in which outcomes can be achieved.3 One is to carry out an attack with the objective of extracting concessions from the government – e.g. to seize hostages by hijacking an aircraft, with the objective of negotiating for a ransom, or the release of previously captured colleagues. We denote this course of action by N (the mnemonic for negotiation, which is the ultimate aim of carrying out the action). The other course of action that the terrorist can adopt is to launch a suicide attack (denoted by S), with the objective of causing large-scale death and destruction.4 By focusing on these two choices, we are implicitly assuming that, for terrorists, these actions are preferable to the employment of non-violent means in the pursuit of their aims.5 For each terrorist, the expected payoff from an attack depends upon: (a) the kind of attack being attempted (whether N or S); (b) the (anticipated) resistance

The economics of high-visibility terrorism

51

or deterrence, ρ, by the immediate victims (e.g. the intended hostages) and the government and security agencies and (c) the terrorist’s own type, θ . For simplicity, we assume that an attack ends in one of only two outcomes: either success or failure, denoted by the subscripts w and l respectively.6 We can write this payoff as: Eu(θ , ρ; N ) = pw (θ, ρ; N ) · uw (θ, ρ; N ) + (1 − pw (θ , ρ; N )) · ul (θ , ρ; N ) where uw (θ , ρ; N ) and ul (θ, ρ; N ) denote the utilities associated with success and failure respectively. A similar expression, with N replaced by S throughout, characterizes the expected payoff from a suicide attack. This general formulation allows for both the probability of success (failure), and the utility associated with it, to vary with θ and ρ. In order to deal with the comparative statics, and to see the intuition more clearly, we can simplify this greatly by working with speciﬁc functional forms. We start by normalizing the utility associated with both types of failed missions to 0.7 For the utility associated with the successful outcomes, we assume that it differs across the different types θ, where 0 ≤ θ ≤ 1, as follows uw (θ , ρ; N ) = ε + λN log (θ + 1) uw (θ , ρ; S) = λS log (θ + 1) where λS > λN . We can simplify this, without signiﬁcant loss of generality, by normalizing λN to 1, and as a further simpliﬁcation, set λN = 2 and 0 < ε < log 2. Figure 4.1 graphs the utility functions. The curves can be interpreted as follows: utility from a successful attack is increasing and concave in θ . For low values of θ, a negotiated outcome yields higher utility than the suicide outcome – in a sense, only the more ‘fanatical’ end of the θ spectrum prefers a successful suicide

Figure 4.1 Utility from a successful attack.

52

Sanjay Jain and Sharun W. Mukand

attack to a successful negotiated outcome, in which their demands are met. We have chosen an extremely simple functional form that accomplishes this. The log term gives us a concave increasing function. By adding ε to that term in one case, and by multiplying it by a constant greater than 1 in the other case, we can get differently shaped functions to accord with the intuition described above. For the probability of success, again, we choose extremely simple speciﬁcations. The respective probabilities of success can be written as: pw (θ , ρ; N ) = 1 − ρ pw (θ , ρ; S) = 1 − γρ where 0 ≤ ρ ≤ 1, and 0 < γ < 1. In both cases, we assume that the probability of success is declining in ρ, i.e. in the resistance faced from the passengers, aircrew, etc., as well as the deterrence from the security agencies, etc. However, this decline is more gradual for the suicide attackers than the negotiators – for example, this might be because suicide attackers are far more willing to kill passengers who resist than are terrorists who intend to use the passengers as bargaining chips in their negotiations. Further, the probability of success is always higher in a suicide attack – as Sprinzak (2000) suggests, suicide attacks are easier to plan and execute, and therefore harder to guard against, partly because they do not need a complicated escape route or rescue operation. Nor do they need the same level of operational support during the course of the attack itself – for example, to guard hostages in the case of negotiations. However, as we see below, the fact that the likelihood of success in a suicide attack dominates that from the N tactic, does not imply that the S tactic will always be chosen. Combining the probabilities and the utility expressions above, we can write: Eu(θ, ρ; N ) = (1 − ρ) · [ε + log (θ + 1)] Eu(θ , ρ; S) = (1 − γρ) · [2 log (θ + 1)] We are now in a position to deﬁne θ ∗ , the terrorist who is just indifferent between N and S, with θ < θ ∗ favoring the N action, while those for whom θ > θ ∗ will favor S. Since we let θ range between 0 and 1, thus θ ∗ is simply the proportion of terrorists who undertake the N action, or (the interpretation we use in the next section) 1 − θ ∗ is the probability that a terrorist is a suicide attacker. This leads immediately to our ﬁrst proposition. Proposition 1. θ ∗ (ρ) is decreasing in ρ. Proof.

To ﬁnd θ ∗ , for a given ρ, set:

(1 − ρ) · [ε + log (θ + 1)] = (1 − γρ) · [2 log (θ + 1)]

The economics of high-visibility terrorism

53

⇔ (1 − ρ) · ε + (1 − ρ) · log (θ + 1)] = (2 − 2γρ) · log (θ + 1) ⇔ (1 − ρ) · ε = [(2 − 2γρ) − (1 − ρ)] · log (θ + 1) ⇔ (1 − ρ) · ε = (1 + ρ − 2γρ) · log (θ + 1) ⇔

(1 − ρ) · ε = log (θ ∗ + 1) 1 + ρ · (1 − 2γ )

Note that log (θ ∗ + 1) is increasing in θ ∗ , and the partial derivative of (1 − ρ) · ε 1 + ρ · (1 − 2γ ) with respect to ρ is given by −ε · [1 + ρ · (1 − 2γ )] − (1 − 2γ )(1 − ρ) · ε −ε − (1 − 2γ ) · ε = 2 [1 + ρ · (1 − 2γ )] [1 + ρ · (1 − 2γ )]2 =

2ε · (γ − 1) [1 + ρ · (1 − 2γ )]2

which is negative, since γ < 1. Hence, by the implicit function theorem, θ ∗ (ρ) is decreasing in ρ. Figure 4.2 shows this graphically. An increase in ρ causes the expected utility from both courses of action to fall, so that the type associated with the terrorist who is just on the margin between choosing N and S falls from θ1∗ to θ2∗ . In other words, an increase in ρ leads the mix of terrorists to change, so that the likelihood of the suicide action being favoured goes up. As mentioned above, if we interpret θ ∗ (ρ) simply as the proportion of terrorists choosing the negotiate

Figure 4.2 Expected utility of N and S attacks.

54

Sanjay Jain and Sharun W. Mukand

option, then that proportion falls as ρ rises. Note that when ρ = 1, i.e. when the level of deterrence and resistance is highest, then θ ∗ = 0, i.e. only suicide missions are launched. At the other extreme, when ρ = 0, then θ ∗ is at its highest, (close to 1 as ε approaches log 2), so that (almost) all terrorists would choose to take the N action. Note that the expected utility from both courses of action falls as ρ rises, since ceteris paribus, the probability of success (from each course of action) falls. This is perfectly intuitive – attacks of both kinds become less attractive to terrorists when greater resistance is anticipated. Hence, our model suggests that although the number of both kinds of attacks is likely to drop, the proportion of extremist terrorists is likely to rise as ρ rises. It is important to note that, in the analysis above, we have used speciﬁc functional forms, chosen to accord with our intuition about the likely shapes of the relationships they capture. Hence the insight offered by the model above, that the mix of terrorist actions might shift adversely as the level of resistance and deterrence goes up, should be viewed as suggestive but not conclusive. In particular, the formulations of the success probabilities play an important role, while the assumptions about the utility functions are relatively innocuous. We develop some intuition for this below. Deﬁne ENS (θ , ρ) as the difference in expected utility between the N attack and the S attack, i.e. as: Eu(θ , ρ; N ) − Eu(θ , ρ; S) = pw (θ, ρ; N ) · uw (θ , ρ; N ) − pw (θ, ρ; S) · uw (θ , ρ; S) so that θ ∗ is deﬁned by ENS (θ ∗ , ρ) = 0. Then dθ ∗ ∂(ENS (θ ∗ , ρ)) =− dρ ∂ρ

∂(ENS (θ ∗ , ρ)) ∂θ ∗

which is negative (positive) if the numerator and the denominator have the same (opposite) signs. The intuition for the sign of the denominator is relatively straightforward. The denominator is negative (positive) if, for a given level of ρ, the expected utility from the N attack, relative to the S attack, is lower (higher) for higher θ types of terrorists. In other words, are higher θ terrorist types more likely to ﬁnd the suicide attack attractive (or at least less unattractive, if uw (θ , ρ; S) < uw (θ , ρ; N ))? Given the way that we have deﬁned θ, as the more sanguinary type, it seems only natural to assume that this should be so, i.e. that the denominator is negative. The sign of the numerator is less obvious. As ρ rises, the expected utility from both kinds of attacks falls, but the sign of the numerator depends on whether the fall in expected utility from the conventional attack is greater than that from the suicide attack for the marginal type θ ∗ . In the analysis above, this is true, but what makes the intuition for this less obvious, and more subtle, is that this need not be true for all θ types, but only at the margin, for the terrorist type θ ∗

The economics of high-visibility terrorism

55

deﬁned by ENS (θ ∗ , ρ) = 0. Here, our assumptions that the success probability of the suicide attack is higher, and is less sensitive to the level of ρ, are crucial although neither assumption is decisive.8 For the reasons discussed earlier, these assumptions appear reasonable to us – more so than the opposite extreme, in which seizing hostages for Negotiations is ‘easier’, and less sensitive to ρ, than a suicide attack. Nevertheless, at least for the sake of completeness, it is important to keep in mind that the analysis above has focused on the more intuitively appealing case.

3.

Information, coordination and mass terror

In order to focus more on the incentives of terrorists with differing motives to choose particular kinds of attacks, we deliberately ignored the important issue of coordination – e.g. among the adversely affected parties, such as potential victims, the security agencies, governments, etc. This omission is potentially important, since one of the more striking aspects of acts of terrorist-hijacking is the ability of a small set of individuals to hold a large number of individuals hostage. Indeed in many such situations, it is quite clear that if all the hostages/victims coordinated their resistance to the hijacker-terrorist, then they are likely to be successful in frustrating the hijacker-terrorist. Given the potentially large gains from coordination, it is particularly important to understand the reasons for the observed coordination failures. In this section we dissect this issue and propose a remarkably simple policy intervention that can serve as a coordinating device. At the outset it is important to emphasize that while our analysis focuses on a very simple stylized hijacker-terrorist example, the analysis is of much more general relevance.9 Indeed our analysis throws light on the role of public pronouncements, and of common knowledge in achieving coordination of anti-terror operations, not just across individuals but also countries and their security forces. As discussed in the previous section, the course of action chosen by the terrorist is either S or N (we can loosely refer to this as the ‘type’ of the terrorist). We begin by taking the proportion of ‘suicide’ terrorist-hijackers as given, and consider the likely responses of the passengers.10 Further, without loss of generality, our analytics will focus on coordination amongst two passengers or potential hostages – each with preferences given by: uj = 0 for lj < 1 and uj = j for lj = 1, where j ∈ {A, B} represent the passengers and lj < 1 is the possibility that the hostage loses his life due to the terrorist action. We allow each hostage to choose from one of two actions – they can either choose to cooperate/comply with the terrorist (aC ), or they can choose to resist (aR ). The key assumption that we make is that the hostages will be successful in resisting and overthrowing the hijacker if and only if they both decide to resist the hijacker, otherwise they will meet with zero probability of success.11 Table 4.1 captures the relevant payoffs.12 We now delineate the structure of the game. On being taken hostage, the passengers have a simple decision to make simultaneously and independently – whether to cooperate with the hijacker-terrorist or whether to resist. Observe ﬁrst that the utility functions are such that if the hijacker-terrorist was known to be a ‘negotiator’, then the unique equilibrium

56

Sanjay Jain and Sharun W. Mukand

Table 4.1 Payoff matrices with only passengers and hijackers B (a) ‘Negotiator’ as hijacker-terrorist (1 − θS ) aC aR (b) ‘Suicide’ hijacker-terrorist (θS ) A

A

aC aR

aC

aR

A , B 0, B

A , 0 A , B

aC

aR

0, 0 0, 0

0, 0 A , B

under weak dominance is for each hostage to cooperate with the terrorist, since the hostage strictly prefers to live. In contrast, if the terrorist is known to be on a suicide mission, then (using strict dominance) we have a unique equilibrium where the hostages prefer to resist, since then (with some probability) they manage to successfully overpower the terrorist. However, what makes decision making difﬁcult for the hostage is that the ‘type’ or intent of the terrorist is not known. Given that lack of information, we can analyse the strategic dilemma faced by the hostages in the form of a simple Bayesian game, where the common prior amongst the passengers that the terrorist-hijacker is on a suicide mission is θS .13 This gives rise to the payoff matrices in Table 4.1. Observe that these payoff matrices describe a simple coordination game. There are two Bayesian-Nash equilibria: if the hostages believe that the other hostages are going to resist, they will all choose to resist; in contrast, if the hostages believe that the others are going to cooperate with the hijacker, they too will strictly prefer to cooperate. In addition, the coordination failure equilibrium where the hostages prefer to cooperate with the hijackers is possibly more likely to arise when the number of agents (i.e. hostages) across whom coordination has to be ensured goes up.14 Finally, there is the issue of common knowledge of the parameters underlying the game, which forms an additional factor that is further likely to result in coordination failure. In particular, the absence of a ‘visible’ event, such as the 11 September attacks, may result in each hostage believing that the other hostage may believe that θS equals zero, and hence is unlikely to choose to resist, since for such a player the dominant strategy is always to cooperate. Therefore, ‘visible’ terrorist attacks by making common knowledge the fact that θS is positive, give rise to the possibility that coordinated resistance by the hostages takes place. However, the actual equilibrium remains indeterminate, and we have no way of pinning down which of the two equilibria is more likely. Suppose now that the government decides to introduce sky marshals in order to prevent hijackings. Such a policy intervention by the government, while likely to be effective in reducing the number of hijackings, is also very costly. However,

The economics of high-visibility terrorism

57

we now demonstrate a striking result – the public announcement of government legislation that introduces even a small percentage of sky marshals to ﬂy on planes would dramatically narrow the set of parameters for which multiplicity of equilibria holds – so that there exists a unique equilibrium where the hostages endogenously coordinate their actions and successfully overthrow the hijacker-terrorist, even if there is no sky marshal around.15 The precise structure of payoffs generated by the preferences of the hostages in the presence or absence of sky marshals is captured in Table 4.2, which is a slightly modiﬁed version of Table 4.1. The modiﬁcations reﬂect the fact that (i) the sky marshal’s preferences are assumed to be slightly different from those of the typical hostage – in particular, the sky marshal strictly prefers resistance;16 and (ii) the probability that there is a sky marshal on board is π .17 If the sky marshal acts alone then he is successful with probability p, where p ∈ [0, 1), in which case he obtains utility M . As described above, hostage A can choose from the set of strategies {aC , aR }, while for hostage B, a pure strategy is one of the four pairs generated by AH × AM , where Aj is the set of actions available to player/hostage B of type tB ( j) ∈ {tB (H ), tB (M )} where tB (H ) and tB (M ) denote a Player B of type ‘ordinary hostage’ and ‘sky marshal’, respectively. In what follows we describe the Bayesian-Nash equilibria to the game described above. If the probability that player B is a sky marshal is below a certain threshold, we have the possibility of two pure-strategy Bayesian-Nash equilibria, in one of which the hostages always resist. If the probability of player B being a sky marshal is high enough, then there is the possibility of a unique equilibrium, where the hostages offer coordinated resistance. Proposition 2. There are two pure-strategy Bayesian-Nash equilibria whenever π ≤ π ∗ or equivalently θS ≤ θS∗ . In the ﬁrst, A chooses aC while B chooses the strategy (aC , aR ), while in the second A plays aR while B plays (aR , aR ).

Table 4.2 Payoff matrices with the possibility of sky marshals (a) Game without sky marshal (1 − π ) tB (H )

A

aC aR

aC

aR

(1 − θS )A , (1 − θS )B 0, (1 − θS )B

(1 − θS )A , 0 A , B

(b) Game with sky marshal (π) tB (M )

A

aC aR

aC

aR

(1 − θS )A , 0 0, 0

(1−θS )A +θS pA ,θS pM +(1−θS )pM A , M

58

Sanjay Jain and Sharun W. Mukand If π > π ∗ , or equivalently θS > θS∗ , then we have a unique equilibrium in which A plays aR and B plays the strategy (aR , aR ). Proof. We begin our analysis by evaluating hostage A’s optimal strategy under the assumption that B’s optimal strategy is to resist the terrorist-hijacker irrespective of type i.e. B chooses (aR , aR ). Observe that since the expected payoff to A from choosing to resist and playing aR equals (1 − π )A + π A = A and the expected payoff from aC equals (1 − π )(1 − θS )A + π [(1 − θS )A + θS pA ], we have aR as a best response if, A ≥ (1 − π )(1 − θS )A + π[(1 − θS )A + θS pA ] ⇔

1 ≥π p

Observe that since p < 1, the above inequality holds ∀π ∈ [0, 1], thereby implying that there exists an equilibrium in which the hostages coordinate and resist the hijacker-terrorist regardless of their information. Now consider the pure strategy Bayes-Nash equilibrium when hostage A plays aC and B plays (aC , aR ). Observe that the expected payoff for A from aC equals, (1 − π )(1 − θS )A + π[(1 − θS )A + θS pA ] In contrast, if hostage A plays aR , he obtains, (1 − π ) · 0 + πA This implies that aC is a best response if the following inequality holds, (1 − π )(1 − θS )A + π[(1 − θS )A + θS pA ≥ π A ] Simplifying, we obtain a simple condition that aC is a best response for hostage A when B plays (aC , aR ) so long as, π∗ =

(1 − θS ) ≥ π. [1 − θS p]

Since p ∈ (0, 1), we always have π ∗ ∈ (1 − θS , 1). This suggests that so long as π ∗ ≥ π , we have two pure-strategy Bayesian-Nash equilibria – one in which (aC , aR ) is an equilibrium, and the other in which (aR , aR ) is another equilibrium. Equivalently, we have two pure-strategy Bayes-Nash equilibria for θS∗ ≥ θS , 1−π where θS∗ = 1−πp . Further it is easy to check that for π > π ∗ , there is a unique equilibrium where A plays aR and B plays (aR , aR ). Observe that the unique equilibrium, where the hostages resist, is guaranteed, even if none of the hostages is a sky marshal. More strikingly, even if there is

The economics of high-visibility terrorism

59

‘mutual knowledge’ (see Osborne and Rubinstein, 1994) that none of the other hostages is a sky marshal, so long as this fact is not common knowledge, there will be coordinated resistance in the unique equilibrium. For instance, hostage A will choose aR , even if he knows that hostage B is not a sky marshal. This might happen if hostage A believes that hostage B believes that hostage A believes that B is a sky marshal, and will therefore choose aR . Therefore, the existence of sufﬁcient uncertainty about the beliefs held by other players may be sufﬁcient to generate a unique outcome in a coordination game.18 The closest parallel to this intuition is Morris and Shin’s (1998) model of currency crises, where ‘almost common knowledge’ about an economy’s fundamentals can generate a unique ‘currency attack’ equilibrium, even when each currency speculator knows individually that the economy’s fundamentals are sound.19 The key difference is that, in our model, the possibility that one of the passengers might be a sky marshal injects uncertainty into the payoffs of the hostages. By contrast, the closest parallel to Morris and Shin’s (1998) model would be to have two identical hostages, each of whom receives a noisy signal about the ‘fundamentals’ – say, the strength of the hijacker.20 If the hijacker were perceived as weak (strong), then all the hostages would resist (not resist). And, in the intermediate range, the uncertainty about the signals received by other hostages would lead to a unique equilibrium with all hostages choosing to resist.21 By contrast, we are interested in considering the implications of the possibility of a sky marshal being on board, and how this changes the potential payoffs of the passengers, as in Table 4.2. And as we show above, if the possibility that there is indeed a sky marshal on board is high enough, (i.e. that π > π ∗ ), then the multiplicity of equilibria disappears. Analogous to Morris and Shin (1998), our model demonstrates that the introduction of uncertainty in the environment can help to eliminate the multiplicity of equilibria, but where the uncertainty in their model concerned the ‘fundamentals’ of the economy, in our model the introduction of sky marshals injects uncertainty regarding the payoffs of each of the hostages.22 In our example, the government, by making a public announcement about the introduction of sky marshals in some ﬂights, injects strategic uncertainty, a remarkably effective policy intervention, which forces potential hijackers to consider the fact that they are likely to face coordinated resistance. Of course, for this coordinated response to come about, θS must be positive and common knowledge – something we can reasonably assume is achieved by events of mass terror such as those of 11 September. Moreover, also observe that, for a given π , if it were perceived that θS had increased sufﬁciently (say because of the reasons analysed in section 2), that would also be sufﬁcient to generate a unique equilibrium in which the hostages coordinated their resistance against the hijacker-terrorist. The above example is simpliﬁed and ignores important considerations relevant to effective policy intervention. However, it throws light on an important policy issue for effective anti-terror legislation. In particular, policy makers have worried about the costs of effective security measures to prevent (among other things) airline hijackings. Indeed, it is frequently argued that, given the volume of air

60

Sanjay Jain and Sharun W. Mukand

travel in the US, it is infeasibly expensive to have sky marshals on every ﬂight. Our analysis says that once we take into account ‘almost common knowledge’ considerations, a public announcement of the introduction of sky marshals in some ﬂights would be a very effective policy measure.23 Remarkably, this policy intervention would be effective even if only a small fraction of the ﬂights had sky marshals. This is because passengers will offer coordinated resistance and choose aR not only in the absence of any sky marshals, but even when they are aware that a sky marshal is not on the ﬂight. While our analysis suggests that public announcements play an important role in coordinating actions, we should be cautious and re-emphasize that our analysis is best treated as a ﬁrst step. This is especially true not just because our example was highly stylized, but also because we have neglected other important aspects of public warnings and terror alerts. Nevertheless, we believe that further exploration of issues of information, common knowledge and coordination, in the nascent literature on the economics of terrorism, may have potentially large payoffs.

4.

Conclusion

We have analysed some implications of visible, discrete, large-scale terrorist actions, such as the 11 September attacks. By making common knowledge the fact that terrorists are willing to commit suicide, such events of mass terror have direct implications on both the supply side of terrorism as well as the response to it. First, this paper has presented results on the implications of increased anti-terror operations on the quantity as well as the mix of terrorists supplied by terrorist organizations. We ﬁnd that while the level of terrorist incidence may decrease, the mix is likely to worsen, with a greater proportion of suicidal terrorist attacks.24 Second, we also present a novel argument for the role of public announcements in endogenously coordinating responses in the face of a terrorist-hijacking. Public announcements, by injecting strategic uncertainty, can give rise to a Pareto-superior outcome, at minimal resource cost to the government. An appealing aspect of our analysis is its simplicity. Nevertheless, it is important to emphasize that our results should be treated as no more than a ﬁrst step. There are several directions which deserve to be explored further. First, we would like to have a more comprehensive general equilibrium analysis of the set of constraints and opportunities facing terrorist organizations. Indeed such an analysis will give us a better understanding of how seemingly irrational ‘suicide’ terrorists are very rational in the pursuit of their aims (see Sprinzak, 2000, for a discussion). Furthermore, such an analysis will throw light on the effectiveness of alternative measures of combating terrorism. Second, we would like to further explore the role of common knowledge and public announcements in combating terrorism. While our example was stylistic and context-speciﬁc, it is perhaps a useful ﬁrst step in the analysis of the role of public ‘warnings’ and terror alerts – so ubiquitous after 11 September.

The economics of high-visibility terrorism

61

Acknowledgements A previous version of this paper was presented at the workshop on ‘The Economic Consequences of Global Terrorism’, 14–15 June, 2002, DIW Berlin, Germany. For helpful comments and discussions, we are grateful to Simon Anderson, Albert Choi, Maxim Engers, Sumit Joshi, Sandeep Kapur, participants at the workshop, and especially our discussant Todd Sandler. For very helpful discussions and assistance, we are grateful to Elio Valladares. The detailed comments of two anonymous referees and the editors, especially Tilman Brück, have signiﬁcantly improved the content and exposition.

Endnotes 1 This point has been forcefully made by Sprinzak (2000), who argues that ‘suicide terrorism has inherent tactical advantages over ‘conventional’ terrorism: It is a simple and low-cost operation (requiring no escape routes or complicated rescue operations); it guarantees mass casualties and extensive damage (since the suicide bomber can choose the exact time, location, and circumstances of the attack); there is no fear that interrogated terrorists will surrender important information (because their deaths are certain); and it has an immense impact on the public and the media (due to the overwhelming sense of helplessness)’. 2 For a survey, see Sandler and Enders (2004). 3 More generally, the terrorist (group) can choose from a range of possible actions. For simplicity, the model focuses on only two strategies. 4 It is important to note that the ‘payoff’ to the terrorist comes not from his own death per se, but from the scale of the death and destruction, with the attendant wide publicity that such an action brings to one’s cause. 5 An alternative statement of this is that we are implicitly considering only that portion of the support of the distribution of θ which does not include the ‘peace-oriented’ types. In the next section, where we consider the likely responses of passengers confronted with an aircraft hijacker, this is a reasonable simpliﬁcation to make – by a revealed preference argument, one can conclude that the hijacker has a ‘high enough’ θ. 6 For the sake of concreteness, it may be helpful to think of the attack under consideration as being the hijacking of an aircraft, so that N and S correspond to the different objectives of using the passengers as hostages in bargaining, versus an attack like those of September 11, where the objective is large-scale death and destruction. In the next section especially, we use aircraft hijacking to motivate the model, but we suggest that the analysis applies more generally. 7 In some sense, we are assuming that a ‘failure is a failure’, regardless of the type of mission. It is possible to extend the model to show that, so long as the difference in the utility from a failed N mission versus a failed S mission is ‘small’, then our result still goes through. Details are available from the authors. ∂pw (θ ∗ ρ; N ) 8 This can be seen by writing out the numerator as: uw (θ ∗ , ρ; N ) − ∂ρ ∂p (θ ∗ , ρ; N ) ∂p (θ ∗ , ρ; S) ∂pw (θ ∗ ,ρ; S) and w ∂ρ are both uw (θ ∗ , ρ; S) and noting that (i) w ∂ρ ∂ρ

negative but the former is larger in magnitude, because the success probability of a suicide attack is less sensitive to ρ; and (ii) uw (θ ∗ , ρ; N ) > uw (θ ∗ , ρ; S), because θ ∗ is deﬁned by pw (θ ∗ , ρ; N ) · uw (θ ∗ , ρ; N ) = pw (θ ∗ , ρ; S) · uw (θ ∗ , ρ; S), and pw (θ ∗ , ρ; N ) < pw (θ ∗ , ρ; S). Note that neither (i) nor (ii) is either necessary or sufﬁcient. 9 As pointed out by a referee, while the example is described in the context of an aircraft hijacking, the insights can be applied more generally to the analysis of coordinated

62

10 11 12

13

14 15 16

17 18 19

20 21 22 23 24

Sanjay Jain and Sharun W. Mukand responses to crimes (for example, on passengers using public transport) when there is a possibility that security personnel are around. For simplicity, we combine all the adversely affected parties, e.g. the passengers, aircrew, air security ofﬁcials, etc., under the rubric of ‘passengers’. This assumption is stronger than is necessary for our results. All that is crucial for our results is the assumption (as in the earlier section) that the probability of success is an increasing function of the number of hostages who simultaneously choose to resist. In Table 4.1, and in what follows, we make a number of simplifying assumptions that are stronger than necessary to establish our results. For example, in Table 4.1, we have assumed that the passengers’ payoff from coordinated resistance against a suicide hijacker is the same as that against a ‘negotiator’ hijacker. This simplifying assumption can be relaxed without qualitatively affecting the analysis. Details are available from the authors. For clarity of exposition, we use θS to denote the perceived probability assigned by the passengers to the event that the terrorist is a suicide terrorist. This is to distinguish it from the actual proportion of suicide terrorists, which is 1 − θ ∗ in terms of the notation of section 2. Of course, in equilibrium, the beliefs held by the passengers will be borne out, so that 1 − θ ∗ = θS in equilibrium. However, in order to make this point in a more systematic way we would need to considerably augment the existing framework. The closest analogue to the intuition underlying our model is that in Morris and Shin’s (1998) model of currency crises. We discuss the parallels and differences between the two models in more detail below. In the analysis that follows, the sky marshal always prefers to resist and prevent a hijacking by assumption. This simplifying assumption captures the general idea that the sky marshal (like ﬁreﬁghters, policemen and soldiers) is trained always to prefer resistance to passive compliance. The precise channel through which this is accomplished – whether by special training, or access to weapons or indeed differences in preferences – is not central to our analysis. For simplicity, the notation in Table 4.2 assumes that passenger B is the possible sky marshal. We ignore the possibility that both passengers may be sky marshals. The fact that a ‘small’ amount of uncertainty can dramatically alter equilibria was ﬁrst pointed out by Rubinstein (1989), and is further explored in Morris and Shin (1997). In that model, each currency speculator independently receives a noisy signal about the fundamentals of the economy. If the fundamentals are ‘weak’ (‘strong’), then there is a unique equilibrium in which the speculators attack (do not attack) the currency. However, if the fundamentals are in the intermediate range, then there may be multiple equilibria, in which the speculators attack or not, depending on their perception of what the other speculators are doing. Morris and Shin (1998) show that, if the true state of the world (‘fundamentals’) is not known, then uncertainty about the signals received by others might cause all speculators to attack, so that the multiplicity of equilibria in the ‘intermediate region’ disappears. Note that this would be distinct from whether the hijacker is a suicide terrorist or not. Loosely speaking, this is analogous to the game described in Table 4.1, with no sky marshal, but with the addition of independent signals about the fundamentals (i.e. strength) of the hijacker, to each of the passengers. We are grateful to an anonymous referee for helping us to distinguish these insights more clearly. Obviously, for the policy to be effective, the precise ﬂights would not be identiﬁed. Hence, somewhat paradoxically, it may be ‘safer’ to ﬂy after the September 11 attacks than it was before, in the sense that the likelihood of hijackings declines. However, as Sandler and Enders (2004) speculate, consistent with the predictions of our model, terrorist attacks are more likely now to be low-probability but high-fatality events.

The economics of high-visibility terrorism

63

References Atkinson, S., Sandler, T. and Tschirhart, J. (1987) ‘Terrorism in a bargaining framework’, Journal of Law and Economics 30(1): 1–21. Becker, G. (1968) ‘Crime and punishment: an economic approach’, Journal of Political Economy 76(2): 169–217. Krueger, A. and Maleckova, J. (2002) ‘Education, poverty, political violence and terrorism: Is there a causal connection’, NBER Working Paper 9074, Cambridge, MA: National Bureau of Economic Research. Kunreuther, H. and Heal, G. (2002) ‘Interdependent security: the case of identical agents’, NBER Working Paper 8871, Cambridge, MA: National Bureau of Economic Research. Landes, W. (1978) ‘An economic study of US aircraft hijackings, 1961–76’, Journal of Law and Economics 21(1): 1–31. Lapan, H. E. and Sandler, T. (1988) ‘To bargain or not to bargain: That is the question’, American Economic Review Papers and Proceedings, 78: 16–20. Lee, D. R. (1988) ‘Free riding and paid riding in the ﬁght against terrorism’, American Economic Review Papers and Proceedings, 78: 22–6. Morris, S. and Shin, H.-S. (1997) ‘Approximate common knowledge and coordination’, Journal of Logic, Language and Information 6: 171–90. Morris, S. and Shin, H.-S. (1998) ‘Unique equilibrium in a model of self-fulﬁlling currency attacks’, American Economic Review 88(3): 587–97. Osborne, M. and Rubinstein, A. (1994) A Course in Game Theory, Cambridge, MA: MIT Press. Rubinstein, A. (1989) ‘The electronic mail game: strategic behaviour under almost common knowledge’, American Economic Review 79(3): 385–91. Sandler, T. and Enders, W. (2004) ‘An economic perspective on transnational terrorism’, European Journal of Political Economy 20(2): 301–16. Sandler, T., Tschirhart, J. and Cauley, J. (1983) ‘A theoretical analysis of transnational terrorism’, American Political Science Review 77(1): 36–54. Sprinzak, E. (2000) ‘Rational fanatics’, Foreign Policy 120, Sept/Oct 2000: 66–73. Selten, R. (1988) ‘A simple game model of kidnappings’, in R. Selten (ed.) Models of Strategic Rationality, pp. 77–93, Boston: Kluwer Press. Townsend, R. (1979) ‘Optimal contracts and competitive markets with costly state veriﬁcation’, Journal of Economic Theory 21(2): 265–93.

5

Decomposing violence: political murder in Colombia, 1946–99 Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

In this paper, we apply the Hodrick-Prescott and Beveridge-Nelson business-cycle decomposition methods to a time-series of homicides in Colombia (1946–99). Separating out ‘permanent’ from ‘cyclical’ murder, we hypothesize that the cyclical part coincides with the periodic political unrest in the country. The results show a good match between the political events in the country and the computed cyclical murder component.

1.

Introduction

The countries of Latin America record, by far, the highest homicide rates in the world, averaging 20 to 30 murders per 100,000 people, i.e. two to three times as many as the next most violent regions of the world (Guerrero, 1998; Londo˜no, 1998:72). Within Latin America, Colombia is known for its extremely high levels of homicidal violence, resulting in one of the highest murder rates in the world. According to Colombian National Police statistics, homicides increased from around 5,000 per year in the 1950s and 1960s to about 10,000 per year by 1980 and to about 25,000 per year by 1990 (Figure 5.1). A further surge to nearly 30,000 murders per year was seen in the early 1990s. This has moderated somewhat but, in absolute numbers, still hovers between 20,000 to 25,000 per year. When adjusted for population growth, i.e. homicides per 100,000 people in the population, a similar pattern emerges. A rapid per capita murder increase occurred from 1946 to the late 1950s, followed by a ten-year period of sustained murder reductions (Figure 5.1). This downward trend reversed in the 1970s and then shows exactly the same pattern as for the absolute numbers. The country’s murder rate varies substantially not only over time, but also from region to region (e.g. Dinar and Keck, 1997:9–10; Guerrero, 1998:96–97; Londo˜no, 1998:76), with rates as low as 16/100,000 (nearly ‘normal’ as compared to the rest of the world) to rates as high as 900/100,000 in the city of Apartado (Guerrero, 1998:97). Frightful as the absolute and population-adjusted numbers are, they underestimate the truth. Following a survey, Rubio (1998a:606) writes that even for murder ‘more than half of the households victimized stated that they had “not done anything”, and only 38 per cent reported that they had made a formal complaint’ to

Decomposing violence: political murder in Colombia, 1946–99 65

Figure 5.1 Total homicides, Colombia, 1946–1999.

the authorities. Incredibly, by comparing separate statistical reporting by the police and by the justice agencies Rubio ﬁnds wide disparities for more than a quarter of Colombia’s municipalities. The disparities are largest in municipalities characterized by the presence of any armed force (military, para-military, drug-gangs, guerrillas; Rubio, 1998a:607). Apparently, victims’ families fear reprisals. Without doubt, Colombia’s murderous violence is related to two of its most salient features, the drug trade and its civil war, that have marred the country for decades. Less well known and appreciated is that these two factors account only for a portion of all murders in the country (Guerrero, 1998:98). For murder, the primary risk factors are alcohol consumption, possession of ﬁrearms, and weekends. For example, a quarter of all murders take place on Sundays, more than half on Friday, Saturday, and Sunday, with disproportionate increases on holidays. Most murders are patently non-political, take place at night, in urban areas, are committed by poor people on poor people, and alcohol is frequently found in the victims (Londo˜no, 1998:75). But Guerrero observes that while alcohol consumption might explain the high levels of murder it cannot explain the drastic murder increase in Colombia in the 1980s and 1990s (1998:98). Also contrary to popular perception, several studies have failed to establish links between murderous violence and poverty rates, unemployment rates, urbanization rates, or rates of economic growth (Londo˜no, 1998:74; Guerrero, 1998:97). Indeed, Rubio (1997) and others have made persuasive arguments according to which the educated and uneducated classes both engage in criminal and violent activity for the simple reason that crime pays well. Income and education are no longer linked, but income and crime are (Rubio, 1997:812). Average annual incomes from crime have been variously estimated at up to $70,000 per person, a huge premium over Colombia’s per capita 1995 GDP of around $1,800 (Bejarano, 1997:12). The break-down of Colombia’s justice system

66

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

also encourages criminal and violent behavior as the probability of being caught, tried, and convicted is becoming smaller over time. By 1994, conviction rates had dropped to below four per cent (Rubio, 1998a:606), and sentences rarely exceeded six months of jail time (Rubio, 1998b:91). There is widespread agreement among analysts that Colombia’s violence is costly, both at the microeconomic level (e.g. Dinar and Keck, 1997) and at the macroeconomic level, estimated at up to 15 per cent of GDP (Bejarano, 1997:10). There is also evidence that major perpetrators of violence – the military and paramilitary forces, the drug-traders, and the various guerrilla groups – act in semi-collusive fashion to keep the spoils of war going (Richani, 1997), evidence almost perfectly in line with the theory suggested by Brito and Intriligator (1992). While ongoing research will have to identify and disentangle the various causes and possible intervention mechanisms of extreme, generalized violence in Colombia, our concern in this paper is much more narrow and limited. On the hypothesis that political unrest and politically motivated murder are cyclical, we apply business-cycle decomposition methods to the murder time-series and then compare the estimated (quantitative) cyclical or ‘transitory’ component to a narrative (qualitative) account of cycles of political violence in Colombia. We ﬁnd a good overlap between our estimates of turning points in the cycle and the narrative of the country’s political ups and downs: when the political context suggests much unrest, the cyclical component of murder increases, and vice versa. The trend or ‘permanent’ component of the series is interpreted as that part of the murder time-series that would have occurred without political violence. The research reported here is a pure time-series study. We make no claim as to the magnitude of ‘political’ murder, but we do suggest that we may have found a way to gauge turning points in the ‘political’ murder cycle. Especially for countries facing domestic terrorism, this would be an important advance, possibly permitting real-time assessment of whether a country ﬁnds itself in a cyclical murder upswing or downswing. Once the time-series is separated into ‘political’ and ‘non-political’ murder, further research can build underlying structural models for each part of the series. The next section presents a discussion of the data and the decomposition methods employed. This is followed by the interpretation of our ﬁndings, i.e. the matching of the cyclical component to the political narrative. The ﬁnal section concludes.

2.

Data and methods

The Colombian National Police has a record of crime statistics that reaches back to 1946 (various issues of Revista Criminalidad). The statistics distinguish among 14 types of crime. Type number 13 (crimes against life and person) contains some 20 categories, such as murder, abortion, and personal injury. The murder categories are murder (homicidio) and aggravated murder (homicidio agravado). Since 1993, there is also an attempt to separate out further types of murder, namely murder with terrorist intent (homicidio con ﬁn terrorista) and death associated with the exercise of ofﬁcial police duties (homicidio con función, razón cargo o ejercicio

Decomposing violence: political murder in Colombia, 1946–99 67 de sus funciones). For our analysis we use the number of murders for these four categories combined, i.e. we collapse all these categories of murder into a single group.1 Time series can be broken into constituent components. Since we use annual data, the seasonal component does not apply to our case. The long-run trend component is often simply modelled as a linear or non-linear trend line over time, and the cyclical component is the remaining variation around this trend. We apply the Hodrick-Prescott (1997) and the Beveridge-Nelson (1981) decomposition techniques. Without going into any detail here, none of a variety of simple time-trend, moving average, and exponential smoothing techniques are appropriate for our purpose.2 For business cycle research, a popular method to smooth a time series and produce its long-run trend component is the Hodrick-Prescott ﬁlter (1997). Technically, consider T t=1

(yt − st )2 + λ

T −1

((st+1 − st ) − (st − st−1 ))2

(1)

t=2

where y is the original series, t is time (t = 1, . . . , T ), and s is the smoothed series. The ﬁlter produces the smoothed series by minimizing equation (1) whose ﬁrst term denotes the squared difference between an original and its smoothed value at time t (i.e. the variance) and whose second term deﬁnes upper and lower bounds of s, where λ = 0 returns the original series and λ → ∞ produces a linear smoothed series. Since we deal with annual data, we use the recommended default value of λ = 100. E-Views 4 contains a Hodrick-Prescott routine that produces the relevant smoothed trend estimates which, along with the original data and the Beveridge-Nelson results, are displayed in (Figure 5.2).3 Deviations from the trend constitute the cyclical component in (Figure 5.3).

Figure 5.2 Original data (total homicides per 100,000 people) and permanent component according to HP and BN methods.

68

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

Figure 5.3 Cyclical component according to HP and BN methods.

Our hypothesis is that, in Figure 5.3, an upward movement of the cyclical component of the homicide time-series corresponds to time-periods with marked political unrest whereas a downward movement of the cyclical component denotes relative political calm in Colombia. We offer a discussion in the ‘Interpretation’ section.4 In 1981, Beveridge and Nelson (BN) introduced a new business cycle decomposition technique. The objective was to produce a technique with superior turning-point performance. Since the estimation of data points in time t completely relies on past values, t − 1, t − 2, . . . , t − n, BN describe their method as a ‘real-time’ technique: plug new data points into the model as data become available, and it will tell you whether the business cycle is turning. The technique requires to ﬁrst ﬁt an ARIMA model on the ﬁrst-differenced natural logs of the dependent variable. Since in the spirit of BN we do not propose any structural model explaining the data movement, nor propose to engage in any forecasting of the data points, we conducted an unabashed best-ﬁt search which resulted in an ARIMA(0,1,13) model with moving average terms at lags 1, 5, and 13. The results are: d(lthompc) =

c

+ ma(1)

+ ma(5)

+ ma(13)

Coefﬁcients t-stats p-values

0.0324 3.2710 0.0005

0.2569 3.2132 0.0023

−0.4911 −5.7861 0.0000

−0.5376 −5.8794 0.0000

Adj R2 = 0.1024; DW = 2.0445; F = 2.9780; p-value(F) = 0.0404

where d(lthompc) is the ﬁrst difference of the natural log of total homicides per 100,000 persons in Colombia, 1946 to 1999. Estimating this model minimized the Akaike Information Criterion (AIC) as well as the Schwartz Criterion (SC).

Decomposing violence: political murder in Colombia, 1946–99 69 We ran the model with RATS 4 as well as with E-Views 4 and, except for rounding, obtained the same estimates.5 Once a model that best ﬁts (or reproduces) the original data is estimated, the permanent and cyclical components can be extracted from the estimated data points. The resulting data are displayed in Figures 5.2 and 5.3.6,7,8

3.

Interpretation of results

It would appear that the Hodrick-Prescott (HP) and Beveridge-Nelson (BN) methods yield substantially different permanent trend information (Figure 5.2). But both result in comparable estimates of the cyclical movement (not in magnitude but in turning points), at least for the early (up to 1965) and the late years (as from 1980) of the series. From 1965 to 1980, the HP method tracks the actual data almost without deviation because there is nothing to be smoothed in the original data series. In contrast, the BN method produces cyclical information throughout the entire time-period. (HP is a smoothing technique whereas BN is a trendextraction technique.) We therefore focus here on interpreting the political events of Colombia in light of the BN cyclical component we extracted from the data series. The chronology and event description is taken from Bushnell (1993).9 It appears, in Figure 5.3, that there are three major periods in the cyclical component of murder in Colombia. The ﬁrst occurs from 1946 to 1958, with a short-term decline in the rate of increase from 1952 to 1957. The second period occurs from 1958 to the late 1970s. One could also argue that the second period lasts from 1965 to 1985, a twenty-year period during which the estimated cyclical (‘political’) murder lies below the permanent trend line. But since we cannot be sure about the actual number of murders, which depends on the quality of record keeping, it is best to focus on turning-points and movements, rather than magnitudes, of the series. The third period occurs as from the late 1970s when an upsurge in murder is observed, until the series peaks in 1991. Thereafter the cycle declines once more. The Conservative Party’s Mariano Ospina Perez assumed the country’s presidency by election in August 1946. On 9 April 1948, Jorge Eliecer Gaitán, a charismatic Liberal Party presidential candidate was murdered by unknown assailants. This murder initiated a violent outburst in Botogá, the Bogotazo, that began the period commonly known as La Violencia, reﬂected in the surge of murders tracked in Figure 5.3. This period lasted until 1957, a period where political violence primarily, but not only, emerges from the confrontation between adherents to the two major political parties. Following Gaitán’s murder, Liberals achieved victory in congressional elections in June 1949, but presidential victory is obtained by the Conservative Party’s Laureano Gómez in November 1949. Political violence with leftist guerrillas, who view both the Liberal and the Conservative Party as establishment parties of land-owners and other vested economic interests, continued to such an extent that the military took power in 1953 under General Gustavo Rojas Pinilla. This resulted in an initial decrease in the pace of violence (see the decline in Figure 5.3 in 1953 and 1954) but picked

70

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

up in the latter years of Rojas Pinilla’s tenure. Murderous violence ﬁrst declined as killings originated by guerrillas were stopped but then increased as the military began to kill increasing numbers of the guerrillas. Rojas Pinilla’s initial success led the national assembly to elect him to a full four-year presidential term in 1954. But by May 1957 Rojas Pinillas’ inability to put an end to La Violencia led to a nation-wide general strike. A military junta took control, and Rojas Pinilla went into exile. The military junta yielded to a bipartisan coalition, called the National Front, that remained in control until 1978. The National Front amounted to an agreement between the Liberal and Conservative Parties to trade presidential terms and to adhere to a quota system in the assignment of ministerial and other government posts. This should reduce the amount of murder due to the clash between the parties’ adherents, as indeed it did (see the cyclical decline in Figure 5.3). The ﬁrst National Front president was Alberto Lleras Camargo (Liberal). President from 1958 to 1962, he was followed by Conservative Party member Guillermo León Valencia (1962–66). The third president (1966–70) was Liberal Carlos Lleras Restrepo (not related to the ﬁrst Lleras), during whose term murder rates fell to 20/100,000, the lowest since the Bogotazo and almost normal by other countries’ standards. The fourth National Front president was Conservative Misael Pastrana Borrero (1970–74). Various old and new guerrilla movements, which had been active for several decades, were formalized in the 1960s. The Fuerzas Armadas Revolucionarias de Colombia (FARC) was formed in 1963, the Ejército de Liberación Nacional (ELN) in 1965, the Ejército Popular de Liberación (EPL) in 1967, and the M-19 (Movimiento 19 de Abril) in 1970. The end of the National Front period occurred between 1974–78, with the presidency of Liberal Alfonso López Michelsen. During his tenure, Colombia experienced considerable economic growth, driven by a commodity export boom (coffee, coal, oil, and marijuana; not yet cocaine). It is generally accepted that Michelsen’s administration suffered from widespread corruption and involvement with the emerging drug trade. It is also the period when guerrilla activity took an upward turn, reﬂected in Figure 5.3 in the cyclical upswing of homicides. The twenty-year period of the National Front ended, as did the relative political calm, when another Liberal won the presidential election (Julio César Turbay Ayala; 1978–82). His term of ofﬁce saw an end to Colombia’s economic boom and an upsurge in revolutionary activity. For example, in December 1979, in a dramatic move that made the world’s news headlines, the M-19 guerrilla group seized the Embassy of the Dominican Republic in Bogotá and held thirteen ambassadors hostage who had met at the embassy. The hostages included the US envoy. Both the HP and the BN cyclical murder series show an upward movement of murder for the Turbay years (Figure 5.3). The next election was won by Conservative Belisario Betancur Cuartas (1982–86). He began peace negotiations between government and guerrillas (except with the relatively small ELN). The cyclical component of our series declines in 1982 and 1983, holding even in 1984 (HP shows a further decline in 1984). But the peace negotiations did not result in

Decomposing violence: political murder in Colombia, 1946–99 71 guerrilla demobilization and weapons surrender. Instead, in 1985, three presidential candidates were assassinated; moreover, the M-19 occupied the Colombian Justice Palace (the seat of Colombia’s Supreme Court) and many judges were murdered. Betancur broke off peace negotiations, and our series shows an up-tick in homicides. The period of the greatest number of murders in Colombia is now ushered in. Liberal Virgilio Barco Vargas, an MIT-trained civil engineer, was elected president (1986–90). Even though a former M-19 member, Bernardo Jaramillo Ossa, founded a new political party, the Unión Patriótica, there was now open violent conﬂict among government, various guerrilla and paramilitary groups, and – increasingly – drug trafﬁckers as cocaine replaced marijuana and the dollar-volume at stake became ever larger. A major break occured when Liberal César Augusto Gaviria Trujillo (1990–94) became president and pushed the legislature to adopt a new constitution in 1991 (to replace the constitution of 1886). A priori, there is no particular reason why a mere change in the country’s constitution should mark the beginning of the pronounced fall in the cyclical murder series (Figure 5.3). But at age 43 Gaviria was the ﬁrst person of the post-La Violencia generation to become president, and the new constitution was received to general acclaim across the country, swept up in a spirit of renewed hope for peace. More substantially, Gaviria did appoint a former guerrilla leader to his cabinet and pushed through large-scale increases in the government’s social expenditure budget, moves that might be interpreted as accommodating rebel demands and leading to declining ‘political’ (i.e. cyclical) murder. At the same time, the increasing dollar-value of the drug-trafﬁc may have pushed up the slope of the permanent murder series (Figure 5.2), perhaps reﬂecting entangled political and economic interests.10 Indeed, Colombian ‘popular’ knowledge across the spectrum of opinion asserts that the administration of Liberal Ernesto Samper Pizano (1994–98) was ‘controlled’ by the predominant drug cartels (Medellín y Cali), with increasing participation in that trade by the main guerrilla groups. If so, there would be no ‘need’ for political violence which switches to become non-political permanent violence. Another Liberal president, Andrés Pastrana Arango (1998–2002), in fact went so far as to grant in 1999 the largest rebel group, the FARC, a demilitarized zone the size of Switzerland (42,000 km2 ) and placed it under the FARCs administrative control. The post–1991 period might be summarized as inclusion (Gaviria), collusive cooption (Samper), and appeasement (Pastrana), all of which would suggest the very decline in cyclical violence we extracted from the data. Despite the cyclical decline in murder in the 1990s, a stable collusive arrangement to jointly exploit the country’s resources (á la the National Front of 1958 to 1978) appears not possible among the major vested interests – the government and its military units, the paramilitary units, the drug-producers, and the various guerrilla groups – and there remains, in our opinion, only the possibility of a renewed surge in political violence and murder. Indeed, Pastrana’s appeasement arrangement with the FARC ended in February 2002 as it became clear that it had no intention of keeping a peace that would leave it restricted to only

72

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

part of the country. It’s ideological concerns not having been met, and unwilling to be locked into a limited territory from which to extract ﬁnancial resources, it has resumed its terrorist attacks on life and property of ordinary Colombians as well as on government property and ofﬁcials. All sides to the conﬂict still appear to believe that each can triumph over the others for a larger share of the spoils. As further data points become available we would expect an upturn in the BN and HP-cyclical component of Colombia’s murder time-series.

4.

Conclusions

This paper is, primarily, a statistical exercise to demonstrate how one might get at turning points in (state and non-state) domestic terrorist activity. Other techniques could be, and will be, tried. It bears repeating, though, that our approach here, in contrast to the political-business cycle literature and other approaches, is a pure time-series and not a structural regression effort. Our narrow interest is not to match quantitative RHS variables with a quantitative LHS variable. In fact, we do not have a LHS variable unambiguously called ‘political murder’. Instead, the effort here is to create a proxy of such a variable – the cyclical murder component – and to learn if we can match it with the generally accepted political narrative of political events in Colombia. It would appear that we can: the cyclical component of Colombia’s homicides strongly coincides with the country’s political events. In future research, we plan to apply other statistical techniques as well as to construct structural models to explain the movement of the permanent and cyclical murder series in Colombia and also to repeat the decomposition exercise with subnational data (i.e. by administrative departamento). Provided that consistently collected ﬁgures are available, we believe that an application of the decomposition method is also possible and worthwhile for countries other than Colombia.

Acknowledgments We thank the Margaret Ann Isely Foundation, Eugenia Almand, and Barbara Bertram for hospitality and ofﬁce space during a research stay in Denver, CO (March 2002). For assistance with literature search, we thank Matthias Spoerle. For comments on a draft paper we thank participants at international conferences in Berlin, London, and Lisbon (all in June 2002) and for written, constructive comments we thank two anonymous referees as well as Manolis Athanassiou, Thad Brown, Kostas Drakos, Charles R. Nelson, Nadir Öcal, Zacharias Psaradakis, Ron Smith, Jülide Yildirim, and Leonid Zarangas. The usual disclaimer applies.

Appendix 5.1:

The Beveridge-Nelson decomposition of a time-series

Let wt be the stationary ﬁrst differences of a non-stationary series zt wt = zt − zt−1

(A1)

Decomposing violence: political murder in Colombia, 1946–99 73 Wold’s (1938) decomposition theorem states that wt = µ + λ0 εt + λ1 εt−1 + · · · , where λ0 ≡ 1

(A2)

and λi are constants and the ε’s are uncorrelated errors. Beveridge and Nelson [BN] (1981) relate each zt to its own future values or ‘forecast proﬁle’, where the proﬁle in time t is interpreted as zt ’s ‘permanent’ component, z¯t . Thus, the estimated zt forecast k periods ahead becomes the expected value of zt+k , conditional on zt ’s past values zˆt (k) = E(zt+k |. . . , zt−1 , zt )

(A3)

which, since w accumulates past z’s, may be written as zˆt (k) = zt + E(wt+1 + · · · + wt+k | . . . , wt−1 , wt ) = zt + wˆ t (1) + · · · + wˆ t (k) (A4) But from (A2), each estimated w, say wt+i , is wˆ t (i) = µ + λi εt + λi+1 εt−1 + λi+2 εt−2 . . . = µ +

∞

λj εt+i−j

(A5)

j=i

Substituting (A5) recursively into (A4) and approximating to an inﬁnite time horizon, we obtain ∞ ∞ zˆt (k) ≈ kµ + zt + λi εt + λi εt−1 + · · · zˆt (k) − kµ ≈ zt +

1 ∞

λi εt +

2 ∞

(A6)

λi εt−1 + · · ·

2

1

Beveridge and Nelson (1981) suggest that it is natural to interpret the LHS of (A6) as zt ’s ‘permanent’ component, denoted as z¯t . The cyclical component, ct , then is ∞ ∞ z¯t − zt = λi ε t + λi εt−1 + · · · (A7) 1

2

The unknown µ and λi ’s in (A5) must be estimated. Beveridge and Nelson suggest an ARIMA procedure of order (p,1,q) with drift µ. wt = µ +

1 − θ 1 L 1 − · · · − θq L q

1 − φ1

L1

− · · · − φp

Lp

εt = µ +

θ (L) εt φ(L)

(A8)

74

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

Cuddington and Winters (1987, p. 127, equation (7)) realized that in the steady state, i.e. L = 1, (A8) reduces to 1 − θ 1 − · · · − θq θ (1) εt = µ + z¯t − z¯t−1 = µ + εt (A9) φ(1) 1 − φ 1 − · · · − φp where µ will be the estimated mean, θi the estimated moving-average terms, and φi the estimated autoregressive terms. Now iterate (A9) recursively, i.e. replace t by (t − 1) and (t − 1) by (t − 2), etc. Then we get z¯t − z¯t−1 = µ + z¯t−1 − z¯t−2 = µ +

θ (1) εt φ(1) θ (1) εt−1 φ(1)

(A10)

.. . z¯1 − z¯0 = µ +

θ (1) ε1 φ(1)

Adding these equations, the terms on the LHS cancel out except for z¯t and z¯0 , and on the RHS µ is added ‘t’ times and the fraction in the second term on the RHS is a constant to be multiplied by the sum of error terms. Thus, we obtain θ (1) εi z¯1 = z¯0 + µt + φ (1) t

(A11)

i=1

This is, except for notation, Newbold’s equation (Newbold, 1990, p. 457, equation (A6)) and the problem reduces to ﬁnding an initial value for z¯0 . Mauricio Cárdenas (1991), in his unpublished dissertation, suggests that z¯0 = z0 , i.e., the very ﬁrst data point of the original series. This makes intuitive sense. Since a forecast proﬁle conditional on past values cannot be computed in the absence of past values, z¯0 = z0 by deﬁnition. Formally, Cárdenas suggests the following (we changed the notation to conform to ours here), where zt refers to the original data series, zt − zt−1 = µ +

p i=1

φi zt−1 +

q

θj εt−j + εt

(A12)

j=1

In words, the series of the ﬁrst differences of zt equals the series mean, adjusted for autoregressive and moving-average terms. Bring the autoregressive term to the LHS to get p q φi zt−1 = µ + θj εt−j + εt (A13) (zt − zt−1 ) − i=1

j=1

Decomposing violence: political murder in Colombia, 1946–99 75 and expand both summation terms

1−φ1 L1 −φ2 L2 −···−φp Lp (zt −zt−1 ) = µ+ 1+θ1 L1 +···+θq Lq εt· (A14) Rearrange (A14) to obtain zt − zt−1 =

θ (L) µ + εt φ(L) φ(L)

(A15)

Now, recursively replace t with (t − 1), and (t − 1) with (t − 2), etc. zt − zt−1 = zt−1 − zt−2 =

θ(L) µ + εt φ(L) φ(L) µ θ(L) + εt−1 φ(L) φ(L)

(A16)

.. . z1 − z0 =

µ θ(L) + ε1 φ(L) φ(L)

which, when added together ‘t’ times, yield θ(L) µ zt − z0 = εi· t+ φ(L) φ(L) t

(A17)

i=1

Rearranged, write µ θ(L) εi· t+ φ(L) φ(L) t

zt = z0 +

(A18)

i=1

In the steady state, i.e., L = 1, equation (A18) readily yields the permanent component of zt , θ(1) µt + εi φ(1) φ(1) t

z¯t = z0 +

(A19)

i=1

which, except for notation, is Cárdenas’ ﬁnal formula (1991, p. 27, equation (A15)). In practice, it iseasy to set upa spreadsheet for (A19) by replacing θ(1) and φ(1) with (1 − θi ) and (1 − φi ), i.e., with the estimated ARIMA coefﬁcients. In our particular case, where the ARIMA did not result in any AR terms at all, the denominators fell out of (A19) altogether, making the

76

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

computation of the permanent, and hence cyclical, components, of the Colombian murder series even easier.

Endnotes 1 If one were to request to re-examine police or justice records in order to classify them case by case into ‘political’ or ‘non-political’ murder categories one would, in Colombia, write one’s own death sentence as one of the chapter’s authors attests (also see Giraldo, 1996; Kirk, 2003). In such cases, our approach of time-series decomposition may be the only feasible alternative to get a statistical handle on political murder. 2 We nonetheless applied the single, double, and Holt-Winter smoothing techniques. In all instances, the estimated series is simply shifted to the right of the original series (all peaks and troughs are right-shifted), thereby consistently missing all turning points of the actual homicide series. 3 In E-Views, double-click on the original data series icon, click on Procs, select HodrickPrescott Filter, and type in the desired smoothing parameter, λ, or accept the default value λ = 100. To test the robustness of the resulting series we varied the smoothing parameter drastically but did not obtain materially different results. For example, according to Ravn and Uhlig (2002), the optimal smoothing parameter should be between 6.25 and 8.25 for annual data, not 100. We used those values, as well as 20, 40, 60, 80, and 800, in addition to the default value of 100. With 54 observations, runs for eight different smoothing parameters result in eight series of estimates with a total of 432 data points. With a mere handful of exceptions, the 432 data points in the eight estimated series of cyclical murder result in identical turning points, and it is the turning points, not the magnitudes of the estimates, that interest us. See Maravall and del Río (2001) for a discussion of the ﬁlter default values. On problems with the mechanical application of the Hodrick-Prescott ﬁlter, see Harvey and Jaeger (1993). 4 There are of course no ‘negative’ murders in Figure 5.3. The negative numbers represent homicides falling below the trend line. In our interpretation, it is the lack of political unrest that accounts for below-trend homicides. 5 E-Views and RATS use different implementations of the Schwartz Criterion (SC). E-Views works off the log likelihood function, whereas RATS works off the sum of squared residuals. We conﬁrmed both formulas by hand-computing the respective criterion to check the computer printouts. Running a variety of ARIMA models, the model with moving average model with lags at 1, 5, and 13 returns the lowest SC in E-Views and RATS, respectively. (RATS does not report the AIC.) 6 Neither E-Views nor RATS contains a Beveridge-Nelson routine (a user-supplied routine is available on the RATS web site at www.estima.com). Using the original BN (1981) method, the extraction of permanent and cyclical components from the original series is computationally very intense. Cuddington and Winters (1987), Miller (1988), and Newbold (1990) provide computationally easier methods. In his dissertation, Cárdenas (1991) provides an exceptionally easy and conceptually appealing way to compute the components, and that is the method we apply. Since we failed to locate a published account of Cárdenas’ method, we provide the mathematical details in appendix 5.1. To ensure correctness we computed the permanent and cyclical components by BN’s original method, by the Cuddington/Winters method, and by the Cárdenas method, using the ﬁrst actual observation as the initial value. All three methods resulted in identical estimates. 7 One reader observed that if the BN method were run for a different time period, say 1946–1991, instead of 1946–1999, the estimates of the permanent and transitory components would change. This is true and changes the estimated magnitudes (we re-ran the underlying ARIMA which also is (0,1,13) with MA lags at 1, 5, and 13) but does

Decomposing violence: political murder in Colombia, 1946–99 77

not change the turning points at all (see Figure in this note). Our interpretation thus remains entirely unchanged. 8 We also ran the BN procedure on the homicide and aggravated homicide subseries (available since 1959) and came up with nearly identical turning points. The BN permanent component in Figure 5.2, incidentally, is not linear. Fitting a linear or quadratic regression to the permanent component data for instance leaves positive autocorrelation in the error terms. Instead, the permanent component is a random walk with the same drift as the original data, plus an innovation proportional to the original data. For mathematical details, contact the authors. 9 The number of possible decompositions of an observed series (homicides) into trend and cycle components is inﬁnite. Whether the extracted cyclical information ‘matches’ the political events in Colombia requires our subjective judgment. We have not found any literature that disagrees with the dating of the country’s major political up and downswings. 10 One might therefore argue that the character of Colombia’s ‘political’ violence changed during the 1990s, i.e. that even though it was cloaked in terms of revolutionary and counter-revolutionary language, the observed violence is linked to the economics of the drug trafﬁc. Greed takes preference over grievance (Collier, 1999, 2000; Sambanis, 2002), and an attempt at collusive exploitation of the country is made. Revolutionaries and counter-revolutionaries become bandits who defend their respective territories and interests with murder. If this is correct, the war in Colombia in the 1990s was essentially an economic war over access to and exploitation of natural resources, not unlike those we observed in Africa in the 1990s (e.g. Sierra Leone, Liberia, Angola). This war is unlikely to cease unless the major sources of funding – US military aid and US drug purchases – cease. This interpretation, even if correct for the 1990s, is now probably moot as the FARC and Colombia’s new president (Álvaro Uribe Vélez, 2002–06) have openly resumed their war.

References Bejarano, J.A. (1997) ‘Inseguridad, violencia y actividad económica’, Lecturas de Economía [Medellín, Colombia], July–December: 7–24. Beveridge, S. and Nelson, C.R. (1981) ‘A new approach to decomposition of economic time series into permanent and transitory components with particular attention to measurement of the “business cycle”’, Journal of Monetary Economics, 7(2): 151–74.

78

Jurgen Brauer, Alejandro Gómez-Sorzano and Sankar Sethuraman

Brito, D.L. and Intriligator, M.D. (1992) ‘Nacro-trafﬁc and guerrilla warfare: a new symbiosis’, Defence Economics, 3: 263–74. Bushnell, D. (1993) The Making of Modern Colombia: A Nation In Spite of Itself. Berkeley, CA: University of California Press. Cárdenas, M. (1991) ‘Coffee Exports, Endogenous State Policy and the Business Cycle’, PhD thesis, Economics. University of California, Berkeley, CA. Collier, P. (1999) ‘On the economic consequences of civil war’, Oxford Economic Papers 51(1): 168–83. —— (2000) ‘Doing well out of war: an economic perspective’, in M. Berdal and D.M. Malone (eds) Greed and Grievance: Economic Agendas in Civil War, pp. 91–111, Boulder, CO: Lynne Rienner. Cuddington, J. and Winters, L.A. (1987) ‘The Beveridge-Nelson decomposition of economic time series’, Journal of Monetary Economics, 19: 125–7. Dinar, A. and Keck, A. (1997) ‘Private irrigation investment in Colombia: effects of violence, macroeconomic policy, and environmental conditions’, Agricultural Economics, 16(1): 1–15. Giraldo, J. (1996) ‘Colombia: The Genocidal Democracy’, Common Courage Press, Monroe, ME. Guerrero, R. (1998) ‘Epidemiology of violence in the Americas: the case of Colombia’, in: Burki, S.J, Aiyer, S.R. and Hommes, R. (eds) Annual World Bank Conference on Development in Latin America and the Caribbean: Poverty and Inequality, pp. 95–100 (Bogotá, Colombia) The World Bank, Washington, DC. Harvey, A.C. and Jaeger, A. (1993) ‘Detrending, stylized facts and the business cycle’, Journal of Applied Econometrics 8: 231–47. Hodrick, R.J. and Prescott, E.C. (1997) ‘Postwar US business cycles: an empirical investigation’, Journal of Money, Banking, and Credit, 29: 1–16. Kirk, R. (2003) More Terrible than Death: Massacres, Drugs, and America’s War in Colombia, New York: Public Affairs. Londo˜no, J.L. (1998) ‘Violence, psyche, and social capital’ in: Burki, S.J, Aiyer, S.R. and Hommes, R. (eds) Annual World Bank Conference on Development in Latin America and the Caribbean: Poverty and Inequality, pp. 71–82, (Bogotá, Colombia) The World Bank, Washington, DC. Maravall, A. and del Río, A. (2001) ‘Time aggregation and the Hodrick-Prescott ﬁlter’, Working Paper No. 108, Madrid: Bank of Spain. Miller, S.M. (1988) ‘The Beveridge-Nelson decomposition of economic time series: another economic computation method’, Journal of Monetary Economics, 21: 141–2. Newbold, P. (1990) ‘Precise and efﬁcient computation of the Beveridge-Nelson decomposition of economic time series’, Journal of Monetary Economics, 26: 453–57. Ravn, M.O. and Uhlig, H. (2002) ‘On adjusting the Hodrick-Prescott ﬁlter for the frequency of observations’, The Review of Economics and Statistics, 84(2): 371–80. Richani, N. (1997) ‘The political economy of violence: the war-system in Colombia’, Journal of Interamerican Studies and World Affairs, 39(2): 37–81. Rubio, M. (1997) ‘Perverse social capital – some evidence from Colombia’, Journal of Economic Issues, 31(3): 805–16. —— (1998a) ‘Violence, organized crimes, and the criminal justice system in Colombia’, Journal of Economic Issues, 32(2): 605–10. —— (1998b) Comment, in: Burki, S.J., Aiyer, S.R. and Hommes, R. (eds), Annual World Bank Conference on Development in Latin America and the Caribbean:

Decomposing violence: political murder in Colombia, 1946–99 79 Poverty and Inequality, pp. 90–2, (Bogotá, Colombia) The World Bank, Washington, DC. Sambanis, N. (2002) ‘A review of recent advances and future directions in the quantitative literature on civil war’, Defence and Peace Economics, 13: 215–43. Wold, H. (1938) A Study in the Analysis of Stationary Time Series, Stockholm: Almquist and Wiksell.

Part III

Economic effects of terrorism

6

The effects of terrorism on global capital markets Andrew H. Chen and Thomas F. Siems*

In this paper, the event study methodology is used to assess the effects of terrorism on global capital markets. We examine the US capital market’s response to fourteen terrorist/military attacks dating back to 1915 and global capital markets’ response to two recent events –Iraq’s invasion of Kuwait in 1990 and the 11 September 2001 terrorist attacks. US capital markets are more resilient than in the past and recover sooner from terrorist attacks than other global capital markets. Evidence suggests that this increased market resilience can be partially explained by a stable banking/ﬁnancial sector that provides adequate liquidity to promote market stability and minimize panic.

1.

Introduction

Prices of individual stocks reﬂect investors’ hopes and fears about the future, and taken in aggregate, stock price movements can generate a tidal wave of activity. Because of their liquidity, terrorist attacks, military invasions and other unforeseen disastrous occurrences can have serious implications for stocks and bonds. Decisions to buy and sell can quickly, easily, and inexpensively be reversed. When information becomes available about a cataclysmic event – like a terrorist or military attack – investors often ﬂee the market in search of safer ﬁnancial instruments and panic selling ensues. This initial panic has the potential to turn

* Chen is Distinguished Professor of Finance, Cox School of Business, SMU, Dallas, Texas 75275, phone: (214) 768-3179, e-mail: [email protected]. Siems (corresponding author) is Senior Economist and Policy Advisor, Federal Reserve Bank of Dallas, Dallas, Texas 75201, phone: (214) 922-5129, e-mail: [email protected]. The authors thank the participants at DIW Berlin’s June 2002 workshop on ‘The Economic Consequences of Global Terrorism’, the participants at the German Federal Foreign Ofﬁce’s program on ‘Terrorism and Economic Vulnerability in an Interdependent World’, and the seminar participants at National Taiwan University. The authors would especially like to thank Tilman Brück, Bengt-Arne Wickström, Valpy FitzGerald, Arye Hillman, Todd Sandler, Mark Chen, David Mauer, Rex Thompson, Ken Robinson, and three anonymous referees for valuable comments on earlier versions of the paper. The views expressed in this paper are not necessarily those of the Federal Reserve Bank of Dallas or the Federal Reserve System. The usual disclaimer regarding errors and omissions applies.

84

Andrew H. Chen and Thomas F. Siems

into chaos and a long-term bear market, or it can be reversed if investors’ hopes return. In this paper, we investigate the response of global capital markets to terrorist and military attacks. We ﬁrst examine the US capital market’s response to terrorist attacks, including government-initiated invasions of other countries, dating back to the sinking of the luxury ocean liner Lusitania by a torpedo in 1915. We then investigate the reaction of several global capital markets to the two recent events with the greatest negative impact on US capital markets: Iraq’s invasion of Kuwait in 1990 and the 11 September 2001 terrorist attacks against the United States. We conclude that terrorist attacks and military invasions have great potential to affect capital markets around the world in a short period of time. In today’s information-oriented world, news travels very fast and contagion can spread quickly. We ﬁnd evidence, however, that US capital markets seem to have become more resilient and are better able to absorb shocks brought on by such events. We also ﬁnd evidence that an economy’s banking/ﬁnancial sector seems to be an important force in returning markets to relative stability. Finally, to increase market stability, policymakers and regulators around the world should be aware of these strong forces and the inter-relatedness in global capital markets and, thus, proactively share information in a timely manner in a move toward greater global cooperation and communication.

2.

Research questions

Within the event study methodology, we can test a number of hypotheses. First, we examine the US capital market’s response to a selection of historical terrorist and military attacks. This is essentially the standard event-study test to determine whether the capital market experienced signiﬁcant abnormal returns in response to any of the past events. Then, we include available market data from actively traded stock exchanges from around the world to examine the global capital markets’ reactions to terrorist and military attacks that occurred in the recent past. And, where data are available, we also test the signiﬁcance of abnormal returns in each market’s banking/ﬁnancial sector.

Research question 1: Are historical terrorist and military attacks associated with signiﬁcant negative abnormal returns in US capital markets? And, if so, to what relative degree?

It appears obvious that a terrorist or military attack would likely have a negative effect on capital markets. Uncertainty about what the future holds and about individual ﬁrms’ abilities and the resources needed to see them through a crisis often clouds judgement, sending many investors into a panic. Moreover, terrorist and military attacks often increase basic costs of doing business as security is enhanced and shipping times lengthened.

The effects of terrorism on global capital markets

85

We identify several historic terrorist attacks and military invasions and compare the depths and statistical signiﬁcance of the negative returns across the various events. Which events resulted in the most negative returns? For which events did the markets seem to rebound the quickest?

Research question 2: Are recent terrorist and military attacks associated with signiﬁcant negative abnormal returns in global capital markets?

Again, for reasons cited above, we expect signiﬁcant negative abnormal returns in global capital markets at, and following, terrorist and military attacks. But which markets reacted more negatively (or more positively) than others, and why? Which events resulted in prolonged negative returns? Which events impacted global capital markets more than US capital markets, and vice versa?

Research question 3: Can the banking/ﬁnancial sector help minimize crises in capital markets resulting from terrorist and military attacks?

Strong economies rely upon strong banking/ﬁnancial sectors, which are in turn inﬂuenced by effective and appropriate monetary policies. In times of crises, it may be necessary for policymakers to add a degree of ﬂexibility in order to provide adequate liquidity to a shaky and panicky market. In most cases, we expect the banking/ﬁnancial sector to respond to news of a terrorist or military attack by generating signiﬁcant negative returns. But if the outlook improves rapidly in this sector – perhaps because the nation’s monetary authority quickly provided adequate liquidity – then we expect the associated capital market to also quickly improve.

3.

Methodology

The event-study methodology is a forward-looking approach that focuses on identifying abnormal returns to ﬁrms from a speciﬁc event. If investors react favourably to an event, we would expect positive abnormal stock returns around the event date. Alternatively, if investors react unfavourably to an event, we would expect negative abnormal stock returns. Hence, when analysed using composite stock indices (or major sector indices), abnormal returns provide a means of assessing the capital market’s (or sector’s) response to speciﬁc events. The event-study methodology is based on the efﬁcient markets hypothesis (Fama et al. 1969). This hypothesis generally states that as new information becomes available (perhaps as the result of some signiﬁcant unexpected event), it is fully taken into consideration by investors assessing its current and future impact. Investors immediately reassess individual ﬁrms and their ability to withstand

86

Andrew H. Chen and Thomas F. Siems

potential economic, environmental, political, societal, and demographic changes resulting from the event. The new assessment results in stock price changes that reﬂect the discounted value of current and future ﬁrm performance. Signiﬁcant positive or negative stock price changes can then be attributed to speciﬁc events. The strength of the eventstudy method lies in its ability to identify such abnormal changes because it is based on the overall assessment of many investors who quickly process all available information in assessing each individual ﬁrm’s market value (Schwert 1981). For example, Abadie and Gardeazabal (2003) use the event-study methodology to examine the economic impact on ﬁrms from terrorist conﬂicts in the Basque Country, Spain. They ﬁnd that stocks of ﬁrms with a signiﬁcant part of their business in the Basque Country showed a positive relative performance when truce became credible, and a negative relative performance at the end of the cease-ﬁre. Because we examine world stock market indices as a portfolio of individual stocks, we follow the excess returns approach as described in Brown and Warner (1985) to measure a market’s (or major market sector’s) abnormal performance. This methodology allows us to statistically test the signiﬁcance of the economic impact of an event on world capital markets as measured by the deviation of index returns from their average. In other words, how different (in size and duration) are the returns from past averages? Since index returns are random variables, they will deviate from their means over any given event window. And when examining these deviations to past average returns and taking into account historical variability, we can draw important conclusions regarding the statistical signiﬁcance (the depth and breadth) of an event. If the return deviation (abnormal return) is small and statistically insignificant on trading days that coincide with an event, then we can conclude that the market saw the event as inconsequential. But if the return deviation is large and statistically signiﬁcant (falling outside the range of returns normally expected), then we can conclude that the market saw the event as important and one that moved it signiﬁcantly.1 Daily excess returns were measured by the mean-adjusted-returns approach; that is, for each day at, and following, the event, we computed: Rj , ARjt = Rjt −

(1)

where ARjt is the abnormal (or excess) return for stock index j at time t, Rjt is the actual observed rate of return for stock index j at time t, and Rj is the mean of stock index j’s daily returns in the (−30, −11) estimation period. Rj is computed as follows: Rj =

−11 1 Rjt . 20 t=−30

(2)

The effects of terrorism on global capital markets

87

The date of the event is t = 0, the mean adjusted returns model is estimated over 20 days, from t = −30 to t = −11 relative to the event date. The main event window under study is the event date itself (t = 0). However, we also examine two longer event windows to see how well and how quickly the market digested the news. Sometimes, the initial uncertainties persist and that keeps stock prices down and volatile, but at other times these fears are reduced because of new information that eases market tensions or policy actions that promote greater market stability. The two longer event windows are from the event date to ﬁve days following the event (t = +5) and from the event date to ten days following the event (t = +10). For these longer event windows, we also compute the cumulative average abnormal returns (CARs). The statistical signiﬁcance of the event period abnormal returns were computed for each sample using the test statistics described in Brown and Warner (1985).

4.

A historical perspective on past terrorist attacks and military invasions on US capital markets

In this section of the paper, we examine several terrorist attacks and military invasions that rocked US capital markets. While our list is subjectively determined, we selected terrorist attacks from the Signiﬁcant Terrorist Incidents list published by the US Department of State (2001) as well as a list from the Constitutional Rights Foundation (2001). We also included signiﬁcant military attacks that seemed to have an element of surprise and resulted in war. For continuity, the events are examined chronologically. We use the Dow Jones Industrial Average (DJIA or Dow) as our market index because of its availability, usefulness and widespread visibility among market participants. There have been a number of past terrorist and military attacks that have resulted in extreme stock price volatility as anxious investors sell to exhaust their panic while buyers remain on the sidelines until they build up enough courage to reenter the market. Table 6.1 lists each event, its abnormal returns over three different event periods, and the number of trading days until the DJIA index returned to its pre-attack level. To indicate whether the index returns deviate from their means by more than one would expect through normal variation, the table includes t-statistics as an indication of how signiﬁcant each event was. The t-statistics essentially test the signiﬁcance of the economic impact of an event on world capital markets as measured by the deviation of index returns from their average. If the event had no consequence, one would expect an insigniﬁcant return deviation. This logic holds for cumulative returns as well as returns on individual days. Also included in the table are the t-statistics and an indication of whether the abnormal returns are statistically different from zero. To see a long-term perspective of the Dow, Figure 6.1 plots the index on the natural log scale. Each event listed in Table 6.1 is highlighted in Figure 6.1 by a vertical line, as well as two additional endogenous events of interest not included in this study: the 1929 stock market crash and the 1987 stock market crash.

Event date

7 May 1915

12 May 1940

7 Dec. 1941

25 June 1950

4 May 1970

23 Oct. 1983

21 June 1985

30 Nov. 1987

Terrorist/military attack

Lusitania torpedoed

Invasion of France

Pearl Harbour attack

North Korea attacks South Korea Kent State shootings

Beirut bombing

Air India bombing

Korean Air bombing

6-day CAR

−16.08%∗∗∗ (−5.04) −16.22%∗∗∗ (−11.12) −3.09% (−1.65) −8.30%∗∗∗ (−7.36) −3.53% (−2.00) −2.69% (−1.36) −0.34% (−0.24) −0.95% (−0.06)

Event-day AR

−5.26%∗∗∗ (−4.24) −4.93%∗∗∗ (−8.69) −2.75%∗∗∗ (−3.78) −4.82%∗∗∗ (−11.00) −2.65%∗∗∗ (−3.87) −0.12% (−0.16) −0.49%∗ (−0.90) −3.34% (−0.52)

−12.30%∗ (−2.85) −21.02%∗∗∗ (−10.65) −6.42% (−2.53) −11.03%∗∗∗ (−7.23) −4.78% (−2.00) −4.19% (−1.56) −2.58% (−1.36) 9.10% (0.41)

11-day CAR

Table 6.1 Average abnormal returns on the Dow Jones Industrial Average Stock Index following terrorist attacks

11

4

1

54

57

232

795

21

Number of trading days to return to pre-attack level

2 Aug. 1990

26 Feb. 1993

19 April 1995

7 Aug. 1998

11 Sept. 2001

Iraq invades Kuwait

World Trade Center bombing

Oklahoma City bombing

Embassy bombing in Kenya

September 11th terrorist attacks

Notes: Standard errors are in parentheses. *** = statistically signiﬁcant at the 0.01 level. ** = statistically signiﬁcant at the 0.05 level. * = statistically signiﬁcant at the 0.10 level.

21 Dec. 1988

Pan Am bombing

−0.13% (−0.14) −1.35%∗∗ (−1.82) −0.06% (−0.10) 0.47% (0.92) 0.23% (0.27) −7.14%∗∗∗ (−7.72)

−0.39% (0.17) −5.79%∗ (−3.04) −0.17% (−0.12) 1.59% (1.22) −1.83% (−0.83) −10.57%∗∗ (−4.45) 0.63% (0.20) −9.35%∗ (−3.63) −0.59% (−0.31) 2.24% (1.27) −0.55% (−0.19) −7.90% (−2.45) 1 1 40

1 1 1

134

3

Figure 6.1 Dow Jones Industrial Average 1915–2002.

The effects of terrorism on global capital markets

91

Overall, the DJIA index has risen an average 5.88 per cent per year from the beginning of 1915 to 2002. Of the fourteen terrorist/military attacks listed in Table 6.1, twelve events experienced negative abnormal returns (ARs) on the day of the event.2 The only events with positive event-day ARs were the bombing of the Alfred P. Murrah Federal Building in Oklahoma City on 19 April 1995 and the Embassy bombing in Kenya on 7 August 1998. Six events – including the ﬁrst ﬁve events on our list – had negative ARs that were signiﬁcantly different from zero at the 0.01 level, one event had a signiﬁcant negative AR at the 0.05 level, and one event had a signiﬁcant negative AR at the 0.10 level. While the event-day ARs are interesting in that they show immediate investor reaction to terrorist and unexpected military attacks, the cumulative average abnormal returns (6-day and 11-day CARs) provide a stronger indication of the capital market’s resilience and ability (or inability) to bounce back from the attacks. Of the fourteen events listed in Table 6.1, eleven experienced negative 6-day and 11-day CARs. Positive CARs resulted after the Oklahoma City bombing, the Korean Air bombing in November 1987, and the Pan Am bombing over Lockerbie, Scotland in December 1988. Three events experienced signiﬁcant negative CARs at the 0.01 level over the 6-day event window, and two of these three events saw signiﬁcant negative CARs persist over the 11-day horizon. The three events with signiﬁcant 6-day negative CARs include the sinking of the luxury cruise ship Lusitania by a torpedo on 7 May 1915, Hitler’s invasion into France on 12 May 1940, and the attack on South Korea by the North Korean People’s Army on 25 June 1950. It is worth noting that for all three events, the CARs for the 11-day event window were all double-digit negative CARs. During this time window, new information continued to have a negative effect on stock prices. Also note that the only other event with a statistically signiﬁcant negative CAR over both of these two event horizons was Iraq’s invasion of Kuwait on 2 August 1990 (both event windows were signiﬁcant at the 0.10 level). The only other attack with a statistically signiﬁcant negative CAR over one of the two longer event horizons occurred after US capital markets reopened following the terrorist attacks against the United States on 11 September 2001. The 6-day CAR was –10.57 per cent, signiﬁcant at the 0.05 level, however, the 11-day CAR dropped to –7.90 per cent and was not statistically different from zero. Figure 6.2 summarizes the 11-day CARs and signiﬁcance levels for these fourteen events. The event CARs are arranged chronologically and indicate a trend toward greater capital market resilience. That is, the CARs generally become less negative and fewer events experienced negative CARs statistically different from zero. The last column of Table 6.1 shows that, over time, fewer trading days in general were required to return the DJIA index to its pre-attack level. Most notably, following Hitler’s invasion into France, the Dow remained down for 795 trading days, or about two-and-a-half years. After the surprise attack on Pearl Harbour, the Dow did not recover for another 232 trading days (about one year). The next

Figure 6.2 US capital markets’ 11-day cumulative abnormal returns (Dow Jones Industrial Average Index).

The effects of terrorism on global capital markets

93

major shock to have a fairly protracted negative impact on the Dow was Iraq’s invasion into Kuwait when the Dow did not fully rebound for 134 trading days (about seven months). After 11 September 2001, the Dow recovered in 40 trading days (about two months) following the most horriﬁc attacks on US soil since the American War of 1812 to 1815. While it is certainly difﬁcult to draw deﬁnitive conclusions when examining the abnormal returns from these events over time, there may be several reasons to explain this increased market resilience. First, improved technology has made communications and information acquisition and transmission more timely and accurate. This has also helped make markets more efﬁcient, along with having more market participants. Additionally, more ﬂexible and appropriate monetary and ﬁscal policies may have been better able to assure markets and promote stability by providing proper levels of liquidity in times of need. Overall, an investigation of the abnormal returns following these fourteen events indicates that US capital markets today appear to be more resilient and are quicker to absorb news of terrorist attacks and military invasions. Increased capital market resilience seems to also be reﬂected by the fact that three of the ﬁrst four events in our series experienced the largest negative 11-day CARs, all of which had signiﬁcant negative CARs at the 0.10 level (with two of the three events signiﬁcant at the 0.01 level). In contrast, only one terrorist/military attack in the last 60 years – Iraq’s invasion into Kuwait – had a signiﬁcant negative 11-day CAR, and its signiﬁcance level barely made the 0.10 cutoff. As horriﬁc as the 11 September attacks were, with planned attacks against US ﬁnancial and military symbols, the capital markets managed to rebound rather quickly.

5.

Global capital markets and recent terrorist/military attacks

In this information age, news (especially bad news) spreads very rapidly around the world with quick spillover or contagion effects, making global capital markets today more tightly inter-linked. See, for example, evidence presented in recent studies by Arshanapalli and Doukas (1993) and Hamao et al. (1990). We now turn our attention to the reaction of global capital markets for two recent terrorist/military attacks – the 11 September terrorist attacks and Iraq’s invasion into Kuwait. These events were selected because they were the only recent events where the Dow experienced signiﬁcant negative abnormal returns and resulted in stock prices remaining below pre-attack levels for 40 or more trading days. We begin with an analysis of the global capital markets’ reaction to the 11 September attacks and then consider the markets’ response to Iraq’s invasion of Kuwait. Global stock market data were obtained from Bloomberg. We used the broadest stock market indexes available in each market, comparable to the New York Stock Exchange (NYSE) composite index in the US. Table 6.2 shows the abnormal returns and statistical signiﬁcance levels for the one-day, six-day, and 11-day event windows following the 11 September terrorist attacks for 33 capital markets (or major indices) located around the world.3

Table 6.2 Average abnormal returns on global capital markets following the 11 September terrorist attacks Global stock market S&P 500 Dow Industrials New York Stock Exchange Nasdaq Toronto Mexico London Frankfurt Europe (Bloomberg) France Spain Switzerland Austria Italy Belgium Amsterdam Portugal Helsinki Norway Sweden Tokyo Hong Kong South Korea

Event-day AR

−4.84%∗∗∗ (−5.48) −7.14%∗∗∗ (−7.72) −4.55%∗∗∗ (−7.14) −6.56%∗∗∗ (−3.70) −4.05%∗∗∗ (−5.67) −5.45%∗∗∗ (−4.24) −5.29%∗∗∗ (−6.46) −7.61%∗∗∗ (−6.73) −6.23%∗∗∗ (−6.71) −7.07%∗∗∗ (−7.26) −4.79%∗∗∗ (−4.43) −7.03%∗∗∗ (−6.59) −0.96%∗ (−1.54) −7.71%∗∗∗ (−9.45) −5.41%∗∗∗ (−10.54) −6.94%∗∗∗ (−7.83) −3.82%∗∗∗ (−5.75) −3.30%∗ (−1.58) −4.53%∗∗∗ (−8.33) −7.65%∗∗∗ (−6.56) −6.20%∗∗∗ (−4.77) −8.45%∗∗∗ (−5.84) −12.42%∗∗∗ (−8.33)

6-day CAR

11-day CAR

Number of trading days to return to pre-attack level

−7.72%∗ (−3.40) −10.57%∗∗ (−4.45) −8.09%∗∗ (−4.93) −10.14% (−2.22) −9.21%∗∗ (−5.01) −13.17%∗ (−3.98) −4.77% (−2.27) −7.98% (−2.75) −6.82% (−2.86) −9.80%∗ (−3.91) −7.64%∗ (−2.75) −5.97% (−2.17) −4.36% (−2.70) −13.51%∗∗∗ (−6.44) −8.51%∗∗∗ (−6.44) −8.52%∗ (−3.74) −6.70%∗ (−3.91) 7.49% (1.40) −9.89%∗∗∗ (−7.08) −4.96% (−1.65) −0.56% (−0.17) −5.57% (−1.50) −11.82%∗ (−3.08)

−3.83% (−1.25) −7.90% (−2.45) −3.98% (−1.79) −9.99% (−1.62) −9.87% (−3.96) −6.54% (−1.46) −9.04% (−3.17) −10.64% (−2.70) −8.30% (−2.57) −10.82% (−3.19) −8.83% (−2.34) −7.29% (−1.96) −7.76% (−3.55) −14.19%∗ (−5.00) −9.22%∗ (−5.15) −10.83% (−3.51) 0.67% (0.29) 15.26% (2.10) −12.39%∗∗ (−6.55) −4.69% (−1.16) −3.05% (−0.67) −5.23% (−1.04) −16.65% (−3.21)

19 40 37 37 12 44 52 22 23 23 31 23 30 97 31 76 42 14 2 78 23 14 20 28 Continued

The effects of terrorism on global capital markets

95

Table 6.2—Cont’d Global stock market

Event-day AR

6-day CAR

India

−5.45%∗∗∗ (−4.24) −3.42%∗∗∗ (−3.16) −4.69%∗∗∗ (−7.80) −4.46%∗∗∗ (−4.89) −4.19%∗∗∗ (−6.50) −4.50%∗∗∗ (−9.28) −3.94%∗∗∗ (−4.28) −4.10%∗∗∗ (−12.05) −1.82%∗∗∗ (−2.62) −2.66%∗∗∗ (−2.69)

−13.17%∗ −6.54% 45 (−3.98) (−1.46) −4.58% −9.31% 89 (−1.65) (−2.47) −12.07%∗∗∗ −16.00%∗∗∗ 59 (−7.80) (−7.64) −12.45%∗∗∗ −15.41%∗∗ 75 (−5.31) (−4.85) −6.81%∗∗ −8.60%∗ 31 (−4.11) (−3.83) −6.66%∗∗∗ −6.22% 33 −(5.35) (−3.68) −11.73%∗∗∗ −15.62%∗∗ 23 (−4.96) (−4.88) −8.19%∗∗∗ −13.82%∗∗∗ 100 (−9.36) (−11.66) −11.27%∗∗∗ −7.77% 45 (−6.30) (−3.21) −11.40%∗∗∗ −12.18%∗ 25 (−4.49) (−3.55)

Jakarta Singapore Kuala Lampur Australia New Zealand Pakistan Saudi Arabia Israel Johannesburg

11-day CAR

Number of trading days to return to pre-attack level

Notes: Standard errors are in parentheses. ***= statistically signiﬁcant at the 0.01 level. ** = statistically signiﬁcant at the 0.05 level. * = statistically signiﬁcant at the 0.10 level.

Also included in the Table is a column showing the number of trading days before each global capital market returned to its pre-11 September level. All 33 global capital markets in our sample experienced signiﬁcant negative ARs the day investors in those markets ﬁrst learned of the 11 September terrorist attacks against the United States. Thirty-one of the 33 capital markets (94 per cent) had signiﬁcant negative ARs at the 0.01 level, and two markets (Helsinki and Austria) declined at the 0.10 level. The one-day ARs are presented graphically in Figure 6.3 for the ten capital markets with the largest market capitalizations. These markets are arranged from the largest to the smallest and show that the decline in the US capital market, though signiﬁcantly different from zero, was not as great as eight of the other nine largest world markets (the exception being Toronto). There also appears to be an inverse correlation between a market’s size and its depth of decline; on the event date, the largest markets appear to drop the least. Over the longer 11-day event window, negative returns persisted in all but two global markets (Portugal and Helsinki). Figure 6.4 shows the 11-day CARs for the largest capitalized markets and again shows that the US market had the

Figure 6.3 Global capital markets’ event-day abnormal returns following the 11 September terrorist attacks.

Figure 6.4 Global capital markets’ 11-day cumulative average abnormal returns following the 11 September terrorist attacks.

98

Andrew H. Chen and Thomas F. Siems

second-smallest decline. While most large global capital markets experienced negative 11-day CARs between ﬁve and eleven per cent, the CAR in the US market was −3.98 per cent. Only the Tokyo market fared better with an 11-day CAR of −3.05 per cent. The last column in Table 6.2 also shows that, for the most part, global capital markets rebounded fairly quickly after the 11 September attacks. Nine of the 33 markets had signiﬁcant negative CARs over the 11-day event window, but none of these markets is generally considered a major global capital market. Moreover, within 20 trading days, 6 of the 33 markets (18 per cent) had returned to their pre-attack levels. Within 40 trading days, 21 markets (64 per cent) had returned to their pre-attack levels, and after 60 trading days, 27 markets (82 per cent) had fully rebounded. It is also interesting to note that the S&P 500 stock index had a negative 11-day CAR of only −3.83 per cent – the fourth best return among the 33 global capital markets. It took just 19 trading days (about one month) for the S&P 500 index to return to its pre-11 September level. Thus, even though the 11 September attacks were targeted directly at the United States, its capital markets displayed amazing resilience by being affected less severely than capital markets located in most of the other areas of the world and by bouncing back quickly following its initial fall. It should be noted, however, that it could be the case that US capital markets recovered better than many global markets because US capital markets were closed for four trading days following the 11 September attacks. While it is unobservable, it is possible that the elapsed time had a calming affect as investors were able to take additional time to absorb the news of the attacks and not panic. It is also possible that US investors boosted stocks higher than they would have been otherwise because of an increased emotional patriotic response. Table 6.3 shows the abnormal returns and number of trading days required for each exchange to return to its pre-event level for 18 global capital markets following Iraq’s invasion into Kuwait. Seventeen of the 18 markets had negative event-day ARs, with nine signiﬁcant at the 0.01 level, four signiﬁcant at the 0.05 level, and one signiﬁcant at the 0.10 level. The same seventeen markets also had negative 11-day CARs, but only nine markets had CARs that were statistically different from zero (six at the 0.01 level, one at the 0.05 level, and two at the 0.10 level). It is interesting to note that the capital markets with the largest negative 11-day CARs are those located in Europe and Asia. For the most part, US capital markets didn’t fall as sharply and recovered faster than other global capital markets. This is displayed in Figure 6.5, and shows the 11-day CARs for the largest global capital markets where data were available. Perhaps this divergence is due to Europe’s greater dependence on foreign oil or its closer proximity to the region, or it may be due to swifter policy responses in America. The results thus far lead to a couple of important questions. Why was there such apparent resiliency in US capital markets? Why did the US capital markets seem to quickly rebound and stabilize? For answers, we now turn to an investigation

The effects of terrorism on global capital markets

99

Table 6.3 Average abnormal returns on global capital markets following Iraq’s invasion into Kuwait Global stock market

Event-day AR

6-day CAR

11-day CAR

S&P 500

−1.22%∗∗ (−1.59) −1.35%∗∗ (−1.82) −1.12%∗∗ (−1.64) −1.61%∗∗∗ (−3.22) −0.70%∗ (−1.16) −1.49%∗∗∗ (−2.69) −1.63%∗∗∗ (−2.28) −1.86%∗∗∗ (−3.69) −2.73%∗∗∗ (−2.07) −1.89%∗∗∗ (−4.13) −1.26%∗∗∗ (−3.33) −1.46%∗∗ (−1.89) −1.98%∗∗∗ (−3.04) −1.08% (−0.72) 2.08%∗ (2.30) −1.95%∗∗∗ (−2.79) −0.31% (−0.21) −0.35% (−0.37)

−4.90% (−2.49) −5.79%∗ (−3.04) −4.71% (−2.67) −5.27%∗∗ (−4.08) −3.06% (−1.98) −4.62%∗ (−3.25) −9.53%∗∗∗ (−5.19) −9.74%∗∗∗ (−7.50) −17.79%∗∗∗ (−5.25) −9.00%∗∗∗ (−7.64) −6.89%∗∗∗ (−7.08) −8.51%∗∗ (−4.28) −13.81%∗∗∗ (−8.24) −3.67% (−0.95) 1.75% (0.75) −13.77%∗∗∗ (−7.68) −3.53% (−0.95) −3.52% (−1.44)

−7.52% (−2.82) −9.35%∗ (−3.63) −7.00% (−2.93) −7.88%∗ (−4.51) −2.99% (−1.43) −5.96% (−3.09) −12.76%∗∗ (−5.13) −11.17%∗∗∗ (−6.36) −28.93%∗∗∗ (−6.31) −11.10%∗∗∗ (−6.96) −9.82%∗∗∗ (−7.46) −8.75% (−3.25) −12.98%∗∗∗ (−5.73) −5.56% (−1.06) 4.24% (1.34) −16.85%∗∗∗ (−6.94) −6.11% (−1.21) −2.68% (−0.81)

Dow Industrials New York Stock Exchange Nasdaq Toronto London Frankfurt Switzerland Austria Amsterdam Sweden Tokyo Hong Kong South Korea India Singapore Pakistan Johannesburg

Notes: Standard errors are in parentheses. *** = statistically signiﬁcant at the 0.01 level. ** = statistically signiﬁcant at the 0.05 level. * = statistically signiﬁcant at the 0.10 level.

Number of trading days to return to pre-event level 131 136 131 131 151 148 796 363 1, 912 178 762 >4,000 and still counting 135 61 1 677 33 8

Figure 6.5 Global capital markets’ 11-day cumulative average abnormal returns following Iraq’s invasion of Kuwait.

The effects of terrorism on global capital markets

101

of a major sector that has great potential in promoting economic stability by providing liquidity: the banking and ﬁnancial sector.

6.

The importance of the banking/ﬁnancial sector

The above analysis suggests that today’s global capital markets are tightly interlinked. We found evidence that US capital markets seem to rebound and stabilize quicker than other markets in the world when surprise terrorist/military attacks shock global capital markets than when they did so in earlier times. There seems to be an added degree of resilience today that was lacking yesterday. So, the question is: what has changed and what is different in the United States? We postulate that the efﬁcient functioning of an economy’s banking/ﬁnancial sector is a key determinant of whether an economy (and hence its capital markets) is able to withstand and quickly absorb exogenous and endogenous shocks. For clues, we investigate the banking and ﬁnancial sector’s abnormal returns in major global capital markets following the 11 September terrorist attacks. If this sector’s abnormal returns in the United States quickly become less signiﬁcant over the longer event windows, and if the abnormal returns are less severe than in other countries, then it could be the case that the banking and ﬁnancial sector is playing a large role in stabilizing the overall market as well as the economy. Table 6.4 shows the abnormal returns following the 11 September terrorist attacks for banking/ﬁnancial sector indices from 14 global capital markets where data were available. The Table shows that global banking/ﬁnancial sectors were negatively impacted, but that the US banking/ﬁnancial sector was impacted less than most other markets. The abnormal returns in the global capital markets are also generally more negative than the overall market abnormal returns reported in Table 6.2. For the ﬁve largest markets where data were available, Figure 6.6 shows the banking/ﬁnancial sector declines and the overall market declines over the 11-day event window following the 11 September attacks. In all of the large global capital markets except the NYSE in the US, the banking/ﬁnancial sector abnormal returns are worse than the overall market abnormal returns. One possible reason these returns are better in the US than in other global capital markets is that immediately following the 11 September attacks, the United States’ Federal Reserve System (Fed) took steps to provide liquidity through the banking and ﬁnancial sector. Perhaps more than anything else, the Fed’s accommodative policy calmed and stabilized the economy through the US banking/ﬁnancial sector. The best 11-day CAR following the 11 September terrorist attacks for major banking/ﬁnancial sectors among 14 global capital markets was in the US at −0.45 per cent. With the exception of the capital market returns in Jakarta, all of the other 11-day CARs were double-digit negative returns, and seven markets had statistically signiﬁcant negative CARs. We also see evidence of a ‘ﬂight to safety’ as rates on US Treasury instruments fell immediately following the attacks. Figure 6.7 shows that ten-year Treasury

102 Andrew H. Chen and Thomas F. Siems Table 6.4 Average abnormal returns on global capital markets’ banking/ﬁnancial sectors following the 11 September terrorist attacks Global stock markets’ banking ﬁnancial sectors

Event-day AR

6-day CAR

11-day CAR

New York Stock Exchange London

−4.79%∗∗∗ (−9.23) −10.09%∗∗∗ (−10.28) −10.06%∗∗∗ (−8.72) −8.54%∗∗∗ (−9.66) −6.17%∗∗∗ (−5.06) −5.79%∗∗∗ (−4.52) −6.50%∗∗∗ (−3.03) −7.87%∗∗∗ (−7.92) −13.33%∗∗∗ (−5.71) −2.83%∗∗∗ (−2.81) −5.20%∗∗∗ (−4.53) −3.98%∗∗∗ (−6.79) −3.67%∗∗∗ (−4.75) −5.27%∗∗∗ (−5.15)

−6.69%∗∗ (−5.01) −8.64%∗∗ (−3.42) −14.54%∗∗ (−4.90) −11.50%∗∗∗ (−5.06) −6.43% (−2.05) −14.18%∗∗ (−4.31) −1.70% (−0.31) −11.02%∗∗ (−4.31) −13.78% (−2.29) −3.73% (−1.44) −13.36%∗∗ (−4.52) −9.46%∗∗∗ (−6.27) −11.39%∗∗∗ (−5.73) −14.43%∗∗∗ (−5.47)

−0.45% (−0.25) −14.14%∗∗ (−4.14) −15.79% (−3.93) −14.82%∗∗ (−4.82) −11.35% (−2.67) −25.55%∗∗ (−5.73) −12.18% (−1.63) −14.34%∗ (−4.14) −19.84% (−2.44) −6.23% (−1.78) −18.68%∗∗ (−4.67) −11.07%∗∗ (−5.42) −14.93%∗∗ (−5.55) −11.00% (−3.08)

Frankfurt Europe (Bloomberg) Helsinki Norway Tokyo Hong Kong South Korea Jakarta Kuala Lumpur Australia New Zealand Johannesburg

Number of trading days to return to pre-attack level 13 22 42 40 31 107 6 30 28 86 65 26 33 162

Notes: Standard errors are in parentheses. *** = statistically signiﬁcant at the 0.01 level. ** = statistically signiﬁcant at the 0.05 level. * = statistically signiﬁcant at the 0.10 level.

Note yields immediately fell by about 25 basis points when the bond markets reopened on 13 September 2001. US government securities with shorter maturities fell by even greater margins, signifying investor demand to hold ‘risk-less’ assets during this time.

7.

Conclusions and policy implications

In conclusion, global capital markets today are tightly inter-linked; news spreads rapidly (especially bad news), with quick spillover, or contagion, effects. We ﬁnd evidence that suggests that modern US capital markets are more resilient than they

Figure 6.6 Global capital markets’ and banking/ﬁnancial sector 11-day cumulative average abnormal returns following the 11 September terrorist attacks.

Figure 6.7 Ten-year US treasury note rates 1982–2002.

The effects of terrorism on global capital markets

105

were in the past and that they recover sooner from terrorist/military attacks than other global capital markets. We also ﬁnd evidence that suggests the possibility that this increased market resilience can be at least partially explained by a banking/ﬁnancial sector that provides adequate liquidity to promote market stability and squelch panic. The policy implications of this research should not be overlooked. First, because global capital markets are closely and tightly inter-related, policymakers and regulators around the world must always be aware of what is going in other parts of the world. Today’s real-time information economy means that news spreads rapidly and has the potential to have serious negative consequences in a very short time. Thus, it is important for regulators and policymakers to cooperate and communicate more with each other on a regular basis. This includes sharing important information – like unusual stock trading or large dollar transactions – that might have consequences elsewhere, and developing disaster recovery plans that can be quickly put into place in case of a cataclysmic event. Second, the importance of a healthy and stable banking and ﬁnancial sector and the efﬁcient execution of monetary policies seems paramount to growing economies. The foundations and regulations that underlie US banking and ﬁnancial markets appear to function efﬁciently and effectively with the ability to absorb tremendous shocks. For the most part, the payments system in the US kept functioning normally with few reported problems. We suggest that capital markets and banking/ﬁnancial sectors in growing economies model themselves after those in the United States. This includes, of course, a quick, effective, ﬂexible and responsive monetary authority. Finally, while many terrorist attacks cannot be avoided, others can be, or their disruption minimized, by sharing important information. This includes sharing information with other policymakers and regulators within a nation’s borders as well as across borders. Terrorist and military attacks generally increase the cost of doing business because of added security and increased risks. Today’s global economy requires the efﬁcient and effective use of real-time information.

Endnotes 1 For most trading days the deviation of the market index return from its past average is fairly small and insigniﬁcant. But for some trading days, the deviation (abnormal return) can be large and signiﬁcant. When large deviations occur, there is typically an underlying reason. Our study examines events where there is some selection bias. As a result, we may have included some events that had no signiﬁcant impact on the market and excluded others that had a signiﬁcant impact. Nevertheless, for the selected events, we can use standard event study analysis to examine the size and duration of market responses. 2 For the 11 September 2001 terrorist attacks, the event date was when US capital markets reopened on 17 September 2001. 3 The event date is 11 September for global capital markets open at the time of the attacks and 12 September for those closed when the attacks commenced. For the US and Mexico capital markets, the event date is 17 September when the markets reopened.

106 Andrew H. Chen and Thomas F. Siems

References Abadie, A. and Gardeazabal, J. (2003) ‘The Economic Costs of Conﬂict: A Case Study of the Basque Country’, American Economic Review, 93: 113–32. Arshanapalli, B. and Doukas, J. (1993) ‘International Stock Market Linkages: Evidence from the Pre- and Post-October 1987 Period’, Journal of Banking and Finance, 17: 193–208. Brown, S.J. and Warner, J.B. (1985) ‘Using Daily Stock Returns: The Case of Event Studies’, Journal of Financial Economics, 14: 3–31. Constitutional Rights Foundation (2001), Signiﬁcant Terrorist Attacks. Available HTTP: Fama, E.F., Fisher, L., Jensen, M.C. and Roll, R. (1969) ‘The Adjustment of Stock Prices to New Information’, International Economic Review, 10: 1–21. Hamao, Y., Masulis, R.W. and Ng, V. (1990) ‘Correlations in Price Changes and Volatility across International Stock Markets’, Review of Financial Studies, 3: 281–307. Schwert, G.W. (1981) ‘Using Financial Data to Measure Effects of Regulation’, Journal of Law and Economics, 24: 121–58. US Department of State (2001) ‘Signiﬁcant Terrorist Incidents, 1961–2001: A Chronology’. Ofﬁce of the Historian, Bureau of Public Affairs, October 31. Available HTTP:

7

Terrorism-induced structural shifts in ﬁnancial risk: the case of airline stocks in the aftermath of 11 September terror attacks Konstantinos Drakos

This paper investigates the effects of terror attacks of 11 September on a set of airline stocks listed at various international stock markets. Utilizing the Market Model as the relevant return generating mechanism, we document a structural break in systematic risk (beta) for airline stocks. Moreover, our empirical evidence shows that, apart from the systematic risk, idiosyncratic risk has also substantially increased. In quantitative terms, conditional systematic risk has on average more than doubled, while the percentage it represents over total risk has shown a considerable increase. These results have implications for portfolio diversiﬁcation and the cost (and ability) of airlines in raising capital.

1.

Introduction We have lost more in one year than we have made in our entire history. This is an industry that is now in a deep hole. We must start looking for footholds and ways to climb quickly out of the ﬁnancial abyss (IATA Director General & CEO Pierre J. Jeanniot, on the opening day of the Airline Financial Summit, New York City, 8 April 2002)

The terrorist attacks of 9/11 had signiﬁcant economic repercussions. The already by then weakening global economy, with the world’s three largest economies slowing down (the US ofﬁcially in a recession since March 2001, Japan and Germany showing virtually zero growth and heading for a recession), was adversely affected. The aviation industry (and especially the passenger carriers) absorbed a large part of the shock.1 The 9/11 events resulted in a variety of economic as well as political effects. One may argue that these events will have enduring or even permanent effects on issues such as national security or international relations. If the terrorist attacks were a one-off anomaly, economic effects will probably be transitory. Focusing on the economic effects, the airline industry has been directly affected by

108 Konstantinos Drakos the terrorist incidents. Terrorism is an act of violence and in this particular case (9/11 attacks) was an exogenous factor exerting an adverse demand shock for the airline industry. The commercial airline industry had already been facing difﬁculties due to the economic slowdown, and 9/11 events acted as catalysts. In the aftermath of 9/11, passenger trafﬁc on the international scheduled services of International Air Transport Association (IATA) airlines declined by 17 per cent in September,2 compared to September 2000, and showed no growth in the ﬁrst nine months of the year.3 Carriers were unable to adjust their seat supply and the passenger load factor fell from 78 per cent in August to 69 per cent in September. Most notably, carriers registered in North America were most seriously affected since their passenger (and freight) trafﬁc fell by more than 30 per cent in September. European and Far Eastern carriers experienced a 12 per cent fall in passenger trafﬁc overall, but carriers with a high US component in their services fared worse. A further 7 point fall – to less than 63 per cent – in the passenger load factor took place in October on the international scheduled services of IATA airlines. The year-on-year fall of 23 per cent in October passenger trafﬁc worldwide made the cumulative change for the ﬁrst 10 months of 2001 negative. North American IATA carriers on average experienced a 33 per cent fall in passenger trafﬁc in October, while for European, Far Eastern and Central and South American carriers on average the fall ranged from 20 per cent to 25 per cent. In December 2001 a similar picture emerged in the global airline industry, through the effects of 9/11. Table 7.1 summarizes the market conditions.4 According to IATA’s estimates its members will collectively record a loss of 15 billion USD in 2001. Focusing on the US aviation industry, which has been hit the most, we report recent developments for a set of performance indicators for 2001. There had been a downward trend in airlines proﬁts in 2001 even before the incidents of 9/11. However, the ﬁnancial results for the third quarter in 2001, which usually is the most proﬁtable quarter for the airline industry, showed a considerable reduction in demand (see operating proﬁt and net income). Additionally, the spread between the break-even load factor and the actual load factor reached an unprecedented level of 15 per cent, which effectively summarizes the losses made

Table 7.1 IATA members’ average (international scheduled services)

Passenger trafﬁc, % change over ’00 (revenue-passenger-kilometres) Passenger seat supply, % change over ’00 (available-seat-kilometres) Passenger load factor (% points) Freight trafﬁc, % change over ’00 (revenue tonne-kilometres)

Dec 2001

Jan–Dec 2001

−12

−4

−11

−1

69 −10

71 −8

Terrorism-induced structural shifts in ﬁnancial risk

109

Table 7.2 Quarterly ﬁnancial review for US passenger airlines (majors); 2001

Operating proﬁt ($ millions) Net income ($ millions) Passenger revenue per RPM (cents) Actual passenger load factor Breakeven passenger load factor Overall RTMs per employee

2001Q1

2001Q2

2001Q3

2001Q4

−858.50 −955.60 13.55 68.40% 73.60% 39.10

−724.40 −3,180.60 −4,316.80 −729.20 −2,429.60 −3,256.1 12.60 11.53 11.44 73.50% 72.40% 66.10% 76.10% 87.70% 89.50% 42.60 40.10 35.60

Change over comparable period 12 months earlier Operating proﬁt ($ millions) −1,573.40 −3,528.30 −5,124.40 −4,352.5 Net income ($ millions) −988.90 −2,385.60 −3,172.80 −2,958.8 Passenger revenue per RPM 1.00% −6.00% −11.40% −15.20% Overall RTMs per employee −2.97% −3.83% −13.39% −14.42% Notes: Alaska Airlines, America West Airlines, American Airlines, American Eagle Airlines, American Trans Air, Continental Airlines, Delta Airlines, Northwest Airlines, Southwest Airlines, Trans World Airlines, United Airlines and US Airways are the airlines constituting the set of majors.

during the third quarter of 2001.5 Of considerable importance for our analysis is the fall in labour productivity, essentially measured by RTMs per employee, that dropped by 13.39 per cent compared to the third quarter of 2000. As expected, the situation was further deteriorated in the fourth quarter of 2001. In particular, labour productivity fell by more than 14 per cent compared to the corresponding quarter of 2000. Moreover, the spread between the break-even and actual load factors was more than 23 per cent. The present paper aims at studying the effects of terrorism on airline stocks, listed on various stock markets around the globe, in terms of their fundamental risk proﬁle. In particular, the analysis by employing a variety of econometric techniques focuses on issues such as structural shifts in volatility and systematic risk of airline stocks. It should be noted that 9/11 is of particular interest as far as its effects on the airline stocks are concerned, for three main reasons: (i) the attacks revealed major deﬁciencies in airport security which called for extra spending (cost), (ii) the attacks led to the war on terrorism that affected travelling plans (trafﬁc), and most importantly (iii) it was the ﬁrst time that aircraft were used by suicide terrorists as means rather than as targets (as in hijackings).

The remainder of the paper is organized as follows. Section 2 provides a description of the dataset used as well as a brief discussion of 9/11 in the light of ﬁnance theory. Section 3 presents the hypotheses to be tested. Section 4 discusses the econometric methodology and the empirical results, while Section 5 concludes.

110 Konstantinos Drakos

2.

A ﬁnance view of 9/11: terrorism, airline stock prices and airline stock returns

Data The dataset employed in our analysis consists of the daily closing prices for 13 airline stocks covering the period 12 July 2000 to 26 June 2002. The airline stocks in our sample are the following: American Airlines (USA), Continental Airlines (USA), Delta Airlines (USA), Southwest Airlines (USA), United Airlines (USA), All Nippon Airlines (Japan), Japan Airways (Japan), British Airways (UK), Air France (France), KLM (Netherlands), Qantas (Australia), Air New Zealand (New Zealand), and Air Canada (Canada). Additionally, we collected data for the following stock market indices to be used as proxies for the market portfolio: S&P500 (New York), FTSE-100 (London), CAC-40 (Paris Bourse), Amsterdam SE All Share Index, Nikkei-225 (Tokyo), Australia Stock Exchange All Ordinaries Index, Toronto SE-35 and New Zealand SE Capital 40. All series were collected from Datastream. The impact of 9/11 on market value The immediate impact of the terrorist attacks can be identiﬁed across the world stock exchanges with virtually all airline stocks’ prices falling sharply on 9/11. Extraordinarily negative (daily) returns for the airline stocks were recorded in all stock markets but most notably in the NYSE traded airline stocks, reﬂecting the drop in stock prices, as a result of the 9/11 terrorist attacks.6 Also recall, that two of the airlines (American Airlines and United Airlines) in our sample had their aircraft hijacked and directed towards the World Trade Centre by the terrorists. Table 7.3 summarizes this information. The average price devaluation (price loss) for the airline stocks in our sample was of the order of 30 per cent, with the New York-listed airlines averaging approximately a 53 per cent loss of market value. Table 7.3 provides some further information regarding the dynamics of price following the 9/11 events. Of particular interest is the number of trading days elapsed until recovery, deﬁned as the number of trading days needed for the stock price to return to its pre-9/11 level. A clear dichotomy is observed between US-based and the rest of the airlines in our sample. Three out of the ﬁve US listed airlines have not been able as yet to recover, while Delta Airlines and Southwest Airlines reached their pre-9/11 level after 127 and 37 trading days respectively. As far as the non-US airlines are concerned, all but one (British Airways7 ) managed to recover. The average time needed for price recovery was 84 trading days (approximately four months). Another interesting fact is the current (point-to-point) price movement compared to 9/11. Since8 the terrorist attacks in the US, nine out of the thirteen airline stocks in our sample have sustained a loss of market value, which is about 68 per cent of the immediate impact of 9/11. This metric, although it combines the 9/11 immediate, and at least the medium-term impact, as well as the effect of expectations and the ongoing structural adjustments in the industry, is still indicative of the

Terrorism-induced structural shifts in ﬁnancial risk

111

Table 7.3 The impact of 9/11 attacks on market value

American Airlines Continental Airlines Delta Airlines South West Airlines United Airlines Average (US airlines only) All Nippon Airlines Japan Airways British Airways Air France KLM Qantas Air New Zealand Air Canada Average (non-US airlines) Average (all airlines) t-statistic for equality of means (US vs. non–US)

Price devaluation # of trading on ﬁrst trading days until day after 9/11 recovery attacks

Price movement relative to 9/11 (up to 26/6/02)

−50.07% −68.16% −59.04% −27.52% −56.59% −53.28% −16.09% −10.98% −23.84% −17.73% −25.46% −9.30% −19.71% −20.07% −17.90% −31.12% 5.87∗∗

−41.24% −59.86% −46.36% −10.74% −63.17% −44.27% −21.90% −5.54% −3.78% 31.56% 44.18% 30.52% −8.45% 10.38% 9.62% −11.11% 4.19∗∗

NR NR 127 37 NR 112 58 NR 49 49 22 177 122 84 -

Notes: Recovery is deﬁned as the stock price reaching its 9/11 level. NR stands for ‘not recovered’. One (two) asterisk denotes signiﬁcance at the 5 (1) per cent signiﬁcance levels. The t-statistic tests the null hypothesis that the relevant mean between US-based airlines is equal to the non-US based airlines.

market conditions suggesting that the industry is still in a critical phase. Further evidence for the above-mentioned dichotomy is provided by the rejection of the t-test for equality of means between the US and non-US airlines. In particular, it is clear that the immediate drop in market value has been on average higher for US airlines. Additionally, US airlines have, on average, sustained higher loss in market value in the period 9/11 to 26 June 2002. The observed sharp fall in airline stock prices can be explained through the fundamental pricing formula, which basically asserts that in an efﬁcient market and in the absence of arbitrage, the current price of an asset is equal to the expected net present value of the future stream of cash ﬂows generated by the asset. Algebraically, that is: T E0 (Ct ) Pi,t = (1 + r)t

(1)

t=1

Where Pi,t is the price for stock i at time t, C is the cash ﬂow (dividend) generated by the stock,9 E0 is the expectations operator conditional on information available at date 0 (the time of decision making), and ﬁnally r is a constant rate of interest.

112 Konstantinos Drakos To the extent that terrorism was perceived as being an adverse demand shock, ceteris paribus, expected future airline proﬁts are lower and consequently dividends will be lower. Hence, the current stock price must be lower in order to reﬂect the path of future stream of dividends. In other words, the observed fall in stock prices can be viewed as the direct result of a corrective process that reﬂects the arrival of new information. Although a vigorous discussion has taken place regarding the short- and medium- term effects of the terrorist activities on the airline industry, no attempt has been made to answer these questions using formal ﬁnancial theory and robust econometric methods. In order to assess the effects of 9/11 on airline stocks, one needs to specify a return-generating process. The most commonly used model is the so-called Market Model according to which, each asset’s overall (total) risk can be decomposed into two components: systematic and idiosyncratic (Sharpe, 1964; Lintner, 1965). The fundamental distinction between the two types of risks is diversifability. In particular, systematic risk cannot be diversiﬁed away, whereas systematic risk can be eliminated by effective diversiﬁcation. As a consequence, the only type of risk that will be priced by rational agents in equilibrium, is systematic risk since investors must be rewarded for bearing it. In contrast, no investor will be rewarded for bearing diversiﬁable risk, since this can be costlessly avoided. Systematic risk is typically deﬁned as an asset return’s covariation with the market (portfolio) return, known as beta. According to the basic form of the Market Model formula, the excess return on an asset, over and above the risk-free rate of return, is generated as follows (for an extensive discussion see Campbell et al. 1997): (Ri,t − Rf ,t ) = βi (RM ,t − Rf ,t ) + εi,t

(2)

Where Ri , RM , Rf stand for the returns on the i-th asset, the market portfolio and the risk-free rate respectively, and εi,t follows a ‘white noise’ process. Each asset’s risk is measured by the parameter βi , which can be recovered by a standard Ordinary Cov[Ri,t , RM ,t ] Least Squares regression and is deﬁned as: βi = Var[R . M ,t ]

3.

Measuring the effects of terrorism on airline stock returns: a set of testable hypotheses

An adverse demand shock that only affects the airline industry is industry-speciﬁc. In other words, airline stocks should not be considered as being riskier in the Market Model sense, since an investor can diversify this type of risk. Thus, according to the model, once the ‘corrective’ action in prices has taken place, returns should be described by the process in equation (1). This is a testable hypothesis: provided that the post 9/11 period developments did not affect the market as a whole, airline stock betas should be unaffected by the terrorist incidents because they did not alter investor perception of assets’ (airline stocks) systematic risk. In contrast, to the extent that terrorism exerts a negative impact on the market as a whole and therefore shifting systematic risk, airline stocks are also affected.

Terrorism-induced structural shifts in ﬁnancial risk

113

Moreover, one would expect airline stocks to exhibit higher sensitivity than the average stock (operating in other than the aviation industry). What is more obvious and rather less controversial is that airline stocks’ total risk has probably increased, implying that at least idiosyncratic risk has increased (which should not be priced in equilibrium). This is another testable hypothesis: airline stocks should exhibit higher total risk, with their returns being more volatile. We test the following three hypotheses: H1: 9/11 attacks did not affect the systematic risk of airline stocks: βi,pre 9/11 = βi,post 9/11 H2: 9/11 attacks did not affect the airline stocks’ volatility (total risk) σi,pre 9/11 = σi,post 9/11 H3: 9/11 attacks did not affect the ratio of systematic risk to total risk:

2 βi2 ∗ σM 2 σi,T

=

pre 9/11

2 βi2 ∗ σM 2 σi,T

post 9/11

Our analysis, although formally not an event study, bears considerable relevance to it and therefore should be viewed as part of such an approach employed earlier in the empirical ﬁnance literature (for applications see Fama et al. 1969; Brown and Warner, 1985; Malatesta and Thompson, 1985; Corrado, 1989; Bhattacharya et al. 2000; Wilson et al. 2000). Typically, an event-study analysis explores whether there is a statistically signiﬁcant reaction in a ﬁnancial market (or an asset in general) to past occurrences of a particular event that is hypothesized to affect prices. The present analysis differs in the sense that instead of focusing on abnormal returns, we are predominantly interested not just in investigating short-term abnormalities, but rather explore the issue of structural shifts in the risk proﬁle of assets whose underlying operation has been affected by terrorism. Furthermore, we employ rather different econometric tools, which are designed to capture shifts in a dynamic framework. Recall that if one had chosen to follow the event-study approach would have not been able to test for such effects. An event-study methodology would only be useful in identifying abnormal returns surrounding the event occurrence. Although such information is useful, it is of limited use to portfolio managers and investors for two main reasons. First, terrorist attacks are random and although one may claim that they follow a certain pattern, this pattern is not linked to developments and trends in ﬁnancial markets (unlike other random events such as mergers or takeovers which are clearly correlated with information in the market). Second, after the immediate impact of the event (in our case the terrorist attack), markets typically factor in the pricing mechanism the new information, leading to the observed abnormalities. However, once this corrective process is completed investors and

114 Konstantinos Drakos portfolio managers are ultimately interested in the portfolio behaviour of (airline) stocks, which ultimately leads to the assessment of their risk proﬁle and other distributional properties (for instance their volatility). For readers interested in the measurement of abnormal returns on the wake of the 9/11 attacks, Chen and Siems (2002) and Carter and Simkins (2002) have focused on these research questions mainly employing an event-study approach. Chen and Siems (2002) have studied a broader set of events (deﬁned as cataclysmic) including 9/11, while Carter and Simkins (2002) have focused on the effects of 9/11 on the US aviation industry. The present study is complementary to Chen and Siems (2002) and Carter and Simkins (2002) since it (i) extends the hypotheses tested, and uses different econometric techniques, and (ii) provides evidence for airline carriers across international markets. One can immediately draw the parallels between the present study and the standard event study framework, 9/11 can be viewed as date 0 (event occurrence), while pre-9/11 (since 12 July 2000) corresponds to the estimation (pre-event) window and ﬁnally 10 September onwards (until 26 June 2002) is the post-event window.

4.

Econometric methodology

In order to explore H1, namely the stability of beta coefﬁcients in the pre- and post- terrorist incidents sub-samples, implying that there is no change in the systematic risk of airline stocks, we estimate betas by employing equation (2) whose parameters are estimated by Ordinary Least Squares. Essentially, our estimation and testing strategy is conditional on the assumption that systematic risk can be correctly measured within the Market Model. The parameter of interest, beta, is estimated for the pre- and post- terrorist incident periods in order to assess its stability by conducting a formal parameter stability test (Chow, 1960). Moreover, to further assess the stability of the beta coefﬁcient we employ a recursive estimation procedure (Brown et al. 1975), which will enable us to trace the evolution of systematic risk. H2 will be tested by applying a set of standard tests of variance equality on the two sub-samples of excess returns (Conover and Johnson, 1981). Finally, H3 will be investigated by decomposing total risk into systematic and unsystematic according to the Market Model. 4.1.

Empirical results: measuring the effects on risk and volatility

Stability of beta In order to assess the effects of 9/11 on the risk proﬁle of airline stocks we investigate whether there has been any signiﬁcant change in their respective betas. This exercise boils down to a stability test of the slope coefﬁcient, obtained from linearly projecting the return of the airline stocks on the corresponding market return. Obviously, the stability test is conducted using 9/11 as the point of reference. The stability test is the well-known Chow (1960) parameter stability test. We estimate

Terrorism-induced structural shifts in ﬁnancial risk

115

the market model (equation (2)) assuming that the error term follows a white noise process.10 The estimation results are reported in Table 7.4.11 The picture that emerges for the airline stocks across stock markets is as follows: before 9/11 the vast majority of airline stocks were associated with betas significantly lower than unity, suggesting that although they ‘carried’ systematic risk they could be thought of as defensive stocks. Standard portfolio theory implies that in a hypothetical scenario of a 1 per cent increase (decrease) in the market portfolio’s return, the return on a typical airline stock would on average increase (decrease) by less than 1 per cent. Consequently, given the defensive nature of these assets according to the Market Model their expected return should reﬂect the fact that they provide investors with a service; that of mitigating ﬂuctuations relative to the market portfolio returns. In contrast, during the post 9/11 period the systematic risk of airline stocks has dramatically increased. From defensive stocks, they have been transformed to aggressive stocks, with betas considerably higher than one. The formal parameter stability tests indicate that the null hypothesis of parameter constancy is overwhelmingly rejected in all but two cases (KLM, Qantas). In other words, according to the Chow test, systematic risk of airline stocks has signiﬁcantly increased as a result of the terrorist incidents of 9/11.12 Our empirical evidence document an apparent shift in the riskiness of airlines stocks after the 9/11 attacks. In other words, we document a sustained terrorism-induced effect on one of the fundamental determinants of airline stocks’ returns. Equality of volatility We employ the returns’ volatility, measured by their sample volatility for the two sub-periods (pre- and post-9/11) and test the hypothesis that the two sub-samples exhibited the same volatility. The results from a battery of tests that were applied are reported in Table 7.5. For the majority of cases (nine out of thirteen) the null hypothesis of equal volatilities between the pre- and post- 9/11 samples was emphatically rejected, implying that unconditional uncertainty during the post-9/11 period has been signiﬁcantly higher. There is no evidence of higher volatility in the post-9/11 era for the two airlines listed in the Tokyo Stock Exchange, while there is relatively weaker evidence for higher volatility for Southwest Airlines and Qantas. In conclusion, the empirical evidence suggest that volatility in the post-9/11 period has signiﬁcantly increased, implying that airline stocks have been trading in a more uncertain environment, which clearly mirrors the uncertainty surrounding their operations. Decomposition of total risk According to the fundamental concept that total risk can be linearly decomposed into systematic and unsystematic risk, one can achieve a numerical decomposition 2 as the i-th asset’s total risk, based on the Market Model. In particular, deﬁning σi,T

0.61 (8.57)∗∗ 0.75 (10.93)∗∗ 1.21 (5.86)∗∗ 0.84 (6.48)∗∗ 0.93 (5.75)∗∗ 0.99 (4.07)∗∗ 2.43 (3.75)∗∗ 0.71 (4.12)∗∗ 1.05 1.13

1.31 (4.59)∗∗ 1.56 (3.95)∗∗ 1.29 (3.84)∗∗ 0.90 (5.49)∗∗ 1.24 (3.67)∗∗ 1.26

-

4%

13%

7%

15%

19%

17%

25%

22%

-

13%

18%

18%

0.51 (5.93)∗∗ 0.65 (6.79)∗∗ 0.82 (2.74)∗∗ 0.62 (5.52)∗∗ 0.71 (5.20)∗∗ 0.82 (4.98)∗∗ 1.45 (3.63)∗∗ 0.65 (3.53)∗∗ 0.77 0.76

0.92 (7.37)∗∗ 0.79 (7.33)∗∗ 0.73 (8.52)∗∗ 0.65 (6.35)∗∗ 0.62 (5.20)∗∗ 0.74

21%

18%

Beta

Adj.R2

Pre-9/11

-

6%

12%

4%

7%

12%

11%

18%

15%

-

10%

13%

17%

0.78 (6.06)∗∗ 0.90 (9.07)∗∗ 1.75 (6.33)∗∗ 1.05 (5.01)∗∗ 1.05 (4.40)∗∗ 1.15 (2.68)∗∗ 4.32 (2.89)∗∗ 0.83 (3.19)∗∗ 1.47 1.73

1.97 (2.88)∗∗ 2.86 (3.05) 2.23 (2.74)∗∗ 1.33 (3.55)∗∗ 2.28 (2.84)∗∗ 2.13

20% 17%

Beta

Adj.R2

Post-9/11

-

2%

17%

13%

20%

25%

25%

36%

36%

-

20%

27%

27%

1.24

1.59

3.76∗∗

7.99∗∗

2.38∗

3.95∗∗

-

2.51*

12.05∗∗

-

17.21∗∗

8.08∗∗

19.98∗∗

26.71∗∗

11.30∗∗

26% 27%

Chow break- point test

Adj.R2

Stability test

Notes: Robust t-ratios based on White (1980). One (two) asterisk denotes signiﬁcance at the 5 (1) per cent signiﬁcance levels.

Average (non-US airlines) Average (all airlines)

Air Canada

Air New Zealand

Qantas

KLM

Air France

British Airways

Japan Airways

Average (US airlines only) All Nippon Airlines

United Airlines

Southwest Airlines

Continental Airlines Delta Airlines

American Airlines

Beta

Whole sample

Table 7.4 Market model for individual airline stocks: estimation results

Terrorism-induced structural shifts in ﬁnancial risk

117

Table 7.5 Tests of equality of volatility

American Airlines Continental Airlines Delta Airlines South West Airlines United Airlines All Nippon Airlines Japan Airways British Airways Air France KLM Qantas Air New Zealand Air Canada

F-test

Siegel-Tukey

Bartlett

Levene

Brown-Forsythe

3.19∗∗ 7.57∗∗ 5.42∗∗ 1.74∗∗ 4.96∗∗ 1.13 1.19 2.13∗∗ 1.52∗∗ 1.34∗ 1.35∗ 11.66∗∗ 3.70∗∗

7.12∗∗ 45.11∗∗ 15.89∗∗ 1.71 27.08∗∗ 0.01 0.34 26.71∗∗ 0.03 10.52∗∗ 1.32 5.65∗ 17.02∗∗

83.38∗∗ 241.06∗∗ 172.24∗∗ 19.24∗∗ 117.42∗∗ 1.05 1.94 35.99∗∗ 11.16∗∗ 5.43∗ 5.44∗ 343.30∗∗ 106.27∗∗

7.22∗∗ 34.28∗∗ 14.48∗∗ 2.14 19.47∗∗ 0.01 0.41 24.80∗∗ 2.12 7.59∗∗ 0.72 25.82∗∗ 26.68∗∗

7.04∗∗ 33.19∗∗ 13.78∗∗ 2.16 19.48∗∗ 0.01 0.35 24.64∗∗ 2.11 7.51∗∗ 0.79 25.93∗∗ 26.30∗∗

Notes: One (two) asterisk denotes signiﬁcance at the 5 (1) per cent signiﬁcance levels.

2 as the market portfolio’s volatility, and σ 2 as the asset’s idiosyncratic risk, the σM i following applies to every asset: 2 2 = βi2 ∗ σM + σi2 σi,T

(3)

We are interested in the percentage of total risk corresponding to systematic risk, as a way to quantify the increase in airline stocks’ systematic risk. So, we compute the following ratio: 2 βi2 ∗ σM 2 σi,T

(4)

Starting with the ratio of the post to pre 9/11 beta coefﬁcients, systematic risk as measured by beta has exhibited an average twofold increase for airline stocks studied in our sample, while for the airline stocks listed in the NYSE a threefold increase is recorded. In relative terms, using the ratio of systematic risk to total risk, the former has shown an average 72 per cent increase in the share it represents in the total decomposition of risk. In particular, the decomposition (equation (4)) showed that in the pre-9/11 period systematic risk accounted for about 10 per cent of total risk while in the post-9/11 era it accounts for about 19 per cent.

5.

Conclusions

According to formal stability tests, the systematic risk of airline stocks has signiﬁcantly increased since the terrorist incident of 9/11. In particular, we provide evidence for a structural break in systematic risk (beta) after the 9/11 attacks. Furthermore, volatility has dramatically increased in the post-9/11 period,

118 Konstantinos Drakos Table 7.6 Decomposition of total risk for individual airline stocks

American Airlines Continental Airlines Delta Airlines South West Airlines United Airlines Average (US airlines only) All Nippon Airlines Japan Airways British Airways Air France KLM Qantas Air New Zealand Air Canada Average (non-US airlines) Average (all airlines)

(Systematic risk/total risk)*100

βpost

Pre-9/11

Post-9/11

βpre

15.83 13.89 13.48 10.02 7.83 12.21 9.52 11.74 9.87 7.54 5.81 5.99 24.35 6.52 10.16 10.95

21.44 22.69 21.89 22.69 16.66 21.07 21.37 20.62 21.88 17.93 14.74 19.58 21.18 2.33 17.45 18.85

2.14 3.62 3.05 2.04 3.67 2.90 1.52 1.38 2.13 1.69 1.47 1.40 2.97 1.27 1.72 2.18

reﬂecting increased uncertainty surrounding the airline industry. Finally, by decomposing total risk into its constituents (systematic and idiosyncratic), our ﬁndings suggest that systematic risk has on average more than doubled (as measured by beta), while the percentage it represents over total risk has shown a considerable increase. This evidence has wide implications for portfolio diversiﬁcation and the cost (and ability) of the listed airlines to raise capital. Managers, whose portfolios include airline stocks, are now faced with a sizeable increase in the undiversiﬁable risk they bear, which might have welfare effects and call for large ﬂows as a result of portfolio reshufﬂing. There is no doubt that interest in the economic consequences of terrorism has increased and also that there is a need for further research into this important issue. Following the line of research presented in the present study, a potentially fruitful extension should focus on the effects of 9/11 across sectors of the economy, in order to explore any potential differences and their underlying causes. Additionally, further international evidence is needed to shed more light on the cross-border economic effects which are bound to be wide given the globalized setting of modern economies.

Acknowledgements I would like to express my gratitude to two anonymous referees for their valuable comments that contributed towards the improvement of the paper. I am also grateful to Roy Bailey, Tilman Brück, Panayiotis Konstantinou, and especially to Tom Siems for their insightful comments. I would also like to thank the organizers and

Terrorism-induced structural shifts in ﬁnancial risk

119

the participants of the DIW (Berlin) Conference on the Economic Consequences of the 11 September Attack for their useful comments. Financial support by the DIW is gratefully acknowledged. Any remaining errors and ambiguities are my responsibility.

Endnotes 1 2 3 4 5 6 7 8 9 10 11 12

The insurance and tourism industries are also sectors facing extraordinary conditions. Source: IATA Monthly International Statistics. Source: IATA Monthly International Statistics. Source: IATA Monthly International Statistics. It should be mentioned, however, that ﬁgures for the third quarter are distorted given the three-day closure of the USA’s airspace in the aftermath of the terrorist attack. It should be noted that the immediate impact on stock markets was initially felt in European stock markets while in the US trading resumed on 17 September. Of course the paths of stock prices represent a combination of the terrorist events and other market conditions. Figures based on 26 June 2002. For simplicity we ignore the resale value of the asset and hence any capital gains. Relaxing the assumption of whiteness for the error term by allowing for a GARCH process does not signiﬁcantly affect the estimation results. The full set of results is available upon request from the author. Estimating beta coefﬁcients using recursive least squares offers a visual testimony to this increase in systematic risk. The time series graphs of the recursive betas show quite clearly an abrupt increase in systematic risk after 9/11. Betas exhibit a ‘jump’, and have remained on a higher path since then. We do not provide the graphs for space conservation reasons, however, they are available from the author upon request.

References Bhattacharya, U., Daouk, H., Jorgenson, B. and Kehr, C. (2000) ‘When an event is not an event: the curious case of an emerging market’, Journal of Financial Economics, 55: 69–101. Brown, R., Durbin, J. and Evans, J. (1975) ‘Techniques for testing the constancy of regression relations over time’, Journal of the Royal Statistical Society, 37: 149–92. Brown, S. and Warner, J. (1985) ‘Using daily stock returns: the case of event studies’, Journal of Financial Economics, 14: 3–31. Campbell, J., Lo, A. and MacKinlay, C. (1997) The Econometrics of Financial Markets, Princeton, New Jersey: Princeton University Press. Carter, D. and Simkins, B. (2002) ‘Do markets react rationally? the effects of the September 11th tragedy on airline stock returns’, unpublished manuscript, Oklahoma: State University. Chen, A. and Siems, T. (2002) ‘An empirical analysis of the capital markets’ response to cataclysmic events’, unpublished manuscript presented at the DIW Conference on the Economic Consequences of September 11th . Chow, C. (1960) ‘Test of equality between sets of coefﬁcients in two linear regressions’, Econometrica, 28: 591–605. Conover, W. and Johnson, M. (1981) ‘A comparative study of tests for homogeneity of variance with applications to the outer continental shelf bidding data’, Technometrics, 23: 351–61.

120 Konstantinos Drakos Corrado, C. (1989) ‘A nonparametric test for abnormal security-price performance in event studies’, Journal of Financial Economics, 23: 385–95. Department of Transportation (USA), ‘Airline quarterly ﬁnancial review’, Ofﬁce for Aviation Analysis, various issues from 1995: Q1–2001:Q3. Fama, E., Fisher, L., Jensen, M. and Roll, R. (1969) ‘The adjustment of stock prices to new information’, International Economic Review, 10(1): 1–21. IATA (2001) ‘Monthly international statistics’, March 2000–December 2001. IATA (2001) ‘Corporate air travel survey and post September 11th supplementary survey’. Lintner, J. (1965) ‘The valuation of risky assets and the selection of risky investments in stock portfolios and capital budgets’, Review of Economics and Statistics, pp. 394–419. Malatesta, P. and Thompson, R. (1985) ‘Partially anticipated events: a model of stock price reactions with an application to corporate acquisitions’, Journal of Financial Economics, 14: 237–50. Sharpe, W. (1964) ‘Capital asset prices: a theory of market equilibrium under conditions of risk’, Journal of Finance, 19: 425–42. White, H. (1980) ‘A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity’, Econometrica 48(4): 817–38. Wilson, B., Saunders, A. and Caprio, G. (2000) ‘Financial fragility and Mexico’s 1994 peso crisis: an event-window analysis of market-valuation effects’, Journal of Money, Credit, and Banking, 32(3): 450–68.

8

Financial markets and terrorism Raﬁ Eldor and Raﬁ Melnick

This paper uses daily data to analyse how Israel’s stock and foreign exchange markets reacted to 639 terror attacks that took place between 1990 and 2003 in which 1,212 people in Israel were randomly killed and 5,726 people were randomly injured. The data distinguishes acts of terror by location, type of attack, type of target, number of casualties, and the number of attacks per day. Suicide attacks had a permanent effect on both the stock and foreign exchange market, as did the numbers of people injured and killed. Location of a terror attack had no effect on either market. Markets did not become desensitized to terror attacks. Our principal conclusion is that ﬁnancial markets continued to efﬁciently perform their economic functions. Also, market-liberalization policies were not disrupted; the conclusions about market efﬁciency suggest that the market liberalization policies contributed to coping with terror. Although based on data on terror against Israel, the results have broader implications for the Western world because of Israel’s democratic regime, its free market, and its well-developed ﬁnancial markets.

1.

Introduction

The population of Israel has been subject to ongoing terror attacks. In particular, in 639 terror attacks in the period 1990 to 2003, terrorists killed 1,212 people and left an additional 5,726 people injured (although the victims were predominantly Jews who were targeted, also included among the victims were non-Jewish citizens of the state of Israel and foreign workers). Such ongoing terror attacks can be expected to have had economic consequences reﬂected in ﬁnancial markets.1 In this paper we use daily data to analyse how the terror affected stock market prices and exchange rates. The terror attacks are identiﬁed with respect to location, type of attack, type of target, the number of casualties and the number of attacks per day. The data is used to address the questions: • • • •

Did the terror at all affect stock and currency markets? If so, are the consequences of all terror attacks alike? Is the effect of a terror attack permanent or transitory? Does market sensitivity to terror diminish over time?

122 Raﬁ Eldor and Raﬁ Melnick The answers have broader relevance for the Western world because of Israel’s democratic regime, its free market, and its well-developed ﬁnancial markets, and the prospect that terror may remain an ongoing threat to western society.2 In the following section we describe the data, which consists of constructed time series for 3,515 daily observations in the years 1990–2003. Section 3 provides background on the stock and the currency markets in Israel. The model and the econometric methodology are set out in Section 4. The hypotheses and the results are presented in Sections 5 and 6. Section 7 summarizes the conclusions.

2.

The anatomy of terror attacks

Terrorism against the civilian population has always been a feature of the Israeli existence, and occurred before the founding of the modern state of Israel in 1948. Terror has, however, particularly been an unfortunate part of the daily experience of the population of Israel in the decade since the Oslo Peace Process began, and more so since 27 September 2000 after which terror heightened. The terror attacks cannot be considered as sporadic events. Hence the economic consequences cannot be analysed as an event study but rather, because of the continuous process, the study of the consequences of terror requires an econometric analysis of the time series type. The central aspects of the time-series dimension of terror in Israel from 1990 until 2003 are presented in Figure 8.1. This ﬁgure indicates that the sample is naturally divided into two periods – before and after 27 September 2000. This date marks the rupture of the peace process and the beginning of massive terror attacks on Israel initiated by the Palestinians. The ﬁgure shows that the peace process was accompanied by a continuous progression of terrorism aimed at preventing a successful resolution of the Israeli-Palestinian conﬂict. After 27 September 2000 the number of terror attacks and the brutality of the attacks intensiﬁed. We constructed time series of the terror attacks and their characteristics; a summary of the data is presented in Table 8.1. The classiﬁcation of the data is: by the geographical location – in and out the green line3 and whether the terror attack took place in a major city, Haifa, Jerusalem, Tel Aviv or elsewhere; by the target type4 – transport, facilities, personnel, military, government and other; by the attack type – killing oneself to kill others (suicide), cold weapon, armed assault, bombing, kidnapping and other; by the total number of people injured and killed; by the number of attacks per day and by the hour of the attack. Table 8.2 presents the damage rates.5 The damage rate for attacks in Haifa is high but there were a small number of attacks in Haifa (only ﬁve). The larger damage rates are for attacks in Tel Aviv, suicide attacks, attacks on facilities,6 and attacks on Jerusalem. Table 8.3 shows the proportion of victims killed, injured, and attacks by categories. These estimates can be interpreted as conditional probabilities. For example: conditional on a terror attack taking place, the probability that the attack was inside the green line is 0.43; conditional on people being killed in the attack, the probability that the attack is inside the green line is 0.61; and so on.

Financial markets and terrorism

123

Figure 8.1 Terror attacks, victims injured, victims killed and suicide attacks.

Before turning to the testing for the impact of terror attacks and their characteristics, we present, in the next section, a short description of the stock and the currency markets in Israel.

3.

Financial markets in Israel

The characteristics of the ﬁnancial markets affected by terror determine market responsiveness. In this section we provide a brief background for the attributes of the stock and foreign exchange markets. The Tel Aviv Stock Exchange (TASE) is the sole securities market in Israel. The market value of the stocks listed on the TASE is in general more than one-half of Israel’s GDP. The TASE, which is an important market for ﬁnancial intermediation and an important source of ﬁnancing for the business sector, is quite sophisticated and includes derivatives such as stock index options (the trading volume of which relative to the underlying assets is the world’s highest). Dual listing of stocks is possible (in 2004, 22 stocks in the 100 index were listed both on the TASE and on one of the US stock markets). The Tel-Aviv 100 index

124 Raﬁ Eldor and Raﬁ Melnick Table 8.1 Summary of terror attacks by categories (January 1990–June 2003) Total Green line1

In Out

Large city1

Jerusalem Tel Aviv Haifa

Target type1

Attack type1

Victims killed2 Victims injured2 Number of attacks

Since 27 September 2000

275 364

140 284

56 18 5

31 11 5

Transport3 Facilities4 Personnel5 Military6 Government7

162 54 231 169 5

120 31 137 118 4

Suicide Cold weapon Armed assault Bombing Kidnapping

109 57 379 69 19

91 11 272 46 3

1,212 5,726 639

807 4,047 424

1 2 3 4

Qualitative categories treated as dummy variables in the econometric analysis. Numerical variables treated as continuous variables in the econometric analysis. Aircraft, Airport, Bus, Cargo Transport, Convoy, Ship, Train/Railway, Vehicle. Bank, Beach/Waterfront, Building, Bus Stop, Entertainment Facility, Garage, Hotel, Marketplace, Ofﬁce, Place of Worship, Restaurant, School/University, Shopping Centre, Store. 5 Businessperson, Civilian, Celebrity/Personality, Religious Figure, Student, Tourist. 6 Checkpoint/Border Crossing, Military Personnel, Police Facility/Personnel. 7 Diplomat, Embassy, Government Building, Government Personnel, Peace-keeping Mission.

(thereafter TA100) used in the paper is a value-weighted index composed of the largest 100 companies in terms of market value traded in the TASE. Figure 8.2 shows a sharp rise and a decline of the stock market in 1993 that has nothing to do with terror attacks but is due to a stock market bubble. The decline in 1993 seems stronger than the decline after the beginning of the terror attacks on Israel in 27 September 2000. The turning point in the TA100, in the year 2000, is identiﬁed at its peak level (544) in March 2000, which preceded the intensiﬁcation of the terror attacks from 27 September 2000. There is a positive trend in the stock market that starts in February 2003, before the beginning of the war in Iraq and before the temporary break in terror attacks (‘Hudna’) during the summer months of 2003 (see also Orgler, 2002). In Figure 8.3 we observe that in the second half of the 1990s the Israeli stock market became closely related to the USA stock market, and that the turning points of the Israeli market appear to be related to the turning points in the USA market.

Financial markets and terrorism

125

Table 8.2 The damage rate1 of attacks by categories2 (January 1990–June 2003)

Green line Large city

Target type

Attack type

In Out Jerusalem Tel Aviv Haifa Other Transport Facilities Personnel Military Government Other Suicide Cold weapon Armed assault Bombing Kidnapping Other

Killed

Injured

2.7 1.3 3.5 5.9 9.6 1.1 2.7 4.2 1.4 1.2 0.8 0.3 4.7 1.1 1.4 1.1 1.1 0.8

17.9 2.2 30.3 37.9 27.6 5.6 11.6 35.8 6.8 1.7 4.6 1.7 34.7 0.4 3.5 9.0 0.0 0.5

1 The damage rate is deﬁned as the number of victims killed or injured per attack by category. 2 See deﬁnitions in Table 8.1.

Table 8.3 Proportion1 of people killed, injured and days of a terror attack by categories2 (January 1990–June 2003)

Green line Large city

Target type

Attack type

In Out Jerusalem Tel Aviv Haifa Other Transport Facilities Personnel Military Government Other Suicide Cold weapon Armed assault Bombing Kidnapping Other

Killed

Injured

Day of attacks

0.58 0.42 0.18 0.09 0.05 0.68 0.33 0.16 0.30 0.20 0.00 0.01 0.38 0.07 0.46 0.08 0.02 0.00

0.70 0.30 0.28 0.12 0.05 0.55 0.30 0.26 0.30 0.11 0.00 0.02 0.51 0.03 0.34 0.11 0.00 0.01

0.45 0.55 0.10 0.03 0.01 0.85 0.25 0.09 0.36 0.26 0.01 0.03 0.18 0.10 0.56 0.12 0.03 0.01

1 The proportion can be interpreted as conditional probabilities. 2 See deﬁnitions in Table 8.1.

126 Raﬁ Eldor and Raﬁ Melnick

Figure 8.2 The Tel Aviv 100 stock market index [January 1990–June 2003 (log)].

Figure 8.3 The Tel Aviv 100 and the S&P500 indices [January 1990–June 2003 (logs)].

The Israeli foreign exchange market is an active inter-bank market that trades continuously in relatively large volumes. Until recently, due to several restrictions imposed by the Central Bank, market participants were mainly corporations who use foreign exchange in their transactions. Due to liberalization of the foreign exchange market, implemented gradually during the last decade, there are no trade

Financial markets and terrorism

127

Figure 8.4 The basket of currencies exchange rate and the exchange rate band [January 1990–July 2003].

restrictions on the foreign exchange market. The banks act as ‘market makers’ since all transactions go through them. Around 3pm of each trading day, except Friday – around 1pm, the central bank publishes a settlement (representative) exchange rate for each currency that is based on the quotes obtained from the major banks. After a stabilization program in 1985 there was a ﬁxed exchange rate. Over time the exchange rate regime became more ﬂexible and a horizontal exchange rate band, around a basket of ﬁve currencies reﬂecting the average international trade between Israel and the ‘basket’ economies, was adopted. In the nineties the horizontal band was transformed into a narrow diagonal band. Further ﬂexibility was applied by widening the band (see Figure 8.4). The central bank is committed to intervene only when the band is breached. Apart from the changes in the slope and the width of the band, the trading regime changed several times during the 1990s. Until July 1994 foreign exchange trade was conducted only between the central bank and commercial banks. In July 1994 inter-bank foreign exchange trading was allowed. Also, until February 1996, the central bank intervened continuously in the foreign exchange market to keep the exchange rate close to the centre of the band, since the exchange rate was used as the anchor to the price level. Central bank intervention ceased in February 1996,7 with a move to an inﬂation-targeting regime with a ﬂexible exchange rate.8 In Figure 8.5 we observe sharp changes, typically upward shocks, of the exchange rate that are closely associated with exogenous shocks unrelated to terror attacks. In the early nineties we observe sharp ﬂuctuations in the exchange rate due to speculative attacks on the domestic currency related to expected changes in the location of the exchange rate band. In 1998 we observe a sharp rise in the

128 Raﬁ Eldor and Raﬁ Melnick

Figure 8.5 The dollar and basket of currencies exchange rate [January 1990–June 2003 (logs)].

exchange rate related to the Asian crises, and the sharp rise in 2002 is associated with the unexpected cut of 200 basis points of the interest rate set by the Bank of Israel. It is clearly seen that the foreign currency market was not profoundly disturbed when the terror attacks started in 27 September 2000. The trends after the beginning of the terror attacks seem a natural continuation of the trends before the attack. This short description of ﬁnancial markets indicates that the impact of terror attacks is not obviously identiﬁed by ‘eye econometrics’. Before turning to the econometric analysis, it is appropriate to remark that the terror attacks were accompanied by a process of reform in both the capital and foreign currency markets, and that at no time did trading cease because of an attack. The markets functioned properly and became more competitive over time. It is possible, although we do not prove here, that the reforms in the ﬁnancial markets and the ﬁnancial liberalization helped the economy to better cope with the uncertainties due to terror attacks.

4.

The theoretical model and the econometric methodology

The basic model is given in (1). xt is a non-stationary I(1) variable with a possible drift ft , and ut is a white noise innovation: xt = ft + xt−1 + ut .

(1)

Financial markets and terrorism

129

Table 8.4 Testing for unit roots (January 1990–June 2003) ADF

1% critical level

5% critical level

Level - TA1001 First difference - TA1002

−0.629 −27.98

−3.97 −3.44

−3.41 −2.86

Level - dollar1 First difference - dollar2

−1.67 −26.75

−3.97 −3.44

−3.41 −2.86

Level - basket1 First difference - basket2

−2.10 −26.91

−3.97 −3.44

−3.41 −2.86

1 In log including constant and trend. 2 Log differences including constant.

In our case xt represents either the log of the exchange rate (the dollar or the basket of currencies) or the log of stock market index (TA100). In Table 8.4, we present the empirical evidence, augmented Dickey Fuller tests for unit roots, using daily Israeli data from 1 January 1990 until 30 June 2003, which strongly support this speciﬁcation. The presence of a unit root cannot be rejected for the log levels and is strongly rejected for the ﬁrst differences. A central question concerns the speciﬁcation of ft . In a rational expectations competitive environment, it is natural to specify the fundamental market inﬂuences ft for the exchange rate model by the interest rate spread, domestic minus foreign interest rates. This assumption implies interest rate parity and the absence of arbitrage in the currency market. The speciﬁcation is justiﬁed by the liberalization of the foreign exchange market and by the foreign exchange regime that is very close and converging, over time, to a pure ﬂoating regime. For the stock market model, ft is speciﬁed by the log difference of the S&P500 index. This is justiﬁed by the liberalization of the domestic capital markets, the openness of the Israeli economy to free capital movements, and the relative large number of stocks that are traded simultaneously in the Israeli and the USA markets. To support this speciﬁcation we present, in Table 8.5, Granger causality tests that indicate strong Granger causality from the S&P500 index to the TA100 index. The residual ut is a white noise innovation. Denoting by D the ﬁrst difference operator, equation (1) can be transformed into: Dxt = ft + ut

(2)

Table 8.5 Granger causality tests (6 lags) (January 1990–June 2003) Null hypothesis:

F-statistic

Probability

Dlog (TA100) does not Granger Cause Dlog (S&P500) Dlog (S&P500) does not Granger Cause Dlog (TA100)

0.65418 21.6941

0.68681 0.00000

130 Raﬁ Eldor and Raﬁ Melnick The ﬁrst natural hypothesis is a test for a structural change in ft at the beginning of the massive terror attacks that commenced on 27 September 2000. A structural change at that date reﬂects the effect of the intensiﬁcation of the terror attacks. In the second hypothesis we ask: are all terror attacks alike? We answer this question by estimating the informational content, the news, of a terror attack. To do that we decompose the ut innovation into two components: ut = ϕt + εt

(3)

where ϕt is the innovation associated with the terror attack and εt is pure noise. The basic regression model to estimate ϕt is: Dxt = α + βft + γ Tt + εt

(4)

where ϕt = γ Tt , γ is a vector of parameters and Tt represents either the occurrence of a terror attack or a set of dummy variables describing the characteristic of the terror attacks.9

5.

The hypotheses

Estimates of the fundamental equations (2) are presented in Table 8.6.10 For the stock market equation we obtain a positive and highly signiﬁcant coefﬁcient of 0.22 for the daily effect of the USA stock market on the Israeli market. The constant is 0.042, indicating an average annual drift of 16.1 per cent. The Chow test on 27 September 2000 indicates a signiﬁcant structural change; this result can be interpreted as the general effect of the terror attacks on the stock market. Because two major causes are in general identiﬁed for the downturn11 in the Israeli

Table 8.6 The fundamental regressions1 [January 1990–June 2003 (3515 daily observations)] Dependent variable

Constant

Fundamental2

Lag3

R2

Dlog (TA100)

0.042 (0.064) 0.066 (0.003) 0.067 (0.001)

0.215 (0.001) −0.006 (0.040) −0.007 (0.011)

0.079 (0.001) 0.037 (0.027) 0.033 (0.049)

0.040

Dlog (Dollar) Dlog (Basket)

0.003 0.003

LM4

Chow5

0.76 (0.47) 1.10 (0.58) 1.14 (0.56)

5.39 (0.001) 2.29 (0.076) 1.61 (0.186)

1 p values in parenthesis. 2 Dlog (S&P500) in NIS terms for the stock exchange equation and corresponding interest rate spread

for the exchange rate equations.

3 Lag-dependent variable. 4 Breuch-Godfrey Serial correlation LM test. 5 Chow test at September 27, 2000.

Financial markets and terrorism

131

Figure 8.6 Tel Aviv 100index – actual and simulated with the pre-27 September 2000 fundamental equation, September 2000–June 2003.

economy that started in October 2000, terror and the end of the high-tech boom,12 it could be argued that our interpretation of the structural change in this equation is biased. We reject this possibility because our fundamental equation includes the S&P500 index that is a reﬂection of the US and world recession. Therefore the second source of the Israeli recession is included in the equation. We can approximate the impact of terror attacks on the stock market by the difference of the actual TA100 index after 27 September 2000, with a simulation of the index using the actual value of the S&P500 index in an equation estimated with data until that date. This simulation is presented in Figure 8.6.13 In the absence of the terror attacks, the value of the TA100 index on 30 June 2003 would have been 604 compared with an actual value of 449, that is, 30 per cent lower. Further analysing the source of the structural change of the equation, we ﬁnd that the only coefﬁcient that changes signiﬁcantly is that of the S&P500 index,14 indicating a decline in the transmission effect between the two markets due to the deterioration of the Israeli economy. We tested for a further change after 9/11, and we found no effect. For the exchange rate equations we obtain a negative and signiﬁcant effect for the interest rate spread on the exchange rate. The very low R2 could be due to the fact that the interest rate spread is measured at the monthly frequency only. Contrary to the stock market equation, the exchange rate equation is stable after 27 September 2000. We do not ﬁnd a statistically signiﬁcant general effect of the intensiﬁcation of the terror attacks on the foreign currency market. It seems that the relatively comfortable balance of payments position15 at the time of the terror attacks prevented a general deterioration of the shekel. It is possible that

132 Raﬁ Eldor and Raﬁ Melnick the impact of terror attacks was reﬂected in the interest rate spread, leaving the interest parity equation unaffected. However, we reject this interpretation, since the average interest rate spread diminished after 27 September 2000.16 We turn now to the estimation of the informational content – the news – of a terror attack. The ﬁrst null hypothesis we test is a simple informational hypothesis: Ho : γ = 0.

(5)

Rejection of Ho will be interpreted as a signiﬁcant informational content of the terror attack or the terror attack characteristics. For the exchange rate model we expect γ > 0, that is, a terror attack weakens the domestic currency (a devaluation of the currency), and for the stock market model we expect γ < 0, that is, a decline in the value of stocks. Since we use daily Israeli data, we have to address the problem of the timing of the terror attack and the measures of the exchange rate and the stock market. Clearly a terror attack can occur at any time17 but the exchange rate and the stock market data are measured at discrete intervals. The representative exchange rate published by the Bank of Israel is announced at 3pm, so the impact of a terror attack early in the day will possibly affect the exchange rate the same day but an attack after 3pm will be captured only the day after. This is also true for the stock market data; we have used closing trade data so that for the stock market model the critical time is 5pm. Another problem is that no exchange rate is announced on Saturdays and Sundays and no trade takes place in the stock market on Fridays and Saturdays. So the response to a terror attack that occurs between Friday afternoon and Monday before 3pm will be captured in the change of the exchange between Friday and Monday. For the stock market, a terror attack that occurs between Thursday after 5pm and Sunday before 5pm will be captured in the stock market change between Thursday and Sunday. The terror attacks data was organized in two sets. For the exchange rate model we constructed a ﬁve-day week starting on Monday and ending on Friday and for the stock market model a ﬁve-day week starting on Sunday and ending on Thursday.18 Although we collected the time of the attack data, given the discrete nature of the dependent variable, the impact of a terror attack could be incorporated in prices either on the same day of the attack or the day after without violating market efﬁciency. To overcome this problem we included in all the regressions the date of the attack and its ﬁrst lag. Ho is tested for both coefﬁcients. In an efﬁcient market the information of a terror attack should be instantaneously incorporated in market prices. Theoretically we could test the efﬁciency hypothesis by adding a second lag to the equation and testing whether the second lag coefﬁcient is statistically different from zero: Dxt = α + βft + γ0 Tt + γ1 Tt−1 + γ2 Tt−2 + εt

(6)

In our case, this is not a clean test for efﬁciency for two main reasons. The ﬁrst reason is technical. Assume that a terror attack has a transitory effect on

Financial markets and terrorism

133

Table 8.7 Interpreting the news of a terror attack Interpretation

No information Transitory information and efﬁcient market Permanent information and efﬁcient market Transitory information with lagged effects2 Permanent information with lagged effects2

Coefﬁcient1 γ0

γ1

γ2

γ0 + γ1 + γ2

− + + + +

− + + + +

− − − + +

− − + − +

1 The + sign means coefﬁcient statistically different from zero, the − sign otherwise. 2 This could indicate market inefﬁciency.

the market, that is, an effect that will not be permanently incorporated in market prices. In this case either γ0 or γ1 will be statistically signiﬁcant from zero, but the transitory effect on the level of the market will be reversed by a signiﬁcant γ2 without violating efﬁciency. The second reason is more substantial. Assume that Israel retaliates or reacts in any way to the terror attack. This will introduce new information that will affect market prices, which could be captured in γ2 . In any case the lag structure of the equation allows us to test for the hypothesis of transitory versus more permanent effects. If either γ0 or γ1 is statistically different from zero, we conclude that the terror attack has informational content. If γ0 + γ1 + γ2 = 0, the effect is transitory. Otherwise it is more permanent. Table 8.7 summarizes the different hypotheses and offers an interpretation of the possible results. The estimated equations presented in Tables 8.8, 8.8a, 8.9 and 8.10 indicate the following: (1)

(2) (3) (4) (5)

All the signiﬁcant effects have the expected signs, negative (a market drop) for the stock market equations and positive (a devaluation of the shekel) for the exchange rate equations. Attacks inside the green line are statistically signiﬁcant in all the equations. With regard to the attack-type category, only suicide attacks are statistically signiﬁcant in all the equations. The number of people injured and killed is statistically signiﬁcant in all the equations. The numbers of terror attacks per day is statistically signiﬁcant in all the equations.

In Table 8.8 we see that, even after adding the different terror dummies or quantitative characteristics, the stock market equations show a structural break in 27 September 2000. Our interpretation of this break is that the general deterioration of the Israeli economy due to the intensiﬁcation of terror attacks is reﬂected in the value of Israeli stocks. It is interesting that this effect is captured by the decline in the

Attack type

Target type

Major city

Kidnap

Bomb

C.5 weapon

Suicide

GOV.

Military

Facilities

Transport

Green line

Tel Aviv

Haifa

Jerusalem

γ1

γ2

γ0 + γ1 + γ2

−0.259 (0.162) −0.396 (0.513) −0.383 (0.228) −0.271 (0.002∗ ) −0.251 (0.029∗ ) −0.065 (0.730) 0.025 (0.829) −0.271 (0.651) −0.379 (0.007∗ ) −0.211 (0.242) −0.212 (0.206) −0.013 (0.965) 0.112 (0.544) −0.043 (0.943) −0.061 (0.848) 0.072 (0.414) −0.044 (0.702) −0.017 (0.929) −0.024 (0.836) 0.218 (0.717) −0.159 (0.252) 0.048 (0.791) 0.178 (0.287) 0.016 (0.959)

−0.238 (0.446) −1.470 (0.161) −0.542 (0.327) −0.193 (0.167) −0.275 (0.141) 0.044 (0.887) −0.097 (0.592) −0.580 (0.577) −0.551 (0.009)∗ −0.225 (0.457) 0.041 (0.882) −0.439 (0.396)

Regressions with dummy variables −0.091 (0.619) −1.031 (0.089)∗∗ −0.098 (0.757) 0.006 (0.944) 0.021 (0.858) 0.125 (0.501) −0.098 (0.392) −0.527 (0.380) −0.013 (0.927) −0.061 (0.735) 0.075 (0.653) −0.441 (0.152)

γ0

Table 8.8 Testing the impact of terror on the stock exchange1,2 (January 1990–June 2003)

2.325 (0.003)∗

2.867 (0.009)∗ 1.686 (0.047)∗

2.396 (0.011)∗

Chow3

0.045

0.042

0.043

0.043

R2

0.667 (0.516)

0.655 (0.519) 0.790 (0.454)

0.674 (0.510)

LM4

1 2 3 4 5

−0.002 (0.295) 0.019 (0.677)

V.5 injured −0.004 −0.002 −0.008 (0.055)∗∗ (0.341) (0.013)∗ −0.136 0.006 −0.111 (0.901) (0.084)∗∗ (0.003)∗

−0.021 −0.019 −0.055 (0.099)∗∗ (0.128) (0.004)∗

p values in parentheses. Coefﬁcients with grey cell are statistically different from zero, ∗ at 5% and ∗∗ at 10%. Chow test at 27 September 2000. Breuch-Godfrey Serial correlation LM test. C. Cold, A. Attackers and V. Victims.

Attacks

−0.015 (0.246)

Regressions with quantitative variables

V.5 killed 2.613 (0.016)∗ 3.129 (0.005)∗

2.594 (0.016)∗

0.043

0.042

0.043

0.787 (0.455) 0.816 (0.442)

0.754 (0.471)

Attack type

Target type

Major city

Kidnap

Bomb

C.5 weapon

Suicide

GOV.

Military

Facilities

Transport

Green line

Tel Aviv

Haifa

Jerusalem

γ1

γ2

γ0 + γ1 + γ2

−0.248 (0.182) −0.347 (0.566) −0.359 (0.258) −0.257 (0.004)∗ −0.238 (0.044)∗ −0.080 (0.669) 0.035 (0.763) −0.243 (0.685) −0.366 (0.011)∗ −0.215 (0.232) −0.211 (0.210) −0.027 (0.929) 0.124 (0.504) −0.101 (0.866) −0.064 (0.841) 0.076 (0.396) −0.049 (0.679) −0.007 (0.972) −0.021 (0.855) 0.262 (0.662) −0.179 (0.210) 0.059 (0.742) 0.174 (0.301) 0.043 (0.888)

−0.195 (0.536) −1.477 (0.161) −0.464 (0.404) −0.155 (0.288) −0.250 (0.216) 0.037 (0.906) −0.088 (0.649) −0.476 (0.648) −0.555 (0.014)∗ −0.228 (0.450) 0.025 (0.932) −0.450 (0.383)

Regressions with dummy variables −0.071 (0.699) −1.029 (0.089)∗∗ −0.041 (0.897) 0.026 (0.771) 0.036 (0.758) 0.124 (0.507) −0.101 (0.384) −0.495 (0.409) −0.010 (0.943) −0.072 (0.690) 0.061 (0.716) −0.466 (0.130)

γ0

0.842 (0.631)

0.710 (0.641) 0.663 (0.821)

0.890 (0.557)

Chow3

Table 8.8a Testing the impact of terror on the stock exchange1,2 (January 1990–June 2003)

0.049

0.046

0.047

0.046

R2

1.322 (0.267)

1.381 (0.252) 1.499 (0.224)

1.416 (0.243)

LM4

−0.002 (0.355) 0.021 (0.659)

V.5 injured −0.004 (0.061) −0.135 (0.006)∗

−0.020 (0.118) −0.002 −0.008 (0.380) (0.024)∗ 0.003 −0.110 (0.946) (0.168)

−0.019 −0.053 (0.146) (0.009)∗ 0.628 (0.708) 1.176 (0.317)

0.819 (0.556)

0.046

0.046

0.046

1.523 (0.218) 1.542 (0.214)

1.454 (0.234)

2 3 4 5

p values in parentheses. Coefﬁcients with grey cell are statistically different from zero, ∗ at 5% and ∗∗ at 10%. Chow test at January 1, 2002 for the post-27 Sept sample. Breuch-Godfrey Serial correlation LM test. C. Cold, A. Attackers and V. Victims.

1 Same speciﬁcation as in Table 6 including a dummy for the post-Sept 27 period and a dummy multiplied by the lag S&P500 log difference.

Attacks

−0.014 (0.287)

Regressions with quantitative variables

V.5 killed

138 Raﬁ Eldor and Raﬁ Melnick coefﬁcient of the change in the S&P500 index and not as a signiﬁcant reduction in the constant. To avoid possible biases in the interpretation of the direct effects of terror attacks, we estimated all the stock market equations by adding a dummy for 27 September 2000 and an interaction of the dummy with the change in the S&P500 index. The results are given in Table 8.8a. Notice that γ2 is not statistically different from zero. This is interpreted as efﬁciency of the stock market, i.e. new information of a terror attack is instantaneously incorporated in the value of stocks. Suicide attacks have a permanent effect on the stock market prices. The same is true for the number of victims injured and the number of victims killed. Terror attacks with a transitory effect on the stock market level are attacks on transport (most usually buses), attacks inside the green line, and the number of attacks per day. The impact of terror attacks on the foreign currency market is presented in Tables 8.9 and Tables 8.10. Since the Bank of Israel stopped its daily intervention in the currency market in February 1996, we ran the regressions starting on that date. Most equations remain stable after 27 September 2000.19 We therefore conclude that no general effect of terrorism is present in the currency market. As in the stock market case, none of the γ2 coefﬁcients is statistically signiﬁcant from zero,20 so indicating an efﬁcient currency market. New information of a terror attack is instantaneously incorporated in the foreign currency market. The results for the dollar and basket of exchange rates are similar. However, it seems that those for the basket of exchange rates are more reliable, since they represent more general trends in the value of the shekel. The dollar exchange rate was inﬂuenced by changes in the cross rates, which were particularly strong in recent years. We estimate permanent effects on the level of the exchange rate of: suicide attacks, attacks inside the green line,21 the number of victims injured, the number of victims killed and the number of attacks per day.22

6.

Testing the diminishing effect hypothesis

In this section we ask whether markets become insensitive to terror attacks over time. That is, do investors expect a particular number of terror attacks and do market prices internalize these expectations? To address this question we used the sample after 27 September 2000 and tested two different hypotheses. The ﬁrst hypothesis tests whether the ﬁrst attack of its kind, i.e. the ﬁrst attack on transport or the ﬁrst suicide attack, had a signiﬁcantly larger effect than the attacks that followed. For the second hypothesis we introduce an interaction between a year dummy (dt = 1 for year t and 0 otherwise) and a terrorattack characteristic to see whether the coefﬁcient of the characteristic diminishes over time. The results are sharp. The ﬁrst-terror-attack-of-its-kind coefﬁcient is not signiﬁcantly different from zero, and thus the effect of the ﬁrst attack type of its kind is not different from the terror attacks that follow, with the exception of the ﬁrst kidnapping case. The kidnapping dummy was not statistically signiﬁcant but the ﬁrst case was highly signiﬁcant, with the stock market declining by 17.1 per cent

Attack type

Target type

Major city

Kidnap

Bomb

C.5 weapon

Suicide

GOV.

Military

Facilities

Transport

Green line

Tel Aviv

Haifa

Jerusalem

γ1

γ2

γ0 + γ1 + γ2

0.064 (0.297) −0.045 (0.808) 0.318 (0.002)∗ 0.077 (0.022)∗ −0.002 (0.960) 0.123 (0.060)∗∗ −0.037 (0.361) −0.088 (0.670) 0.057 (0.212) 0.065 (0.463) −0.020 (0.727) 0.036 (0.832)

0.003 (0.957) 0.601 (0.001)∗ −0.067 (0.518) 0.019 (0.561) 0.034 (0.404) −0.004 (0.949) 0.018 (0.659) −0.023 (0.909) 0.060 (0.182) −0.204 (0.018)∗ −0.029 (0.609) 0.004 (0.979)

0.025 (0.807) 0.495 (0.126) 0.472 (0.009)∗ 0.065 (0.223) 0.048 (0.467) 0.150 (0.187) −0.009 (0.896) −0.099 (0.782) 0.210 (0.003)∗ −0.220 (0.139) −0.025 (0.797) 0.144 (0.621)

Regressions with dummy variables −0.042 (0.493) −0.061 (0.741) 0.221 (0.032)∗ −0.031 (0.349) 0.016 (0.687) 0.031 (0.638) 0.011 (0.790) 0.012 (0.952) 0.093 (0.040)∗ −0.081 (0.361) 0.024 (0.670) 0.104 (0.535)

γ0

2.650 (0.001)∗

1.244 (0.281) 2.567 (0.002)∗

3.006 (0.001)∗

Chow3

Table 8.9 Testing the impact of terror on the dollar exchange rate1,2 (February 1996–June 2003)

0.015

0.009

0.009

0.020

R2

Continued

1.484 (0.227)

2.789 (0.062) 0.258 (0.773)

1.237 (0.290)

LM4

1 2 3 4 5

0.001 (0.206) −0.011 (0.469)

V.5 injured

γ2

γ0 + γ1 + γ2

0.001 (0.088)∗∗ 0.013 (0.410)

−0.001 (0.804) 0.001 (0.212) 0.028 (0.075)∗∗

0.007 (0.125) 0.003 (0.012)∗ 0.029 (0.199)

0.012 (0.072)∗∗

p values in parentheses. Coefﬁcients with grey cell are statistically different from zero, * at 5% and ** at 10%. Chow test on 27 September 2000. Breuch-Godfrey Serial correlation LM test. C. Cold, A. Attackers and V. Victims.

Attacks

γ1 Regressions with quantitative variables

0.006 (0.139)

γ0

V.5 killed

Table 8.9—Cont’d

1.471 (0.184) 1.535 (0.163)

1.748 (0.106)

Chow3

0.008

0.009

0.009

R2

2.001 (0.135) 1.551 (0.212)

0.695 (0.498)

LM4

Attack type

Target type

Major city

Kidnap

Bomb

C.5 weapon

Suicide

GOV.

Military

Facilities

Transport

Green line

Tel Aviv

Haifa

Jerusalem

γ1 0.081 (0.234) −0.096 (0.642) 0.364 (0.002)∗ 0.089 (0.017)∗ −0.027 (0.542) 0.143 (0.050)∗ 0.009 (0.844) −0.143 (0.535) 0.063 (0.215) 0.053 (0.588) 0.025 (0.689) 0.121 (0.516)

γ0 −0.013 (0.849) −0.091 (0.660) 0.195 (0.090)∗∗ −0.018 (0.619) 0.042 (0.345) 0.025 (0.732) 0.029 (0.522) 0.231 (0.317) 0.129 (0.011)∗ −0.093 (0.343) 0.038 (0.537) 0.112 (0.548)

−0.021 (0.754) 0.724 (0.000)∗ (0.025) (0.828) 0.041 (0.270) 0.027 (0.554) 0.034 (0.635) 0.007 (0.875) 0.013 (0.955) 0.076 (0.131) −0.197 (0.040)∗ −0.032 (0.613) −0.085 (0.648)

γ2 0.047 (0.684) 0.537 (0.136) 0.584 (0.004)∗ 0.112 (0.062)∗∗ 0.041 (0.573) 0.202 (0.110) 0.046 (0.535) 0.101 (0.801) 0.268 (0.001)∗ −0.237 (0.152) 0.031 (0.769) 0.148 (0.647)

γ0 + γ1 + γ2

Regressions with dummy variables

2.063 (0.009)∗

0.800 (0.570) 2.302 (0.007)∗

2.271 (0.016)∗

Chow3

Table 8.10 Testing the impact of terror on the basket exchange rate1,2 (February 1996–June 2003)

0.012

0.007

0.006

0.016

R2

Continued

1.387 (0.250)

1.537 (0.215) 0.276 (0.759)

0.131 (0.877)

LM4

1 2 3 4 5

0.001 (0.192) 0.005 (0.756)

V.5 injured

γ2

γ0 + γ1 + γ2

0.002 (0.017)∗ 0.017 (0.334)

0.004 (0.427) 0.001 (0.248) 0.033 (0.059)∗∗

0.006 (0.211) 0.004 (0.004)∗ 0.055 (0.034)∗

0.018 (0.017)∗

Regressions with quantitative variables

γ1

p values in parentheses. Coefﬁcients with grey cell are statistically different from zero, ∗ at 5% and ∗∗ at 10%. Chow test on 27 September 2000. Breuch-Godfrey Serial correlation LM test. C. Cold, A. Attackers and V. Victims.

Attacks

0.008 (0.098)∗∗

γ0

V.5 killed

Table 8.10—Cont’d

0.812 (0.561) 1.298 (0.255)

1.219 (0.294)

Chow3

0.005

0.007

0.005

R2

1.587 (0.205) 0.912 (0.402)

0.372 (0.689)

LM4

Financial markets and terrorism

143

and the value of the dollar and the basket of currencies increasing by 4.1 per cent and 3.6 per cent respectively. None of the year interaction dummies is signiﬁcantly different from zero, indicating that the news content of an additional terror attack does not diminish over time. Our results indicate that one cannot talk about a routine of terror attacks. The informational content of a terror attack is a factor for each and every attack and it does not diminish over time. The ﬁnancial markets continued to incorporate efﬁciently the news of terror attacks as the attacks occurred.

7.

Conclusions

This paper has analysed the impact of Palestinian terror attacks on stock market prices and exchange rates in Israel using daily time-series data from 1990 until 2003. Intensiﬁed Palestinian terror attacks after 27 September 2000 had a permanent negative effect on the stock market but not on the foreign currency market. The stock market decline indicates that, beyond the loss of life and personal injuries to vicitms, the terror attacks had real economic costs that reduced ﬁrms’ expected proﬁts. Given the position of the balance of payments, the terror did not affect the value of the shekel. We have proposed a methodology for decomposing the innovations in the market into the news from a terror attack and other white noise. The results indicate that suicide attacks have a permanent effect on both markets; other types of attack do not. Terror attacks on transport have a transitory effect on the stock market and terror attacks on other targets do not. The target type does not affect the currency market. Terror attacks that took place within Israel’s 1948–1967 armistice lines (the green line) have a transitory effect on both markets, while other attacks do not. Attacks on major cities have no special effect. The number of victims injured and killed has a permanent effect on both markets. The number of attacks per day has a transitory effect on both markets. We ﬁnd that markets are efﬁcient in incorporating the news of terror attacks and ﬁnd no evidence that markets became desensitized to terror attacks over time. Our principal conclusion is that ﬁnancial markets continued to perform their economic functions in an efﬁcient way. Economic policies were also not affected; in particular, given the results about market efﬁciency, we can conclude that the process of market liberalization contributed to coping with terror.

Acknowledgements We thank the Caesarea Edmond Benjamin de Rothschild Center for Capital Markets & Risk Management at the Interdisciplinary Center Herzliya (IDC) for ﬁnancial support. We also thank Amir Barnea and Oded Sarig and Boaz Ganor– the director of the International Policy Institute for Counter-Terrorism at IDC. We thank Eran Gal for his excellent research assistance in the econometric analysis and Liat Goldstein, Sharon Granot, Don Radlauer and Baruch Shadmi for assistance in the preparation of the data for econometric analysis.

144 Raﬁ Eldor and Raﬁ Melnick

Appendix 8.1:

Exploring the source of the structural change in the fundamental stock market equation

Dependent variable Dlog (TA100) (included observations: 3515 after adjusting endpoints) Variable

Coefﬁcient

Std. error

t-Statistic

Prob.

C 27 September Dlog (S&P500(-1)) 27 September* Dlog (S&P500(-1)) Dlog (TA100(-1)) 27 September* Dlog (TA100(−1)) R-squared Adjusted R-squared SE of regression Sum squared resid Log likelihood Durbin-Watson stat

0.049476 −0.060935 0.269688 −0.151368

0.025299 0.055743 0.025587 0.042993

1.955664 −1.093139 10.54024 −3.520718

0.0506 0.2744 0.0000 0.0004

0.084865 −0.049713

0.018108 0.046006

4.686639 −1.080579

0.0000 0.2800

0.044628 0.043267 1.33212 6,226.87 −5,992.568 1.987415

Mean dependent var SD dependent var Akaike info criterion Schwarz criterion F-statistic Prob (F-statistic)

0.057432 1.361908 3.413125 3.423648 32.78303 0.0000

Endnotes 1 See Abadie and Gardeazabal (2003), who found that terrorism reduced per capita GDP in the Basque Country by some 10 per cent relative to the synthetic control region and that spikes in terrorist activity widen this gap. They also studied the effect of a 1998 truce on stock prices of ﬁrms with a signiﬁcant part of their business in the Basque Country and ﬁnd a positive relative performance as the truce became credible. Chen and Siems (2004) have studied the link between ﬁnancial markets and 9/11. On terror in Israel, see Eckstein and Tsiddon (2004), who attribute a total 5 per cent reduction in GDP over the years 2000–2003 to Palestinian terror, and Fleischer and Buccola (2002), who show how terror adversely affected tourism. Other effects of terror have also been investigated. See for example Trajtenberg (2003) who analyses the nature of the terrorist threat following 9/11 and derives implications for US defence R&D policy. 2 See Bernholz (2004). 3 The green line was the armistice line from the creation of the state in 1948 until the Six Day War in June 1967. 4 See deﬁnitions in Table 1. 5 Deﬁned as the number of victims people or injured per attack by category. 6 This category includes the attack on the Dolphinarium at the seashore of Tel Aviv where young women waiting to enter a nightclub were targeted and the attack on the Park Hotel in Natanya where people gathering to celebrate the Passover holiday were targeted. 7 Brenner et al. (2001) show that, at that time, the use of currency options become popular in Israel.

Financial markets and terrorism

145

8 The central bank continues to be committed to the exchange rate band but given that the band is widening, over time, the system is converging to a ﬂoating regime. 9 For each terror characteristic we deﬁne a dummy variable that equals 1 if the attack belongs to the characteristic and 0 otherwise. 10 The fundamental equations include the lagged dependent variable to account for serial correlation due to the use of index data in the stock market equation and possibly lagged adjustments in the foreign exchange market. 11 See business cycles dates obtained by the ‘Melnick’ index (Melnick, 2002). 12 See annual report of the Bank of Israel for 2001. 13 The constant of the equation is corrected for the lower inﬂation rate in this period. 14 See the appendix. 15 Low current account deﬁcit, large amount of foreign exchange reserves and low and diminishing net foreign debt. 16 The average interest rate spread for the dollar and the basket of currencies before 27 September was 7.2 and 6.8 per cent, and after 27 September, 4.4 and 4.3 per cent respectively. 17 The terror attacks are evenly distributed among the days of the week. 18 The Jewish holidays since they do not appear at a constant date; in the estimation we ignore this problem. 19 From the relevant equations, that with a suicide attack is an exception. This is probably due to the small number of suicide attacks before 27 September. 20 Except for the Haifa location, ignored due to the small number of cases, and the cold weapon dummy that is signiﬁcant but with the wrong sign. 21 Only for the basket exchange rate. 22 Only for the basket exchange rate.

References Abadie, A. and Gardeazabal, J. (2003) ‘The Economic Costs of Conﬂict: A Case-Control Study for the Basque Country’, American Economic Review, 94: 113–132. Bernholz, P. (2004) ‘Supreme values as the basis for terror’, European Journal of Political Economy, 20. Brenner, M., Eldor, R. and Hauser, S. (2001) ‘The price of options Illiquidity’, The Journal of Finance, 56: 789–805. Chen, A.H. and Siems, T.F. (2004) ‘The effects of terrorism on global capital markets’, European Journal of Political Economy, 20. Eckstein, Z. and Tsiddon, D. (2004) ‘Macroeconomic consequences of terror: theory and the case of Israel’, Discussion Paper No. 3–2004, The Pinhas Sapir Center for Development, Tel-Aviv University. Fleischer, A. and Buccola, S. (2002) ‘War, terror, and the tourism market in Israel’, Applied Economics, 34: 1335–43. Melnick, R. (2002) ‘Business Cycles in Israel’, The Economic Quarterly, 49: 219–44 (Hebrew). Orgler, Y.E. (2002) ‘The Tel-Aviv Stock Exchange (TASE): Functioning in a turbulent Period’, Journal of Global Financial Markets, (winter), pp. 44–56. Trajtenberg, M. (2003) ‘Defense R&D Policy in the anti-terrorist era’, NBER Working Paper 9725, Cambridge MA: National Bureau of Economic Research.

9

Global terrorism and the insurance industry: new challenges and policy responses1 Michael Wolgast

1.

Introduction

The events of 11 September have hit the insurance industry world-wide in more than one way. For one thing, they provided the necessity to cope with one of the largest damages caused by any single event in the history of insurance. At the same time, ﬁnancial markets were reacting, providing another additional challenge for insurance undertakings both in their role as investors and in their role as listed companies relying on favourable access to the markets. Short-term adjustment following 11 September has by now been largely accomplished: the issue of terrorism is now taken account of in many insurance contracts, premiums had to be re-calculated, and new capacity has been created in global reinsurance at a breath-taking pace. However, the implications of global terrorism for the insurance industry go far beyond this short-term adjustment: up to now, there is hardly any consensus whether or to what extent the protection against ‘terrorism’ can be provided by insurance undertakings just in the same way as the insurance against damage from say trafﬁc accidents or even natural disasters. Indeed, this question touches both on the theoretical foundations regarding the question of insurability, one of the cornerstones of insurance economics, and on the deﬁnition of the nature of ‘terrorism’. Even if new contracts offered in the market have been able to again (partly) include the risk of terrorism even after the 11 September events, this issue is far from being resolved. The most obvious corollary of discussing the insurability of the risk of terrorist acts is an assessment of the (potential) role of the state in this ﬁeld. The existence of insurance provides positive spill-overs for economic activity and especially entrepreneurial risk-taking which is crucial for innovation, structural change, growth and employment. Therefore, if the markets cannot (fully) provide coverage against terrorism, other institutional set-ups possibly including the state must be discussed in order not to fall back in terms of economic efﬁciency. Model solutions had been established earlier e.g. in Britain and France, and are currently being set up in the US and in Germany. The aim of this paper is to give a broad overview over these issues.

Global terrorism and the insurance industry

2.

147

The impact of 11 September on the markets for insurance

A ﬁrst obvious implication of the attacks of 11 September for the insurance industry was that since a large part of the economic damage was covered by insurance contracts, the events caused major insurance industry losses. Even if even now it is still impossible to be speciﬁc about the precise volume of claims in the context of 11 September, due to administrative complexities and pending legal issues,2 earlier estimates of the potential losses ranged from USD 32 to 56 billion.3 Other estimates predicted a volume of insured losses of roughly USD 40 billion.4 11 September is thus at the same time by far the worst terrorist act which occurred in recent years in terms of fatalities and insured property losses (see Table 9.1) and also – including natural catastrophes – probably the most costly single event in the history of insurance, even if in terms of the insured loss in non-life business, according to Swiss Re, Hurricane Andrew in 1992 might still have hit a higher number (see Table 9.2).5 Before 2001, the largest man-made property insurance loss – as opposed to losses triggered by natural catastrophes – was the Piper Alpha oil platform explosion in 1989 with a loss amount of USD 3 billion in 2001 values, which in terms of the loss caused by the World Trade Center event was more than 10 times less. Most of the losses caused by 11 September are linked to property insurance and insurance against business interruption, however, liability and to a lesser extent life insurance and workers’ compensation are also affected. Due to the global nature of the insurance markets, the losses were not conﬁned to the insurance industry in the United States: insurance undertakings in many countries were concerned, and in the reinsurance market, which is to a high degree globally integrated, European companies (Lloyd’s, Munich Re, Swiss Re) were among those which suffered the highest losses.6 For an economic assessment of these losses, it is important to realize that claims of this magnitude cannot easily be absorbed by the private insurance sector. According to calculations from Tillinghast-Towers Perrin for single lines of business, the 11 September losses vary between 9–15 per cent (workers’ compensation), 66–84 per cent (property and business interruption), 8–10 per cent (life, accidental death and disability) or 14–55 per cent (liability) of the industry’s normal annual losses.7 In aviation insurance, the estimated losses even amount to 400 to 500 per cent. Intuitively, it should therefore be clear that 11 September came as a severe blow for the underlying calculations in the insurance business and had a strong immediate impact on the industry’s proﬁts and capital reserves, even if the insurance sector was still able to absorb this one single terrorist event without major disruptions. On the other hand, a multitude of events of this scale would deﬁnitely put the stability and soundness of major insurance companies and of the insurance industry world-wide at risk. An important short-term implication of the volume of losses was that since the (prospective) compensation of claims signiﬁcantly reduced the capital available for writing business in the insurance markets (the ‘capacity’ of the market), new coverage was either hardly available or only available at extremely high prices in the reinsurance markets in an immediate reaction to the events of 11 September.

USA

Lebanon

11 Sep 01

23 Oct 83

USA

Sri Lanka

USA

UK UK UK

24 Jul 01

26 Feb 93

10 Apr 92 09 Feb 96 24 Apr 93

Event

91 100 54

1000

15

0 0 1500 147

0

467

1100 0 5075

100

2250

2874 271

1311

757

169

90000

Fatalities Injured Estimated economic loss USD m (2001 price levels)

NYC, Wash. DC Terror attacks against WTC and Pentagon 3122 by hijacked airliners Beirut Bombing of US Marine barracks and 300 French paratrooper base Bombay Series of 13 bomb attacks 300 Lockerbie PanAm Boeing 747 crashes due to bomb 270 Nairobi Two simultaneous bomb attacks trigger 253 damage in and around US embassy complex Oklahoma City Truck bomb attack on government building 166 in Oklahoma City Indian Ocean Hijacked Ethiopian Airlines Boeing 767-260 127 ditched at sea Moscow Bomb explosion destroys apartment building 118 Addis Ababa Arson in arms warehouse 100 Colombo Bomb attack on Ceylinco House 100 Buenos Aires Bomb attack on building of Jewish 95 organisation Colombo Intl. Rebels destroy 3 airliners, 8 military aircraft 20 Airport and heavily damage 3 civilian aircraft New York Bomb explodes in garage of world Trade 6 Center London Bomb explodes in ﬁnancial district 3 London IRA bomb attack in South Key Docklands 2 London Bomb explodes near NatWest tower (City) 1

Location

*Including business interruption and aviation hull losses. Source: Swiss Re / W.Schaad, (2002, p. 3)/Sigma catastrophe database.

Russia Ethiopia Sri Lanka Argentina

13 Sep 99 04 Jun 91 31 Jan 96 18 Jul 94

23 Nov 96 Comoros

19 Apr 95

12 Mar 93 India 21 Dec 88 UK 07 Aug 98 Kenya

Country

Date

Table 9.1 The worst terrorist acts in terms of casualties and insured property losses

671 259 907

725

398

6

45

145

6 138

19000

Insured property loss USD m (2001 price levels)*

38 3000 60 51 95 80 61 22 64 26 600 33 167 6425 45

20185 19000 16720 7338 6221 6164 5990 4674 4323 4293 3833 3150 2994 2872 2551

23.08.1992 11.09.2001 17.01.1994 27.09.1991 25.01.1990 25.12.1999 15.09.1989 15.10.1987 25.02.1990 22.09.1999 20.09.1998 05.06.2001 06.07.1988 17.01.1995 27.12.1999

Date

Source: Swiss Re (2002, p. 23).

1 Excluding liability losses. 2 Dead and missing. 3 Premiums in 2000, calculated at 2001 prices.

Victims2

Insured loss1 (in USD m, indexed to 2001) Hurricane Andrew Terrorist attack on WTC, Pentagon etc. Northridge earthquake Typhoon Mireille Winterstorm Daria Winterstorm Lothar over Western Europe Hurricane Hugo Storm and ﬂoods in Europe Winterstorm Vivian Typhoon Bart hits south of country Hurricane George Tropical storm Allison; rain ﬂoods Explosion on platform Piper Alpha Great-Hanshin earthquake in Kobe Winterstorm Martin over South-West France and Spain

Event

Table 9.2 The most costly insurance losses 1970–2001

US, Bahamas US US Japan France, UK et al. France, CH et al. Puerto Rico, US et al. France, UK et al. Western/Central Europe Japan US, Caribbean US UK Japan France, E, CH

Country

_

0.7% 5.5% 1.9%

_

3.9%

_ _ _ _ _

5.0% 4.1% 4.0% 6.5%

Loss as % of Non-lifepremium volume3

150 Michael Wolgast Of course, one can argue that the price mechanism should lead to an efﬁcient reallocation of capital, and that, as a consequence, this situation could be regarded as a temporary, exceptional phenomenon only. In the short run, however, since new capital is never instantaneously available, the events of 11 September provoked a substantial ‘hardening’ of the insurance markets – or, more precisely, the markets for reinsurance – world-wide. Beside the huge volume of losses and the short-term near collapse of the global reinsurance markets, a second strong impact of 11 September on the global insurance industry was due to the reaction of the ﬁnancial markets. The dramatic almost immediate world-wide decline in stock market levels provided an additional challenge for insurance undertakings both in their role as investors (in particular in life and – to a lesser extent – also in non-life business) and in their role as listed companies relying on favourable access to the markets (especially in the situation caused by the events of 11 September). Moreover, 11 September hit the ﬁnancial markets in a period where they were already going through a critical development, as the prices for many assets, in particular shares and above all technology shares, had severely deteriorated even before 11 September. In this environment, the dramatic further decline of security prices triggered by 11 September, in many countries, led to a reconsideration of those supervisory, ﬁscal and accounting rules for insurers which might have brought about the threat of a massive selling of their shares by insurance companies – the largest institutional investor with pension funds in OECD countries – in order to fulﬁl regulatory requirements. Indeed, in the aftermath of 11 September, insurance companies in their role as investors were confronted with a market situation which would have forced them to sell large parts of their portfolio under current regulatory provisions. To avoid major disruptions, regulators adopted a wide spectrum of temporary measures. Forbearance measures ranged from an easing of solvency/funding requirements, changes in accounting standards,8 changes in tax regulations, to a relaxation of investment rules.9 Not least as a consequence of these short-term policy measures, by the end of 2001, the markets had recovered, and both the major indices and the value of most single insurance stocks were back at their pre-11 September levels (see Table 9.3). Also, with respect to the destruction of capacity and the subsequent hardening of the global markets for reinsurance, the price mechanism indeed led to an adjustment process towards a new economic equilibrium, and, moreover, Table 9.3 Development of major indexes on the stock market in 2001 Date

DAX

DAX insurance

STOXX

STOXX insurance

Dow Jones industrial average (US)

Dow Jones insurance (US)

02.01.2001 10.09.2001 21.09.2001 28.12.2001

6290 4670 3787 5160

5016 3633 2955 3797

4702 3441 2878 3806

459 321 248 322

10646 9606 8236 10137

408 347 306 365

Source: Handelsblatt, Bigchart.

Global terrorism and the insurance industry

151

Table 9.4 Global insurance markets/inﬂow of new capital (USD million) (as of 12 April 2002) CGNU Swiss Re Axis (Marsh) Ace Ltd. AIG Montpelier Converium Allied World XL Capital Endurance

1,700 D 1,600 V 1,600 S 1,150 S 1,000 CD 1,000 PE 985 S 959 S 819 S 800 V

Arch Capital XL Capital Chubb St. Paul Wellington All others Total completed Pending GRAND TOTAL

763 S 600 D 600 D 575 TP 564 S 7,281 24,584 8,889 33,473

Type of issuance: CD = Convertible Debt; D = Debt; PE = Private Equity; S = Stock; TP = Trust Preferred; V = Various. Source: Insurance Information Institute/Morgan Stanley. See also supplementary sheet for proposed depiction.

adjustment occurred at a breath-taking pace. The scarcity of capital and the premium increase in reinsurance, on account of the expected related yields, resulted in a substantial inﬂow of new capital into that market (see Table 9.4). In total, new capacity in the order of some USD 30 billion had been or was about to be established, mainly through capital increases (new stock), the placement of debt paper or the creation of entirely new reinsurance companies.10 Whether this will be sufﬁcient in order to fully balance the initial effects of the events of 11 September remains to be seen.11 In any case, the almost immediate, substantial inﬂow of new capital into the market in reaction to a situation of severe shortage of supply and high prices can be regarded as a model of the extremely efﬁcient functioning of the price mechanism in a global market which has not yet been subject to major regulation. In the meantime, short-term adjustment following 11 September has thus to a large extent been accomplished. However, despite an efﬁcient settlement of claims not least due to the ability of the insurance industry to absorb this one huge single loss without major disruptions and in spite of the comparatively easy recovery both of the stock markets and the markets for reinsurance following the immediate impact of September 11, the implications of global terrorism for the insurance industry go far beyond this short-term management of the event. Rather, the devastating attacks of 11 September have led to a fundamental reassessment of existing assumptions and scenarios. Historically, many insurance contracts covered damages regardless of their cause, with the exception of damage caused by war, civil war or civil commotion. Since terrorism in most countries12 was not part of the war exclusion clause, damages resulting from a terrorist attack were covered.13 By contrast, the staggering, entirely unexpected loss amounts of 11 September and the new political dimension14 have led insurers and reinsurers world-wide to reconsider the question whether the risks associated with national or global terrorism can at all be insured by means of standard industry techniques. Consequently, in a ﬁrst, immediate reaction, insurers and reinsurers world-wide

152 Michael Wolgast either cancelled or abstained from renewing existing contracts upon expiration, especially in the ﬁeld of aviation insurance and industrial property insurance,15 or imposed new terrorism exclusion clauses,16 leaving parts of the risk uninsured.17 Even if in the meantime new contracts offered in the market have again been able to (partly) include the risk of terrorism even after the 11 September events, there is hardly any consensus whether or to what extent the protection against ‘terrorism’ can be provided by insurance undertakings just in the same way as the insurance against damage from say trafﬁc accidents or even natural disasters. Obviously, one important consequence of discussing the insurability of terrorism risk is an assessment of the (potential) need for public-sector intervention in this ﬁeld. These questions will be discussed in the remainder of this paper.

3. 3.1.

Insurability of the risks of terrorism The limits of insurability

The basic principle of insurance consists in an institutionalized pooling or sharing of risks which from the perspective of the insured allows for a partial or complete elimination of these risks against a certain payment towards the community of the insured. (In this view, the insurance company acts as the organizer of such a collective arrangement.) Insurance therefore provides the possibility to share risks and thus to almost completely eliminate them from an individual perspective. From the standpoint of risk-averse consumers or entrepreneurs, insurance thus brings about a signiﬁcant increase in economic welfare (the reduction of risks leads to an improvement in terms of the utility function) or entrepreneurial activity (parts of the risks associated with business activities are taken over by the insurance system). More growth and employment are the consequence. The diversiﬁcation of risks which is the basic underlying principle of insurance takes place across the community of the insured, but also over time. Important extensions of the basic model include transaction costs, the degree of risk aversion and asymmetric information possibly leading to adverse selection or moral hazard problems.18 Beside these basic tools of Insurance Economics, the theoretical conditions under which the insurance principle can actually be applied in practice (the ‘insurability conditions’ or ‘insurability criteria’) have been the subject of an intense discussion in the economic literature on insurance.19 Conversely, it was established that beyond the limits of insurability, risks could not be insured, i.e. they could not be eliminated through a process of diversiﬁcation across individuals and over time.20 In an attempt to summarize the literature, the most important insurability criteria would include:21 1. 2. 3. 4. 5.

Randomness (of the loss occurrence) Deﬁnability of the (maximum possible) loss Assessibility of the probability and severity of losses Independence (of loss occurrences) Size of the (maximum possible) loss

Global terrorism and the insurance industry

153

The ﬁrst condition (Randomness) implies that the time at which the insured event occurs must be unpredictable, and the occurrence itself must be independent of the will of the insured. Those events which will occur with certainty cannot be insured, and it must be ruled out that the insured can manipulate the occurrence of the event. (However, for the latter condition, the borderline towards all forms of moral hazard is blurred.) The second and third condition (Deﬁnability and Assessibility) also belong to the actuarial-mathematical area. In advance, it should be possible beyond any reasonable doubts to determine the possible (probable) maximum loss (the PML) and to calculate the probabilities of certain loss events on the basis of statistical information from the past and additional knowledge of the technical determinants of the risk in question. A prerequisite for this is the existence of an unambiguous, precise deﬁnition of the losses to be covered by the insurance arrangement in the case of the event. The fourth and ﬁfth conditions (Independence and Size) refer to the economic side of insurance. For the practical implementation of the insurance principle, the correlation between single events should not be too high, and the size of the potential single losses must remain within certain borders, since, otherwise, it could become economically unfeasible to settle the claim or all claims in the case of the insured event. From this short discussion of the insurability criteria, it should have become clear that they cannot be viewed as clear-cut, precise deﬁnitions. Rather, they provide an important guideline for economic policy-makers and the management of insurance companies. Still, if any benchmark can be used in order to assess the insurability of the risk of terrorism, it is these criteria which have emerged from the literature. 3.2.

The nature of terrorism

The phenomenon of terrorism is difﬁcult to deﬁne in unambiguous terms. In any case, 11 September has brought to light a new dimension of global terrorism, with a staggering, previously inconceivable scale of threat scenarios and loss potentials. Political, economic and social developments have combined to bring terrorism into the focus of international attention. What was formerly predominantly a national concern has now become a global threat. Whereas in the past, terrorism was primarily a national phenomenon (e.g. in the UK, Spain, Israel), terrorism has assumed a new international dimension since 11 September. Coupled with these geopolitical factors is the fact that terrorist organisations now have greater access to extremely effective and lethal weapons. Also, the size, complexity and vulnerability of certain targets – such as densely overbuilt downtown areas, ﬁnancial centres or industrial or nuclear plants – and, in general, the growing concentration of wealth in economic centres around the world enable terrorists to trigger damage of astounding dimensions with relatively simple concentrated attacks. The unprecedented attacks of 11 September have illustrated that both the severity and frequency of loss exposure have become virtually immeasurable. The potential extent of fatal and disruptive effects is also due to the fact that terrorists will be

154 Michael Wolgast ready to sacriﬁce their own lives to maximize damage, disruption, horror and the number of fatalities. Terrorism risk bears some similarity with the so-called ‘new risks’ which have been taken over by the insurance industry in recent years22 or with natural catastrophe risks such as earthquakes, storms and ﬂoods. In both cases, enormous inherent loss potentials make diversiﬁcation difﬁcult to achieve; individual events can affect entire economies and many different insurance lines of business. And yet, there are also differences: unlike terrorist attacks, natural hazards or technical accidents – irrespective of their dimension – occur randomly and without intent, and their probabilities and consequences can be modelled with scientiﬁc data and methods. On the contrary, 11 September not only stands for a new dimension of the probable damage, but also for a very different origin of that damage. With the political motives behind this kind of ‘man-made disaster’, the factor ‘intentional conduct’ provides a striking difference between the risks of terrorism on the one hand and e.g. natural catastrophes on the other hand. In that respect, terrorism has to be associated with other political risks like war or civil unrest rather than with accidents or the forces of nature.23 In terms of the ‘insurability conditions’, it is hard to conceive full compliance of the risk of terrorism with all ﬁve insurability criteria. Of course, terrorist attacks meet with the criterion of Randomness (of the loss occurrence), since they cannot be foreseen and are hardly subject to the will or the inﬂuence of the insured (apart from the standard moral hazard problem). Also, the condition of Deﬁnability of the (maximum possible) loss might be fulﬁlled, even if this remains subject to some doubt. Terrorism exposure would have to be estimated to arrive at reasonable premiums. While such estimates might take into account location and types of risk, the attacks of 11 September – unprecedented and unthinkable before that date – demonstrate that it is hard to think of all possible atrocious scenarios, which is of course – on the other hand – a standard problem e.g. in the context of insurance against third-party liability. In any case, however, the other three conditions are hardly entirely fulﬁlled in the case of terrorism risks. For example, in the attempt of actuarial calculations of the risks of terrorism, insurers would be confronted with the problem that terrorism hardly complies with the criterion of the Assessibility of the probability and severity of losses, since historical data cannot be relied on, and technical or scientiﬁc approaches fail in the assessment of the likelihood of terrorist events. In general, occurrences linked to ‘intentional conduct’ can be calculated and are thus insurable provided that historical experience and the assumption of a continued social behaviour consistent with earlier time spans can be relied on. (One example of this is burglary/theft.) By way of contrast, it becomes extremely difﬁcult for an insurer to calculate the actions of Osama bin Laden or the Al-Qaida network. Moreover, it is the very nature of terrorist incidents that very often they should come as a complete surprise, implying that they would not follow historical patterns. Due to the absence of any statistically reliable loss history and the inability to project future losses, private insurance coverage of these risks runs into problems, at least according to theoretical considerations.24 In addition, the criteria of Independence (of loss occurrences) and Size of the

Global terrorism and the insurance industry

155

(maximum possible) loss – aiming at the economic capacity of the insurance industry – seem to be violated in the case of global terrorism. To the contrary, in the event of terrorist attacks, a series of occurrences, each affecting many lines of business (property/casualty, business interruption, life, liability, etc.), is highly probable, and the PML might go beyond reasonable calculations. In sum, from the standard textbook insurability criteria, the conclusion would be that in the case of terrorism, like in the case of war or other types of social disruption, the associated risks can hardly be (entirely) covered by private sector insurance,25 so that the limits of insurability might be reached. 3.3.

Insurance contracts offered in the markets

One important observation, which is often referred to as counter-evidence against the assertion that the risks of terrorism were not insurable, is that several weeks after the initial shock of 11 September, and in contradiction to ﬁrst statements, insurers began to re-include the risk of terrorism in their contracts or to offer new, separate insurance contracts covering the risk of terrorism.26 However, examining this phenomenon more closely, the observation is by no means in contradiction with the assessment that terrorism risks are basically uninsurable. Similar to other types of insurance which are difﬁcult to estimate, such as natural catastrophes or product liability, in the case of terrorism, ways to assume at least a certain proportion of the risk are subject to careful underwriting and exposure control. In practice, this means that upper limits are imposed on the maximum loss to be borne by the insurance industry,27 and that, moreover, short cancellation periods will allow for the necessity of being able to react almost instantaneously in case the risk should materialize. Also, the way the risks of terrorism are being dealt with still varies a lot between different lines of business or by countries or by whether the ﬁeld of primary insurance or reinsurance is being considered.28 At the end of the day, the forms of private insurance can therefore hardly be regarded as a full protection against the threats of terrorism. Moreover, from a more general perspective, as has been pointed out before, the theoretical concept of insurability must not be confounded with the actual availability of the corresponding insurance contracts in real-world markets.29

4. 4.1.

The case for public-sector intervention The social cost of uninsured risks

So far, it has been established that (full) protection against the threats of terrorism will not be available in the private insurance markets. In itself, however, this conclusion would not yet provide an argument for public-sector intervention since obviously the state cannot be charged with the provision of any good which was not available on the private markets. Still, in the case of the provision of insurance coverage against the risks of terrorism, there is a strong case for the state to intervene.30 The main reason for this assertion is that the

156 Michael Wolgast reduction of risks brought about by insurance is not in the ﬁrst line a consumption good. Rather, the economic importance of insurance lies in the reduction of risks in the process of production. Through a reduction or absorption of some of the risks associated with the process of production, insurance enables riskadverse entrepreneurs to take on new entrepreneurial risks to a much larger extent than would be the case in the absence of insurance.31 The existence of insurance thus provides positive spill-overs for economic activity and especially entrepreneurial risk-taking which is crucial for innovation, structural change, growth and employment. Since this line of reasoning also applies in the case of insurance against the risks of terrorism, intuitively, welfare gains should be feasible if the state would help to remove this risk and thus alleviate the burden on private enterprise. In this context, it should be brought to mind that in addition to the direct detrimental economic effects of 11 September, there are also lasting effects due to a severe impact on the risk perception of economic actors. The decline of the main consumer and investor conﬁdence indicators has rapidly accelerated in September 2001 (OECD 2002, p. 121). These psychological processes have negative real economic effects, since they tend to paralyse economic activities. Moreover, the persistent perception of threat induces actors to intensify their security and protection efforts in terms of money and time. The increasing level of security measures has, of course, adverse effects on trade, particularly in industries that are characterized by just-in-time production regimes. According to OECD estimations small increases in transaction costs can have strong impacts: the elasticity of trade ﬂows with respect to transaction costs amounts to between 2 and 3 per cent (OECD 2002, p. 71). In this context, the chief economist of the International Monetary Fund, Kenneth Rogoff, has recently compared the implications of 11 September on trade in goods and on the migration of the labour force to the effects of a ‘terror tax’ (Rogoff, 2002). In addition to the effects of extended security measures on growth and trade, without the sufﬁcient availability of insurance coverage further negative effects on economic activity can be expected. One striking example of the economic consequences of a lack of insurance against the risks of terrorism was the airline industry. After the termination of the liability cover applicable to date shortly after the terrorist attacks, it became clear that entire airﬂeets would have had to stay on the ground due to the lack of insurance coverage. As a response, the governments in most countries offered the airlines temporarily – at ﬁrst restricted to one month, later renewed several times – state third-party liability (TPL) coverage for loss or damage caused by war or terrorism in order to ensure the functioning of national and international air trafﬁc. Moreover, it is not hard to imagine that without sufﬁcient insurance coverage being available, the amount and volume of risks investors are ready to carry will decrease. At the core of the problem is the dilemma for investors either to allow their risk exposure to increase or to abstain from certain economic activities. This holds not only for potential equity capital holders but also for potential debtors. In order to protect the lender, debt ﬁnancing usually implies sufﬁcient insurance of the project. Since a reduction in the protection of assets by

Global terrorism and the insurance industry

157

means of insurance has negative effects on the credit status of their owners, many economic activities will be abstained from as a consequence. Exact knowledge on the extent of these negative effects of missing insurance or underinsurance on GDP is hard to get. Still, there is some anecdotal evidence on the situation in the aftermath of 11 September 2001. At least the US creditors seem to consider insurance against the risks of terrorism to be indispensable for giving a loan. The US bond market association estimates that about 10 per cent of the commercial mortgage-backed securities market has been suspended or cancelled due to lacking insurance coverage. Missing availability of insurance coverage has negative impacts on both the credit sector and the industries relying on debt ﬁnancing. Due to hardening of the mortgage-backed security markets, e.g. the output of the US commercial construction sector has shrunk by 20 per cent during the ﬁrst quarter of 2002. Industrial projects located in central business districts of major metropolitan areas are jeopardized not only by missing property insurance but also by missing liability coverage. Hillman (2000, pp. 9–14) reports several examples of investors’ unavailing attempts to satisfy their lenders’ requirements for insurance coverage.32 The case for public-sector intervention can also be stated in terms of the theory of public goods. Undisputedly, even in a market economy, it is the ‘classical’ role of the state to ensure that political stability and the rule of law provide an adequate framework for private enterprise and an efﬁcient allocation of resources through private markets. ‘External stability’ and ‘internal stability’, i.e. the absence of any form of violence or coercion and full respect of private property rights are indeed among the cornerstones of any market economy. Of course, if these conditions are violated e.g. by criminal activities, the state – producer of these ‘public goods’ – cannot be made liable for any damage or loss which is caused by the state’s failure to fully ensure that these basic principles of a market economy are completely adhered to in practice. However, in the strive to minimize deviations from the principles of full external and internal stability, removing the risks associated with global terrorism through public-sector intervention in order to help set up adequate insurance schemes can no less be regarded as the provision of a public good than the original provision of internal and external security which is undoubtedly the task of the state in a market economy.33 The claim that the economic consequences of adverse random occurrences linked to global terrorism should not be borne by the individual, but – at least to some extent – by the community of all citizens (i.e. the state) also corresponds to historical precedents. Very often, in the case of wars, civil unrest or large natural catastrophes, the state organized some redistribution schemes for the beneﬁt of the victims of these events. Politically, this form of implicit solidarity can also be regarded as an important constituent principle of a market economy.34 Since individuals and entrepreneurs can be conﬁdent that in the case of extreme events, they will be able to rely on some form of burden-sharing among the citizens, they can assume more risks and will not have to (inefﬁciently) reallocate resources for the sake of self-protection.35 Finally, it is hard to understand why the devastating effects of terrorism should be borne by a random sample of individuals whereas

158 Michael Wolgast the terrorists were striving to attack and possibly destroy Western society – their war-like ideological enemy – as a whole.36 In sum, there is overwhelming evidence that the claim that – if (full) protection against the threats of terrorism will not be available in the private insurance markets – protection against the risks associated with global terrorism should at least to some extent be organized by the state is in full accordance with the principles of a market economy. 4.2.

Model solutions

In theory, model solutions for public-sector intervention can take various forms, with subsidies ranging from a complete absorption of the risks of terrorism by the state to minor support for corresponding private-sector solutions, as e.g. credit guarantees or ﬁscal incentives. (However, the latter instruments might turn out to be insufﬁcient remedies in view of the underlying strong economic reasons for the lack of (full) private insurance coverage of terrorism risks.) Of course, the economic disadvantages of leaving the risks associated with global terrorism without adequate insurance coverage would vanish by deﬁnition if the state promised to settle all claims connected to any damage or loss caused by terrorist attacks. However, while at ﬁrst sight this would provide a straightforward solution of the problem, several reservations apply: (1) the management and settlement of claims through public-sector administrative bodies might be inefﬁcient, due to the lack of professional experience in the ﬁeld of insurance. This argument would call for some form of public-private partnership, at least in the realm of practical management of the government guarantees. (2) It might be argued that a full and cost-free absorption of the risks of terrorist attacks by the state could lead to a distortion of incentives and hence imply inefﬁcient economic outcomes. However, in the case of externalities of self-protection market solutions would not be efﬁcient, either.37 Moreover, even if no externalities prevail, self-protection can be more costly and absorb more economic resources or imply more welfare losses than compensation of the victims of terrorism by the state. Also, from a more general perspective, one should bear in mind that the reference scenario can hardly be a situation in which there was terrorism, but no state guarantees, even if in this case the economic distortions would be obvious. Rather, the reference scenario must be a situation in which there was no terrorism at all. Both arguments would call for some form of burden-sharing between private insurance (fed through premium income from the insured) and the state (ﬁnanced by tax revenues). An obvious way of organizing such a model would be a solution in which terrorist damages beyond a certain threshold would (partially) be borne by the state, whereas smaller losses would entirely be covered by private-sector insurance. (3) A third warning applies to the potential scale of the terrorist threat. If losses become too high, it is conceivable that at some point they cannot be ﬁnanced by the state (i.e. the taxpayer) anymore, either. While this argument implies that some limitations must be in place for the government guarantees, it also provides a strong backing for the assertion that the risks of terrorism cannot be

Global terrorism and the insurance industry

159

borne by private-sector insurance, since the (potential) ﬁnancial resources of the state go far beyond the ﬁnancial means of any private-sector insurance scheme.38 In addition to these three qualiﬁcations, it has been discussed whether the insurance against the risks of terrorism must be made mandatory and whether the state could restrict public intervention to a certain time period (both arguments only apply in the case of positive premiums to be paid by the insured). The ﬁrst qualiﬁcation (mandatory insurance) leads over to tax-ﬁnanced reimbursement schemes organized by the state.39 Indeed, the borders between these two extremes are blurred, and it is often hard to tell whether mandatory insurance is not in fact a hidden tax (depending also on the way premiums are being calculated).40 The second qualiﬁcation (state guarantees for a limited time period only) refers to the idea that over time, due to the accumulation of premium income, reserves in private-sector terrorism insurance can be built up, so that the ﬁnancial capacity of the system is ﬁnally enhanced to the effect that government guarantees can be completely removed. However, this idea rests on the assumption that damage and losses due to terrorist events will not be too high, and that there is an upper limit on the size of potential claims, since, otherwise, state intervention could hardly ever be renounced. Model solutions along these theoretical lines are in place in many countries, and in all the various possible forms which have just been discussed.41 In Northern Ireland and in Israel, i.e. countries that are highly exposed to local terrorist activities, purely public schemes have been set up for the compensation of terror victims. These schemes are fully ﬁnanced by tax revenues. In other countries, private/public partnerships are given preference. Table 9.5 presents a synopsis of selected national schemes providing insurance against the risks of terrorism.42 Among other aspects, the schemes differ with respect to a)

whether they had been created earlier in reaction to former terrorist attacks (like the Pool Re in the UK or the CCS in Spain) or whether they have been newly founded in reaction to the 11 September attacks (Austria, France, United States); b) whether they are set up for a limited time period (like GAREAT in France or the ‘Terrorism Risk Insurance Act’ in the US) or whether their duration is conceived as unlimited (like in the case of Pool Re or the CCS); c) whether participation for the insurance companies or the insured is voluntary (like in Great Britain) or made mandatory (like in the United States or in France); d) whether they offer primary insurance as in the United States or reinsurance as in France and in the UK. The existing reinsurance schemes all rely more or less on pool models which partition the risks of terrorism between several companies and the national government.43 Every participating insurance company makes a certain capacity available up to which it carries liability in the case of loss. Among the participants risks and premiums are allocated according to a layer system.

France

•

•

•

Reinsurance pool specialized in the insurance of terrorism risks Obligatory membership for French insurance companies Obligatory cession of terror risks above an insurance sum of E6 million

GAREAT (Gestion de l’Assurance et de la Réassurance des Risques Attentats et Actes de Terrorisme)

Primary insurance pool Voluntary membership

Österreichischer Versicherungspool zur Deckung von Terror-Risiken

Austria

• •

Name and nature of national scheme

Country

Layer system 1st layer (E50 million): Austrian property insurers 2nd layer (E150 million): international reinsurers Negotiations between Austrian insurance industry and Austrian government concerning a third layer (E600 million) provided by the Austrian government. Renewed Layer system (implemented 1.1.2003) 1st layer (E400 million) retention by primary insurers (GAREAT) 2nd layer (E1.1 billion) by international reinsurers 3rd layer (E250 million) by international reinsurers 4th layer: above E1.75 billion unlimited state guarantee

1.1.2003 Set up for 1 year but with the option to renew

1.1.2002 Set up for 1 year but with the option to renew.

Structure of the scheme/role of the government

Date of start of business and limitation of the scheme

Table 9.5 Synopsis of international model solutions for the insurance of terror risks Restrictions on terror acts included

War, theft and GAREAT insures fraud as a terrorism risks in consequence of property and transport terrorism are insurance above excluded E6 million. These risks are excluded in standard Hacker attacks and nuclear attacks reinsurance contracts are excluded

Property insurance limited Acts of terrorism occurring in to a maximum loss of Austria E5 million per event per risk; above this limit: individual contracts possible

Losses covered

State-owned ‘Consorcio de Compensación de Seguros’ (CCS) Coverage of risks of terrorism linked to a base policy is provided by the public sector

United Pool Re Kingdom • State-organized reinsurance pool (mutual reinsurance company) • Voluntary participation for insurance companies

Spain

1993

1941 last modiﬁcation: 1.1.2002

Continued

Insurance of the risks of Terror acts having Provision of subsidiary coverage, terrorism in ﬁre and national and serves as a guarantor in case of property insurance and international insolvency of primary insurer; personal accident background coverage is ﬁnanced by a insurance surcharge on all policies in speciﬁc risk classes. Backed by state guarantee should the reserves prove inefﬁcient • Provision of property Acts of terrorism Renewed layer system (implemented 1.1.2003) occurring in and business 1st layer: industry wide retention England, Wales, interruption insurance and Scotland which will be increased from coverage. • Pool Re provides (not in Northern 60 Mio. GBP p.a. (2003) to Ireland) 200 Mio. GBP p.a. (2006) (to its members) the possibility to reinsure War and hacker 2nd layer: Pool Re’s current capacity attacks are the terrorist exposure 3rd layer: unlimited state guarantee excluded

Source: own compilation from selected sources.

Layer system Risks of terrorism in 1st layer: Insurers deductible. The commercial property and business annual volume depends on a interruption insurance comparison of an insurer’s and in workers covered losses in that year to its compensation direct earned premiums for lines insurance of business covered by the programme in the prior year. The deductible is progressive in time. 2nd layer: Above this deductible, the US government pays 90 per cent of each insurer’s property-casualty losses up to an annual cap of USD 100 billion, the other 10 per cent must be raised by private sector insurance

Terrorism Risk Insurance Act 25.11.2002 Inclusion of terrorism risks is Programme ends 2005 made obligatory for insurance companies operating in the US, government guarantees beyond a certain threshold

USA

Losses covered

Structure of the scheme/role of the government

Name and nature of national Date of start of scheme business and limitation of the scheme

Country

Table 9.5—Cont’d

•

•

Coverage of acts of terrorism having an international background No coverage of domestic terrorism

Restrictions on terror acts included

Global terrorism and the insurance industry 4.3.

163

Foundation of the ‘Extremus AG’ in Germany to cover terrorism risks

In Germany, considerations on involving the state in the insurance of terrorism risks in April 2002 led to an agreement between the Federal Government and the insurance industry on the basic conditions of a private/public partnership solution to cover terrorism risks.44 Among the most essential conditions stipulated by the partners in April 2002 were: •

• • •

• • •

On top of a capacity to be procured by the insurance industry amounting to EUR 3 bn, the Federal Government would provide a state guarantee to the extent of EUR 10 bn to cover the risk of terrorist attacks which had so far not been insurable in high-sum property business. The German government’s decision to limit its guarantee to EUR 10 bn contrasts with the unlimited guarantees given by the UK or the French government. According to estimations, however, the probable maximum loss of terror attacks in Germany would amount to some EUR 3 bn,45 so that the capacity made available seemed to be sufﬁcient. Contrary to the situation in some of the other countries, the participation is voluntary for insurance companies. The arrangement comprises damage to buildings and property as well as business interruption losses resulting from these. The insurance cover provided by this arrangement applies to large risks with an insured value of more than EUR 25 million. Below this threshold, the terrorist risk is borne by private insurers alone. The same is the case also in other lines of business. For instance, in non-life insurance there is no exclusion of terrorist risks for the entire private and the so-called commercial business, i.e. property risks beyond a certain threshold. In the TLP, transport and engineering insurance classes this applies to industrial risks as well. Third party liability insurance for airlines is excluded from the agreement. Maximum annual beneﬁts are limited to EUR 1.5 bn per policy-holder. There is a time limit for the state guarantee, which will for the time being be warranted until the end of 2005. This is to enable the state to consider its gradual withdrawal from its guarantee commitment. The solution refers only to losses occurring in Germany since, for comprehensible reasons, the German government was reluctant to take responsibility for terror attacks outside of Germany.

Based on these general conditions, in September 2002, 16 insurance companies founded the ‘Extremus AG’, a special insurer for terrorism risks. Ofﬁcial approval from the German supervisiory authority (the Bundesanstalt für Finanzdienstleistungsaufsicht) was available in October, and, in November 2002, Extremus could ﬁnally start its business operations. Extremus’ operates as a primary insurer, it carries direct liability for the claims of the policy-holders. Simultaneously, it provides reinsurance coverage in

164 Michael Wolgast certain cases. Prospective customers interested in full insurance coverage exceeding the amount of EUR 25 million are given the possibility to buy, in addition to the basic policy where the terrorism risks are excluded, a second one from Extremus. The new terror insurance comes relatively close to an all risk cover since it includes not only risks speciﬁcally mentioned (such as ﬁre, explosion, collision or crash of ﬂying objects etc.), but also a wide range of ‘other malicious damage’. Among the risks and losses which are nonetheless not insured (exclusions) are war and warlike events, but also attacks by B and C weapons. Extremus’ total capacity of EUR 13 billion is divided into three layers: the ﬁrst one amounting to EUR 1.5 billion is provided by German primary insurance companies, the second one, equally amounting to EUR 1.5 billion, is provided by international reinsurance companies,46 whereas the third one consists of the EUR 10 billion guarantee granted by the German government. Between the layers, premiums are subdivided according to the respective risks. For the German insurance industry, however, the premiums to be paid by German trade and industry to the new special insurer do not nearly constitute any improvement of earnings. Apart from payments to the state47 and to international reinsurance a provision for large risks, which has been set up before, will have to be serviced. The tax treatment of these provisions in connection with the foundation of ‘Extremus AG’ has at present not yet been deﬁnitively settled, even if a political consensus has been reached. Earlier, the German insurance industry had made clear to the Federal Government that without any recognition of these provisions for tax purposes the new special insurer would hardly be able to fulﬁl its role. Overall, the role of the government is rather limited within this framework. Nevertheless, due to governmental engagement in the foundation of Extremus a gap in the availability of insurance cover could be ﬁlled. Since the negative impacts of missing insurance coverage on economic development are well known, the importance of this intervention goes far beyond the insurance industry.

5.

Summary/conclusions

The devastating events of 11 September 2001 abruptly forced the insurance industry to address the urgent issue of global terrorism. Short-term adjustment following 11 September has by now been largely accomplished. However, despite an efﬁcient settlement of claims not least due to the ability of the insurance industry to absorb this one huge single loss without major disruptions and in spite of the comparatively easy recovery both of the stock markets and the markets for reinsurance following the immediate impact of 11 September, the implications of global terrorism for the insurance industry go far beyond this short-term management of the event. Rather, the attacks demonstrated that this type of threat has become virtually immeasurable in terms of both the severity and frequency of exposure – making it difﬁcult for the private insurance sector to adequately cover this risk. In terms of the ‘insurability conditions’ – a checklist for assessing the insurability of a certain risk known from the Insurance Economics literature – terrorism risk

Global terrorism and the insurance industry

165

hardly complies with most insurability criteria. The conclusion therefore is that in the case of terrorism, like in the case of war or other types of social disruption, the associated risks can hardly be (entirely) covered by private-sector insurance. While insurers and reinsurers are willing and able to provide limited cover for terrorism risk, a viable long-term approach towards coping effectively with this risk will thus necessarily rely on some form of public-sector involvement. In terms of economic welfare, leaving the risks associated with global terrorism without adequate insurance coverage would be the worst solution. The consequence would be a severe set-back in terms of economic efﬁciency and especially entrepreneurial risk-taking which is crucial for innovation, structural change, growth and employment. Therefore, since the markets cannot (fully) provide coverage against terrorism, other institutional set-ups including the state must be relied on. In order to generate sufﬁcient capacity in the markets for insurance against the risks of global terrorism, risks need to be spread among the insurance industry and the state. In conclusion, the risk-bearing community comprising the insured, insurers and governments must reach agreement on state involvement for covering the risks of terrorism in the medium and long term. Not least the surge in demand for terrorism coverage at a time where the insurance industry is coping with the aftermath of the largest insurance loss ever calls for solutions with an equitable private/public partnership. In that sense, governments must act as an ‘insurer of last resort’. At the end of the day, the availability and limits of the coverage of the risks of terrorism will depend on the willingness of governments to assume terrorism risk.

Endnotes 1 A preliminary version of this paper was presented at the DIW workshop ‘The Economic Consequences of Global Terrorism’, Berlin, 14–15 June 2002. The author wishes to thank Peter Walkenhorst (OECD) and other workshop participants for many good comments and suggestions. Also the author would like to thank Prof. Walter Karten (University of Hamburg) and Prof. Thomas Straubhaar (HWWA Hamburg Institute of International Economics) for valuable comments on an earlier version of this paper and Stefan Richter (Managing Director Property/Casuality, German Insurance Association) for many important and instructive discussions. Earlier versions of (parts of) this paper have also been published in different places (Wolgast, 2002c, 2002d, 2002e). 2 One prominent example in this context was the question whether the terrorist attacks on the two towers of the World Trade Center (WTC) constituted one single occurrence or two logically independent events. In the ﬁrst case, due to a clause in the property insurance contract imposing an upper limit of USD 3.5 billion in maximum compensation per event, the insured loss exactly amounts to this sum, in the second case, the amount would have to be doubled. Also, the implications of the Victims’ Compensation Fund set up by the government for private-sector insurance contracts are still under discussion. In the case of earlier events (Hurricane Andrew, Northridge Earthquake), it took also more than one year before the ﬁnal amount of losses actually emerged. 3 Source: Tillinghast-Towers Perrin. 4 Source: Insurance Information Institute (III), New York. Swiss Re’s best estimate for the overall insured loss was ‘some USD 38 billion’ (Schaad, 2002, p. 4). Also, Swiss Re came out with an overall economic loss estimate of USD 90 billion. However, it is even

166 Michael Wolgast

5

6

7 8

9

10

11

12 13

14 15

16

much more difﬁcult to quantify the overall economic loss due to the tragic events of 11 September than the insured loss. One reason for this is the high number of fatalities in a highly industrialized country. A second reason is the high potential of liability suits which constitutes by far the most important single source of uncertainty about the entire volume of losses caused by 11 September. According to statistics compiled by the III, the loss estimate for Lloyd’s amounted to USD 2,800 million, Munich Re USD 2,368 million, Swiss Re USD 1,300 million, compared to USD 2,280 million for Berkshire Hathaway or USD 1,740 million for Ace (updated to 31 December 2001). Guinn, 2002. For example, in Germany, on account of special accounting rules applicable only to them, the German insurance companies would have been forced to sell substantial share and investment holdings in an adverse stock market situation, in order to minimize ﬁscal disadvantages and take precautionary accounting measures. The German government has removed this pressure by adopting new accounting rules (through a modiﬁcation of article 341 b of the German Handelsgesetzbuch). For instance, in the United Kingdom, the FSA temporarily eased the ﬁnancial investment rules for life insurance companies, amending the ‘resilience test’, which tests the ability of a fund to withstand major falls in asset prices (such as equities and ﬁxed interest securities). One prominent example is the creation of a high number of new reinsurance companies in Bermuda, ﬁnanced mainly through private equity investment from the United States. In total, the six largest new companies founded in Bermuda have attracted some USD 7 billion in fresh capital (source: Fortune, 10 June 2002). One aspect in this context is the ‘ﬂight to quality’, i.e. the augmented demand for ﬁrstclass reinsurance cover due to a stronger differentiation between reinsurance companies in terms of their rating and ﬁnancial strength, which provides an additional factor in the market getting tighter. In the medium term, some observers even expect a new trend to over-capacities and a decline in reinsurance premiums (‘cyclicality’ of reinsurance markets). Special regulations or pool solutions with state support were in place to cover terrorism risk even before 11 September, but only for a few particularly exposed countries (e.g. UK, Spain, South Africa, Israel). We shall come back to these models later in the paper. Otherwise the statistics provided in Table 9.1 would not exist. However, the risk of terrorism was calculated on a different basis, and accordingly no speciﬁc provisions were allocated to meet these exposures. In the case of the World Trade Center insurers considered the possibility of an aeroplane crashing into one of the buildings – not to speak of both towers – as so highly improbable that it was not even taken into account in the calculation of the Probable Maximum Loss (PML). The outstanding character and novelty of the occurrence of 11 September therefore calls for a re-examination both of the probability of occurrence and of the maximum loss (the PML) and thus requires at least a recalculation of premiums needed to cover these risks (if at all possible). In the past, terrorism was primarily a national phenomenon (e.g. in the UK, Spain, Israel). Terrorism has assumed a new international dimension since 11 September. One important aspect in this context is the divergence between primary insurance and reinsurance. After 11 September, many insurers had to realise that part of their portfolio – and especially exposure to the risk of terrorism – was not covered by reinsurance anymore. According to Heck (2002), as from 1 January 2002, reinsurance arrangements between primary property/casualty insurance companies and reinsurance companies contain a large number of new restrictions on top of the exclusion clauses for terrorist acts. Kollar (2002) provides examples of terrorism exclusion clauses. Other examples for the direct or indirect exclusion of terrorism were under discussion in Germany for life or private

Global terrorism and the insurance industry

17

18 19 20

21 22

23

24 25

167

accident insurance or motor liability insurance (removal of the unlimited liability which is currently in place). In aviation insurance, the liability cover applicable to date to the amount of USD 1 to 1.5 billion per aeroplane was removed, taking advantage of the very short seven days’ cancellation period. As a response, the governments in most countries offered the airlines temporarily – at ﬁrst restricted to one month, later renewed several times – state third party liability (TPL) coverage for loss or damage caused by war or terrorism because otherwise the aeroplanes would not have been able to take off. Not least these developments renewed in the public at large the awareness of the economic functioning and importance of insurance. See e.g. McCall (1987, 1992). Standard textbooks include Borch (1989), Gollier (2001) or the handbook edited by Dionne (2000). Gollier (2002) also provides a concise summary of the main arguments. See e.g. Karten (1972, 1988, 1997), Berliner (1982, 1985, 1988) or Gollier (2000, 2002). It is important to note that this theoretical deﬁnition of ‘insurability’ will not necessarily coincide with the availability of insurance in real-world markets. Deviations are conceivable in both directions: (1) Even if a risk was insurable in theoretical terms, the institutions for implementing the corresponding insurance system might not yet have emerged in practice. (This would be equivalent to saying that proﬁt opportunities for insurance companies have not yet been identiﬁed and exploited.) (2) Even if a risk was not insurable in theoretical terms, insurance cover might be available in the markets. One example for this would be a situation in which some form of gambling is involved on the part of the insurance company. In that situation, since the system will not be able to settle all claims in the case of an extreme event, some residual risk will in fact be left with the insured, even if in legal – though not in economic – terms the insurance contract seems to fully cover the risk in question. For insurance managers, decisions concerning the insurability of certain risks are among the hardest strategic choices, especially since full information will hardly ever be available. The following enumeration mainly draws on the discussion of the insurability criteria provided by W. Karten (2000, pp. 128–135). During the past, the insurance industry has been ready to cover emerging new risks though these were often far larger and more complex than any risk covered before (e.g. nuclear reactors, space satellites, barrages, offshore oilrigs, huge oil tankers, etc.). The increase of so-called large risks can also be attributed to rising wealth and technical progress. As a result, ever-increasing values must be insured which, in addition, often appear in a massive concentration. This is necessarily accompanied by an increasing vulnerability to large claim occurrences. This observation also corresponds to the standard deﬁnitions of terrorism applied by the insurance industry. ‘Terrorism’ is always deﬁned in terms of the motives behind the causes of a certain loss or damage. E.g. Swiss Re (2002, p. 16) uses the following deﬁnition: ‘Terrorism means an act or threat of violence or an act harmful to human life, tangible or intangible property or infrastructure with the intention or effect of coercing any government or putting the public or any segment of the public in fear’. A similar deﬁnition is quoted by Kollar (2002, p. 22). However, in practice, it might turn out extremely difﬁcult, especially for smaller incidents, to determine whether a certain event can be linked to an act of terrorism, and costly litigation procedures cannot be ruled out. However, this is a standard problem in the case of the so-called ‘new risks’. Gollier (2002, pp. 16–26) argues that it would be superior from an efﬁciency standpoint if these risks were shared by private individuals world-wide. However, since (mandatory) world-wide insurance cover and unlimited access to the capital markets – two of the main elements in his model – are hardly feasible through private insurance (alone), his conclusion is that in view of the resulting insurability problem associated with terrorism

168 Michael Wolgast

26

27 28

29 30 31 32

33

34

35

coverage, the state must play a role, in order to avoid important adverse effects on welfare. The same conclusion – that terrorism cannot be covered by private insurance alone, especially after the extreme events of 11 September – is reached by Jaffee and Russell (2002) or Kunreuther (2002). A different view seems to be expressed – at least by ﬁrst sight – by Munich Re (see Munich Re, 2001) or by Swiss Re (see. Schaad, 2002, p. 7, Swiss Re, 2002, p. 20). Also, on the internet, Swiss Re maintains that ‘while the parameters have changed, terrorism is, nonetheless, an insurable risk . . .’ (Swiss Re, Research & Publications, Swiss Re top topics, Terrorism, Terrorism – a new dimension (available under www.swissre.ch)). However, it would be entirely wrong to take these statements as counter-evidence against the assertion that terrorism was not (fully) insurable. Rather, the statements from the industry rely on the assumption of limited coverage clauses and short cancellation periods which cannot be regarded as a full protection against the threats of terrorism. E.g. in the ﬁrst half of 2002, under the name of Special Risk Insurance and Reinsurance Luxembourg S.A. (SRIR), ﬁve major European insurers and one company from Bermuda founded a special insurer for terror risks. Similarly, Allianz and Berkshire Hathaway or AIG developed concepts for insuring aeroplanes against the risks of terrorism. A counter-example, however, is provided by the attempt of leading German industrial companies to create an own (captive) reinsurance company charged with covering the risks of terrorism, which after months of preparations ﬁnally collapsed in late May 2002. For example, in Germany, any damage caused by terrorist acts was not excluded from private insurance subject to an upper limit on property losses of no more than E25 million. For example, even the risk of war used to be included (in some cases) in the realm of marine insurance. Nevertheless, coverage against the risk of terrorism in that line of business seems to follow the patterns of property insurance rather than providing a straightforward inclusion of terrorism risk. See endnote 21. The following arguments were also brought forward in more detail in Wolgast (2001). This important role of insurance in the economy has recently been pointed out again by Sinn (1986, 1988). The negative effects of the lack of insurance coverage against terrorism in the United States were also described by Heck (2002). The OECD (2002, pp. 124–128) pointed out that to the extent that it increases uncertainty related to investment decisions, reduced insurance coverage may have a negative effect on growth. In general, the OECD warned against the possible negative medium term consequences of 11 September through various indirect channels even though the short term recovery from the terrorist attacks had been faster than expected (OECD 2002, p. 136). An additional aspect is the international competition between producers or the international competition between states or countries as a location for international investment. Clearly, if protection against the risks of terrorism is offered by the state in one country, this imposes a disadvantage on competitors from other countries where such protective schemes are not in place, and makes this country a more attractive location for investment. Beyond this rational approach, the absence of implicit solidarity would also bring about a severe set-back in psychological terms. Conﬁdence that the state will provide a certain basic framework of existence has always been regarded as an important prerequisite for optimism and economic activity. Self-protection against the risks of terrorism will take much more effort anyway compared to self-protection against many other risks, since terrorists are motivated by a high level of destructiveness and will resort to ever new means in achieving their objectives. A further argument why self-protection will be inefﬁcient compared with government

Global terrorism and the insurance industry

36 37 38

39

40

41 42

43

169

intervention in the market for terrorism insurance was brought forward by Lakdawalla and Zanjani (2002). Self-protection by one target encourages terrorists to substitute towards less fortiﬁed targets. Investments in self-protection thus have negative external effects in the presence of rational terrorists. Government subsidies for terror insurance can discourage self-protection and limit the inefﬁciencies associated with these externalities. In the literature, sometimes, the argument that government subsidies would reduce the quest for self-protection and therefore be equivalent to severe adverse incentives due to a moral hazard problem has been relied on in order to reject the case for public-sector intervention (Rees, 2001; Nell, 2001). The paper by Lakdawalla and Zanjani as well as the high costs and the low level of effectiveness of self-protection against terrorism imply that this simple line of reasoning can hardly be maintained anymore. Rather, similar to corresponding situations in the ﬁelds of ﬂood insurance and self-protection or anti-theft insurance and self-protection, strong externalities imply that the market outcome without state intervention will be inefﬁcient. Of course, this last statement implies some form of ethical value-judgement, since distributional issues in most cases cannot be discussed under the mere postulate of efﬁciency. See endnote 36. Arguing that the ﬁnancial potential of the public sector would go beyond the means of private-sector insurance is of course contingent upon the possibility of (massive) tax increases, if necessary. In any case discussing the appropriate answer of society to the threats of terrorism under the restrictions of short-term budgetary consolidation issues would seem totally inappropriate. Straubhaar (2003, p. 12–13) argues in favour of a model in which insurance cover against the risks of terrorism is made mandatory since in his view this would be sufﬁcient in order to overcome the uninsurability problem. His model leaves scope for two alternative interpretations. One alternative is that premiums would be ﬁxed as a certain percentage of other insurance premiums in the ﬁeld of property risks, and would not be calculated individually according to the actual likelihood and severity of losses caused by terrorist attacks. While in this model, the risks would be spread over a large community of insured, it is not quite clear why it should be superior to tax-ﬁnanced subsidies towards a private insurance solution in which premiums would be calculated according to individual exposure. In both cases, highly exposed risks would beneﬁt from some form of (hidden) subsidy ﬁnanced by all taxpayers (participants), and the potential for inefﬁcient outcomes due to a distortion of incentives would be identical under both models. In an alternative interpretation, premiums would be calculated according to individual risks. However, this solution would not necessarily be sufﬁcient in order to overcome the uninsurability problem, especially if no other form of government intervention is involved. In an extreme case, the government would even collect the ‘premiums’. If in addition these ‘premiums’ to be paid by the insured do not correspond to the individual risk, it becomes obvious that they should rather be considered as some form of additional tax payments. See. Jaffee and Russell (2002, p. 27–29) or Schaad (2002, p. 4). For a description of the Pool Re in the UK see also Tillinghast-Towers Perrin (2001). For a description of GAREAT in France see also Kessler and Lamère (2002a, 2002b). Another example is South Africa where private insurers withdrew from the market in 1976. This led to the establishment of a government backed pool of 15 of the largest insurance companies known as SASRIA (South African Strikes and Riots Insurance Association). The South African government is the Pool’s insurer of last resort. Pool models do not necessarily imply a governmental participation. Insurance pools are always considered when in the case of rare risks, which are hardly assessable and exposed to large claims and cumulative claim potentials, a distribution between a great

170 Michael Wolgast

44

45 46 47

number of insurers is appropriate for reasons of policies followed in respect of risks, and when this distribution cannot be left to free competition, but must be organized in a predetermined form. In these cases it seems to be inevitable that a pool restrains competition insofar that the form of the insurance cover and the premiums are settled in a contract. Otherwise the risks concerned would not be covered at all on account of the lack of coverage capacities. In Germany, for several types of risks pool models have been created. The German nuclear insurance pool (Deutsche Kernreaktor-Versicherungsgemeinschaft (DKVG)) is a community of insurers who offer, within certain limits of liability, TPL and property insurance cover for hazards in relation with the erection and the operation of nuclear reactors. The DKVG operates mainly as a reinsurance pool. A reinsurance exchange exists with foreign nuclear pools in particular in non-compulsory reinsurance. In TPL, the DKVG covers losses up to an amount of EUR 255.6 million. For losses above this sum insured, the operators of German nuclear power plants are mutually liable up to a limit of EUR 2.5 bn. In addition, the government is involved: in case of DKVG and the operators of German nuclear power plants becoming insolvent the German government carries liability up to a total amount of EUR 2.5 bn. Another insurance pool is the German aviation pool (Deutscher Luftpool) which provides reinsurance for the aviation and aerospace insurers. A third one is the pharmaceutical pool (Pharma-RückversicherungsGemeinschaft) which provides reinsurance coverage for pharmaceutical liability risks. Pool members are not only German but also foreign insurance companies having subsidiaries in Germany. Whereas the state is indirectly involved in the nuclear insurance pool to some extent, taking the role of insurer of last resort, both the aviation pool and the pharmaceutical pool operate without any formal state involvement. The German Insurance Association (GDV) had called for a solution involving publicsector intervention as early as in November 2001 (see Wolgast, 2002a, pp. 88–90, 2002b, pp. 78–80). Talks between the insurance sector and the government had also involved the association of German industry (the BDI). These estimations were based on the scenario of the Frankfurt bank quarter being attacked. Thus, no net retention is to remain with the new special insurer. In exchange for the state guarantee the state will claim its share in premium income, which could amount to more than 10 per cent of total premiums. At ﬁrst, this percentage is to be slightly lower in order not to put too great a burden on the new special insurer during its start-up and development period.

References Berliner, B. (1982), Limits of Insurability of Risks, Englewood-Cliffs. —— (1985), Large Risks and Limits of Insurability, in: The Geneva Papers on Risk and Insurance, 10 (no. 37, October 1985), pp. 313–329. —— (1988), Versicherbarkeit, in: Farny, D. et al. (eds.), Handwörterbuch der Versicherung, Karlsruhe, pp. 951–958. Borch, K. (1989), Economics of insurance, Amsterdam. Dionne, G. (ed.) (2000), Handbook of Insurance, Boston, Dordrecht, London. Gollier, Ch. (2000), Towards an economic theory of the limits of insurability, in: Assurances, January, pp. 453–474. —— (2001), The economics of Risk and Time, Cambridge. —— (2002), Insurability, mimeo, 18 January 2002, available under www.nber.org/ ∼confer/2002/insw02/insurprg.html

Global terrorism and the insurance industry

171

Guinn, P. (2002), Why Are Insurance Prices Going up?, in: Leaders Magazine, vol. 25, number 2 (April 2002), p. 150, also available under www.leadersmag.com Heck, Warren W. (2002), Lack of Terrorism Reinsurance Threatens Economic Recovery, Washington Legal Foundation, Legal Backgrounder vol. 17, No. 10, Washington DC, 22 February 2002. Hillman, R. J. (2002), Terrorism Insurance, Rising Uninsured Exposure to Attacks Heightens Potential Economic Vulnerabilities, Testimony before the Subcommittee on Oversight and Investigations, Committee on Financial Services, House of Representatives, General Accounting Ofﬁce, Washington DC, 27 February 2002. Jaffee, D. and Russell, Th. (2002), Extreme Events and the Market for Terrorist Insurance, mimeo, 11 January 2002. Karten, W. (1972), Zum Problem der Versicherbarkeit und zur Risikopolitik des Versicherungsunternehmens – betriebswirtschaftliche Aspekte, in: Zeitschrift für die gesamte Versicherungswissenschaft, vol. 61, pp. 279–299. —— (1988), Existenzrisiken der Gesellschaft – Herausforderung für die Assekuranz, in: Zeitschrift für die gesamte Versicherungswissenschaft, vol. 77, pp. 343–362. —— (1997), How to Expand the Limits of Insurability, Geneva Papers on Risk and Insurance, No. 85, pp. 515–522. —— (2000), Versicherungsbetriebslehre: Kernfragen aus entscheidungstheoretischer Sicht, Karlsruhe. Kessler, D. and Lamère, J.-M. (2002a), The Financial Impact of September 11 on the French Insurance Market, in: Études et Dossiers, Working Paper Series of the Geneva Association, No. 251, Special Issue on the Economic Consequences of the September 11 Attacks on National Insurance Markets, Geneva, pp. 59–76. —— (2002b), The Financial Impact of September 11 on the French Insurance Market, in: Liedtke, P., and Courbage, C. (eds.), Insurance and September 11 – One Year After, Impact, Lessons and Unresolved Issues, International Association for the Study of Insurance Economics, Geneva, pp. 56–69. Kollar, John J. (2002), Terrorism Insurance Coverage in the Aftermath of September 11th , A Public Statement by the Extreme Events Committee of the American Academy of Actuaries, mimeo, 17 April 2002, available under www.actuary.org Kunreuther, H. (2002), The Role of Insurance in Managing Extreme Events: Implications for Terrorism Coverage, December 2001, Wharton Financial Institutions Center Working Paper No. 02-07, Wharton School, University of Pennsylvania, Philadelphia. Lakdawalla, D. and Zanjani, G. (2002), ‘Insurance, Self-Protection, and the Economics of Terrorism’, NBER Working Paper 9215, Cambridge MA, also available under www.nber.org/papers/w9215 McCall, John J. (1987), Insurance, in: Eatwell, J., Milgate, M., Newman, P. (eds.), The New Palgrave Dictionary of Economics, vol. 2, E-J, Basingstoke, New York, pp. 868–870. —— (1992), Insurance, in: Newman, P., Milgate, M., Eatwell, J. (eds.), The New Palgrave Dictionary of Money and Finance, vol. 2, F-M, London, New York, pp. 424–427. Munich Re (2001), Press Release of 29th November 2001, www.munichre.com Nell, M. (2001), Staatshaftung für Terrorrisiken?, in: Zur Diskussion gestellt: Risikoübernahme – sollte der Staat bestimmte Versicherungsgarantien übernehmen?, ifo Schnelldienst, vol. 54, No. 24, pp. 6–9. OECD (2002), Economic Consequences of Terrorism, in: Economic Outlook 71, Spring 2002, pp. 117–140.

172 Michael Wolgast Rees, R. (2001), Kein Grund für staatliche Interventionen, in: Zur Dikussion gestellt: Risikoübernahme – sollte der Staat bestimmte Versicherungsgarantien übernehmen?, ifo Schnelldienst, vol. 54, No. 24, pp. 5–6. Rogoff, K. (2002), ‘Opening Remarks on the World Economic Outlook Press Conference’, available under www.imf.org Schaad, W. (2002), Terrorism – dealing with the new spectre, Swiss Re Focus Report, Zurich. Sinn, H.-W. (1986), Risiko als Produktionsfaktor (Risk as a Factor of Production), in: Jahrbücher für Nationalökonomie und Statistik (Journal of Economics and Statistics), vol. 201/6, pp. 557–571. —— (1988), Gedanken zur volkswirtschaftlichen Bedeutung des Versicherungswesens, in: Zeitschrift für die gesamte Versicherungswissenschaft, vol. 77, pp. 1–27. Straubhaar, Th. (2003), Die Weltwirtschaft nach dem 11. September: Mehr Regulierung oder mehr Markt?, in: Zeitschrift für Politikwissenschaft, vol. 13, forthcoming, available under www.hwwa.de/hwwa.html Swiss Re (2002), Natural catastrophes and man-made disasters in 2001: man-made losses take on a new dimension, sigma No. 1/2002, Zurich. Tillinghast-Towers Perrin (2001), Pool Re and Terrorism Insurance in Great Britain, Update, October 2001. Wolgast, M. (2001), Zweckgerichtete Arbeitsteilung zwischen Versicherungswirtschaft und Staat ist erforderlich, in: Zur Diskussion gestellt: Risikoübernahme – sollte der Staat bestimmte Versicherungsgarantien übernehmen?, ifo Schnelldienst, vol. 54, No. 24, pp. 12–14. —— (2002a), The implications of 11 September for the German insurance industry, in: Études et Dossiers, Working Paper Series of the Geneva Association, No. 251, Special Issue on the Economic Consequences of the September 11 Attacks on National Insurance Markets, Geneva, pp. 77–97. —— (2002b), The implications of 11 September for the German Insurance Industry, in: Liedtke, P. and Courbage, C. (eds.), Insurance and September 11 – One Year After, Impact, Lessons and Unresolved Issues, International Association for the Study of Insurance Economics, Geneva, pp. 70–86. —— (2002c), Global terrorism and the insurance industry: The case for public-sectorintervention, in: Insurance Economics, No. 46, International Association for the Study of Insurance Economics, Geneva, July 2002, pp. 7–10. —— (2002d), Global Terrorism and the Insurance Industry: New Challenges and Policy Responses, in Liedtke, P. and Courbage, C. (eds.), Insurance and September 11 – One Year After, Impact, Lessons and Unresolved Issues, International Association for the Study of Insurance Economics, Geneva, pp. 243–257. —— (2002e), ‘Insurability of the Risks of Terrorism and the Role of the State’, in: Asia Insurance Review, December, pp. 33–35.

10 Terrorism and international trade: an empirical investigation Volker Nitsch and Dieter Schumacher

This paper examines the effect of terrorism and warfare on international trade. We investigate bilateral trade ﬂows between more than 200 countries over the period from 1960 to 1993. Applying an augmented gravity model that includes several measures of terrorism and large-scale violence, we ﬁnd compelling evidence that terrorist actions reduce the volume of trade; a doubling in the number of terrorist incidents is associated with a decrease in bilateral trade by about 4 per cent.

1.

Outline

In this paper, we examine the impact of terrorism and warfare on international trade. To identify the effect of terrorism on trade empirically, we apply a gravity model of trade and add several measures of terrorist activity, internal instability and external conﬂict. We ﬁnd that violence strongly affects the pattern of trade; countries that are plagued by a larger number of terrorist attacks trade signiﬁcantly less with each other than otherwise similar countries that do not suffer from terrorism. This ﬁnding is robust for a number of modiﬁcations, including a large and diverse set of alternative violence measures. The remainder of the paper is in four parts. Section 2 discusses the association between terrorism and trade. Section 3 describes the empirical approach and the data. Section 4 presents the results, and section 5 concludes.

2.

Discussion on terrorism and trade

A large and diverse body of literature (mainly in political science) has documented the relationship between political variables and international trade. One line of research focuses on the impact of trade on political conﬂict. Polachek (1980), for example, argues that mutual trade dependencies diminish hostility and promote cooperation. Another line of research studies the impact of political factors on trade. Pollins (1989), for example, argues that diplomatic cooperation increases bilateral trade. Morrow et al. (1998) ﬁnd that trade ﬂows are greater between countries with common interests and similar democratic institutions. Recent theoretical

174 Volker Nitsch and Dieter Schumacher contributions that explore issues of the protection of property rights, security and trade include Grossman (2001) and Skaperdas and Syropoulos (2001). The empirical literature is surveyed, among others, in Barbieri and Schneider (1999) and Reuveny (1999–2000). This paper is closely related to the literature that assesses the political determinants of international trade ﬂows. In contrast to previous work, however, that mainly analyses more general measures of conﬂict and political cooperation, we focus directly on terrorist activity and internal and external instability. While the impact of terrorism on trade may vary across time and place, violence and warfare generally imply additional costs for transactions so that, if anything, we would expect a negative association between terrorist activity and the volume of trade. More speciﬁcally, there are at least three principal ways in which warfare may be a hindrance to international trade. First, terrorism leads to insecurity and thereby raises the costs of doing business. Buckelew (1984, p. 18) deﬁnes terrorism as ‘violent, criminal behavior designed primarily to generate fear in the community, or in a substantial segment of the community, for political purposes’ (emphasis added). Depending on the dimension of terrorist attacks, people may become increasingly confused, become nervous, or feel generally less safe. For instance, after ‘18 months of random violence that has killed close to 400 citizens, injured thousands, and distressed millions, [. . .] the atmosphere in Israel [is] a mix of deﬁance and despair.’ (The Economist, 30 March 2002). The collapse of the twin towers on 11 September 2001 was a laming shock to the US economy, whose full impact remains to be measured. Terrorism may lead to changes in a country’s consumption and production patterns (for example affecting tourism and shopping habits, and willingness to travel by public transport), thereby affecting production and consumption patterns and so international trade. However, terrorism-induced insecurity may also have a direct negative effect on trade. Since the economic impact of terrorist bombings and shootings are hardly predictable, existing business plans may quickly become obsolete so that ﬁrms generally face larger risks. Such insecurity may reduce the attractiveness of this market for international producers. Second, the typical response to an increase in terrorist activity is an increase in security measures. Stronger security regulations, however, imply that trade becomes more expensive, such as by increasing delivery times. For instance, after the terrorist attacks on 11 September 2001, US borders were temporarily closed; trucks on the border between Canada and the US had to wait up to 20 hours for a crossing that normally takes minutes.1 El Al, Israel’s airline, generally holds goods for a day for security checks (The Economist, 22 September 2001). Third, there is the risk of a direct destruction of traded goods. Terrorists may target a country’s trade when countries appear to be particularly vulnerable to the disruption of industry supply chains or to the destruction of particular transport modes.2 Despite identiﬁable channels, the overall magnitude of the effect of terrorism on trade remains a priori unclear. In fact, it is possible that terrorism has almost no measurable effect on trade since the overwhelming majority of terrorist actions

Terrorism and international trade: an empirical investigation

175

are operations with only local implications. Also, terrorists rarely target freight directly and physical losses are, at least in principle, insurable. In the following sections, we explore the association between terrorist activity and international trade empirically.

3.

Methodology and data

Our empirical approach to identify the effect of terrorism on trade can be summarized as follows: Tradeij = α + βTerror + γ Z + ε

(1)

where we regress (the log of) bilateral trade between countries i and j (Tradeij ) on our measure of terrorist activity (Terror) and a set of other conditioning variables Z that have the potential to affect the bilateral volume of trade, with ε being a well-behaved residual. The control variables in vector Z are from the gravity equation, which is a longestablished and empirically successful framework to model trade ﬂows. Hence, our basic regression framework is quite conventional:3 Tradeij = α + β Terror + γ1 Dij + γ2 Yi Yj + γ3 Yi Yj /Popi Popj + γ4 Language + γ5 Border + γ6 Colonizer + γ7 Nation + γ8 Colony + ε

(2)

where D is distance, Y is real GDP, Pop is population (all in natural logs), Language is a dummy that takes the value of one if i and j share a common language, Border is a common border dummy, Colonizer is a common colonizer dummy, Nation is a common nation dummy (e.g. for French overseas départements), and Colony takes the value of one if i colonized j or vice versa.4 Our data come from a number of different sources. The data for the dependent and the Z variables are taken from Glick and Rose (2002), who have constructed a comprehensive data set that covers (real) bilateral trade between 217 countries and territories between 1948 and 1997 and also provides information on all the other standard gravity variables.5 To this data set, we add four types of controls for Terror. In a ﬁrst exercise, we explore direct measures of terrorism; this is our main variable of interest. Raw data of terrorist activity are available from Mickolus (1980) who provides a detailed chronology of terrorist events around the world for the period from 1968 to 1979.6 Mickolus (1980, p. xiii) deﬁnes terrorism as ‘the use, or threat of use, of anxiety-inducing extra-normal violence for political purposes [. . .] when such action is intended to inﬂuence the attitudes and behavior of a target group wider than the immediate victims and when [. . .] its ramiﬁcations transcend national boundaries’. The types of politically motivated incidents included in his data set are: kidnapping, barricade-hostage taking, occupation, letter bombing, incendiary bombing, explosive bombing, missile attack,

176 Volker Nitsch and Dieter Schumacher armed attack, aerial hijacking, non-aerial takeover, assassination or murder, sabotage, exotic pollution, nuclear weapons threat, theft or break-in, conspiracy, hoax, sniping, shootouts with police and arms smuggling. Based on this information, we construct three variables of terrorist activity: the yearly number of terrorist events, the total number of terrorist events between 1968 and 1979, and a dummy variable that takes the value of one for at least one terrorist action. Following Mickolus, we treat each incident equally; we do not distinguish between either the types of incidents or their severity or any other characteristic (such as the number of casualties).7 It should also be noted that Mickolus’s data is on transnational terrorism. Mickolus (1980, p. xix) notes that: ‘[While] the bulk of politically violent incidents may occur in third world nations [. . .], nearly half of all attacks are recorded in westernised democracies. Transnational attacks are very infrequent in Asia and Africa.’ Table 10.1 shows the ﬁve countries that, according to our measures, suffered most strongly from terrorism over the period. A second group of variables focuses on measures of internal instability other than terrorism. These measures include the number of politically motivated murders or attempted murders of high government ofﬁcials or politicians (assassinations); the number of any armed activities, sabotage, or bombings carried out by independent bands of citizens or irregular forces and aimed at the overthrow of the present regime (guerrilla activity); the number of jailings or executions of political opposition within the ranks of the regime or the opposition (purges); the number of violent demonstrations or clashes of more than 100 citizens involving the use of physical force (riots); and the number of any illegal or forced changes in the top governmental elite, any attempt at such a change, or any successful or unsuccessful armed rebellion whose aim is independence from the central government (revolutions). The original source for this information is Banks (1979), but the data are also available in an updated version from Easterly and Sewadeh (2002). In a third set of controls, we analyse structural variables that aim to capture the extent to which a country’s resources are devoted to the reduction of military conﬂicts. In particular, we use defence expenditures as a share of GDP (defence) taken from the International Monetary Fund’s Government Financial Statistics and the size of armed forces as a fraction of the country’s total population (military) obtained from Easterly and Sewadeh (2002). The idea is that, unless countries are outright aggressors, a larger share of resources spent on national defence should indicate a greater risk to internal and external threats.8 Finally, we explore measures of external conﬂict. In this respect, Barro and Lee (1994) provide two variables that appear to be particularly appropriate to identify the impact of warfare on trade and are therefore used in our analysis: a dummy variable for countries that participated in at least one external war over the period 1960–1985 and the fraction of time over 1960–1985 each country was involved in an external war. Descriptive statistics for the data and bivariate correlations are provided in the working paper version available at www.diw.de.

Terrorism and international trade: an empirical investigation

177

Table 10.1 Description of data on terrorist activity Five countries that suffered most strongly from terrorism; period: 1968–79 a) Number of terrorist actions Country

Year

Number

United States United States United Kingdom Argentina Argentina France Argentina United States United States United States United States United States France Italy Argentina

1975 1976 1974 1973 1976 1976 1974 1978 1977 1974 1968 1979 1975 1976 1970

41 35 34 34 33 31 31 30 29 28 27 27 27 26 26

b) Total number of terrorist actions Country

Number

United States Argentina France United Kingdom Israel

288 192 163 149 135

c) Number of terrorist actions per capita Country

Number

OECD countries

Lebanon Israel Cyprus Djibouti Barbados

45.5 40.9 27.9 21.0 20.6

Ireland Greece Switzerland Netherlands Belgium

10.6 8.4 4.0 3.6 3.3

Source: calculated from Mickolus (1980).

4.

Results

We begin our investigation by estimating equation (2), augmented with our measures of terrorist activity. The results are presented in Table 10.2. For the sake of completeness, we report the estimates on the entire set of regressors, but later we will focus exclusively on our variables of interest, since all coefﬁcients on the standard gravity variables take the expected sign and are statistically and

178 Volker Nitsch and Dieter Schumacher Table 10.2 The impact of terrorism on trade Dependent variable: (log) bilateral trade period: 1968–1979 (Log product) 1+ number of terrorist actions (Sum dummy) at least one terrorist action, 1968–79 (Log product) 1+ total number of terrorist actions, 1968–79 (Log) distance (Log product) real GDP (Log product) real GDP per capita Common language Common land border Common colonizer Same nation Colonial relationship No. of observations SER Adj. R‘2

(1)

(2)

(3)

−0.041∗∗ (0.007) −0.098∗∗ (0.018)

−1.054∗∗ (0.010) 0.804∗∗ (0.004) 0.532∗∗ (0.006) 0.315∗∗ (0.020) 0.336∗∗ (0.047) 0.761∗∗ (0.031) 1.310∗∗ (0.277) 1.832∗∗ (0.044)

−1.053∗∗ (0.010) 0.800∗∗ (0.004) 0.550∗∗ (0.006) 0.312∗∗ (0.020) 0.361∗∗ (0.047) 0.783∗∗ (0.031) 1.221∗∗ (0.280) 1.795∗∗ (0.044)

−0.065∗∗ (0.005) −1.055∗∗ (0.010) 0.825∗∗ (0.004) 0.547∗∗ (0.006) 0.320∗∗ (0.020) 0.339∗∗ (0.047) 0.745∗∗ (0.031) 1.254∗∗ (0.279) 1.867∗∗ (0.043)

61,013 1.84 0.63

59,780 1.84 0.63

61,013 1.84 0.63

Notes: OLS estimation with (unreported) year effects. White heteroskedastic-consistent standard errors are in parentheses. ∗∗ denotes signiﬁcant at the 1% level. Data sources are described in the text.

economically signiﬁcant. For instance, trade increases with both higher GDP and higher GDP per capita (for the country pairing) and falls the greater the distance between two countries. In a ﬁrst exercise, we enter the (log interacted) number of terrorist attacks as (an additional) explanatory variable for the bilateral volume of trade. The estimated β coefﬁcient is indeed negative and, with a t-statistic of about 6, statistically highly signiﬁcant. The effect is also economically large; the point estimate of −0.04 implies that a doubling in the number of terrorist incidents (a rise by 100 per cent) is associated with a decrease in bilateral trade by about 4 per cent, holding all other things constant. This is in our view a very strong result since it implies that trade is

Terrorism and international trade: an empirical investigation

179

already sizeably reduced in the same (calendar) year in which a terrorist incident occurs. In the next two columns of Table 10.2, we explore our two alternative measures of terrorist activity and ﬁnd similar results. The coefﬁcient on the (additively linked) dummy of at least one terrorist action (that can take values of 0, 1, or 2) implies that the ﬁrst terrorist incident in a country pair reduces bilateral trade by almost 10 per cent; a pair of countries in which one country suffers from terrorist attack(s) trades only about 91 per cent (exp[−0.10] = 0.91) of what the two countries would trade if they were completely free from acts of terrorism. Also, replacing the number of terrorist actions per year with the total number of terrorist actions over a 12-year period has little effect on the results. The estimated β coefﬁcient is even somewhat larger, probably capturing the trade effects that are not visible in the same year of an attack.9 We have performed additional sensitivity analysis in order to establish the robustness of our results. For instance, we have dropped OECD countries. We have also analysed different time periods separately. Our estimates were robust to these perturbations (as were all other ﬁndings reported in this paper). To summarize, there is compelling evidence that terrorist activity negatively affects bilateral trade ﬂows. Countries targeted by terrorism trade signiﬁcantly less with each other than countries unaffected by terrorism. In the following, we explore the effect of other measures of internal and external conﬂict on international trade. We begin with alternative measures of internal instability; results are reported in Table 10.3. For each of the ﬁve main variables assassinations, guerrilla activities, purges, riots, and revolutions, we follow our approach for the terrorism variable and construct three separate measures. A ﬁrst measure again gives the number of relevant incidents for each year. As before, we consider this test to be very strong since a signiﬁcant coefﬁcient on this variable would imply that any additional incident already has a measurable effect on trade in the same year it occurs. A more reliable measure therefore might be the average number of relevant incidents, our second variable. Finally, we have constructed frequency measures, based on the number of years in which relevant incidents have occurred. In Table 10.3, each line gives the results of a separate regression; the regressions include the full set of gravity variables, but, to save space, we report only the estimated β coefﬁcient and some diagnostic statistics. Reviewing the results, all of the coefﬁcient estimates on our internal instability measures are indeed negative and statistically highly signiﬁcant. Thus, all kinds of domestic violence appear to sizeably reduce trade, irrespective of whether it takes the form of assassinations, guerrilla activities, purges, riots, or revolutions. While we consider these results as particularly encouraging, the estimates appear to be unusually strong. Could it be that our variables, though different, ﬁnally capture various aspects of the same event? We can convincingly reject this hypothesis. The bivariate correlations (though consistently positive) are often quite low. Also, if one enters the different instability measures jointly, the estimated coefﬁcients are all negative and statistically signiﬁcant (results are available from the authors on request).

180 Volker Nitsch and Dieter Schumacher Table 10.3 The impact of internal instability on trade Dependent variable: (log) bilateral trade period: 1960–1993

Coefﬁcient

No. of observations

SER

Adj. R2

(Log product) 1 + number of assassinations (Log Product) 1 + average number of assassinations, 1960–93 (Log product) 1 + number of years with at least one assassination, 1960–93 (Log product) 1 + number of guerrilla activities (Log product) 1 + average number of guerrilla activities, 1960–93 (Log product) 1 + number of years with at least one guerrilla activity, 1960–93 (Log product) 1 + number of purges

−0.160∗∗ (0.010) −0.622∗∗ (0.014) −0.224∗∗ (0.004) −0.183∗∗ (0.011) −0.416∗∗ (0.018)

148,574

2.09

0.63

163,327

2.06

0.63

170,725

2.06

0.62

148,574

2.09

0.63

163,327

2.07

0.62

−0.120∗∗ (0.004)

170,725

2.07

0.62

−0.163∗∗ (0.014) −1.021∗∗ (0.029) −0.229∗∗ (0.005) −0.091∗∗ (0.007) −0.096∗∗ (0.010) −0.132∗∗ (0.005) −0.145∗∗ (0.016) −0.837∗∗ (0.031) −0.219∗∗ (0.005)

148,574

2.09

0.63

140,594

1.99

0.65

147,028

1.99

0.64

148,574

2.09

0.63

140,594

1.99

0.65

147,028

2.00

0.64

148,574

2.09

0.63

140,594

1.99

0.65

147,028

1.99

0.65

(Log product) 1 + average number of purges, 1960–93 (Log product) 1 + number of years with at least one purge, 1960–93 (Log product) 1 + number of riots (Log product) 1 + average number of riots, 1960–93 (Log product) 1 + number of years with at least one riot, 1960–93 (Log product) 1 + number of revolutions (Log product) 1 + average number of revolutions, 1960–93 (Log product) 1 + number of years with at least one revolution, 1960–93

Notes: OLS estimation with (unreported) year effects. Other regressors not shown in the table: constant, log distance, log GDPs, log per capita GDPs, common language, common border, common colonizer, colonial relationship. White heteroskedastic-consistent standard errors are in parentheses. ∗∗ denotes signiﬁcant at the 1% level. Data sources are described in the text.

In the next step, we explore two structural variables that should capture the potential risk of military conﬂict. Armed forces that make up a relatively large share of a country’s total population as well as defence expenditures that represent a large fraction of GDP could both be indications of a country’s higher propensity to conﬂicts and warfare (including terrorism). Our results appear in Table 10.4. In line with our intuition, the relative size of the military has a negative effect on bilateral trade.

Terrorism and international trade: an empirical investigation

181

Table 10.4 The impact of military personnel and defence expenditures on trade Dependent variable: (log) bilateral trade period: 1972–1982 (Log product) size of military/population (Log product) defence expenditures/GDP

−3.637∗∗ (0.994)

No. of observations SER Adj. R2

46,105 1.87 0.64

0.148∗∗ (0.011) 16,394 1.70 0.71

−15.120∗∗ (1.841) 0.218∗∗ (0.015) 13,876 1.62 0.72

Notes: OLS estimation with (unreported) year effects. Other regressors not shown in the table: constant, log distance, log GDPs, log per capita GDPs, common language, common border, common colonizer, colonial relationship. White heteroskedastic-consistent standard errors are in parentheses. ∗∗ denotes signiﬁcant at the 1% level. Data sources are described in the text.

The coefﬁcient is statistically and economically signiﬁcant, implying that countries with large armed forces trade signiﬁcantly less with one another than otherwise similar countries with relatively small armies. For defence expenditures, however, the estimated coefﬁcient is positive and also highly signiﬁcant – a ﬁnding that survives extensive robustness checks such as including the (log product of the) land areas as an additional explanatory variable or splitting the sample into OECD and non-OECD countries. A potential explanation for this result is that large defence expenditures may indeed provide better security. If both variables enter our regression jointly, the results are essentially unchanged from the default speciﬁcations. Our ﬁnal exercise aims to identify the impact of external conﬂict on trade. As shown in Table 10.5, there is strong evidence that a country’s participation in an external war reduces its volume of international trade. The estimated β coefﬁcients are negative and statistically and economically signiﬁcant. The most interesting result, however, is recorded in the ﬁnal column of Table 10.5. When estimated jointly, the coefﬁcient on the fraction of time involved in external wars is (signiﬁcantly) positive, suggesting that it is mainly whether a country has ever participated in a war that matters and not the length nor the frequency of wars.

5.

Summary

It is plausible to expect that terrorism and large-scale violence have a negative effect on international trade. Higher risks, additional security measures and direct destruction raise the transaction costs and should thereby lower the volume of international trade. In this paper, we have presented evidence conﬁrming this intuitive claim. Applying an augmented gravity model and analysing bilateral trade ﬂows between more than 200 countries over the period from 1960 through

182 Volker Nitsch and Dieter Schumacher Table 10.5 The impact of war on trade Dependent variable: (log) bilateral trade period: 1960–1985

(1)

(Sum dummy) participation in at least one external war, 1960–85 (Log product) 1 + fraction of time involved in external war, 1960–85

−0.338∗∗ (0.010)

No. of observations SER Adj. R2

95,187 1.80 0.64

(2)

(3)

−0.395∗∗ (0.028)

−0.368∗∗ (0.012) 0.152∗∗ (0.034)

95,187 1.81 0.64

95,187 1.80 0.64

Notes: OLS estimation with (unreported) year effects. Other regressors not shown in the table: constant, log distance, log GDPs, log per capita GDPs, common language, common border, common colonizer, colonial relationship. White heteroskedastic-consistent standard errors are in parentheses. ∗∗ denotes signiﬁcant at the 1% level. Data sources are described in the text.

to 1993, we ﬁnd compelling evidence that terrorist actions reduce the volume of trade; this result is robust to alternative measures of political instability and military conﬂict. Our estimates suggest that a doubling in the number of terrorist incidents in a year is associated with a decrease in bilateral trade by about 4 per cent already in the same year.

Acknowledgements We thank Paul Gregory, Sanjay Jain, participants at the DIW workshop on ‘The Economic Consequences of Global Terrorism’, two anonymous referees, and an editor (Arye Hillman) for helpful comments and suggestions. This paper was written while Nitsch was an economist at Bankgesellschaft Berlin.

Endnotes 1 For a detailed account of the impact of the terrorist attacks on 11 September 2001 on international trade, see Walkenhorst and Dihel (2002). 2 For instance, according to the terrorism database of the US State Department (2002), there were not less than 178 bombings against a multinational oil pipeline in Colombia in 2001 alone. Terrorists in Iraq also targeted pipelines after the fall of the regime of Saddam Hussein. 3 The approach is similar to other recent work in empirical international trade where augmented gravity models have been applied to identify the impact of regional trading blocs (e.g. Frankel, 1997), national borders (e.g. McCallum, 1995), and currency unions (e.g. Rose, 2000) on trade. More importantly for our purposes, this approach is also widely used in the political science literature to assess the impact of political factors on trade (e.g. Morrow et al. 1998). 4 Note that our dependent variable is the total volume of bilateral trade so that we are unable to distinguish between an exporting and an importing country and our explanatory

Terrorism and international trade: an empirical investigation

5

6

7

8 9

183

variables enter the regression jointly for country i and j (usually as a product). So, to be more precise, our terrorism variable takes the form Terrori Terrorj . In principle, however, one might also make the case for decomposing trade between exports and imports, which allows investigation of whether terrorism has a greater impact in the exporting or the importing country. The data set is graciously provided by Andrew Rose on his website (http://faculty. haas.berkeley.edu/arose). In our actual analysis, we use different subsets of this large (>400,000 observations) panel data set according to the availability of data for our variable of interest. One of Mickolus’s collaborators, Todd Sandler, has informed us that the published data for the years 1978 and 1979 may be incorrect. We have checked the robustness of our results and ﬁnd that none of our results is affected when information for the last two years is excluded. Mickolus’s (1980) data does not allow us to weight severity of terrorist incidents. Ideally, one would distinguish terrorist attacks by the amount of media coverage since many terrorists seek public attention; see Mickolus and Simmons (1997). Also, there is no separation between state-sponsored terrorism and non-state-sponsored terrorism. Although there is evidence that, in particular in poor countries, military spending takes place for reasons related to corruption. See Gupta et al. (2001). In unreported results, we ﬁnd that the total number of terrorist actions is indeed the clearly dominant variable. If one enters the different terrorism measures jointly, only the estimate on the total number of attacks remains negative and statistically signiﬁcant, though the results may suffer from multicollinearity. Nevertheless, this ﬁnding suggests that it is mainly the long-run average of terrorist activities that affects international trade rather than some brief periods of isolated attacks or the fact that a country has suffered any terrorist action.

References Banks, A. S. (1979) Cross National Time-Series Data Archive, updated, Center for Social Analysis, State University of New York at Binghampton. Barbieri, K. and Schneider, G. (1999) ‘Globalization and peace: assessing new directions in the study of trade and conﬂict’, Journal of Peace Research, 36: 387–404. Barro, R. J. and Lee, J.-W. (1994) Data Set for a Panel of 138 Countries, Department of Economics, Harvard University. Buckelew, A. H. (1984) Terrorism and the American Response, San Rafael, CA: Mira Academic Press. Easterly, W. and Sewadeh, M. (2002) Global Development Network Growth Database, Washington, DC: World Bank. Frankel, J. A. (1997) Regional Trading Blocs, Washington, DC: Institute for International Economics. Glick, R. and Rose, A. K. (2002) ‘Does a currency union affect trade?: The time series evidence’, European Economic Review, 46: 1125–51. Grossman, H. (2001) ‘Creation of effective property rights’, American Economic Review, 91: 347–52. Gupta, S., de Mello, L. and Sharan, R. (2001) ‘Corruption and military spending’, European Journal of Political Economy, 17: 749–77. McCallum, J. (1995) ‘National borders matter: Canada-US regional trade patterns’, American Economic Review, 85: 615–23. Mickolus, E. F. (1980) Transnational Terrorism, Westport, CT: Greenwood Press.

184 Volker Nitsch and Dieter Schumacher Mickolus, E. F. and Simmons, S. L. (1997) Terrorism, 1992–1995, Westport, CT: Greenwood Press. Morrow, J. D., Siverson, R. M. and Tabares, T. E. (1998) ‘The political determinants of international trade: the major powers, 1907–90’, American Political Science Review, 92: 649–61. Polachek, S. W. (1980) ‘Conﬂict and trade’, Journal of Conﬂict Resolution, 24: 55–78. Pollins, B. M. (1989) ‘Does trade still follow the ﬂag?’, American Political Science Review, 83: 465–80. Reuveny, R. (1999–2000) ‘The trade conﬂict debate: a survey of theory, evidence and future research’, Peace Economics, Peace Science and Public Policy, 6: 23–49. Rose, A. K. (2000) ‘One money, one market: the effect of common currencies on trade’, Economic Policy, 30: 7–45. Skaperdas, S. and Syropoulos, C. (2001) ‘Guns, butter, and openness: on the relationship between security and trade’, American Economic Review, 91: 353–57. US Department of State (2002) Patterns of Global Terrorism 2001. Available at www.state.gov/s/ct/rls/pgtrpt/2001/pdf. Walkenhorst, P. and Dihel, N. (2002) Trade impacts of the terrorist attacks of 11 September 2001: a quantitative assessment. Paris: OECD.

11 Fiscal consequences of armed conﬂict and terrorism in low- and middle-income countries Sanjeev Gupta, Benedict Clements, Rina Bhattacharya, and Shamit Chakravarti

1.

Introduction

Contrary to expectations, the end of the Cold War has not been a harbinger of peace. There has been a proliferation of armed conﬂicts around the world over the past dozen years. In particular, terrorist groups have become increasingly sophisticated, daring and destructive. More than four million people are estimated to have perished in violent conﬂicts between 1989 and 2000, and 37 million people have been displaced as refugees, either inside or outside their countries (World Bank, 2000). In 2000, there were 25 major armed conﬂicts around the world, of which 23 were intrastate conﬂicts (SIPRI Yearbook 2001).1 International terrorist attacks increased from an average of about 342 a year between 1995 and 1999 to 387 a year between 2000 and 2001.2 Most of the armed conﬂicts and terrorist activities have taken place in low- and middle-income countries. Between 1996 and 2000, almost 70 per cent of the major armed conﬂicts, more than 20 per cent of all international terrorist attacks, and over 70 per cent of all casualties due to such attacks took place in Asia and Africa. While the literature has documented the economic costs of armed conﬂict and terrorism, a cross-country examination of their ﬁscal consequences is yet to be undertaken. Armed conﬂict and prolonged terrorist activities can strongly inﬂuence the revenues and expenditures of countries, and in turn affect their economic growth. Although armed conﬂict and terrorism are often treated as distinct phenomena, experience from different parts of the world shows that there is a close link between the two. This paper analyses the effects of armed conﬂict and terrorism on ﬁscal balances and economic growth in low- and middle-income countries. The remainder of this paper is structured as follows. Section 2 provides a brief overview of the literature, followed in Section 3 by a description of the channels through which armed conﬂict and terrorism can affect the ﬁscal accounts and economic growth. Section 4 sets out the methodology for the empirical analyses presented in the paper. Section 5 compares the evolution of various macroeconomic variables and socioeconomic indicators before, during, and after 22 episodes of armed conﬂict in a number of low- and middle-income countries. Section 6 estimates an integrated system of equations for real per capita income growth,

186 Sanjeev Gupta et al. government revenue, and government spending, to highlight the main channels through which armed conﬂict and terrorism affect the ﬁscal accounts. Section 7 concludes.

2.

Review of the literature

Several studies have assessed the economic costs of armed conﬂicts.3 Richardson and Samarasinghe (1991) estimate that the total accumulated economic cost of the armed conﬂict in Sri Lanka in the ﬁve years between 1983 and 1988 was about 4.2 billion USD, or 68 per cent of Sri Lanka’s GDP in 1988. Arunatilake, et al. (2000) perform a similar exercise for a longer period and estimate that the conﬂict between 1983 and 1996 cost Sri Lanka about twice the country’s 1996 GDP. In a similar vein, several empirical studies, based on different techniques, approaches and data, have found an inverse relationship between different measures of political instability and violence on the one hand, and growth or investment on the other (Venieris and Gupta, 1986; Barro, 1991; Alesina and Perotti, 1993 and 1996; Alesina et al., 1996; and Rodrik, 1999). Armed conﬂict impacts on a country’s ﬁnancial development. Addison et al. (2002) conclude that conﬂict can (1) adversely affect the process of ﬁnancial deepening by undermining conﬁdence in the domestic currency due to fear of inﬂation and depreciation; (2) encourage the movement of funds away from productive assets (bank deposits, capital) to nonproductive assets (gold); and (3) affect the regulation and supervision of the ﬁnancial system. Their model, applied to 79 countries, shows that conﬂict signiﬁcantly reduces ﬁnancial development, and that the negative effect increases as conﬂict intensiﬁes. Prolonged terrorist activities, like armed conﬂict, also lower growth, both directly and indirectly. Abadie and Gardeazabal (2001) ﬁnd that after the outbreak of terrorism in the 1970s, per capita GDP in the Basque region of Spain declined by about 10 per cent relative to a ‘synthetic’ control region, and that this gap widened in response to spikes in terrorist activity. Some studies have empirically assessed the impact of terrorism on tourism, both domestic and regional, and have found the expected negative effect (Drakos and Kutan, 2001; Enders and Sandler, 1991; and Enders, et al., 1992). For example, in a study covering Greece, Israel, and Turkey, and using Italy as a ‘control variable’, Drakos and Kutan (2001) found that the intensity (measured by number of casualties) of terrorist incidents has signiﬁcant domestic and cross-country effects on the market shares of the affected countries, and that there are signiﬁcant contagion effects from terrorism within the region. Terrorist threats raise the transaction costs of doing business and trade. Nitsch and Schumacher (2002) show that terrorist acts and large-scale violence adversely affected bilateral trade ﬂows for more than 200 countries for the period 1960–93. A doubling of the number of terrorist incidents is associated with a decrease in bilateral trade by about 6 per cent. Moreover, additional security measures put in place to deter terrorist attacks can impede the ﬂow of goods and services. Walkenhorst and Dihel (2002) estimate the global welfare losses due to

Fiscal consequences of armed conﬂict and terrorism

187

tighter security precautions which have been put in place following the attacks of 11 September 2001 at about 75 billion USD. As noted earlier, a cross-country examination of the ﬁscal consequences of armed conﬂicts is yet to be undertaken. However, recent case studies and related empirical studies of military spending and growth suggest channels through which armed conﬂict and terrorism can have an effect on ﬁscal accounts and economic growth.

3.

Fiscal effects of armed conﬂict and terrorism: potential channels

Armed conﬂict and terrorism can affect the ﬁscal accounts by disrupting economic activities, eroding the tax base, lowering the efﬁciency of tax administration, and distorting the composition of public spending. Tax receipts, for example, vary with the health of the economy. Economic downturns due to insecurity and violence can lead to a decline in tax revenues. Beyond their effects on real activity, armed conﬂict and terrorism (especially if prolonged) can destroy part of the tax base (through the destruction of business ﬁrms, for example) and weaken the efﬁciency of tax administration. For example, Ndikumana (2001) notes that, following the outbreak of armed conﬂict in two countries in Africa, not only did the tax base collapse, but tax administration was also hampered. With the return of peace and the resumption of normal production in one of the two countries, tax revenues recovered progressively, and by 1998 exceeded the preconﬂict level. Military expenditures typically increase in response to conﬂict and terrorism, and tend to remain high even after cessation of violence.4 Higher spending for security can also affect the composition of public spending by decreasing outlays for education, health, and other productive items. Moreover, the destruction of physical infrastructure and human capital due to violence, and the indirect effects on trade, tourism, and business conﬁdence, all weaken the ﬁscal position and adversely affect economic growth, as noted earlier. Defence spending can affect the long-run sustainable growth rate both negatively and positively (Shieh et al., (2002)). First, there is a ‘crowding out effect’, whereby an increase in defence expenditures by the government reduces the resources available to the economy for private investment and for public spending on sectors that have a strong and positive impact on growth. Second, there is a ‘spin-off ’ effect from the positive supply-side spillover effects of defence expenditure on the nondefence sectors of the economy. This effect is likely to be small in low- and middle-income conﬂict-affected countries, since the majority of defence spending tends to be on imported armaments. Third, there is a ‘resource mobilization’ effect on savings and investment: defence spending provides both internal and external security, and hence, boosts private savings and investment and attracts foreign investment. This has a positive effect on growth.5 Earlier studies have suggested that defence spending has a positive effect on economic growth in less-developed countries (Benoit, 1978). However, more

188 Sanjeev Gupta et al. recent empirical research shows that cutting military spending fosters economic growth (Arora and Bayoumi, 1993; Bayoumi et al., 1993; and Knight et al., 1996). These papers argue that lower military spending can encourage growth by increasing capital formation and improving the efﬁciency with which resources are utilized in the economy. Cessation of conﬂict and terrorism can result in a ‘peace dividend’, releasing ﬁscal resources to be used for deﬁcit reduction, lowering taxes, or raising the allocation for spending in social sectors.6

4.

Empirical methodology

The empirical analysis in this paper is based on two approaches. The ﬁrst approach assesses the impact of armed conﬂict within conﬂict-affected countries, by examining the evolution of macroﬁscal and socioeconomic variables before, after, and during 22 episodes of conﬂict in 20 low- and middle-income countries.7 The sample includes those episodes of armed conﬂict that either began or were ongoing in 1985 or later, and which ended by 1999, based on SIPRI’s deﬁnition of major armed conﬂicts.8 SIPRI draws data on armed conﬂicts from the Uppsala Conﬂict Data Project of the Department of Peace and Conﬂict Research, Uppsala University, Sweden. The Uppsala Conﬂict Data Project divides armed conﬂicts into the following three categories based on the level of casualties: • • •

Minor armed conﬂict: At least 25 battle-related deaths a year and fewer than 1,000 battle-related deaths during the course of the conﬂict. Intermediate armed conﬂict: At least 25 battle-related deaths a year and an accumulated total of at least 1,000 deaths, but fewer than 1,000 in any given year. War: At least 1,000 battle-related deaths a year.

SIPRI’s characterization of a major armed conﬂict covers the two most severe levels of conﬂict, i.e. ‘intermediate’ armed conﬂict and war (Gleditsch et al., 2001). This paper does not include ‘minor’ armed conﬂicts, since these are unlikely to have measurable effects on the ﬁscal accounts and the economic growth of the affected countries. One shortcoming of the SIPRI index is that it applies an absolute criterion for the number of battle-related deaths. Thus, a country with a large population will be classiﬁed as being in conﬂict even though the number of deaths may be small relative to its population. Moreover, the number of battle-related deaths may not adequately capture the economic impact of armed conﬂict; it is possible that a number of sporadic, low-intensity incidents affecting mainly the local population will have a different impact on business and consumer conﬁdence and international perception of risk in the country concerned than a single dramatic event affecting mainly the tourist sector or key sectors linked to foreign trade. Despite these drawbacks, the SIPRI index is broadly consistent with the conﬂict index produced by the Heidelberg Institute for International Conﬂict Research (HIIK).9

Fiscal consequences of armed conﬂict and terrorism

189

The second approach followed compares the economic consequences of armed conﬂict and terrorism across countries by estimating an integrated system of equations for real per capita income growth, government revenue, and government spending. The International Country Risk Guide (ICRG) ratings on internal conﬂict are used as a proxy for the combined risk from terrorism and conﬂict.10 The ICRG ratings provide an overall assessment of violence in a country due to civil war, terrorism, and civil disorder, and the actual or potential impact on governance. The highest rating is given to those countries ‘…where there is no armed opposition to the government and the government does not indulge in arbitrary violence, direct or indirect, against its own people’. The lowest rating is given to a country embroiled in an ongoing civil war and/or facing terrorist attacks. Given the difﬁculty of reaching a consensus on a universally acceptable deﬁnition of terrorism as well as of measuring terrorist activities, separate risk ratings for terrorism are not available. One advantage of the ICRG ratings is that they provide ratings of risk due to internal conﬂict and terrorism for a wide range of countries, and not just for those which have had major armed conﬂicts as deﬁned by SIPRI.11 The SIPRI index of armed conﬂicts (proportion of each ﬁve-year period during which there were armed conﬂicts) is used to check the robustness of the results. The SIPRI index has been used in other empirical studies, such as Davoodi et al. (2001).

5.

Macroeconomic and ﬁscal variables and socioeconomic indicators: preconﬂict, conﬂict, and postconﬂict periods

The results from comparing the conﬂict, preconﬂict, and postconﬂict phases of 22 episodes of armed conﬂicts in lower- and middle-income countries are presented in Figures 11.1–5 and Table 11.1. The data on real GDP are consistent with the hypothesis of a signiﬁcant pickup in growth in the immediate postconﬂict years. There is a dramatic pickup in inﬂation during the conﬂict period, followed by a signiﬁcant decline in the immediate postconﬂict period (see Figures 11.1 and 11.2). The data show a notable increase in the share of gross ﬁxed-capital formation to GDP in the immediate postconﬂict years, particularly in the private sector (see Figure 11.3). Figures 11.4–5 show the evolution of ﬁscal variables over the preconﬂict, conﬂict and postconﬂict periods. Due to data constraints, government revenue and foreign grants are used as a proxy for government revenue.12 The available data for the sample of countries show that the share of government revenue in GDP tends to fall during the conﬂict period and to pick up somewhat in the immediate postconﬂict period. On the expenditure side, there appears to be a signiﬁcant increase in government expenditure and net lending as a per cent of GDP during the conﬂict period compared with the preconﬂict period, followed by a notable decline in the immediate postconﬂict period. In particular, the available data suggest high government spending on defence during the conﬂict period and in the period immediately preceding it, followed by a signiﬁcant fall in the immediate

190 Sanjeev Gupta et al.

Figure 11.1 Real GDP growth in conﬂict countries1 (average annual per cent changes). 1

Based on a sample of 12 countries. The real GDP per capita growth corresponding to the preconﬂict, conﬂict, and postconﬂict periods are −3,7,−3,5, and 1.2 percent per annum, respectively. Sources: IMF, World Economic Outlook; World Bank, World Bank, World Development Indicators 2001; and IMF staff calculations.

postconﬂict period. However, high defence spending during the conﬂict period and in the years immediately preceding it tends to be at the expense of macroeconomic stability (as reﬂected for example in higher budget deﬁcits and a pickup in inﬂation) rather than at the cost of lower spending on education and health as a share of GDP. Nevertheless, since conﬂict is associated with lower real GDP growth, the implication is lower growth in real per capita government spending on education and health during conﬂict periods. Turning now to the socioeconomic indicators, Table 11.1 shows a signiﬁcant decline in the rate of improvement of life expectancy at birth during the conﬂict period, but the trend for improvement in life expectancy picks up again in the immediate postconﬂict period. There is also a signiﬁcant deterioration in the rate of improvement of infant mortality during conﬂict years, but the deterioration continues into the immediate postconﬂict period. The available data also show a marked improvement in gross enrolment rates (at all three levels – primary, secondary, and tertiary) following the end of armed conﬂict. While a useful exercise, the conclusions drawn from the before-during-after analysis should be interpreted with caution. This analysis does not control for other factors that affect macroeconomic and ﬁscal outcomes, independent of armed

Fiscal consequences of armed conﬂict and terrorism

191

Figure 11.2 Consumer price inﬂation in conﬂict countries1 (average annual per cent change). 1

Based on a sample of 9 countries. Sources: IMF, World Economic Outlook; World Bank, World Development Indicators 2001; and IMF staff calculations.

conﬂict and terrorism, which may have also changed over the periods of violence. To isolate more rigorously the effects of conﬂict and terrorism, the following section presents the econometric estimation of a system of interlinked equations covering a wider range of countries, including those not affected by conﬂict and terrorism.

6.

Econometric estimates

As mentioned earlier, there are three main ways in which armed conﬂict and terrorism can affect the ﬁscal accounts: by inﬂuencing real economic activity (GDP) and therefore, government revenues; by adversely affecting both the tax base and the efﬁciency of the tax administration; and by changing the composition of government spending. These ﬁscal consequences can have repercussions on economic growth, which would further affect the public ﬁnances. To capture all these effects, a structural model with three equations is speciﬁed: the ﬁrst for economic growth, the second for the ratio of government revenue to GDP, and the third for the composition of government spending measured by the share of defence spending in total government expenditure.

192 Sanjeev Gupta et al.

Figure 11.3 Capital formation in conﬂict countries1 in per cent of GDP. 1

Based on a sample of 17 countries for gross ﬁxed capital formation, and on 11 countries each for gross public and private capital formation. Sources: IMF, World Economic Outlook; World Bank, World Development Indicators 2001; and IMF staff calculations.

In the structural model, the equations for per capita income growth (equation (1)), government revenue to GDP (equation (2)), and defence expenditure as a share of total government spending (equation (3)) are speciﬁed as follows: GRPCYit = αr + α1 PCYINIi,t + α2 GSECINIi,t + α3 DEFEXPDi,t + α4 AGEDEPi,t + α5 CONFi,t + α6 INVGDPi,t + µ1i,t

(1)

GREVGDPit = βr + β1PCYi,t + β2NONAGRXi,t + β3CONFi,t + β4AGRVAi,t + β5URBPOPi,t + µ2i,t DEFEXPDit = λr + λ1 DEFGDPNi,t + λ2 CONFi,t + µ3i,t , where GRPCY = growth of real per capita income (GDP). PCYINI = real per capita income (GDP, in USD) in the initial year of the sample period.

(2) (3)

Fiscal consequences of armed conﬂict and terrorism

193

Figure 11.4 Fiscal aggregates in conﬂict countries1 in per cent of GDP. 1

Based on a sample of 14 countries. Sources: IMF, World Economic Outlook Bank, World Development Indicators 2001; and IMF staff calculations.

= gross secondary school enrolment rate in the initial year of the sample period. DEFEXPD = share of defence expenditure in total government spending. AGEDEP = age-dependency ratio. CONF = a conﬂict variable (discussed below). INVGDP = total investment in per cent of GDP. GREVGDP = government revenue as a ratio of GDP. PCY = real per capita income (in USD). NONAGRX = share of nonagricultural exports in GDP. AGRVA = agriculture value added in per cent of GDP. URBPOP = urban population as a share of total population. DEFGDPN = (unweighted) average of neighboring countries’ ratio of defence spending to GDP.

GSECINI

αr , βr , and λr are region-speciﬁc factors, and µ1it , µ2it and µ3it are the usual error terms. The subscript (it) for the main explanatory variables refers to country and time period, respectively. The endogenous variables in the system are the three dependent variables and the investment ratio. The model is estimated using ﬁve-year averages of annual data for each country over four time periods: 1980–4,

194 Sanjeev Gupta et al.

Figure 11.5 Composition of government spending in conﬂict countries1 in per cent of GDP. 1 Based on a sample of 12 countries for defence expenditure, and on 6 countries each for education and health spending. Sources: IMF, World Economic Outlook; World Bank: World Development Indicators 2001; and IMF staff calculations.

1985–9, 1990–4, and 1995–9. Region and time dummies were included in the estimated equations. Some authors have argued that conﬂict and terrorism are, in a sense, endogenous due to the possibility of reverse causation, i.e., that prolonged poor growth performance may help engender conﬂict. Violence and unrest may not only be a cause but may also arise from ﬂuctuations in economic variables. Indeed, instrumental variable techniques have been used in some of the studies to correct for reverse causation, but the validity of instruments in crosscountry regressions has been questioned by some authors (Abadie and Gardeazabal, 2001). However, given the difﬁculty of empirically modelling conﬂict and terrorism, and in ﬁnding suitable instruments, they are taken to be exogenous in line with a number of other studies (e.g. Davoodi et al. 2001; Gupta et al. 2001; Hess, 2003). The above structural model was estimated using the Generalized Method of Moments (GMM) estimation technique so as to address the underlying problems of autocorrelation and heteroscedasticity that typically arise in estimating a structural panel model with endogenous variables. Following the standard approach, the

Fiscal consequences of armed conﬂict and terrorism

195

Table 11.1 Selected social indicators in countries experiencing armed conﬂictsa (average annual rates of change, per cent) Preconﬂict b Conﬂict b Postconﬂict b Number of countries for which data are available Life expectancy at birth, total (years) Mortality rate, infant (per 1,000 live births) c Gross primary enrolment rate Gross secondary enrolment rate Gross tertiary enrolment rate

0.4

−0.5

0.4

5

3.8

0.6

0.0

7

2.6 1.1 −1.5

3.2 2.1 2.1

9 9 9

.. .. ..

Notes:

a Combines all the low-income, lower-middle-income, and upper-middle-income countries affected by armed conﬂict as discussed in the paper. Countries are classiﬁed into income categories based on the World Bank’s criteria in terms of level of 1998 GNP per capita – low-income, $760 or less; lower-middle-income, $761 to $3,030; and upper-middle-income, $3,031 to $9,360. b Conﬂict period refers to the period over which a country experienced armed conﬂict (as deﬁned by SIPRI); preconﬂict refers to the average of three years preceding the conﬂict, and postconﬂict refers to the average of three years following the conﬂict (depending upon availability of data). c Positive rates of growth signify an improvement in the variable.

Sources:

IMF, World Economic Outlook; World Bank, World Development Indicators 2001; and authors’ calculations.

instruments used in the estimation were all the exogenous variables of the structural model – i.e. all the variables in the system except for the three dependent variables and the investment ratio – plus an ICRG corruption index and inﬂation volatility. The latter two are used as proxies for the investment climate to instrument for the investment ratio.13 All of the results presented below pass the Sargan test for validity of the instrument set. The data used in estimation of the structural model were taken from the IMF’s World Economic Outlook, the World Bank’s World Development Indicators 2001, Yearbooks of the Stockholm International Peace Research Institute, and the International Country Risk Guide. Due to the limited availability of time series data on tax revenues, data on revenues and foreign grants are used as a proxy for domestic government revenues. Model 1a (Table 11.2) uses the ICRG measure of internal conﬂict and terrorism. Note that a higher value of the ICRG conﬂict rating implies a lower risk of internal conﬂict and terrorism.14 As in the standard Barro growth equations (Barro, 1991), the coefﬁcient on the initial level of per capita income is negative and statistically signiﬁcant at the 5 per cent level. However, the coefﬁcient on the initial stock of human capital (proxied by the gross secondary school enrolment rate) is not statistically signiﬁcant. The implication is that, at least for the sample

Table 11.2 Regression results Dependent variable

Model 1a

Model 1b

Growth of real per capita income Per capita income, initial Gross secondary enrolment, initial Ratio of defence spending to government expenditure

−0.0004 (−2.33)∗∗ −0.017 (−1.22) −0.37 (−7.01)∗∗∗

SIPRI rating for major armed conﬂicts

R−squared

−0.15 (−0.91) −4.04 (−1.30) 0.17 (1.10) 0.18

Dependent variable

Model 1a

ICRG internal conﬂict rating (civil wars and terrorism) Age dependency ratio Total investment

−0.0005 (−3.20)∗∗∗ −0.02 (−1.40) −0.28 (−5.65)∗∗∗ −1.72 (−1.90) −6.84 (−2.22)∗∗ −0.01 (−0.05) 0.24 Model 1b

Revenue (in per cent of GDP) Real per capita income Ratio of non agricultural exports to GDP

0.001 (2.76)∗∗∗ 0.23 (4.85)∗∗∗

SIPRI rating for major armed conﬂicts ICRG internal conﬂict rating (civil wars and terrorism)

R−squared

0.06 (0.30) −0.07 (−1.09) 0.02 (0.41) 0.58

Dependent variable

Model 1a

Agriculture value added Urbanization

0.001 (2.70)∗∗∗ 0.22 (4.75)∗∗∗ −0.34 (−0.30) −0.09 (−1.38) 0.02 (0.42) 0.58 Model 1b

Defence spending (in per cent of government spending) Average defence spending of neighbours (in per cent of GDP)

0.97 (2.61)∗∗∗

SIPRI rating for major armed conﬂicts ICRG internal conﬂict rating (civil wars and terrorism) R−squared Number of observations p−values a

−0.80 (−3.87)∗∗∗ 0.46 137 0.70

1.26 (3.38)∗∗∗ 2.68 (1.31) 0.43 127 0.78

Notes: White’s heteroscedastic consistent t−statistics are in parentheses; (***), (**), and (*) denote signiﬁcance at the 1%, 5%, and 10% levels, respectively. a The p−values refer to the test of overidentifying restrictions implied by the exogeneity of instruments. The instruments used are: corruption, inﬂation volatility, and all the exogenous variables in the system (i.e., all variables except for the three dependent variables and the ratio of total investment to GDP).

Fiscal consequences of armed conﬂict and terrorism

197

of countries included in this study, convergence toward a common level of real per capita income is not dependent on the initial stock of human capital. The age dependency ratio is also not statistically signiﬁcant, and neither is the ICRG rating for internal conﬂict and terrorism. Consistent with our hypothesis, the ratio of defence spending in total government expenditure has a negative and statistically signiﬁcant effect on growth. The structural equation for the government revenue-to-GDP ratio is based on studies such as Bahl (1971), Tanzi (1992), and Ebrill et al. (2001). The estimates are consistent with their ﬁndings that the share of government revenue in GDP in developing countries is a function of the level of development (proxied by real per capita income) and of the openness of the economy (proxied by the ratio of nonagricultural exports to GDP). However, the internal conﬂict and terrorism variable does not have any signiﬁcant effect on the government revenue-to-GDP ratio, and neither does the structure of the economy (proxied by the ratio of valueadded in agriculture to GDP). One reason why stronger results are not obtained for this equation could be the inclusion of foreign grants in the measure of revenues; some of the structural variables explaining government tax revenues, for example, may not have an impact on grants in the same way. The third equation for the share of defence in government expenditure is consistent with the ﬁnding in Davoodi et al. (2001) that higher spending on defence by neighbouring countries – which could be interpreted as a measure of regional tensions – is associated with a signiﬁcantly higher share of defence in total government spending. Moreover, the coefﬁcient for internal conﬂict and terrorism is positive and statistically signiﬁcant at the 1 per cent level.15 In summary, the empirical results using the ICRG rating for internal conﬂict and terrorism suggest that violence and insecurity raise the share of defence spending in total government expenditure, which in turn has a negative effect on growth by diverting resources away from spending on sectors (education, health infrastructure) that promote economic growth over the long term. The risk from conﬂict and terrorism does not seem to have any additional negative impact on growth, over and above its impact on the composition of government spending. Moreover, conﬂict and terrorism do not seem to have any impact on government revenue. To assess the robustness of the results, the above model is reestimated using a different measure of conﬂict: the proportion of years during each ﬁve-year period when the country was in conﬂict according to the SIPRI index. The results (Model 1b) tell a somewhat different story from the Model 1a estimates; the ratio of defence spending to government expenditure still has a statistically signiﬁcant and negative effect on growth, but (unlike in Model 1a) the SIPRI-based conﬂict variable does not have a statistically signiﬁcant impact on the composition of government spending. However, the SIPRI-based measure of armed conﬂict has a direct negative effect on growth which is statistically signiﬁcant at the 10 per cent level. Also, the age dependency ratio now becomes statistically signiﬁcant as well. In short, the results using the SIPRI-based conﬂict variable suggest that conﬂict

198 Sanjeev Gupta et al. has a direct negative impact on growth, rather than an indirect effect through the composition of government spending. The results using the SIPRI-based conﬂict index may differ from those using the ICRG rating for internal conﬂict and terrorism because the former is discrete for any given year (either 0 or 1). By contrast the ICRG rating for internal conﬂict and terrorism is a more continuous variable, and varies from 0 to 12 with changes in the perceived risk from violence and insecurity. Some authors have argued that ethnic fractionalization also has an impact on growth (e.g. Easterly and Levine, 1997). A variable measuring fragmentation, however, is not found to have a statistically signiﬁcant effect (Table 11.3). Furthermore, the SIPRI-based measure of armed conﬂict remains a statistically signiﬁcant determinant of growth, while the results using the ICRG measure of internal conﬂict and terrorism are broadly unchanged.

7.

Conclusions

The empirical literature on economic costs of armed conﬂicts and terrorism has yet to provide a comprehensive, cross-country examination of their ﬁscal consequences. This study provides a cross-country examination using two approaches. First, the evolution of various macroeconomic and ﬁscal variables and socioeconomic indicators during 22 episodes of conﬂict, and in the years immediately preceding and following the conﬂicts, was analysed. Second, an integrated system of equations for real per capita income growth, government revenue, and government spending was estimated to examine the main channels through which armed conﬂict and terrorism affect the ﬁscal accounts. The empirical results using the ICRG measure for internal conﬂict and terrorism are consistent with the hypothesis that armed conﬂict and terrorism lead to a higher share of defence spending in total government expenditure, which has a negative effect on growth by diverting resources away from spending on socially and economically productive sectors that promote economic growth. The results using the SIPRI-based conﬂict measure, however, suggest that conﬂict has a direct and signiﬁcant negative impact on growth, rather than an indirect effect through its impact on the composition of government spending. The results using the SIPRIbased conﬂict index may differ from those using the ICRG rating for internal conﬂict and terrorism because the former is discrete for any given year (either 0 or 1). By contrast, the ICRG rating for internal conﬂict and terrorism is a more continuous variable, and varies from 0 to 12 with changes in the perceived risk from violence and insecurity. The ﬁndings from the econometric estimation are generally consistent with the conclusions of the before-during-after conﬂict analysis. The share of government revenue in per cent of GDP tends to fall during the conﬂict period, and to pick up somewhat in the immediate postconﬂict period. This analysis also suggests that armed conﬂict leads to higher government spending on defence, but this tends to be at the expense of macroeconomic stability (reﬂected, for example, in signiﬁcantly higher budget deﬁcits and a pickup in inﬂation) rather than at the cost of lower

Table 11.3 Regression results: robustness test Dependent variable

Per capita income, initial Gross secondary enrolment, initial Ratio of defence spending to government expenditure SIPRI rating for major armed conﬂicts

Growth of real per capita income Model 1a

Model 1b

−0.0004 (−3.52)∗∗∗ −0.001 (−0.07) −0.36 (−8.74)∗∗∗

−0.0006 (−4.76)∗∗∗ −0.008 (−0.51) −0.27 (−5.73)∗∗∗ −2.17 (−2.16)∗∗

Ethnic fragmentation R-squared

−0.13 (−0.86) −3.58 (−0.92) −0.005 (−0.042) 0.01 0.19

Dependent variable

Revenue (in per cent of GDP)

ICRG internal conﬂict rating (civil wars and terrorism) Age dependency ratio Total investment

Model 1a Real per capita income Ratio of non-agricultural exports to GDP

0.0008 (1.81)∗ 0.18 (4.74)∗∗∗

SIPRI rating for major armed conﬂicts ICRG internal conﬂict rating (civil wars and terrorism) Agriculture value added Urbanization R-squared Dependent variable

−0.29 (−1.75)∗ −0.08 (−1.33) 0.08 (1.70)∗ 0.51

ICRG internal conﬂict rating (civil wars and terrorism) R-squared Number of observations p-values a Notes:

Model 1b 0.0008 (1.91)∗ 0.18 (4.48)∗∗∗ 0.60 (0.60) −0.09 (−1.49) 0.05 (1.24) 0.49

Defence Spending (in per cent of government spending) Model 1a

Average defence spending of neighbours (in per cent of GDP) SIPRI rating for major armed conﬂicts

−6.79 (−1.80)∗ −0.08 (−0.51) 0.01 0.24

0.79 (2.30)∗∗ −0.67 (−2.58)∗∗ 0.44 126 0.74

Model 1b 1.12 (2.79)∗∗∗ 1.56 (0.57) 0.34 127 0.96

White’s heteroscedastic consistent t-statistics are in parentheses; (***), (**), and (*) denote signiﬁcance at the 1%, 5%, and 10% levels, respectively. a The p-values refer to the test of overidentifying restrictions implied by the exogeneity of instruments. The instruments used are: corruption, inﬂation volatility, and all the exogenous variables in the system (i.e. all variables except for the three dependent variables and the ratio of total investment to GDP).

200 Sanjeev Gupta et al. spending on education and health – at least when measured as a per cent of GDP. However, since conﬂict is associated with lower real GDP growth, the result is lower growth in real per capita government spending on education and health during conﬂict periods. Not surprisingly, the data are consistent with an increase in the share of investment in GDP in the immediate postconﬂict period, and in the share of private sector investment. The available data also show a dramatic pickup in inﬂation during the conﬂict period, followed by a signiﬁcant decline in the immediate postconﬂict period. The results suggest sizeable economic gains in terms of economic growth, macroeconomic stability, and the generation of tax revenues to support povertyreducing spending, for countries that end conﬂicts and tackle terrorism. Ending violence and restoring security can be expected to lower the share of the budgetallocated to military spending. These results conﬁrm those of earlier studies, underscoring the potential for the ‘peace dividend’ to contribute to economic development. For example, a recent study by Hess (2003) ﬁnds that the pure economic welfare losses from conﬂict are quite large. The authors estimate that these losses are typically four times larger than the welfare costs of business cycles as calculated by Lucas (1987), and that, on average, individuals would give up over 6 per cent of their current annual level of consumption as a one-time payment in order to live in a world of perpetual peace. Successful reconstruction after conﬂict involves rebuilding damaged institutions and infrastructure, renewing the social contract, generating a sense of trust among the warring parties, and ensuring that grievances due to economic disparities or perceived biases in ﬁscal policies are addressed. All this takes time. The continued involvement (and not just one-shot assistance) of the donors and the international community is therefore critical, especially in countries that have experienced prolonged conﬂicts.16 International institutions (such as the IMF) have been involved in lending for reconstruction to postconﬂict countries. As part of its emergency assistance facility to help members emerging from conﬂicts rebuild capacity and recover economic stability, the IMF, for example, has provided 300 million USD over the period 1995–2000 to seven postconﬂict countries. The ﬁndings of this paper have implications for the design of macroeconomic and ﬁscal policies for countries emerging from conﬂicts. In particular, the results suggest that conﬂict- and terrorism-affected countries are likely to experience a pickup in government tax revenues and a reduction in military spending (albeit with a lag) following the cessation of violence, and this would help in restoring macroeconomic stability.

Appendix Sample countries For the preconﬂict, conﬂict, and postconﬂict analysis, a sample of 20 countries (22 episodes of major armed conﬂicts based on SIPRI data) where conﬂict began or was ongoing after 1985, but ended by 1999, is used. The sample

Fiscal consequences of armed conﬂict and terrorism

201

includes 15 low-income countries (Armenia, Azerbaijan, Bangladesh, Cambodia, Chad, the Republic of Congo, Georgia, GuineaBissau, the Lao People’s Democratic Republic, Mozambique, Nicaragua, Senegal, Tajikistan, Uganda, and the Republic of Yemen), 3 lower-middle-income countries (Albania, El Salvador, and Guatemala) and 2 upper-middle-income countries (Croatia and Lebanon). Where ICRG country ratings on internal conﬂict are also available for the corresponding episodes (for 14 of the 20 countries), there is a broad match between low ICRG ratings of 8 or less (the lower the ICRG rating, the higher the risk of internal conﬂict) and countries that have been classiﬁed as conﬂict-affected by SIPRI. On the 012 ICRG scale, 0 denotes Very High Risk of Conﬂict and 12 denotes Very Low Risk. For example, Liberia had an average ICRG (Internal Conﬂict) rating of 2.1 between 1990 and 1994.The average ICRG internal conﬂict score (where available) for these 20 countries is 3.7 between 1984 and 1989, 6.4 between 1990 and 1994, and 8.2 between 1995 and 1999. This is a reﬂection of the fact that in most of these 20 countries, the conﬂicts took place mainly during the 1980s (or before) and during the ﬁrst half of the 1990s. For the econometric analysis, a larger set of 66 countries including conﬂict and nonconﬂict, low- and middle-income countries is used (see list on the following page). Of these 66, the following countries – Armenia, Azerbaijan, Croatia, El Salvador, Guatemala, Mozambique, Nicaragua, Senegal, and Uganda – were classiﬁed by SIPRI as countries experiencing major armed conﬂicts. Problems of data availability, which are particularly severe for countries affected by armed conﬂict, constrained the sample considerably. Sample of countries used in the econometric analysis 1 Albania 2 Angola 3 Argentina 4 Armenia 5 Azerbaijan 6 Belarus 7 Bolivia 8 Brazil 9 Bulgaria 10 Cameroon 11 Chile 12 Colombia 13 Côte d’Ivoire 14 Croatia 15 Czech Republic 16 Ecuador 17 Egypt 18 El Salvador 19 Estonia

20 Ethiopia 21 Gabon 22 Gambia, The 23 Guatemala 24 Honduras 25 Hungary 26 India 27 Indonesia 28 Iran 29 Jamaica 30 Jordan 31 Kazakhstan 32 Kenya 33 Latvia 34 Lithuania 35 Madagascar 36 Malawi 37 Malaysia 38 Mali

39 Mexico 40 Moldova 41 Morocco 42 Mozambique 43 Nicaragua 44 Niger 45 Nigeria 46 Oman 47 Pakistan 48 Paraguay 49 Philippines 50 Poland 51 Romania 52 Saudi Arabia 53 Senegal 54 Slovak Republic 55 South Africa 56 Sri Lanka 57 Syrian Arab Republic

202 Sanjeev Gupta et al. 58 Tanzania 59 Thailand 60 Tunisia

61 Turkey 62 Uganda 63 Uruguay

64 Venezuela 65 Zambia 66 Zimbabwe

Endnotes [*] Reprinted from the European Journal of Political Economy, Vol. 20, Sanjeev Gupta, Benedict Clements, Rina Bhattacharya, and Shamit Chakravarti, ‘Fiscal Consequences of Armed Conﬂict and Terrorism in Low- and Middle-Income Countries’, © 2003, with permission from Elsevier. The authors would like to thank Emanuele Baldacci, Hamid Davoodi, Stefano Fassina, Hong-Sang Jung, Mansoob Murshed, Erwin Tiongson, an anonymous referee, and participants of the DIW workshop for useful comments and suggestions. 1 The Stockholm International Peace Research Institute (SIPRI) publishes a yearly review of armaments, disarmament, and international security. A major armed conﬂict is deﬁned in the SIPRI Yearbook 2000 as ‘a contested incompatibility that concerns government and/or territory over which the use of armed force between the military forces of two parties, of which at least one is the government of a state, has resulted in at least 1000 battle-related deaths over the duration of the conﬂict’. 2 Data on terrorist activities and casualties are drawn from a report prepared by the US Department of State (2002). There is no consensus regarding how terrorism should be deﬁned. Title 22 of the United States Code, Section 2656f(d) deﬁnes terrorism as ‘premeditated, politically motivated violence perpetrated against noncombatant targets by sub-national groups or clandestine agents, usually intended to inﬂuence an audience’. The Columbia Encyclopedia, 6th Edition, 2001, deﬁnes terrorism as ‘the threat or use of violence, often against the civilian population to achieve political ends. Terrorism involves activities such as assassinations, bombings, random killings, hijackings, and skyjackings. It is used for political, not military purposes, and by groups too weak to mount open assaults’. 3 Over and above the economic costs, prolonged armed conﬂicts can impose signiﬁcant social and political costs that are difﬁcult to estimate. For example, it is not possible to quantify the intangible costs of violence and insecurity, the human suffering and trauma, the breakdown in law and order, the animosity and mistrust that are created among warring parties, and the adverse effects of the reduced stock of health and education endowments on the long-run growth prospects of a country. 4 In Sri Lanka, for example, between 1983 and 1996, defence spending increased from 1.4 per cent to 6 per cent as a share of GDP, and from 4.4 per cent to 21.6 per cent as a share of total government spending (Arunatilake et al., 2001). 5 This effect is likely to be highly nonlinear: up to a certain basic level of spending on defence, there is a positive impact on savings and investment, but after this threshold is passed, higher government spending on defence is unlikely to promote further private sector savings and investment. 6 Conﬂict and violence can itself be affected by the perceived inequities in the distribution of the tax burden and in the pattern of public spending (Addison and Murshed, 2001). 7 Because of the problems of deﬁning terrorism and of the sensitivity involved in classifying countries as victims or as perpetrators of terrorism, the preconﬂict, conﬂict and postconﬂict analysis is restricted only to countries that have experienced armed conﬂicts as deﬁned by SIPRI. See Appendix. 8 See footnote 1 for the deﬁnition of armed conﬂict used in this paper. The Appendix lists the sample countries for this as well as for the subsequent econometric analysis. 9 Unlike SIPRI, the Heidelberg Institute does not consider a cut-off level of 1,000 conﬂict-related deaths to classify a country as being affected by conﬂict. It deﬁnes

Fiscal consequences of armed conﬂict and terrorism

10

11

12

13 14 15

16

203

conﬂict broadly as ‘the clashing of overlapping interests (positional differences) around national values and issues (independence, self-determination, borders and territory, access to or distribution of domestic or international power); the conﬂict has to be of some duration and magnitude of at least two parties (states, groups of states, organizations, or organized groups) that are determined to pursue their interests and win their case’. Since more than 90 per cent of all major armed conﬂicts since 1990 have been internal (SIPRI Annual Yearbooks), only the ICRG internal conﬂict rating is used in the econometric estimation. The ICRG ratings are compiled by a US-based consultancy service, the Political Risk Services Group. Details are available via the Internet: http://www.prsgroup.com/index.html Where ICRG internal conﬂict ratings are available for the corresponding episodes of the conﬂict, preconﬂict, and postconﬂict analysis (for 14 of the 20 countries), there is a broad match between low ICRG ratings (of about 8 or less) and countries that have been classiﬁed as conﬂict-affected by SIPRI and HIIK. The average ICRG internal conﬂict score for these 14 countries is 3.7 between 1984 and 1989, 6.4 between 1990 and 1994, and 8.2 between 1995 and 1999. Grants, on average, are much lower than revenue. For example, for a sample of 31 low-income countries with programs supported by the IMF since 1999, grants were only 3.5 per cent of GDP, compared with revenue of about 18 per cent of GDP (Gupta et al., 2002). See, for example, Brunetti and Weder (1998), who ﬁnd that economic volatility and corruption are detrimental to investment. For the sake of brevity, the estimates of the time dummy and regional dummy coefﬁcients are not presented in Table 11.2. Collier and Hoefﬂer (2002b) ﬁnd, based on data for the period 1960–1999, that military expenditure by a country is strongly inﬂuenced by the level of military expenditure of its neighbours. They estimate that an initial exogenous increase in military expenditure by one country is more than doubled in both the originating country and its neighbours. Potentially, there is an offsetting public good effect if rebellions are deterred by military expenditure. However, instrumenting for military expenditure, Collier and Hoefﬂer ﬁnd no deterrence effect of military spending on the risk of internal conﬂict. Hence, there appears to be no regional public good effect offsetting the public bad arising from a neighbourhood arms race. This is emphasized by Collier and Hoefﬂer (2002a), who ﬁnd that during the ﬁrst three postconﬂict years, absorptive capacity on average is no greater than normal, but that in the rest of the ﬁrst postconﬂict decade, it is approximately double its normal level. Thus, ideally, aid and donor involvement should be phased over several years following the end of the conﬂict. Collier and Hoefﬂer ﬁnd that historically, aid has not been higher on average in postconﬂict societies, and indeed it has tended to taper off over the course of the decade following the cessation of conﬂict.

References Abadie, A. and Gardeazabal, J. (2001) ‘The economic costs of conﬂict: a case study of the Basque country’, American Economic Review, 93: 113–32. Addison, T. and Murshed, S.M. (2001) ‘The ﬁscal dimensions of conﬂict and reconstruction’, WIDER Discussion Paper No. 2001/49, World Institute for Development Economics Research, Helsinki: United Nations University. Addison, T., Chowdhury, A.R. and Murshed, S.M. (2002) ‘By how much does conﬂict reduce ﬁnancial development?’, WIDER Discussion Paper No. 2002/48. World Institute for Development Economic Research, Helsinki: United Nations University.

204 Sanjeev Gupta et al. Alesina, A. and Perotti, R. (1993) ‘Income distribution, political instability, and investment’, NBER Working Paper No. 4486, Cambridge, MA: National Bureau of Economic Research. —— (1996) ‘Income distribution, political instability, and investment’, European Economic Review, 40: 1203–28. Alesina, A., Ozler, S., Roubini, N. and Swagel, P. (1996) ‘Political instability and economic growth’, Journal of Economic Growth, 1: 189–212. Arora, V. and Bayoumi, T. (1993) Economic beneﬁts of reducing military expenditure, Annex II in World Economic Outlook, Washington, DC: International Monetary Fund. Arunatilake, N., Jayasuriya, S. and Kelegama, S. (2001) ‘The economic cost of the war in Sri Lanka’, Research Studies: Macroeconomic Policy and Planning, Series No. 13, Colombo: Institute of Policy Studies of Sri Lanka. Bahl, R.W. (1971) ‘A regression approach to tax effort and tax ratio analysis’, IMF Staff Papers, 18: 570–612. Barro, R.J. (1991) ‘Economic growth in a cross section of countries’, Quarterly Journal of Economics, 106: 407–43. Bayoumi, T., Hewitt, D. and Schiff, J. (1993) ‘Economic consequences of lower military spending: some simulation results’, IMF Working Paper No. 93/17, Washington, D.C: International Monetary Fund. Benoit, E. (1978) ‘Growth and defense in developing countries’, Economic Development and Cultural Change, 26: 271–80. Brunetti, A. and Weder, B. (1998) ‘Investment and institutional uncertainty: a comparative study of different uncertainty measures’, Weltwirtschaftliches Archiv/Review of World Economics, 134: 513–33. Collier, P. and Hoefﬂer, A. (2002a) ‘Aid, policy and growth in post-conﬂict economies’, paper presented at a Joint World Bank-IMF Seminar, Washington, DC Collier, P. and Hoefﬂer, A. (2002b) ‘Military expenditures: threats, aid, and arms races’, paper presented at a Joint World Bank-IMF Seminar, Washington, DC Davoodi, H., Clements, B., Debaere, P. and Schiff, J. (2001) ‘Military spending, the peace dividend, and ﬁscal adjustment’, IMF Staff Papers 48 (2), 290–316. Available at www.imf.org/External/Pubs/FT/staffp/2001/01/index.htm. Drakos, K. and Kutan, A.M. (2001) ‘Regional effects of terrorism on tourism: Evidence from three Mediterranean countries’, Center for European Integration Studies (ZEI), Working Paper No. 26, Bonn: Rheinische Friedrich-Wilhelms-Universität. Easterly, W. and Levine, R. (1997) ‘Africa’s growth tragedy: policies and ethnic divisions’, Quarterly Journal of Economics, 112: 1203–50. Ebrill, L., Keen, M., Bodin, J.P. and Summers, V. (2001) The Modern VAT, Washington, DC: International Monetary Fund. Enders, W., Sandler, T. and Parise, G.F. (1992) ‘An econometric analysis of the impact of terrorism on tourism’, Kyklos, 45: 531–54. Enders, W. and Sandler, T. (1991) ‘Causality between transnational terrorism and tourism: the case of Spain’, Terrorism, 14: 49–58. Gleditsch, N.P., Strand, H., Eriksson, M., Sollenberg, M. and Wallensteen, P. (2001) ‘Armed conﬂict 1946–99: a new dataset’, paper presented at the Euroconference on: Identifying Wars: Systematic Conﬂict Research and Its Utility in Conﬂict Resolution and Prevention. Uppsala University, Uppsala, Sweden. Gupta, S., de Mello, L. and Sharan, R. (2001) ‘Corruption and military spending’, European Journal of Political Economy, 17: 748–77.

Fiscal consequences of armed conﬂict and terrorism

205

Gupta, S., Plant, M., Dorsey, T. and Clements, B. (2002) ‘Is the PRGF living up to expectations?’, Finance and Development, 39: 17–20. Hess, G.D. (2003) ‘The economic welfare cost of conﬂict: an empirical assessment’, Working Paper, Department of Economics, Claremont McKenna College, CA. Knight, M., Loayza, N. and Villanueva, D. (1996) ‘The peace dividend: military spending cuts and economic growth’, IMF Staff Papers, 43: 1–37. Lucas Jr., R.E. (1987) Models of Business Cycles, Oxford: Blackwell. Ndikumana, L. (2001) ‘Fiscal policy, conﬂict, and reconstruction in Burundi and Rwanda’, WIDER Discussion Paper No. 2001/62, World Institute for Development Economics Research, Helsinki: United Nations University. Nitsch, V. and Schumacher, D. (2002) ‘Terrorism and trade’, paper presented at the German Institute for Economic Research (DIW) workshop, The Economic Consequences of Global Terrorism. Available at www.diw.de/deutsch/service/ veranstaltungen/ws_consequences. Richardson Jr., J.M. and Samarasinghe, de A.S.W.R. (1991) ‘Measuring the economic dimensions of Sri Lanka’s ethnic conﬂict’, in Samarasinghe, de A. and Coughlan, R. (eds) Economic Dimensions of Ethnic Conﬂict, New York : St. Martin’s Press. Rodrik, D. (1999) ‘Where did all the growth go? external shocks, social conﬂict, and growth collapses’, Journal of Economic Growth, 4: 385–412. Shieh, J.-Y., Ching-Chong, L. and Wen-Ya, C. (2002) ‘Endogenous growth and defense expenditures: a new explanation of the Benoit hypothesis’, Defense and Peace Economics, 13: 179–86. Stockholm International Peace Research Institute, SIPRI Yearbook, 2001, Armaments, Disarmament and International Security, Oxford: Oxford University Press. Tanzi, V. (1992) ‘Structural factors and tax revenue in developing countries: a decade of evidence’, in Goldin, I. and Winters, L.A. (eds) Open Economies: Structural Adjustment and Agriculture, Cambridge, UK: Cambridge University Press. US Department of State (2002) ‘Patterns of global terrorism–2001’ (Washington, D.C.). Available at www.state.gov/s/ct/rls/pgtrpt/2001/html. Venieris, Y.P. and Gupta, D.K. (1986) ‘Income distribution and sociopolitical instability as determinants of savings: a cross-sectional model’, Journal of Political Economy, 94: 873–83. Walkenhorst, P. and Dihel, N. (2002) ‘Trade impacts of the terrorist attacks of 11 September 2001: a quantitative assessment’, paper presented at the German Institute for Economic Research (DIW) workshop, The Economic Consequences of Global Terrorism. Available at www.diw.de/deutsch/service/veranstaltungen/ws_consequences. World Bank (2000) ‘Economic causes of civil conﬂict and their implications for policy’, press brieﬁng, June 15, Washington, DC Available at www.worldbank. org/html/extdr/extme/pr061500.htm.

Part IV

Anti-terrorist policies

12 Rights and citizenship in a world of global terrorism Dennis C. Mueller

11 September 2001 has been interpreted by many as an attack upon ‘Western civilization’, or upon civilization in general. To protect civilized society steps have been taken that infringe upon the freedoms that many people associate with Western civilization. These range from allowing the police more freedom in citizen surveillance, detention and interrogation, to greater restrictions on immigration and travel. These actions raise fundamental questions about the proper designations of citizen rights in a civilized society, and about the deﬁnition of citizenship itself. These questions form the focal point of this essay. It employs the methodological tools of public choice to analyse the properties of rights and citizenship, and the consequences of global terrorism for the optimal deﬁnitions of rights and citizenship. It also discusses the sense in which 11 September signals a ‘clash of civilizations’.

1.

Introduction

11 September 2001 has been interpreted by many opinion and political leaders in the West as an attack upon ‘our civilization’, or upon civilization in general. To protect our civilized society steps have been taken that infringe upon freedoms that many people associate with Western civilization. These range from allowing police more freedom in citizen surveillance, detention and interrogation, to greater restrictions on immigration and travel. Understandably, the United States has taken the lead in introducing these changes, but similar moves have also been taken in other countries. These actions raise fundamental questions about the proper designations of citizen rights in a civilized society, and about the deﬁnition of citizenship itself. These questions form the focal point of this essay. To address them we must ﬁrst deﬁne what we mean by rights and citizenship, and discuss their roles in a civilized society. To do so, I employ the methodological tools of public choice and constitutional political economy. Namely, I analyse the properties of rights and citizenship for a society of rational self-interested individuals who deﬁne a set of rights and criteria for citizenship as part of a constitution written to advance their collective interests (Sections 2 and 3). Sections 3 and 4 examine the implications of

210 Dennis C. Mueller the rise of terrorism for the choice of deﬁnitions of rights and citizenship. We argue that global terrorism could and indeed should lead to signiﬁcant rethinking of the proper declinations of rights and criteria for citizenship in a democratic society.1 In Section 7, I take up the issue of whether 11 September signals a ‘clash of civilizations’. I argue that it does, but not simply one between Islam and the West, as it has sometimes been characterized, but a more fundamental clash between those individuals who think that society should be organized along principles of rational behavior, and those who reject these principles. Having identiﬁed what I believe to be the most fundamental implications of 11 September I close the essay with some speculations about the future of democratic societies, and indeed of democracy itself, in a world of global terrorism.

2.

The nature of constitutional rights

Imagine a group of people, say occupants of an island, writing a constitution, which deﬁnes a set of democratic institutions under which they shall live. Among the many elements that must go into a constitution is a voting rule to be used for making future collective decisions. Although the unanimity rule would ensure that no future collective action would harm any citizen, the decision-making costs associated with it argue that some qualiﬁed majority rule is likely to be optimal for many decisions. Now consider the decision calculus of an individual at the constitutional convention. He or she must look into the future and envisage all of the possible issues that might come up and then decide on the optimal voting rule for each. Given the uncertainties at the constitutional stage this is an impossible task. It is, however, reasonable to assume that an individual can envisage broad categories of issues and choose a voting rule for deciding these. On any particular issue, the citizen will be on either the winning or the losing side. Let s be her gain if she is on the winning side, and u(s) her utility from this gain. Let t be her loss if she is on the losing side, and v(t) her utility loss. The probability that she is on the winning side, p(m), is an increasing function of the required majority to pass an issue, m, reaching 1.0 under the unanimity rule.2 An individual at the constitutional stage would then maximize her expected utility by balancing the gains from increasing the required majority and thus her chances of being on the winning side against the increased decision-making costs accompanying a rise in m. It is reasonable to assume that these decision-making costs, d(m), not only increase with m, but increase at an increasing rate (d (m) > 0, and d

(m) > 0) . An individual’s expected gain from a future collective decision is then G = p(m)u(s) − [1 − p(m)] v(t) − d(m)

(1)

Maximizing (1) with respect to m yields p (m) [u(s) + v(t)] = d (m)

(2)

Rights and citizenship in a world of global terrorism

211

as a ﬁrst-order condition. The left-hand-side of equation (2) is the marginal gain from increasing the required majority and thereby reducing the chance that one is on the losing side, the right-hand-side is the marginal cost of increasing m and thereby increasing decision-making costs. Different collective decisions will have different gains and loss functions and thus to get more of a handle on which voting rule is optimal for which types of decisions, we need to assume something about these gains and losses. A simple way to approach this issue is to assume that the loss to someone on the losing side is proportional to the gain to a winner, t = bs, b ≥ 0. It is then easy to show that the majority satisfying equation (2), m∗ , increases with b. Some possibilities are illustrated in Figure 12.1. To the left of m = 0.5, the outcome under a qualiﬁed majority rule is undeﬁned, as mutually inconsistent proposals can win, and so the curve has only been drawn starting at m = 0.5. For many decisions the marginal gains lines are likely to resemble g1 and g2 , and the simple majority rule is optimal. For a high value of b, however, the marginal gains curve looks like g3 , and some qualiﬁed majority greater than 0.5 is optimal. When the expected loss to the loser under a collective decision becomes very large relative to the gain to a winner, the marginal gains curve looks like g4 , and the unanimity rule becomes optimal.3 Consider ﬁrst the simple action of wiggling one’s toes. This action carries with it a small gain for the actor, and harms no one else. If the community had to vote on whether a person wishing to undertake this action should be allowed to or not, the simple majority rule would certainly be the optimal rule, and one assumes any proposal to allow someone to wiggle their toes would achieve the required majority. There are an inﬁnite number of such actions giving a small

Figure 12.1 Possible optimal majorities.

212 Dennis C. Mueller utility gain to the actor at no loss to the community (wiggling one’s ears, scratching one’s toe, etc.), however, and thus the transaction costs of voting on all of them would be immense. Citizens at the constitutional convention would minimize future decision-making costs by allowing individuals to undertake any action that is not speciﬁcally prohibited. When an action creates a negative externality, say burning trash, the community will want to be able to prohibit it, and thus the optimal constitution will allow future collective decisions to prohibit certain actions creating negative externalities. Trash burning might conceivably fall into the category of actions for which the simple majority rule is optimal. Now consider the action of practising one’s religion. Religions often require their members to undertake actions that, for whatever reasons, irritate some people in the community – that is, religious practices can be sometimes seen as creating negative externalities. As such, one must anticipate that at some time a majority of the community might choose to prohibit a religious practice of a minority, if this were possible under the simple majority rule. Such a collective action might be expected to impose a large loss in welfare on members of the religious minority, however. If the externality caused by the religious practice were modest, the characteristics of this collective action would ﬁt those for which the unanimity rule is optimal. If all citizens at the constitutional stage perceived the loss from being prevented from practising one’s religion as very large relative to any loss to those experiencing a negative externality from this practice, and they were uncertain over whether they would be in the religious minority subject to a future prohibition, all citizens might well vote to protect the freedom to practise one’s religion by requiring that any prohibitions of religious practices obtain the unanimous support of the community. If those experiencing a loss of utility from a minority’s religious practices were rich enough, they might be able to offer the minority a sufﬁciently large bribe that it would willingly give up the practice in question, and a ban on it would pass even under the unanimity rule. But if the constitution drafters were correct in choosing religious practices as a set of actions to be prohibited only with unanimous agreement, i.e., the loss to those prevented from acting is very large relative to the externality it causes, any bribes offered are unlikely to be large enough to produce unanimity. When placing bans on religious practices under the protection of the unanimity rule, therefore, the constitution framers must anticipate that much time and energy will be wasted in the future over idle debates and votes on proposals to ban certain religious practices that in the end fail to achieve unanimity. Realizing this, the constitution framers can economize on future decision-making costs by placing a right to practise a religion into the constitution, whereby a constitutional right is deﬁned as a prohibition against any person or group of persons – including the entire community – interfering with an individual’s freedom to undertake the protected action. Two features of constitutional rights under this theory need to be noted. First, explicit rights will be deﬁned only for actions capable of generating sufﬁciently strong negative externalities to elicit efforts by some members of the community

Rights and citizenship in a world of global terrorism

213

to restrict them. Even if wiggling one’s toes is expected to give great enjoyment, no constitutional protection in the form of an explicit right to act will be afforded, if it is deemed unlikely that this action will ever generate a negative externality. Many actions that provide considerable beneﬁts for the actor will never be challenged, and need not be protected. Second, there is an inherent tension between constitutional rights and majoritarian democracy. When the institutions of explicitly deﬁned rights and the simple majority rule are both found in a constitution to deal with situations where individual interests conﬂict, these situations should differ dramatically in the perceived losses imposed on the different sides from curtailing the action. The simple majority rule is optimal for resolving a negative externality, when an individual at the constitutional stage expects the utility gain from undertaking the action to equal the loss it causes. Rights are deﬁned precisely where the simple majority rule is not optimal, because the expected gains and losses from a ban are dramatically different, and the constitution framers wish to preclude its use. Because rights will be deﬁned only when signiﬁcant losses are expected for those prevented from acting relative to the losses imposed on others, disputes over rights are likely to be emotionally charged, as they pit a perhaps substantial majority that feels harmed by an action against an intense minority that beneﬁts from it.4

3.

Deﬁning citizenship

The normative justiﬁcation for the state in the public choice literature is to correct market failures and thereby achieve a Pareto-optimal allocation of resources. The properties of a public good require that all members of the community consume the same public good and in the same quantity, and this must of course be true for any public goods provided in our island community. Assume ﬁrst that all residents of the island have identical preferences and incomes. There is but one public good that needs to be provided and the community decides on quantity GI , which maximizes the joint utilities of the islanders. Under the assumptions made, this decision would be reached even under the unanimity rule. Now assume a group from outside of the island sails to its shores and wishes to enter and join the community. The preferences of members of this group differ from those of the indigenous islanders, however, and the newcomers prefer the quantity GN = GI of the public good. If the collective choice process used were such that the participation of the newcomer would result in a compromise choice of public good quantity, GC , the indigenous islanders would be worse off as a result of the entry of the newcomers. If the islanders are self-interested utility maximizers, there are four possible outcomes: (1) (2) (3)

The islanders do not allow the newcomers to enter. The newcomers are allowed to enter but not to participate in the collective choice process. The newcomers are allowed to enter and participate in the collective choice process, but a voting rule is used such that the newcomers cannot affect

214 Dennis C. Mueller

(4)

the outcome of the process, as for example, the simple majority rule, if the newcomers are a minority. The newcomers are allowed to enter and to participate in the collective choice process, and to affect the outcome of the process, because there are compensating advantages to the islanders that offset the loss imposed upon them by the change in the quantity of public good.

The latter possibility might arise for several reasons. For example, the island faces a shortage of workers, or more magnanimously, the newcomers might be asylum seekers on whom the islanders take pity. Similar issues arise when the islanders ﬁrst write their constitution, if we drop the assumption that all islanders have homogeneous preferences. Although the option of not granting one group entry to the island does not exist in this situation, the possibility of limiting citizenship to only certain groups does, and one group might choose to form a polity and exclude the other from having voting rights in it, should the two groups have sufﬁciently heterogeneous preferences. The possibility that both groups obtain citizenship increases, if we assume that constitutional decisions are made from behind the veil of ignorance. This likelihood increases still further if we expand the number of post-constitutional collective choices and allow for more groups, so that every group has a chance of being part of the majority on some future issues. Although such assumptions increase the likelihood of inclusive deﬁnitions of citizenship, they do not guarantee it, when individuals choose deﬁnitions of citizenship that maximize their expected utility. With sufﬁcient preference heterogeneity, the expected utility of an individual at the constitutional will be maximized – even from behind the veil of ignorance – by excluding some groups from affecting future collective choices. This can always be accomplished by excluding these groups from citizenship. When these groups constitute a minority of the population, they can be effectively excluded even when granted citizenship, by choosing the simple majority rule as the community’s voting rule.5

4.

The impact of growing terrorism on the deﬁnitions of rights

The analysis of both rights and citizenship can be regarded as normative in that it describes the constitutional choices that individuals would make when they are uncertain of their future positions. No constitution has ever been literally written by all of the individuals who would live under it, and no participant in a constitution drafting process is totally uncertain about his future position under the constitution or of the positions of those whom he represents. Nevertheless, the long-run nature of a constitution does introduce uncertainty over the consequences of certain of its provisions, and participants in a constitutional convention will be cognizant of the fact that a constitution is something more than a horse trade over a public works budget. If individuals are ever motivated by Rawlsian arguments

Rights and citizenship in a world of global terrorism

215

to place themselves behind a veil of ignorance, they are likely to be so moved at a constitutional convention. Thus, it is reasonable to assume that the analysis of rights and citizenship in Sections 2 and 3 would describe essential features of actual constitutions, whenever they have been written by people who might be uncertain about the consequences of some provisions of the constitution. And I believe it does. Many early colonists in America had ﬂed religious persecution in Europe. Those writing the Constitution in Philadelphia knew that it would be possible for some members of a society to attack others for their religious beliefs, and they were probably unable to predict which religions might become the target of an attack. The protection of religious freedom provided by the US Constitution accords with the predictions of the above analysis. Only a decade before the Constitutional Convention, the United States had been engaged in a war to free itself from what was perceived to have been a tyrannous regime. In the years leading up to the war some leaders of the revolutionary movement had been jailed for considerable time without being charged with a speciﬁc crime. Those writing and ratifying the US Constitution might well have feared that a future government might incarcerate members of political movements without due cause.6 The protection against such actions afforded by the US Constitution also accords with the predictions of the above analysis. Additional evidence supporting the theory’s predictions is provided by the most glaring civil liberty not protected in the US Constitution. There were no slaves represented in Philadelphia, and no one present at the Convention had any fears that he or any of his decedents might someday be slaves. Slave owners were represented. Thus, the constitution of a people who had fought a long and bloody war to free themselves from tyranny not only did not prohibit slavery, it actually contained provisions to protect it. These examples from the US Constitution underscore the importance of uncertainty over future position to the delineation of a right in a constitution. Examples from elsewhere illustrate the importance of the other features of rights contained in the theory. There is no writ of habeas corpus protection in Israel, and people in Israel are sometimes imprisoned for prolonged periods without being charged with a speciﬁc crime. Why? The most obvious answer is, of course, that the Jews and Arabs in Israel have never gathered together to write a constitution. I suspect, however, that even if they were to do so they would not be able to agree on a constitutional provision providing writ of habeas corpus protection. First of all, it would be extremely difﬁcult for representatives of the Jewish and Arab communities to step behind the veil of ignorance and imagine that they are members of the other community. I would argue further, however, that even if members of both communities could step behind the veil of ignorance, they would not include habeas corpus protection in the constitution. Such rights protection should be included, if the costs imposed on someone prevented from acting are very large relative to any costs imposed on others from the action. The costs imposed upon someone imprisoned for a long period can be assumed to be large. The costs on the rest of the community of letting

216 Dennis C. Mueller this person go free, if he has not committed a crime, are that he might do so if free. If the probability of this happening is small, and the likely crime that he would commit is not serious, the conditions for a constitutional right affording habeas corpus protection are met. The history of violence and terrorism in Israel suggests, however, that the probabilities that members of certain groups at certain points of time will commit crimes are not low, and the crimes that they commit are very serious. Under these conditions a constitutional right to be set free if not charged with a speciﬁc crime within a short period of time may not maximize the expected welfare of the community, even if all members consider this right from behind a veil of ignorance. I would not expect a constitution written in Israel or Northern Ireland to contain such a right, even when written under the best of circumstances imaginable. A right to free speech can be defended under the assumption that the loss to a person exercising such a right would be extremely large relative to any negative externalities this action imposed on the rest of the community. In Germany and Austria fascist political parties are banned, and a public ﬁgure can be prosecuted for making statements that are interpreted as pro-fascism. Even an American lawyer would not bother to sue a public ﬁgure in the US for making similar statements, because she would know that they are fully protected by the First Amendment. Why are the constitutions of the seemingly liberal democracies of Germany and Austria so unprotective of free speech rights in this area? The answer is obvious. The costs on society from people espousing fascism are perceived in these countries to be so large that no rights are afforded them. 11 September changed the perceptions of many Americans, and non-Americans, of the potential external costs of certain actions. Should the member of a terrorist organization caught in a crowded building with dynamite tied to his body be afforded the right to remain silent? Should he be allowed to have the best lawyer that money can buy, if his terrorist organization is rich from illegal activities and donations from those who espouse its cause? How long should the police be allowed to detain a suspected terrorist without charging him with a crime, and what means should they be allowed to use when interrogating him? These are difﬁcult questions and I shall offer no answers. What I wish to point out is merely that the answers that were deemed optimal (right) before 11 September may no longer be correct. Many people think of rights as being bestowed upon us by nature or God, as being absolute. One cannot compromise when it comes to ‘the rights of man’. No tradeoffs are allowed. These are not the implications of the theory of rights sketched above. One of the most important lessons economics has to teach is that virtually all decisions involve tradeoffs, and this is an important lesson of my political economy theory of rights. Rational individuals choosing a set of rights to include in their constitution would weigh the likely beneﬁts to individuals guaranteed the freedom to act against the possible costs imposed on others of these actions. Tradeoffs are inherent in the choice of rights. 11 September has changed the dimensions of these tradeoffs, it may call for changes in the deﬁnitions of rights.

Rights and citizenship in a world of global terrorism

5.

217

The impact of growing terrorism on the criteria for citizenship

Rights are of importance only in communities with heterogeneous interests. A right to practise one’s religion is unnecessary in a community with a single religion. Heterogeneities also ﬁgure prominently in a community’s decision to admit various groups as citizens. In addition to not having any slaves in attendance at the Philadelphia Convention, it failed to invite any representatives of the indigenous ‘Indian’ population. Nor where members of these communities immediately granted citizenship and attendant rights to vote. The cultural gap between those living in America of European ancestry and the native Americans was too great at the founding of the republic for anyone to think of a common citizenship for the two groups. The United States was founded by Europeans and has had a liberal policy toward immigrants throughout most of its history. And this policy has served the country well. The United States has been able to skim off the most adventurous and industrious people from other countries for over two centuries, and this ‘immigration policy’ has certainly contributed to its great economic success. When one sees how successful the American ‘melting pot’ has been in assimilating immigrants from diverse cultures, one might question the relevance of the discussion in Section 3 of the potential costs to a community from admitting people with ‘heterogeneous preferences’. By and large preference, or perhaps more aptly, cultural heterogeneity has not been a problem in the United States, because most immigrants have chosen to leave their countries and their cultures behind and adopt that of the United States. Signiﬁcant cultural heterogeneity was a transitory phenomenon. The two major exceptions to this happy scenario are the blacks, who did not voluntarily go to the United States in pursuit of ‘the American dream’, and the native Americans who, perhaps out of resentment from having their land stolen from them, do not for the most part seem to have shared this dream with the immigrants. Once-white Britain and France have admitted large numbers of people from their former empires and have created the same problems for themselves that the United States created because of slavery. Both countries, like the United States, have large black ghettos with high poverty, unemployment and crime rates. I know of no society with a sizeable black minority where the blacks have been successfully integrated in the sense that they have the same average incomes, occupy the same positions of authority, etc., as members of the white (non black) community. Why this is so I do not know, but it does seem to be a fact.7 In addition to race, language differences have often been a source of social stratiﬁcation and unrest as, of course, have also religious differences. The United States and many other countries stand as outstanding examples of the great beneﬁts that a country can reap from immigrants, when they are successfully assimilated. The United States and several other countries also bear witness to the heavy costs that can be incurred when minority groups remain unassimilated. As with rights

218 Dennis C. Mueller there are tradeoffs to be faced, when a country admits as residents or citizens people with a different cultural background. 11 September has again changed the perceived magnitudes involved in this tradeoff. A common pattern for taking up citizenship in another country is ﬁrst to arrive as a student, worker or even tourist, later apply for permanent residence, and eventually for citizenship. The host country has time to evaluate the desirability of granting residency/citizenship, the applicant has time to decide whether she wants residency or citizenship in this country. The terrorists who carried out the attacks on 11 September came to the United States on student and tourist visas – some came to learn how to pilot a plane. A reevaluation of policies with respect to granting such visas is an obvious response to the attacks. American-born Timothy McVeigh did not kill as many people as died on 11 September but he managed to kill quite a few. McVeigh reminds us that hatred for a country’s people and institutions can be home-grown. Hatred on the order of that exhibited on 11 September seems more likely to develop outside of the United States than within in it, however, and much more likely to develop in countries that do not share its culture and prosperity than in those that do. The logical response to 11 September is a selective tightening of restrictions on travel and immigration. Although I have referred mostly to the United States, since it was the target of the 11 September attacks, these attacks confront all developed democracies with similar tradeoffs, for the attacks appear to be directed not only at the United States, but also at its way of life – or so it has been claimed.

6.

A clash of civilizations?

Some people have interpreted 11 September as conﬁrming Samuel Huntington’s forewarning of a clash between Islam and Western culture, others have gone to pains to say that this is not the case. I too see 11 September as signalling an attack on Western institutions, but see it as both broader and more fundamental than just between Islam and the West, for it comes from both outside and within the countries that share ‘Western culture’. Consider the following places around the globe that have ﬁlled the headlines over the recent past – Northern Ireland, Kosovo, Macedonia, Israel, India/Pakistan, Afghanistan and, of course, Manhattan and Washington on 11 September. What do these trouble spots around the world have in common? Religion is an important cause for each conﬂict. Islam has been a component of several conﬂict situations, but not all, and even where it has been present the conﬂict has not always been between Islam and ‘the West’. Conﬂicts over religion are not new. Indeed, parts of Europe were involved in religious wars of one form or another for much of the second millennium. Putting an end to religious wars – with the important exception of Northern Ireland – must be regarded as one of the great achievements of Western civilization over the past millennium. The course of events that led to this triumph of reason over madness might be traced to the awakening of intellectual thought that started in the Renaissance.8

Rights and citizenship in a world of global terrorism

219

Probably no single sentence in all philosophy is better known than Descartes’ ‘I think, therefore, I am’ (Eaton, 1927, p. xxv). In placing the individual and his thought process at the center of his proof of God’s existence, Descartes embraced the revolutionary idea that evidence of God’s existence was not to be found in declarations of the church, or the king, nor by consulting a holy book or a holy man, but in the reasoning process of the individual contemplating God’s existence before the ﬁreplace in the solitude of his study. If a thinking man could construct a logical argument for God’s existence that convinced himself, then God existed. No higher authority was needed than the rational logic of an individual’s own mind. Descartes helped to advance the transformation in Western thought that began in the Renaissance and culminated in the Enlightenment, a transformation that saw the individual move to centre-stage in social and philosophical analysis, and which manifested increased conﬁdence in the individual’s power to reason. Democracy was also rediscovered during Europe’s intellectual reawakening in Renaissance Italy. It was the thinking of the Enlightenment, however, with its optimistic trust in human reasoning that provided democracy its intellectual raison d’être, a trust that was most eloquently expressed in the language of the United States Constitution. A similar evolutionary process was set into motion during the Renaissance that led to the justiﬁcation of Western society’s other great institutional pillar, capitalism. From the somewhat mundane but nevertheless revolutionary invention of double-entry book-keeping, through the rejection of the belief that charging interest is sinful, a series of important victories for rational thinking over superstition and ignorance can be traced reaching an optimistic apex again at the peak of the Enlightenment with the publication of Adam Smith’s Wealth of Nations, with its optimistic faith in man’s capability to ‘truck and barter’ and the invisible hand of market competition that guided these activities. One of the most important and certainly earliest contributions to public choice was the Marquis de Condorcet’s ‘jury theorem’.9 Written at the peak of the Enlightenment it is at once a justiﬁcation for both the use of the simple majority rule to make collective decisions and for democracy itself. The theorem concerns a community’s choice between two alternatives, one is correct (true) and the other is not. The example Condorcet ﬁrst used involved the guilt of a person accused of a crime, hence the theorem’s name. All citizens are assumed to have the same goal – to make the correct decision – and each has an independent probability, 0.5 < p < 1.0, of making the correct choice. The theorem states that the probability that the community makes the correct choice, if it votes using the simple majority rule, increases with the size of the community approaching certainty in the limit. Consider now the import of each of the theorem’s assumptions. The community shares a common goal. No single person ‘knows the correct answer’, nevertheless, the right answer can be found in the collective judgements of the community. Note here the importance of the assumption that the probability of any person being correct exceeds one-half. Citizens do not merely ﬂip coins to decide how

220 Dennis C. Mueller to vote, they are assumed to invest enough time and effort studying the question to increase their probability of being correct to above one-half. Moreover, the theorem implicitly contains the optimistic assessment of citizens’ capabilities, that if they do invest time trying to determine the correct answer, they will increase their probability of being correct. Thus, this normative case for democracy rests on the assumptions that we routinely make in public choice: people engage in a collective decision process to advance their common interests, and they behave rationally.10 It is sobering in this regard to recall that Condorcet was put to death by individuals who placed less stock in rational behavior than he. Similar assumptions underlie a normative defence of capitalism. The consumer must be deemed sufﬁciently self-interested and intelligent to be trusted to make his or her own consumption, savings and work choices.11 11 September calls these assumptions about individuals into question. Piloting an airplane into a building is difﬁcult to reconcile with our usual notions of rational, self-interested behavior. One way to reconcile the two is to assume a strong belief in a life after death and a heavy weight being placed on the enjoyment experienced in the afterlife. But such strong religious beliefs are themselves incompatible with democracy and the assumptions that justify its use. Descartes began his celebrated demonstration of God’s existence with the words, ‘I think’; the Apostles’ Creed begins with the words ‘I believe’. Religion substitutes belief for thought, and this makes all religions fundamentally at odds with those core Western values that place trust in man’s ability to reason, and underpin the institutions of democracy and free markets. The prescriptions of various religions have often been at odds with the principles of market exchange, of course. What is perhaps less widely appreciated is that religious beliefs are also often fundamentally at odds with the basic principles underlying democratic institutions. If there are members of the community who represent God, and claim to speak with God, why consult the average citizen to determine the correct action for the community? How can citizen or consumer sovereignty take precedence over the will of God? For the person fully committed to a religious faith, theocracy must be preferred to democracy. Thus, 11 September does symbolize a fundamental and signiﬁcant ‘clash’, but it is not simply a clash between the United States or the West and Islam. It is a clash between those people and those countries that are oriented to improving the welfare of individuals today, and place trust in the judgements of individuals as expressed in the market place and democratic institutions to achieve this goal, and those people who place their trust in a God and are willing to sacriﬁce the welfare and even the lives of individuals today to comply with God’s will. Fundamentalist Islam represents one component of this latter group, but it is not the only component. Fundamentalist Christianity, fundamentalist Hinduism and any other fundamentalist ideology that makes the individual and his or her welfare secondary to some overarching ideological goal are equally serious challenges. Groups espousing such fundamentalist beliefs exist both inside and outside the family of Western countries.

Rights and citizenship in a world of global terrorism

7.

221

Conclusions

A survey of educated Saudis aged 25 to 41, conducted in October of 2001, found that 95 per cent of them supported Osama bin Laden’s cause.12 Scary. What is to be done? In 1996 I proposed that everyone in a democracy should be required to pass a test demonstrating that they know something about its democratic institutions and to take an oath that they support the constitution before being allowed to vote (Mueller, 1996, Ch. 20). The purpose of this proposal was to elevate the act of voting from a right to a privilege, and to reduce the negative externalities imposed on the community by the participation in the political process of those who did not take the time to become informed or did not support the community’s basic democratic institutions. At the time this seemed to some as a radical proposal. Today it perhaps seems less radical. 11 September has taught us that the actions of other members of the community can impose signiﬁcant costs on us. Indeed, today it is obvious that much more must be done to protect our democratic institutions. Each citizen must be educated and taught to think and behave rationally, and brought to understand the principles underlying our democratic and market institutions. If another country educates its people to oppose democratic and capitalist institutions, and hate the countries of the West, there is little we in the West can do, other than protect ourselves against those who hold these beliefs. On the other hand, much can be done with the people born within our own countries. Inculcating the ‘values of the community’ has always been thought to be an important justiﬁcation for state-provided education. 11 September has both underscored the importance of this goal, and highlighted just what values need to inculcated. Here, Europe is in a much better position than the United States. The notion that the state should provide education and instill ‘community values’ into its citizens is more widely shared in Europe than in the United States. Recent trends in the United States have seen parents removing their children from the public schools and placing them into private ones, and even removing them entirely from schools to be educated at home. While this development has been sparked in part by the poor job American schools do teaching reading, writing and arithmetic, quite often the parents’ goal is to inculcate fundamentalist Christian values into their children. Despite the constitutional separation of church and state, these values are also taking over in some publicly funded school systems. In a growing number of states the scientiﬁc knowledge accumulated over the last 150 years concerning the origin of the universe and the evolution of life on earth is being taken out of public school curricula and replaced by stories invented some three millennia ago to account for the same phenomena.13 The rise of religious fundamentalism in the United States is so ﬁrmly integrated into the political process, that acts of violence by religious fanatics are widely tolerated if not applauded. Why is the woman who waits to ambush a doctor as he emerges from an abortion clinic not just as much of a terrorist as the perpetrators of the attacks on 11 September? Why is she not recognized as a threat to civil society and democratic order? How sadly ironic it

222 Dennis C. Mueller was to see America’s political leaders rushing to church immediately following 11 September. Fortunately, religious fundamentalism has yet to take hold in Europe. But one should not forget that it has not been that long since many Europeans fell prey to the secular fundamentalism of fascism and communism, and many Europeans still fall all too readily for the latest ideological fad. What differentiates a farmer’s blowing up a McDonald’s outlet in protest against ‘American capitalism’ from 11 September other than the scale of the damage? That this man becomes a national hero instead of being immediately imprisoned as a terrorist should give every European who cherishes democracy and civil society pause for thought. Man’s ability to think and reason is obviously what sets him apart from the other animal species. The democratic, legal and economic institutions that we associate with Western societies have brought us such economic wealth that for the ﬁrst time in mankind’s history old age and obesity are more of a problem than the reverse. The democratic and legal institutions created in Europe appear ﬁnally to have brought an end to its wars. These triumphs are the culmination of a long struggle in which those who spoke for progress and rational thought have always been attacked and resisted by those defending ignorance and the status quo. The attacks of 11 September remind us that the struggle is not over. There continue to be those both within and outside of the West who wish to impose their beliefs on others, and are willing to kill if need be to do so. To respond to this challenge we need to strengthen both our democratic institutions and our citizens’ understanding and support for these institutions.

Endnotes 1 Since my focus is on democratic institutions and the implications of terrorism for these institutions, I shall for most of this essay use the narrower expression ‘democratic society’ instead of the more grandiose and ambiguous one of ‘civilized society’. 2 Speciﬁcally, I assume p (m) > 0, p

(m) < 0, and (m = 1) ←→ (p(m) = 1). 3 This discussion reproduces in a slightly different way the classic treatment of the choice of a voting rule by Buchanan and Tullock (1962, pp. 63–91). See also, Mueller (2001). 4 For further discussion of these issues, see either Mueller (1991) or Mueller (1996, Ch. 14). 5 For further discussion of these issues, see Mueller (2002) and Mueller (1996, Ch. 20). 6 The protection of these rights was added after the Philadelphia Convention and thus reﬂects the perceived uncertainties of not only some of its members, but also of those not represented at the Convention, most importantly the Anti-federalists. 7 I used to think that Cuba was an exception to this generalization until a Cuban-American friend of mine pointed out that the Cuban army ﬁghting in Angola consisted of European-decendent ofﬁcers leading mostly black ground troops. I have been told that Brazil is an exception, and confess not to know enough about Brazil to say whether this is true or not. 8 The following arguments were ﬁrst put forward in Mueller (2000). 9 For a proof and discussion, see Young (1997). 10 It also rests on two additional assumptions that have often been contested in the modern literature – that people become informed about the issues and that the preference aggregation process produces ‘the correct answer’, or a Pareto-optimum. For recent

Rights and citizenship in a world of global terrorism

223

justiﬁcations for this optimism by leading public choice scholars, see Wittman (1995) and Breton (1996). 11 The classic statement of this optimistic assessment of consumer sovereignty is still Friedman (1962). 12 Economist (February 2, 2002, p. 15) reporting on an article in the New York Times. 13 ‘USA: Neuer Kreationisten-Vorstoß’, Der Standard, 12 Feb, 2002, p. 35

References Breton, A. (1996) Competitive Governments, Cambridge: Cambridge University Press. Buchanan, J.M. and Tullock, G. (1962) The Calculus of Consent, Ann Arbor: University of Michigan Press. Condorcet, M. de (1785) ‘Essai sur l’application de l’analyse á la probabilité des décisions rendues á la probabilité des voix. De l’imprimerie royale’, Paris; English translation: ‘Essay on the Application of Mathematics to the Theory of Decision Making’, in: Baker, K. (ed.) (1976) Condorcet, Selected Writings, Indianapolis: Boobs-Merrill. Eaton, R.M. (ed.) (1927) Descartes Selections, New York: Charles Scribner’s Sons. Friedman, M. (1962) Capitalism and Freedom, Chicago: University of Chicago Press. Mueller, D.C. (1991) ‘Constitutional rights’, Journal of Law, Economics, and Organization, 7: 313–33. —— (1996) Constitutional Democracy, Oxford/New York: Oxford University Press. —— (2000) ‘Capitalism, democracy and rational individual behavior’, Journal of Evolutionary Economics, 10: 67–82. —— (2001) ‘The importance of uncertainty in a two-stage theory of constitutions’, Public Choice, 108: 223–58. —— (2002) ‘Deﬁning citizenship’, Theoretical Inquiries in Law, 3: 151–66. Wittman, D. (1995) The Myth of Democratic Failure: Why Political Institutions are Efﬁcient, Chicago: University of Chicago Press. Young, H.P. (1997) ‘Group choice and individual judgments’, in Mueller, D.C. (ed.) Perspectives on Public Choice, pp. 181–200, Cambridge: Cambridge University Press.

13 Decentralization as a response to terror Bruno S. Frey and Simon Luechinger

Anti-terrorism policy concentrates almost exclusively on deterrence. It seeks to fend off terrorism by raising the cost of undertaking terrorist acts. This paper suggests an alternative anti-terrorism policy that is based on reducing the expected beneﬁts of undertaking terrorist acts to prospective terrorists rather than raising the costs of doing so. Speciﬁcally, it is argued that strengthening decentralised decision-making in the polity and economy may be an effective antidote against terrorist attacks.

1.

Introduction

Politics focuses almost exclusively on deterrence in its ﬁght against terrorism. Terrorists must be dissuaded from attacking by threats of heavy sanctions and by using police and military forces to ﬁght them. This strategy has also been central in rational choice analyses, most of which start with the model of subjective expected utility maximisation.1 In striking contrast to the prominence given to deterrence, the evaluation of this strategy by many renowned terrorism experts is unfavourable.2 Hoffman (1998, p. 61), for example, claims, that countless times ‘attempts by the ruling regime to deter further violence [. . .] backﬁred catastrophically’. Despite this failure of deterrence, there are no systematic attempts to consider alternative anti-terrorism policies. We suggest an alternative counter-terrorism strategy; the amount of terrorist acts may be diminished by reducing the expected beneﬁts of terrorist acts to the terrorists instead of raising the costs. Speciﬁcally, we suggest that strengthening decentralised decision-making may be an effective antidote against terrorist attacks. We do not think that decentralisation is the only effective strategy, nor that it works in every case. Moreover, as decentralisation as a counter-terrorism strategy has been rarely discussed and has never been applied in reality, our propositions are to some extent tentative. Nevertheless, given the frequently established failure of deterrence, we believe that systematic reasoning about possible alternatives is desirable, if not necessary. In a previous paper, we discuss another alternative to a deterrence policy (Frey and Luechinger, 2003). In contrast to the proposition presented here, in the previous paper we suggest strategies aiming

Decentralization as a response to terror

225

at increasing expected opportunity cost of terrorism, such as principal witness programs, broad access to the media and normal political process as well as reintegration measures. Section 2 presents a simple rational choice model of terrorism and focuses on the marginal beneﬁts and marginal costs of undertaking terrorist acts. Section 3 discusses anti-terrorism policy and Section 4 concludes that decentralization of the polity and the economy is an effective antidote against terrorism.

2.

The calculus of terrorists

Terrorists can be characterised as rational actors who rely on violence or its threatened use to promote their political goals (see e.g. Lichbach, 1987). As discussed in greater detail below, terrorists derive utility and incur costs from undertaking terrorist acts. According to a model of subjective expected utility maximization, the optimal level of terrorist activity from the point of view of prospective terrorists, is where the marginal beneﬁt from undertaking an additional terrorist act equals the marginal cost of doing so. Rational choice theory suggests that this corresponds to the amount of terrorism observed in reality. In Section 3, we look more closely at what determines this equilibrium. Terrorism is often seen as a speciﬁc form of political participation: ‘Terrorists attempt to assert their interests in the complicated process of deciding “who gets what, when, how”, the process of politics’ (Badey, 1998, p. 96). The ultimate aims of terrorism are – among others – the redistribution of power and property rights and the extortion of rents. For this, terrorists seek to attain three main tactical goals, all aimed at imposing the maximum possible cost on the country under attack (see e.g. Schelling, 1991, and Enders and Sandler, 1995): (1)

(2)

(3)

Terrorists seek the attention of the media in order to make their cause more widely known. Laqueur (1977, p. 106) goes so far as to claim that ‘terrorist action is nothing, publicity is everything’. Terrorists seek to destabilize the polity. When the government loses power and, more importantly, when the political system’s legitimacy is eroded, the terrorists’ chances of achieving their goal improve. Terrorists seek to damage the economy. They want to impose material costs on the population in order to force them to comply with their demands. Empirical research has shown that terrorist acts can have substantial negative economic consequences (for a survey see Frey et al. 2006).

In order to achieve these goals, the terrorists undertake various types of attacks (see e.g. Hoffman, 1998). One possibility is a targeted attack, for example the assassination of a powerful political leader. A second possibility is an attack on a target with a high symbolic value. A third possibility is to disperse fear and panic among the population by attacking civilians (seemingly) at random. The antiterrorism policy developed in the next Section concentrates on the ﬁrst two types of terrorist acts. According to Hoffman (1998), left-wing terrorists and, to a lesser

226 Bruno S. Frey and Simon Luechinger extent, ethno-nationalist/separatist groups, rely particularly on the ﬁrst two types of attacks, while religious terrorists are more indiscriminate about attacking civilians. Therefore, the proposed counter-terrorism policy is not an antidote against all terrorist attacks, but against a signiﬁcant number of them. The marginal beneﬁt reﬂects the beneﬁt to terrorists of undertaking additional terrorist acts. It is assumed to diminish with increasing levels of terrorist activity. This is the case if additional terrorist acts have less and less effect on the targeted country and population. The marginal cost, the costs of engaging in an additional terrorist act, increases with increasing levels of terrorist activity, because it is increasingly costly to undertake terrorist acts. To begin with, the easiest and most conveniently available targets are chosen. Thereafter it becomes increasingly difﬁcult to undertake terrorist acts.3 Furthermore, the relationship between marginal cost and the level of terrorist activity depends on the motivation of the terrorists. Highly intrinsically motivated terrorists, or fanatics, are convinced they are doing the right thing, irrespective of incentives from outside.4 In this case, terrorists hardly react to either the costs or rewards of their actions. The available evidence, however, suggests that most prospective terrorists react to external incentives (see e.g. Enders and Sandler, 1995). Both extrinsic and intrinsic motives play a role.

3.

Anti-terrorism policies

Within the framework of the simple model of the marginal costs and marginal beneﬁts of terrorist acts, the options of a counter-terrorism policy are identiﬁed: the marginal costs to potential terrorists to undertake terrorist acts must be raised and the marginal beneﬁts derived by terrorists must be lowered. Deterrence policy aims at raising the cost of terrorist acts by making them more difﬁcult to undertake and by punishing the actors more severely. The exogenous increase in the cost of undertaking terrorist acts results in a smaller equilibrium amount of terrorism. In the (unlikely) event that terrorists are solely intrinsically motivated, they would not respond to negative incentives and a deterrence policy would have less effect. Such a policy is also ineffective if it is offset by rewards. This would be the case if a larger punishment is interpreted as an indication to terrorists that their cause is particularly worthwhile and will be rewarded accordingly in the afterlife. But deterrence policy does not solely depend on the incentives of the potential terrorists. It also seeks to prevent terrorist acts by making them more difﬁcult to undertake. A case in point is tightening up security measures to prevent the hijacking of planes. It follows that deterrence policy in many cases is able to shift the perceived cost of terrorism upwards and to therewith reduce the intensity and number of terrorist acts. This is the reason why deterrence is at the forefront of anti-terrorism policy. As already mentioned, the equilibrium amount of terrorism may also be diminished by reducing the marginal beneﬁts of terrorist acts to the actors. The basic idea

Decentralization as a response to terror

227

is to make terrorism a less attractive option to potential terrorists. This second antiterrorism policy option has been rarely discussed and seldom applied in reality.5 Speciﬁcally, we suggest that an effective way to immunize a country against terrorist attacks is to decentralize activity, both with respect to the polity and the economy. A polity with many different centres of decision-making and implementation is difﬁcult, if not impossible, to destabilize. If one of the centres is hit and destroyed by a terrorist attack, the other centres can take over. That this is indeed possible has been demonstrated by a recent incident in Switzerland. In September 2001, a man ran amok in the parliamentary building of the Swiss canton of Zug and shot dead no less than three of the seven members of the government council, as well as eleven members of parliament. Nevertheless, within a very short period of time, the government was functional again, not least because the heads of the partly autonomous communes took over. A similar incidence in Armenia plunged the country into a political crisis. In October 1999, ﬁve gunmen burst into Armenia’s parliament, assassinating the Prime Minister, Parliamentary Speaker and seven other government ofﬁcials. Armenia’s Defence Minister stated that the situation which had been created was fraught with uncertainty, and that the internal and external security of the state were in danger (Freedom House, 2000). Because of the centralized nature of Armenia, the killings left a power vacuum and lesser federal levels were not able to take over. Decentralizing political power, or polyarchy, takes two forms: (1)

(2)

Political power is distributed between a number of different political actors. The classical division of power between government, legislature and courts, as well as democracy and the rule of law, are the most important. Political power must also be divided up between various levels of government. In federal (i.e. spatially decentralized) countries, there is usually the federal, state/provincial/cantonal, and the communal level. But it is possible to go one step further still by introducing a fourth, regional level, or to grant far-reaching autonomy to all kinds of functional, overlapping and competing jurisdictions (for the idea of FOCJ, see Frey and Eichenberger, 1999).

Strengthening political decentralization via the division of power and federalism contributes considerably to a country being less vulnerable to terrorist attacks. The attraction of such actions for terrorists is diminished. The marginal beneﬁt of terrorism falls and the equilibrium amount of terrorism is reduced. A market economy is based on an extreme form of decentralization of decisionmaking and implementation. The terrorist attacks on the United States in 2001 dramatically demonstrate that economic activity was quickly resumed, despite the high number of people killed and whole ﬁrms being wiped out. This was possible because there were a substantial number of competitors, which could quickly be substituted, rather than just one monopolistic supplier of the respective economic services. The more an economy functions according to market principles, the less vulnerable it is to terrorist attack. The terrorists’ marginal

228 Bruno S. Frey and Simon Luechinger beneﬁts diminish as terrorist activities prove to have little or no effect on aggregate economic activity.

Conclusion As the preceding discussion suggests, strengthening decentralized decisionmaking may be an effective antidote against terrorist attacks. Political economy suggests, however, that a deterrence policy nearly always tends to be favoured by governments, because it demonstrates politicians’ determination to ﬁght terrorism. In contrast, seeing and conserving the value of decentralization has more the character of a public good, whose maintenance is not directly attributed to the government in power. It is therefore all the more important to safeguard political and economic decentralization at the constitutional level.

Acknowledgements We wish to thank Matthias Benz, Reto Jegen, Stephan Meier, Todd Sandler, Alois Stutzer, Friedrich Schneider, Ronald Wintrobe and two anonymous referees for helpful comments.

Endnotes 1 See e.g. Landes (1978), Kirk (1983), Sandler et al. (1983) and Lichbach (1987). Only a small part of the vast literature on terrorism, however, uses a rational choice approach and, even then, usually concentrates on speciﬁc aspects. An overview on important contributions based on economic methods is provided by Enders and Sandler (1995) and Sandler and Enders (2002). For further references, see also Frey and Luechinger (2003), and Frey (2004). 2 In particular, several unintended or even counterproductive consequences of deterrence are discussed in the literature. In the political science literature it is often stressed that government repression of dissidents sparks off feelings of frustration and anger in larger fractions of the population. Deterrence may therefore increase the future mobilization of dissidents (Lichbach, 1987). It has often been argued that terrorists aim at provoking such a repressive overreaction (Lake, 2002). According to Wilkinson (2002), there is abundant evidence that such responses play into the hands of terrorists and become totally counterproductive. Whenever a deterrence policy is directed against a certain type of attack, the terrorists react by substituting this type of attack with another, often more deadly, one (see e.g. Sandler and Enders, 2002). Therefore, anti-terrorist policies are far less effective and more costly than anticipated. Deterrence also has unintended consequences for initially uninvolved countries. A deterrence policy of a potential target has negative externalities for other potential targets because it diverts terrorist attacks to these other targets. Therefore, without cooperation among the potential targets, each target expends too much effort on deterrence (Sandler and Lapan, 1988). 3 In the unlikely cases of diminishing marginal cost or increasing marginal beneﬁt with an increasing level of terrorism, the propositions presented in this paper do not hold if the absolute value of the slope of the marginal cost curve is greater than the absolute value of the slope of the marginal beneﬁt curve (i.e. if δMC/δT < δMB/dT < 0 or 0 < δMB/δT < δMC/δT ). Within a certain range, the marginal beneﬁt curve may slope upwards or the marginal cost curve may slope downwards. The latter is the case if there

Decentralization as a response to terror

229

are economies of scale in planning and executing a certain number of operations (see e.g. Sandler and Enders, 2002). But over the whole range of operations, and in the long run, this is unlikely to be the case. 4 For a precise deﬁnition and further discussion of intrinsic and extrinsic motivation, see e.g. Frey (1997). In the realm of terrorism, such behaviour could also be explained by intertemporal utility considerations. A suicide bomber may place a very high value on the life hereafter and trade off current consumption for the delights of 72 white virgins in the afterlife. 5 Other anti-terrorism policies based on reducing the expected beneﬁts of terrorism are discussed in Frey (1988) and Lapan and Sandler (1988). Frey (1988) suggests an information policy that would sharply reduce terrorists’ beneﬁts in the form of publicity. Lapan and Sandler (1988) analyse the often proposed and applied strategy never to negotiate with terrorists in hostage-taking incidents.

References Badey, T.J. (1998) ‘Deﬁning international terrorism: a pragmatic approach’, Terrorism and Political Violence, 10: 90–107. Enders, W. and Sandler, T. (1995) ‘Terrorism: theory and applications’, in Hartley, K. and Sandler, T. (eds) Handbook of Defense Economics, Vol. 1. Handbooks in Economics, Vol. 12, Amsterdam, New York and Oxford: Elsevier, 213–49. Freedom House (2000) Freedom in the World. The Annual Survey of Political Rights and Civil Liberties, 1999–2000, New York: Freedom House. Frey, B.S. (1988) ‘Fighting political terrorism by refusing recognition’, Journal of Public Policy, 7: 179–88. ——(1997) Not Just for The Money. An Economic Theory of Personal Motivation, Cheltenham, UK, and Brookﬁeld, USA: Edward Elgar. ——(2004) Dealing With Terrorism: Sticks or Carrots?, Cheltenham, UK: Edward Elgar. Frey, B.S. and Eichenberger, R. (1999) The New Democratic Federalism for Europe: Functional Overlapping and Competing Jurisdictions, Cheltenham, UK: Edward Elgar. Frey, B.S. and Luechinger, S. (2003) ‘How to ﬁght terrorism: alternatives to deterrence’, Defence and Peace Economics, 14: 237–49. Frey, B.S., Luechinger, S. and Stutzer, A. (2006) ‘Calculating tradegy: Assessing the costs of terrorism: Journal of Economic Surveys, forthcomming. Hoffman, B. (1998) Inside Terrorism, New York: Colombia University Press. Kirk, R.M. (1983) ‘Political terrorism and the size of government: a positive institutional analysis of violent political activity’, Public Choice, 4: 41–52. Lake, D.A. (2002) ‘Rational extremism: understanding terrorism in the twenty-ﬁrst century’, Dialog-IO (Spring), 15–29. Landes, W.A. (1978) ‘An economic study of US aircraft hijackings, 1961–1976’, Journal of Law and Economics, 21: 1–31. Lapan, H.E. and Sandler, T. (1988) ‘To bargain or not to bargain: that is the question’, American Economic Review, 78: 16–21. Laqueur, W. (1977) Terrorism, Boston: Little Brown. Lichbach, M.I. (1987) ‘Deterrence or escalation? The puzzle of aggregate studies of repression and dissent’, Journal of Conﬂict Resolution, 31: 266–97. Sandler, T. and Enders, W. (2002) ‘An economic perspective on transnational terrorism’, Mimeo, Los Angeles: University of South California.

230 Bruno S. Frey and Simon Luechinger Sandler, T. and Lapan, H.E. (1988) ‘The calculus of dissent: an analysis of terrorists’ choice of targets’, Synthese, 76: 245–61. Sandler, T., Tschirhart, J. and Cauley, J. (1983) ‘A theoretical analysis of transnational terrorism’, American Political Science Review, 77: 36–54. Schelling, T.C. (1991) ‘What purposes can “international terrorism” serve?’, in Frey, R.G. and Morris, C.W. (eds) Violence, Terrorism, and Justice, pp. 18–32, Cambridge: Cambridge University Press. Wilkinson, P. (2002) Terrorism Versus Democracy: The Liberal State Response, London: Frank Cass.

14 Global threats and the domestic struggle for power Michelle R. Garﬁnkel1

This paper considers an economy where groups compete in a contest for power to redistribute future income in their favour. An increased external threat of terrorism – either an increase in the likelihood of a successful terrorist attack or a greater loss of income in the event of a successful attack – would tend to reduce the expected value of the contest prize and thus lessen the severity of the conﬂict at home. However, unless the marginal return from guarding against terrorism is not too large or diminishes at a sufﬁciently fast rate, such a shock could imply, in equilibrium, both a greater sense of security among the groups against external threats and a greater conﬂict between them in the domestic struggle for power.

1.

Introduction

In the past, wartime crises have typically sparked an increased awareness by Americans of their common national identity and a desire to work together on local and national fronts in response.2 Thus, from a historical perspective, the effect of terrorist attacks to evoke solidarity and patriotism, such as that observed in the United States following the events of 11 September 2001, would not appear to be especially unusual.3 But what can economics tell us about the relation between increased threats to national security and domestic politics? Does the outbreak of war necessarily weaken domestic conﬂict? The analysis of this paper aims to address these and related issues, building on a simple, one-nation model of domestic politics. Speciﬁcally, in the spirit of the emerging literature on conﬂict and appropriation, the analysis envisions groups within a single nation as competing in a contest in the current period for power to redistribute future income in their favour.4 An additional layer of conﬂict – namely, global terrorism – motivates their collective action.5 While each group struggles to secure a share of future income for itself, together they can guard against terrorism to protect the income available for everyone in the future.6 The analysis highlights the effect of an increased threat of terrorism on the groups’ overall sense of security and, thus, their expected payoffs from participating in the domestic contest; this effect, in turn, inﬂuences the groups’

232 Michelle R. Garﬁnkel current production, guarding and contest activities. A terrorist attack, if successful, destroys a ﬁxed fraction of the nation’s future income. Otherwise, the nation’s future income is left intact. An increased threat – i.e. either a greater likelihood of a successful terrorist attack or a greater fraction of income destroyed in the event of a successful attack – given the groups’ guarding choices, tends to reduce their overall sense of security, inducing them to discount the contest prize of power by more. But the groups’ guarding decisions are likely to change too. Unless the increased threat of terrorism undermines the marginal effectiveness of guarding, the incentive to guard necessarily increases. Such a response would tend to offset the direct effect of the increased threat on their sense of security. Thus, whether the degree of domestic conﬂict is ampliﬁed or dampened depends on the groups’ collective resolve to restore their overall sense of security. In the context of the model of this paper, provided that the marginal return from guarding is not too large or diminishes at a sufﬁciently fast rate, the equilibrium sense of security would remain lower. The analysis predicts in this case, consistent with the apparent shift in US voters’ attitudes as was documented by the news media in much of the year following the 11 September terrorist attacks, a weakening of the degree of conﬂict within the nation.7 The predicted effect on current income, however, is ambiguous. But, the increased threat need not weaken the degree of conﬂict at home. If the marginal return from guarding is too large or does not diminish at a sufﬁciently fast rate, an increased threat of terrorism could induce a large enough increase in protection to induce groups to increase their valuation of the contest prize. That is to say, the increased external threat could imply, in equilibrium, both a greater sense of security among the groups within this nation against future acts of terrorism and a greater degree of conﬂict between them. In this case, current income necessarily falls. These ﬁndings suggest perhaps that, as America continues its ﬁerce battle against terrorism and the heightened sense of vulnerability dissipates, ‘politics as usual’ will make a much stronger comeback. In what follows, the next section presents the analytical framework: a model of domestic conﬂict, which is essentially a rent-seeking model modiﬁed to allow for exogenous shocks that reduce the total amount of rents to be divided and, at the same time, for the allocation of resources by groups to reduce the likelihood that such shocks occur.8 Section 3 characterizes the solutions for each group’s participation in the domestic conﬂict and their incentives to guard against terrorist attacks. Given that characterization, Section 4 identiﬁes the conditions under which domestic conﬂict would be expected to become more or less severe with an increased external threat of terrorism and examines the corresponding implications for output. Section 5 concludes.

2.

Analytical framework

Consider a two-period economy in which a single consumption good is produced. Normalizing its price at unity, the consumption good is taken as the numeraire.

Global threats and the domestic struggle for power

233

The economy is populated by n + 1 risk-neutral groups, each of size 1, having identical preferences deﬁned over current and expected future consumption.9 The groups, indexed by k, compete in a nonviolent way for power. The beneﬁt of being in the position of power is to be able to extract resources from all others. In particular, the group in power in period t, t = 1, 2 collects a ‘tax’ from the other n groups in that period.10 The identity of the ﬁrst-period group in power is exogenously given. Thus, the analysis considers only that domestic conﬂict which emerges towards the end of the ﬁrst period, resulting possibly in a transfer of power between the ﬁrst and second periods. To ﬁx ideas, suppose that the ﬁrst n groups, k ∈ O = {1, 2, . . . , n}, are those who are not in power in t = 1. The n + 1th group is the group in power in period t = 1 and is indicated by k = I . In the ﬁrst period, whether in or out of power, each group has one unit of time to allocate among three different activities: earning income, contesting political power for the second period, and guarding against terrorist activity. Possibly having a direct impact on second-period consumption by all groups as described in more detail below, terrorist activity reﬂects the second layer of conﬂict in this framework. Suppose that a group k, allocates ek units of effort towards the struggle for political power and gk units of effort towards guarding against external threats. Then that group’s earned income is w(1 − ek − gk ), where w denotes the wage rate. Assume that w is exogenously given and ﬁxed.11 Actual consumption by each group in the ﬁrst period will differ from that group’s earned income due to the redistribution of goods as the group initially in power, k = I , collects a lump sum tax from the other n groups. Let τ denote this tax, which is given exogenously.12 Then, ﬁrst-period consumption by each group not in power initially, k ∈ O, is w(1−ek −gk )−τ ; and, consumption by the incumbent group k = I is w(1 − eI − gI ) + nτ . The analysis assumes that the outcome of the domestic struggle for power in the second period depends on the efforts put forth by each of the groups {ek }k . In particular, the probability that group k wins the contest is πk =

ek /E λek /E

if k ∈ O if k = I

(1)

where E = k∈O ek + λeI and λ ≥ 1, when E > 0; otherwise, πk = 1/(n + 1) for all k. This speciﬁcation of the contest success function, ﬁrst introduced by Tullock (1980), admits the possibility that the group in power in period t = 1 has an advantage relative to the other groups: λ > 1.13 That is, the incumbent group’s efforts are naturally more effective at the margin in manipulating the system to secure political support.14 Since individuals live only two periods, there are no choices to be made in the second period, t = 2.15 Each group earns w implying an aggregate output of (n + 1)w. As in the ﬁrst period, however, this income will be redistributed through taxation. Speciﬁcally, the group in power in the second period collects τ from each

234 Michelle R. Garﬁnkel of the other n groups, giving that group an after-tax income of VI ≡ w + nτ , while leaving each of the other groups with an after-tax income of VO ≡ w −τ . Thus, the expected after-tax income for any group k is the weighted sum, πk VI + (1 − πk )VO or equivalently w − τ + πk (n + 1)τ . But, the acts of terrorism launched against this economy imply that VI and VO represent only potential payoffs. Likewise, the difference between these potential payoffs, VI − VO = (n + 1)τ , represents at most the potential prize of the contest for power. It is realized by the winner in period t = 2 only in the event that all attempted acts of terrorism are thwarted, an event which occurs with probability θ . Thus, the weighted sum, πk VI + (1 − πk )VO , represents the expected payoff, conditional on the terrorists’ failure. In the event that an act of terrorism is successfully launched against this economy, which occurs with probability 1 − θ , a ﬁxed fraction of all income is destroyed. Let this fraction be denoted by ρ ∈ (0, 1]. In this event, the payoff to the group that holds power in period t = 2 is (1−ρ)VI , and that to all others is (1 − ρ)VO , implying a smaller prize to the winner of the contest for power: (1 − ρ)[VI − VO ]. And, the expected second-period payoff for any group k, conditional on a successful attack, is only (1 − ρ) [πk VI + (1 − πk )VO ]. Assume that the probability of avoiding such destruction, given the extent of terrorist activity which is denoted by X > 0, is strictly increasing in the total time n+1 allocated to guarding by all groups in the initial period t = 1, G = k=1 gk , but at a diminishing rate: θ (G, X ) ∈ [0, 1),

(2)

for G ≥ 0, where ∂θ (G, X )/∂G ≡ θG > 0, ∂ 2 θ (G, X )/∂G 2 ≡ θGG < 0, and ∂θ(G, X )/∂X ≡ θX < 0.16 Note that this speciﬁcation does not preclude the possibility that, without protective measures taken against terrorism in period t = 1 (G = 0), the terrorists’ attacks might be entirely undermined with a strictly positive probability; however, eliminating the possibility of terrorist success (ex ante) is not feasible.

3.

Equilibrium analysis

In the ﬁrst period, each group k chooses a labour allocation {ek , gk }k so as to maximize the expected present discounted sum of its consumption over the two periods, Uk : Uk = w(1 − ek − gk ) − (1 − sk )τ + sk nτ + β(G, X , ρ) × [w − τ + πk (n + 1)τ ], where sk =

0 if k ∈ O 1 if k = I

(3)

Global threats and the domestic struggle for power

235

β ∈ (0, 1] denotes the common discount factor; reﬂecting the groups’ common sense of security against terrorism, (G, X , ρ) is given by (G,X ,ρ) ≡ θ (G,X )+[1−θ(G,X )](1−ρ) = 1−ρ[1−θ(G,X )] < 1;

(4)

and, as previously deﬁned, G ≡ k gk . The groups’ optimizing choices are made independently, subject to the contest success function (1) and the guarding technology (2) and given the external threat of terrorism X . Suppose further, as implicitly assumed here, that the inequality constraint, 1 − ek − gk ≥ 0, is not binding for any group k. Participation in the domestic conﬂict Consider ﬁrst each group’s optimal choice of effort allocated to the contest for power, ek , given its choice of guarding, gk , and the choices made by all other groups, {ej , gj }j=k . The payoff maximizing effort levels allocated to this contest, ek , satisfy the following conditions: β(G, X , ρ)

E − ek R−1≤0 E2

λβ(G, X , ρ)

E − λeI R−1≤0 E2

for k ∈ O for k = I ,

(5) (6)

where R ≡ [VI −VO ]/w = (n+1)τ /w denotes the potential prize from the domestic contest normalized by the (constant) marginal opportunity cost of participating in that contest w. For each group not initially in power, k ∈ O, the ﬁrst condition is met as a strict equality when ek > 0. Similarly, the second condition is satisﬁed as a strict equality when eI > 0. Focusing on the quasi-symmetric solution where ek = eo for all k ∈ O, it follows that E = neO + λeI . Provided (G, X , ρ) > 0 given G ≥ 0 and X > 0, the conﬂict technology as speciﬁed in equation (1) implies that E > 0 in equilibrium.17 Combining the optimality conditions for the groups not initially in power, (5) for k ∈ O each a as strict equality, with the optimality condition for the incumbent, (6) for k = I also as a strict equality, yields the following solutions: ∗ = eO

β(G, X , ρ)λnR (1 + λn)2

(7)

eI∗ =

β(G, X , ρ)nR [1 + (λ − 1)n] (1 + λn)2

(8)

∗ E ∗ = neO + λeI∗ =

β(G, X , ρ)λnR , 1 + λn

(9)

with (G, X , ρ) as deﬁned in (4) given G and X . As revealed by a close inspection of these solutions, if either n = 1 so that there are only two groups in total or λ = 1

236 Michelle R. Garﬁnkel so that there is no incumbency advantage, all groups allocate the same effort to ∗ = e∗ . But, only if λ = 1, will all groups have an equal the contest for power: eO I chance of success in that contest. If neither of these conditions is satisﬁed (n > 1 and λ > 1), then those groups not in power will devote less labour resources to ∗ < e∗ . the domestic conﬂict than the incumbent group: eO I ∗ ∗ When evaluated at eO , eI given G as well as X , equation (1) for k = I equals the equilibrium degree of political stability, πI∗ – i.e. the probability that the one group holding power in the ﬁrst period k = I will continue to hold power in the second: πI∗ =

λeI∗ n . =1− ∗ ∗ λeI + neO 1 + λn

(10)

This probability is at least as large as the probability that another given group k ∈ O will seize the position of power, or πO∗ =

∗ eO 1 ∗ = 1 + λn , λeI∗ + neO

(11)

since λ ≥ 1. Henceforth, the analysis assumes that the group initially in power has a strictly positive advantage in the domestic contest (λ > 1), implying that πI∗ > 1/(n + 1) > πO∗ . In addition, from equation (10), one can see that the equilibrium degree of political stability, πI∗ , is decreasing in the number of groups k ∈ O, n, and is increasing in the incumbent group’s advantage, λ > 1, but is independent of the normalized potential prize, R. Furthermore, the equilibrium degree of political stability is independent of the external threat of terrorism, as reﬂected in (G, X , ρ). Regardless of its effect on the equilibrium allocation of resources to protect the economy G, terrorism inﬂuences each group’s labour allocation to the domestic conﬂict proportionately so as to leave the equilibrium degree of political stability, πI∗ , unchanged.18

4.

Equilibrium guarding against terrorist attacks

To proceed, now consider the groups’ labour allocation to guarding. Each group’s choice maximizes its expected payoff (3) subject to (1) and (2) with (4), given X and the equilibrium winning probabilities πI∗ and πO∗ . The ﬁrst-order condition to this maximization problem for each group implies

1 − 1 ≤ 0, (12) k (G) = βρθG (G, X ) 1 + R πk∗ − n+1 for k = 1, 2, . . . , n + 1, where as previously deﬁned G ≡ k gk and R = (n + 1)τ /w; for k = I , πI∗ is given by equation (10); and, for k ∈ O, πO∗ is given by (11). For each group k, when k (G −k ) > 0, where G −k ≡ j=k gj , given gj , j = k, the condition in (12) must be satisﬁed as a strict equality and gk∗ > 0.

Global threats and the domestic struggle for power

237

In what follows, the analysis focuses on the quasi-symmetric solution wherein all groups k ∈ O choose the same guarding allocation. Let that optimizing allocation be indicated by gO∗ . Thus, G ∗ = ngO∗ + gI∗ . However, under the maintained assumption that the incumbent group has a relative advantage in the contest for power (λ > 1), the n + 1 conditions in (12) cannot all be satisﬁed as strict equalities. That is, gk > 0 is not possible for all k. But guarding against terrorist attacks is not ruled out entirely. Since πI∗ > πO∗ , the quasi-symmetry assumption implies that G ∗ > 0 holds if and only if gI∗ > 019 . Speciﬁcally, we have gO∗ = 0 and G ∗ = gI∗

= >

0

if

k (0)

≤ >

0.

(13)

When the marginal return from guarding (θG ), the discount factor (β) and/or the potential damage from terrorist attacks (ρ) are relatively small such that I (0) ≤ 0, not even the group with the greatest potential interest in the protection of the future tax base against terrorist attacks (k = I ) ﬁnds guarding to be optimal. In this case, no resources are allocated to guarding in equilibrium: gI∗ = G ∗ = 0. But, otherwise, only the group in power in the ﬁrst period (k = I ) allocates some resources to guarding. In this case, where I (0) > 0, the equilibrium value of aggregate guarding is implicitly deﬁned by (12) for k = I as a strict equality: G ∗ = gI∗ = gI∗ (X , ρ, πI∗ ).20 Then, using (7)–(9), the equilibrium allotment to the contest for power by groups k ∈ O can be written as a function of X , ρ and πI∗ . Aggregating over the ∗ = ne∗ (X , ρ, π ∗ ). Similarly, we can write the incumbent groups, we have neO O O group’s equilibrium allotment to the contest for power as eI∗ = eI∗ (X , ρ, πI∗ ). The total allocation to production by all groups is simply the residual: n + 1 − ∗ − e∗ − g ∗ , which is strictly positive under the maintained assumption that the neO I I inequality constraint, 1 − ek∗ − gk∗ ≥ 0, is not binding for any group.

5.

Threats of terrorism and domestic conﬂict

The analysis now turns to study the effects of global threats of terrorism on domestic conﬂict. Although threats of terrorism have no direct effect on the equilibrium degree of political stability πI∗ , the solutions shown in (7)–(9) reveal that they do inﬂuence the groups’ incentive to participate in the domestic struggle for power through their effect on the groups’ sense of security, (G, X , ρ). Indeed, an increased threat of terrorism, such as that perceived following the 11 September 2001 attacks, could be represented by •

•

a positive shock to ρ, or an increase in the fraction of income destroyed in the event of a successful attack given θ (G, X ), as the horriﬁc events on that day shattered the belief that (transnational) terrorists could do relatively little harm within the borders of the United States; or a positive shock to X , or an increase in the likelihood of a successful terrorist attack θ(G, X ), given G, as the events themselves and the information

238 Michelle R. Garﬁnkel that quickly surfaced thereafter revealed a greater resolve, cohesion and overall strength of the terrorist organization than had been previously perceived by most. Focusing on the total amount of labour allocated to the contest for power, or ∗ + e∗ , as the relevant measure of the severity of the conﬂict at home,21 F ∗ ≡ neO I the potential inﬂuence of such shocks is formally represented as dF ∗ βnR 1+ = 1 + λn dρ βnR dF ∗ = 1+ dX 1 + λn

n(λ − 1) d(G ∗ , X , ρ) . 1 + λn dρ n(λ − 1) d(G ∗ , X , ρ) . 1 + λn dX

(14) (15)

Hence, to identify the effects of an increased threat of terrorism on the severity of conﬂict at home, we need only evaluate the sign of the derivatives d(G ∗ , X , ρ)/dρ and d(G ∗ , X , ρ)/dX . From (4), given the amount of guarding, G, either sort of shock would reduce the sense of security among the groups against terrorism () and, thus, would imply more discounting of the future prize from the domestic contest for power: d(G ∗ , X , ρ) = − 1 − θ (G ∗ , X ) < 0 dρ dG=0 ∗ d(G , X , ρ) = ρθX < 0. dX dG=0

Of course, any perceived change in the threat of terrorism would be expected to inﬂuence the groups’ incentive to guard. Factoring in these indirect effects, we have dG ∗ d(G ∗ , X , ρ) = − 1 − θ(G ∗ , X ) + ρθG dρ dρ ∗ dG d(G ∗ , X , ρ) . = ρ θX + θ G dX dX

(16) (17)

The two subsections that follow analyse the implications of shocks to ρ and X , in turn, considering both the indirect and direct effects and identifying the conditions under which shocks to each would lessen the severity of conﬂict at home.22 When the potential destruction from terrorism increases Consider ﬁrst the effect of an increase in the fraction of income destroyed in the event of a successful terrorist attack, ρ. Suppose initially that gI∗ = G ∗ > 0.

Global threats and the domestic struggle for power

239

Then, applying the implicit function theorem to (12) for k = I with equation (2), the second-order condition and the envelope condition shows dG ∗ θG > 0. =− dρ θGG ρ

(18)

Since the marginal return to guarding is diminishing (θGG < 0) by assumption, the sign of this expression is determined by the sign of the numerator, which represents the marginal effect of guarding on the likelihood of thwarting all terrorist attacks. By assumption, this effect is positive. An exogenous increase in the potential damage caused by a successful terrorist attack (as measured by ρ) induces an increase in guarding, which tends to offset the direct effect of the shock on (G, X , ρ). Combining equations (18) and (16) gives the full effect: θG2 d(G ∗ , X , ρ) ∗ . = − 1 − θ(G , X ) + θGG dρ

(19)

With equation (14), this expression implies Proposition 1. An exogenous increase in the income destroyed in the event of a successful terrorist attack (ρ) reduces the severity of domestic conﬂict (F ∗ ) when d(G ∗ , X , ρ)/dρ < 0, which holds if and only if 1 − θ (G ∗ , X ) > −θG2 /θGG . Although an exogenous increase in the potential harm caused by terrorism would induce more guarding, the condition for that shock to reduce the severity of conﬂict at home, as stated in the proposition, ensures that the increase in guarding is not so large as to cause the groups’ sense of security, (G, X , ρ), to increase on net. This condition is simply that the marginal return from guarding is not too large or that it diminishes at a sufﬁciently fast rate. If, for example, θ(G, X ) = G/(G + X ), an increase in the potential damage caused by terrorism ρ would induce increased protection against external threats, but necessarily would result in more discounting of the prize from the contest for power. While the degree of domestic stability, πI∗ , would be unaffected, the intensity of domestic conﬂict as measured by F ∗ would fall as a result. Alternatively, if θ(G, X ) = G α /(1 + X ) where α ∈ (0, 1), an increase in the potential damage caused by terrorism would amplify the conﬂict between groups, unless α were sufﬁciently small.23 Note that, when the condition stated in Proposition 1 is satisﬁed, the implications for output are generally ambiguous. The amount of labour allocated to the domestic conﬂict falls, while that allocated to guarding rises. However, when that condition is not satisﬁed, the effect on output is clear. Speciﬁcally, since the amount of labour devoted to both the domestic contest for power and guarding rises, the remaining allotment for production necessarily falls.

240 Michelle R. Garﬁnkel When the likelihood of a successful terrorist attack increases Now consider the effect of an exogenous increase in the likelihood of a successful terrorist attack, as measured by an increase in . Assuming initially that gI∗ > 0, an application of the implicit function theorem to (12) for k = I with (2) shows dG ∗ θGX =− 0 dX θGG

if

θGX 0

(20)

Since θGG < 0, equation (20) reveals that the sign of the effect of an exogenous increase in the external threat (as measured by X ) equals the sign of θGX , which could be positive or negative. In the case that an increase in the external threat decreases the marginal return from guarding (θGX < 0), the shock induces less self-protection. In this case, the indirect and direct effects of such a shock on (G ∗ , X , ρ) would reinforce each other. Otherwise, the indirect effect would tend to offset the direct effect. Combining equations (20) and (17) shows θGX d(G ∗ , X , ρ) . = ρ θX − θG dX θGG

(21)

With (15), this expression, in turn, implies Proposition 2. An exogenous increase in the likelihood of a successful terrorist attack (X ) reduces the severity of the conﬂict between groups (F ∗ ) when d(G ∗ , X , ρ)/dX < 0, which holds if either (i) θGX < 0 or (ii) θX < θG θGX /θGG . As stated in this proposition, a positive shock to the likelihood of success of terrorist activity could lower the incentive to guard (θGX < 0), in which case the groups’ sense of security, (G ∗ , X , ρ), would necessary fall, implying a negative effect on domestic conﬂict. In this case, the shock would induce a greater allocation of labour to production in period t = 1. Even if the incentive to guard were to rise (θGX > 0), the effect of the shock on domestic conﬂict would be negative provided again that the marginal return from guarding was not too large or diminished at a sufﬁciently fast rate. In this case, however, the effect on period t = 1 production would be ambiguous. Returning to the examples introduced above, if θ(G, X ) = G/(G + X ), an increase in X could induce more or less guarding, depending on whether initially G was less than or greater than X . By contrast, if θ (G, X ) = G α /(1 + X ) where α ∈ (0, 1), the effect on guarding would unambiguously be negative. But, under either speciﬁcation, the effect on the groups’ overall sense of security would be the same – that is, negative.24 Hence, for both speciﬁcations, the effect of the increased threat of terrorism would be to reduce the severity of the conﬂict at home.

Global threats and the domestic struggle for power

6.

241

Concluding remarks

This paper has examined some of the implications of an increased threat of global terrorism on domestic politics, highlighting the direct and indirect effects of such threats on the nation’s overall sense of security and thus the expected payoffs from participation in a domestic contest for power. That is, an increased threat of terrorism can reduce the gains that special interest groups would expect from rent-seeking activities, and so weaken their incentive to engage in this sort of activity, resulting in less internal conﬂict. But, even in the context of the simple model presented here where the only source of conﬂict concerns the distribution of income, the ﬁnding that domestic conﬂict becomes less severe with an increased threat of terrorism requires that the marginal return from guarding not be too large or that it diminish at a sufﬁciently fast rate. Unless this condition holds, one might reasonably predict the launching of a more extensive campaign against terrorism to increase the sense of security enjoyed by the groups in the economy. While detracting from current production, the increased threat would amplify the struggle for power at home.

Appendix This appendix provides some details concerning the effects of terrorism on the nation’s overall sense of security, (G ∗ , X , ρ) = 1 − ρ[1 − θ (G ∗ , X )], based on two speciﬁcations for the guarding technology, θ(G, X ), mentioned in Section 4. Example 1. Suppose θ(G, X ) = G/(G + X ) where G ≡ k gk . Using (12) with (13) given X , one can ﬁnd the following solution for guarding:

∗

G =

gI∗

0 = √

z1 − X

√ if z1 < X otherwise,

(A1)

where

z1 ≡ ρβX 1 + R πI∗ −

1 n+1

√ √ In turn, this solution (when G ∗ = gI∗ > 0) implies θ (G ∗ , X ) = ( z1 − X )/ z1 . Furthermore, one can easily see, using (18) or simply by differentiating (A1), √ that dG/dρ = 12 z1 /ρ > 0. In addition, we have θG (G ∗ , X ) = X /z1 . These ﬁndings together with (16) yield d(G ∗ , X , ρ)/dρ = − 12 [1 − θ(G ∗ , X )] < 0. √ From either (A1) or (20), one can also verify that dG/dX = 12 z1 − X /X , √ which could be positive or negative, and that θX (G ∗ , X ) = − z1 − X /z1 < 0. Then, with (17), one can verify the following: d(G ∗ , X , ρ)/dX = − 12 ρ[1 − θ(G ∗ , X )]/X < 0.

242 Michelle R. Garﬁnkel

Example 2. Suppose θ(G, X ) = G α /(1+X ) where α ∈ (0, 1) and G ≡ k gk . Then, from (12) and (13) given X , one can ﬁnd the following solution for guarding: G ∗ = gI∗ = z2

1/(1−α)

(A2)

where

z2 = ρβ 1 + R πI∗ −

1 n+1

[1 + X ].

For this example, the inequality constraint for k = I , 1 − eI − gI ≥ 0, is not binding. In addition, one can easily verify, using (A2) or (18), that 1/(1−α) /(1 − α)ρ > 0 and θG (G ∗ , X ) = α/(1 + X )z2 . In turn, dG/dρ = z2 these ﬁndings with (16) imply d(G ∗ , X , ρ)/dρ = −1 + θ(G ∗ , X )/(1 − α), which is negative if and only if α < 1 − θ (G ∗ , X ). Using either (A2) or 1/(1−α) /(1 − α)(1 + X ) < 0 (20), one can also conﬁrm that dG/dX = −z2 ∗ ∗ and θX (G , X ) = −θ(G , X )/(1 + X )< 0. With (17), these results imply d(G ∗ , X , ρ)/dX = −ρθ(G ∗ , X )/(1 − α)(1 + X ) < 0.

Endnotes 1 Department of Economics, University of California-Irvine, Irvine CA 92697-5100, USA, Tel.: +1–949–824–3190; fax: +1–949–824–2182. Email address: mrgarﬁn@ uci.edu. This paper was originally prepared for the DIW Berlin Workshop on the Economic Consequences of Global Terrorism, June 2002 and was subsequently published in the European Journal of Political Economy 20, June 2004. I am grateful to Stergios Skaperdas, the organizers of the workshop and the editors of the special issue of the EJPE, Tilman Brück and Bengt-Arne Wickström, and two anonymous referees for their helpful suggestions. 2 See, in particular, Skocpol (2002) who draws on qualitative evidence in making the relevant comparisons. She argues further, however, that sudden changes in civic attitudes today are not likely to translate into increased civic participation as they had in the distant past, unless the government engages in mass mobilization, since there are relatively few civic institutions currently in place through which any newfound energy can be channelled. 3 As many Americans reached out to help those in need through charitable contributions of their money and time, Democrats and Republicans in Congress alike joined forces with President George W. Bush to wage war against the terrorists in Afghanistan. See, for example, ‘Getting to Grips with Evil,’ The Economist, September 22, 2001, pp. 28–9. 4 This literature, which abandons the traditional paradigm that treats the existence and costless enforcement of property rights as given, can be traced back to the pioneering work of Haavelmo (1954, pp. 91–8). More recent contributions include Hirshleifer (1991, 1995), Skaperdas (1992), and Grossman and Kim (1995). (See Garﬁnkel and Skaperdas (2000) who provide a brief overview.) 5 ‘Terrorism’ in this paper could be interpreted more broadly as an act of war launched against the nation. In any case, it is treated here as exogenous. See Sandler and Enders (2004) who survey the small but growing theoretical and empirical literature on the economics of terrorism.

Global threats and the domestic struggle for power

243

6 In considering two layers of conﬂict of this sort, the present analysis is related to the literature on collective rent-seeking. (See Nitzan (1994) for an excellent survey.) But, in contrast, there is no presumption here that binding commitments (e.g. sharing rules) among the groups within the nation are possible. 7 In the weeks immediately following the attacks, domestic dissent was nowhere to be found in the political arena. Although partisan divisions on issues other than terror remained strongly in place throughout the year following the attacks, voters’ heightened sense of vulnerability to terrorism and their relatively high approval ratings of President George W. Bush appeared to make Democrats and even Republicans reluctant to challenge the President on any issue. See for example, Herbert, Rob, ‘As Bush’s Stature Rises ...’, The New York Times, 31 January, 2002, p. 25; Balz, Dan and Broder, David S. ‘On Issues Except War, Voters Still Disengaged; Politics’ Value Wants for Many After Sept. 11’ The Washington Post, 24 February, 2002, p. 1; and Brownstein, Ronald. ‘The Nation; A Year After; Attacks Produced a Seismic Shift in the Political Agenda’, The Los Angeles Times, 3 September, 2002, p. 1. 8 The model could be applied to analyse a variety of shocks, other than those arising from terrorist activity – e.g. ﬂoods, nuclear power accidents, and global warming. 9 Since the analysis abstracts from issues of collective action at this level, groups could be thought of alternatively as individuals. The qualitative results derived below would follow even if we were to allow for the strategic interaction that naturally arises within groups as well as that which arises between groups (thereby adding a third layer of conﬂict), but at a considerable notational and computational expense. 10 Focused on a somewhat reﬁned form of conﬂict within the nation, the basic model has a ﬂavour much like those in the rent-seeking literature – e.g. Hillman and Riley (1989). In a related analysis of domestic politics, Garﬁnkel (1994) assumes that preferences deﬁned over the composition of peaceful production (private versus public consumption goods) vary across individuals within the nation. But with equal (lump sum) taxation, the conﬂict reﬂected in this variation similarly results in a redistribution of resources. 11 Since wages are independent of k, one might think of the groups as forming along ethnic lines, though their objectives are purely economic. See Robinson (2001) for a very interesting analysis of class and ethnic conﬂict. Groups divide along ethnic lines; but within-group variation in income may, in turn, translate into inequality of income across groups. In this setting, Robinson ﬁnds that ethnic differences might result in more conﬂict than differences in classes alone even when there is social mobility. 12 The analysis assumes a lump sum tax only for convenience. Assuming instead a distortionary tax does not change the qualitative results, but does make the analysis considerably less revealing. 13 See Hirshleifer (1989) who discusses the properties of this speciﬁcation (without an advantage for the incumbent group) and related ones. 14 For example, the group in power might be able to make use of state controlled resources (e.g. the army and police) to prevent others from voicing any opposition. At the same time, the incumbent group might have at its disposal the use of a state-owned broadcasting system to communicate its mission to others. (See Konrad (2002) for an interesting analysis that considers both this and a ‘head-start’ advantage for the incumbent in the struggle for power.) 15 Accordingly, there is no need to introduce additional notation whereby the group in power in period t = 2 can be distinguished from all others. 16 One might suppose that the group in power in the initial period is relatively more effective in guarding against terrorism. In a more fully articulated model, this comparative advantage might reﬂect increasing returns to scale in the technology of defense enjoyed by the state. Or, as suggested by Grossman (2002), it could reﬂect the state’s ability to

244 Michelle R. Garﬁnkel

17

18 19

20

21 22 23 24

‘enforce a collective choice’. As will become obvious below, in this model with lump sum taxes, the incumbent has a greater incentive to contribute to guarding even without any such advantage. Otherwise, given ej = 0 for j = k, any group k could secure a victory in the contest with an inﬁnitesimally small amount of effort. Since no rational, forward-looking group would leave such an opportunity unexploited, ek = 0 for k = 1, 2, . . . , n + 1 cannot be an equilibrium outcome. Hence, political instability (or the frequency with which power changes hands) measured by 1 − πI∗ need not be a good indicator of the severity of the domestic conﬂict or of its implications for resource allocation. When taxes are distortionary, however, the incumbent group need not have a greater incentive to guard, since such a tax increases the relative opportunity cost of this group’s ∗ even when λ > 1. time at the margin, which also raises the possibility that πI∗ < πO Nevertheless, as suggested earlier, the qualitative nature of the central results derived below do not depend on whether taxes are lump sum or distortionary. Note that, consistent with the results in Garﬁnkel (1994), an increase in the degree of political stability πI∗ , caused by exogenous increase in the incumbency advantage, λ, would induce more guarding G ∗ against external threats. As one can easily verify, F = E + (1 − λ)eI . Since, by assumption, λ > 1, F < E. Nonetheless, the signs of dF ∗ /dρ and dF ∗ /dX are identical respectively to those of dE ∗ /dρ and dE ∗ /dX . An appendix provides details under two particular speciﬁcations for the guarding technology, θ (G, X ), which are discussed brieﬂy in the main text. In particular, α < 1 − θ (G ∗ , X ) must hold. See the appendix for details. Again, see the appendix for details.

References Garﬁnkel, Michelle R. (1994) ‘Domestic politics and international conﬂict’, American Economic Review, 84: 1292–1309. Garﬁnkel, Michelle R. and Skaperdas, Stergios (2000) ‘On the consequences of a broader view of self-interest in economics’, American Economist, 40: 5–16. Also reprinted in Szenberg Michael and Ramrattan Lall (eds) New Frontiers in Economics, Cambridge, UK: Cambridge University Press, 2004, pp. 261–80. Grossman, Herschel I. (2002) ‘ “Make us a king”: anarchy, predation, and the state’, European Journal of Political Economy, 18: 31–46. Grossman, Herschel I. and Kim, Minseong (1995) ‘Swords or plowshares? A theory of the security of claims to property’, Journal of Political Economy, 103: 1275–88. Haavelmo, Trygve (1954) A Study in the Theory of Economic Evolution, Amsterdam: North-Holland. Hillman, Arye L. and Riley, John G. (1989) ‘Politically contestable rents and transfers’, Economics and Politics, 1: 17–39. Hirshleifer, Jack (1989) ‘Conﬂict and rent-seeking success functions: ratio vs. difference models of relative success’, Public Choice, 63: 101–12. —— (1991) ‘The paradox of power’, Economics and Politics, 3: 177–200. —— (1995) ‘Anarchy and its breakdown’, Journal of Political Economy, 103: 26–52. Konrad, Kai A. (2002) ‘Investment in the absence of property rights: the role of incumbency advantages’, European Economic Review, 46: 1521–37. Nitzan, Shmuel (1994) ‘Modeling rent seeking contests’, European Journal of Political Economy, 10: 41–60.

Global threats and the domestic struggle for power

245

Robinson, James A. (2001) ‘Social identity, inequality, and conﬂict’, Economics of Governance, 2: 85–99. Sandler, Todd and Enders, Walter (2004) ‘An economic perspective on transnational terrorism’, European Journal of Political Economy, 20: 301–16. Skaperdas, Stergios (1992) ‘Cooperation, conﬂict and power in the absence of property rights’, American Economic Review, 82: 720–39. Skocpol, Theda (2002) ‘Will 9/11 and the war on terror revitalize American civic democracy?’, Political Science and Politics, 35: 537–40. Tullock, Gordon (1980) ‘Efﬁcient rent seeking’, in Buchanan, James M., Tollison, Robert D., Tullock, Gordon (eds), Toward a Theory of the Rent Seeking Society. Texas A&M University Press, College Station, pp. 3–15.

15 Global ﬁnancial information, compliance incentives and terrorist funding Valpy FitzGerald

Interdiction of terrorist funds is a priority for intergovernmental cooperation. However, multilateral and unilateral attempts to ensure timely reporting of transactions made by targeted individuals or groups, and to deny them access to the international ﬁnancial system, have had limited success. This is mainly due to economic disincentives for the disclosure of the identity and purpose of transacting agents, particularly those using correspondent banking services, informal money transfer networks and offshore ﬁnancial centres. This paper proposes that solutions should be based on positive incentives for disclosure, and could include trans-border withholding taxes on transactions with unregulated clients and the provision of affordable transfer systems for emigrants.

1.

Introduction

Since the terrorist attack of 11 September 2001 on the USA, the interdiction of terrorist funds has occupied the attention of international organizations and national governments throughout the world, and has been deﬁned as a central element of the ‘war on terror’. During the succeeding twelve months trans-border cooperation on the provision of ﬁnancial information was strengthened, led by Washington and London in their double role as strategic security allies and regulators of the two major global ﬁnancial centres. In this period, some 166 countries issued orders blocking the terrorist assets, but only 112 million USD of funds were frozen.1 However, closer examination reveals that it proved very difﬁcult to ensure timely reporting of the ﬁnancial transactions of targeted individuals or groups, or to deny them access to the international ﬁnancial system.2 For the purposes of this paper, terrorism is understood to constitute violent actions by non-state groups against civilian targets for the purpose of achieving political ends. Although the important issue of who determines which individuals or groups are to be internationally deﬁned as terrorists lies beyond the scope of this paper, our concern with cross-border terrorism (i.e. acts against the citizens of countries other than those of the terrorist) implicitly excludes self-determination movements concerned with violent conﬂict at the domestic level at least.

Global ﬁnancial information, compliance incentives and terrorist funding

247

Two economic characteristics of cross-border terrorism are important for this paper. First, the need for considerable and continuous funding, and the need to transfer these funds internationally, imposes an effective budget constraint on terrorists – that can reduce the number and intensity of attacks as effectively as direct police action (Sandler et al. 1983). Second, there are considerable externalities arising from the interdiction of international terrorism in any one territory, while the costs are local, so that there are both global gains from international cooperation in this ﬁeld and a temptation – particularly for small states – to engage in free-riding (Lee, 1988). Within the limitations of publicly available information,3 this paper sets out to explore this problem from the viewpoint of the economics of ﬁnancial regulation. The key issue is the geostrategic need for more information on the identity and activities of transactors and the ability to freeze or conﬁscate their assets. However, the regulators’ problem is that the large unregulated ﬂows generated by broadly legal activities such as tax avoidance and migrant remittances provide cover for crime-related transactions (Johnson, 2001). To the extent that the loss of secrecy and increased reporting requirements increases transactions costs, this must be set against the less tangible global welfare gain of increased security. The obvious parallel here is with the logic of the prudential regulation of ﬁnancial markets where in order to prevent systemic risk banks are required to provide proprietary information to the authorities in exchange for the licence to operate (Goodhart et al. 1998). In economic theory, ﬁnancial intermediaries, banks undertake the task of monitoring the ﬁrms to whom they lend on behalf of the households who deposit funds with them (Diamond, 1984). These intermediaries enjoy scale and learning economies in this activity, and possess privileged information about their debtors. In consequence they are in a unique position to provide accurate information on their clients to the regulatory authorities without undue cost. Further, ﬁnancial intermediaries possess considerable information on their creditors, particularly when they provide a wide range of ﬁnancial services and thus build up a picture of clients’ assets and activities. Traditionally, therefore, ﬁnancial intermediaries in OECD countries have worked closely with national law enforcement authorities in the detection of criminal activities – particularly the transfer of criminal proceeds and the evasion of tax obligations. However, over the last decade the competitive forces associated with ﬁnancial globalization have tended to undermine this relationship: on the one hand, banks have increasingly moved away from direct lending towards the management of capital market assets, and so there is much less non-ﬁnancial information available on borrowers; on the other hand, high net worth individuals regard secrecy as a key service, and in the absence of capital controls can move funds rapidly to another jurisdiction if this is not provided. Further, the increased transaction costs implicit in the increasing regulation of international banks make the use of unregulated transfer systems even more attractive. The demand for secrecy ranges from concealing assets from other family members or business associates, through tax evasion to laundering the proceeds

248 Valpy FitzGerald of criminal activities. International law enforcement, however, requires information about depositors and users of ﬁnancial services. This information asymmetry is precisely the inverse of that analysed in the ﬁnancial literature, where it is assumed that the lender has insufﬁcient information to properly monitor the borrower (Hiller, 1997). Further, in the case of interdiction of cross-border terrorist funds, there is the additional complication that the origin of the funds can be fully legitimate (e.g. a scholarship fund) and their use also (e.g. hotel bills), until the attack takes place. This paper is structured as follows. Attempts to detect and interdict illegal ﬁnancial ﬂows related to the funding of trans-border terrorism are built on existing systems of reporting the use of ﬁnancial systems for illegal purposes. This foundation is examined in Section 2, which covers multilateral coordination on ﬁnancial interdiction, the role of unregulated money transfer networks (MTNs) and the problem of offshore ﬁnancial centres (OFCs).4 The limitations of this approach are shown to derive from the bias of economic incentives towards less rather than more disclosure. Section 3 examines the multilateral and unilateral attempts during the twelve months after the attack of September 2001 to strengthen regulatory controls on international banking and to regulate informal transfer systems. This leads to the conclusion in Section 4 that unless market incentives work with – rather than against – regulatory disclosure, there is little prospect of substantial progress in the interdiction of terrorist funds. A possible policy solution could combine trans-border withholding taxes on transactions with unregulated high-wealth clients, supported by the establishment of a bank-based system for low-cost money transfers by emigrants.

2.

Antecedents for the interdiction of ﬁnancial transfers

Mechanisms to interdict terrorist ﬁnancing are inevitably built upon the foundations of those already established for the prevention of ﬁnancial crime including money laundering. Existing multilateral regulation bodies such as the Financial Action Task Force (FATF) of the G7, Interpol and the United Nations have long been involved in attempts to coordinate reactions to all forms of trans-border ﬁnancial crime (Johnson, 2001). In the immediate aftermath of September 2001, it thus appeared that the most effective way to counter this threat was to redirect and intensify the efforts of already-existing multilateral monitoring and coordination bodies. In the ‘formal sector’ of international banking, the weakest link appears to be the essentially self-regulated international network of correspondent banks. International correspondent banking exists in order that banks may provide a wide range of alternate services for their clients in territories where they have no established branches. This in turn makes these formal ﬁnancial institutions vulnerable to unwitting collusion in money laundering activities. This problem is most evident when international correspondent banks engage partners domiciled in poorly regulated

Global ﬁnancial information, compliance incentives and terrorist funding

249

emerging market countries. These arrangements allow the transfer of both illegally and legally derived money from the unregulated into the regulated ﬁnancial institutions, thus ‘allowing funds in through the backdoor of the regulated institutions to commence or continue the laundering process’ (Johnson, 2001, p. 130). For banks, attempts to detect and interdict terrorist ﬁnancing are dealt with through existing mechanisms to counter money laundering. However, while most money laundering funds derive from other criminal offences, such as the sale of illegal drugs, the sources of terrorist funds are frequently legal. The banking sector is aware that any action which would be able to limit terrorist ﬁnancing would require cooperation between banks, between governments and between banks and governments. This cooperation would take the form of information sharing and the creation of new mechanisms to track suspicious behaviour. A group of twelve leading international banks agreed to the Wolfsberg Principles for self-regulation in October 2000.5 These principles were established with assistance from Transparency International (an anti-corruption NGO), and focus on increased due diligence on the part of the banks. In particular, they are expected to be more aware of their customers and their actions. The Wolfsberg Group of banks has extended its principles to meet the new concern of terrorist ﬁnancing, increasing the due diligence expected from its members. Various United Nations agencies were also previously involved in combating ﬁnancial crime in general and terrorist funding in particular. Existing UN resolutions, such as the 1999 International Convention for the Suppression of the Financing of Terrorism that became effective in April 2002, encouraged many states to pass anti-terrorist ﬁnancing legislation. The perception of new threats gave greater importance to the ratiﬁcation of existing legislation. Additionally, the UN Security Council and General Assembly have adopted resolutions calling upon all nations to act to stop terrorism and its supporters in all ways. The most powerful of these resolutions, detailing measures to be taken to suppress the ﬁnancing of terrorist acts, is Security Council Resolution 1373. The UN called for increased cooperation from all member states to stop acts of terrorism and the means by which terrorist actions were supported. Multilateral monitoring and policy proposing organizations such as the FATF were already concerned with ﬁnancial crime, in particular money laundering. The FATF was founded in 1989 to coordinate policies against the misuse of the ﬁnancial system by money laundering.6 Now including 31 member states and several regional afﬁliated bodies, it monitors the actions of member states in implementing anti-money laundering policy and promoting such policies around the world. Its ‘Forty Recommendations’ delineate measures for law enforcement agencies, banks and governments. These recommendations are designed to be widely applied, providing ﬂexibility for individual states to apply them in their own jurisdictions. The FATF recommendations are composed of several sections, including the criminalization of laundering money from crime, the requirement of banks to

250 Valpy FitzGerald know their clients, and reporting any suspicious transactions to their governments. The FATF not only contributes to monitoring but also provides a platform for information sharing among law enforcement and regulatory ofﬁcials from member states and international organizations. At these annual meetings, member regulators discuss changes in money laundering practice and new ways to counteract them. Multilateral bodies such as the FATF, although principally concerned with the integrity of the international banking system, have also expressed considerable concern about global money transfer networks (MTNs), particularly unregulated or ‘informal’ MTNs such as the hawala system. The regulatory problems they raise are complex, both because it is very difﬁcult to trace the activities of these networks accurately, and because it is unclear how to regulate them even in principle. The apparent use of informal MTNs by terrorists has increased the pressure to bring them under regulatory control.7 All informal MTNs share a common set of operational characteristics, a ‘lack of records, customer identiﬁcation or regulatory oversight, and the potential for misuse by criminals’.8 Those most commonly referred to are the Chinese chit or chop system of East and Southeast Asia, the black market peso exchange system of Latin America, and the hawala system, with its offshoot the hundi system, in South Asia. Mirroring established migrant labour patterns, these networks provide unregulated (and thus cheaper) remittance services between developed and developing countries. Unregulated small-scale money transfer networks are also used to transfer funds between commercial parties both within and across national borders.9 The ﬁnal settlement between the associate hawaladars may be made through various mechanisms such as bank transfers, reciprocal remittances, manipulation of trade invoices, smuggling of gold or drugs, or through the physical movement of cash by trusted couriers. Informal MTNs are often held to be ‘archaic’ or ‘underground’ ﬁnancial systems (Gilligan, 2001). However, such networks are better viewed as a structural response to information asymmetries in the ﬁnancial services market place and a necessary feature of segmented ﬁnancial systems (Nawaz et al. 2002). This explains why these networks thrive: they are used predominantly by those consumers, particularly the poor, who are not served by formal ﬁnancial institutions because of cost and risk. Speciﬁcally, they provide a rapid, reliable and relatively cheap means for migrant workers to remit cash to poor and illiterate families. Given that migrant remittances provide the ‘base load’ for such systems, informal MTNs have two further advantages over the formal ﬁnancial system. First, they avoid the additional costs imposed by regulation of banks, for prudential as well as policing purposes. The introduction of international codes and standards by the G7 through the Financial Stability Forum, as well as increased consumer protection, has led to increased costs for banks, which are passed on as transactions fees.10 This makes the international transfer of relatively small sums highly unattractive to banks. Second, hawala networks circumvent the record keeping that formal banking involves. Thus they provide the highly valued ‘ﬁnancial’ service of

Global ﬁnancial information, compliance incentives and terrorist funding

251

customer conﬁdentiality similar to that offered by offshore ﬁnancial centers – particularly for the transactions of wealthy individuals with highly regulated OECD countries. Although hawala networks rely on traditional forms of personal trust rather than written contracts, they also use formal banking systems and continually respond to the ongoing process of ﬁnancial globalization. For example, they are well suited to new e-commerce platforms and form a key element of international ‘cyber-laundering’ business along with activities such as gambling (Philippsohn, 2001). International capital mobility makes criminal and terrorist funds transfer easier, particularly since these transactions can be obscured by the quasi-legal ﬂows related to tax evasion, which require a similar degree of secrecy. The role of the tax factor in determining business location gives rise to wasteful tax competition for investment, particularly between small developing countries with a small business sector of their own, for whom the positive externalities of such investment are a signiﬁcant source of national income, but who do not bear the externalities involved in terms of tax loss to other countries (UNCTAD, 1995). The deregulation of cross-border capital ﬂows also reduces or even eliminates the information about investors generated by licensing systems. Liberalization also reduces the transactions cost of the use of offshore ﬁnancial centres (OFCs) to channel cross-border capital ﬂows through the incorporation of offshore holding companies. The object here is not so much to attract foreign investment as such, but rather the administration of assets and tax revenue. The use of these schemes is detrimental to both the home and host country through reduced tax revenues and distorted investment inﬂows (OECD, 1998). The number of double taxation agreements has thus increased rapidly in recent decades, and there are now some 2,000 such treaties in existence. The principle of information exchange between national tax authorities has been intrinsic to such treaties since the last century; and was central to the draft tax convention prepared by the League of Nations in 1935 (IBFD, 1998). These treaties become ineffective if offshore centres are used as transfer pricing points as well as for tax avoidance (JCT, 1999). The OECD Multilateral Convention on Mutual Administrative Assistance in Tax Matters established in 1988 ensures that information on such holding is shared between OECD members – but not with non-members. Extensive transfer pricing and tax-induced capital restructuring (e.g. intra-ﬁrm debt) take place despite the OECD Model Tax Convention and the OECD Transfer Pricing Guidelines (Bartelsman and Beetsma, 2000). A signiﬁcant problem associated with this issue is that of offshore ﬁnancial centres. By allowing multinationals, through transfer pricing, to declare the majority of their proﬁts in shell companies registered in OFCs, the tax revenues resulting from investment and production in (non-OFC) countries are lost to the hosts. The Financial Stability Forum (FSF, 2000) has addressed the issue of OFCs from the point of view of systemic instability in international capital markets.

252 Valpy FitzGerald Those centres unable or unwilling to adhere to international supervisory standards of cooperation and transparency pose two serious problems. First, there are prudential concerns for the effective supervision of international ﬁnancial intermediaries in order to forestall a bank collapse from putting the payments system in danger. Second, there are market integrity concerns relating to the effectiveness of international enforcement efforts in respect of illicit activity and abusive market behaviour. There are thus substantial global welfare losses from OFCs to be set against the marginal beneﬁts to small economies with no other legal source of high income other than perhaps tourism. These losses are composed of the loss of ﬁscal revenue to other jurisdictions where ﬁrms’ and individuals’ economic activities take place on the one hand, and the cover that tax evasion ﬂows give to money laundering, narcotics funding – and now terrorist ﬁnance – on the other (UN, 1998). IMF research on the prudential issues raised suggests that ‘OFCs attract funds because they promise anonymity and the possibility of tax avoidance or evasion. A high level of bank secrecy is almost invariably used as a selling point by OFCs, some of which have been (and are) exploited also for activities related to money laundering’ (Errico and Musalem, 1999, p. 10). The Financial Action Task Force (OECD, 2000) initially deﬁned some thirtyﬁve jurisdictions as ‘non-cooperative countries or territories’ in key areas of transparency and information exchange required to meet international anti-money laundering standards for – a list which included the Bahamas, Cayman Islands, St. Kitts & Nevis, St Vincent & the Grenadines, Dominica and Panama in the region as well as Israel, Lebanon, Liechtenstein, Philippines and Russia. Six, including Bermuda and the Caymans, rapidly made commitments to eliminate harmful practices by the end of 2005. In the wake of global anti-terrorist measures after September 2001, the remaining Caribbean states and UK dependencies (and the Channel Islands) in the region had signed by February 2002; embracing international standards for transparency, exchange of information and fair tax competition (FSF, 2000). It remained to be seen, however, how effective the information exchange is in practice, because this would mean that OFCs would lose most of their ﬁnancial services income.

3.

Interventions in the international ﬁnancial system during the ﬁrst year of the ‘war on terrorism’

Most of the actions taken in the immediate aftermath of 11 September involved the restructuring of existing bodies to meet this new threat. In October 2001, the Financial Action Task Force expanded its mandate from narcotics-related money laundering to include terrorist ﬁnancing. Its existing mandate apparently made it particularly suitable to prevent the misuses of the world ﬁnancial system by terrorists.

Global ﬁnancial information, compliance incentives and terrorist funding

253

Following this, the FATF issued a series of special recommendations to be adopted by member states. These recommendations drew upon the content of the ‘Forty Recommendations’ but were more closely concerned with terrorist funding. The seven parts of the new FATF guidelines include the criminalization of terrorist funding and connected money laundering, followed by the seizing and conﬁscating of terrorist assets.11 States should also increase cooperation, sharing information and assisting others in bringing ﬁnancial criminals to justice. However, relatively little thought has apparently been given yet to implementation, with decisions instead being left up to individual member states for the time being. The FATF has yet to develop interpretations based on the experience of member states: effectively it is still at the data gathering stage. The FATF recommendations are intended to separate the legal and necessary actions of remittances and the donation of money to non-proﬁt organizations on the one hand from the illegal acts of ﬁnancial crime and support of terrorist activities on the other. Unlike more traditional money laundering, in which crime has already taken place before the money is transferred, the origins of terrorist ﬁnancial support may be completely legal, until they enter terrorist ﬁnancial channels. Informal channels such as hawala are used legally by many to transmit their remittances to their home countries, as well as providing a conduit for terrorist support. As a result, these transactions are both more difﬁcult to track and risk impeding legal ﬁnancial transfers. Since these transfers are by nature informal, there is no ‘paper trail’ for investigators to follow. In an attempt to curtail an activity that seems unregulatable, the FATF recommendations suggest promoting alternative means for transferring money and increasing the number of documents kept on transactions. Additionally, wire transfer operators have been asked to keep additional records of transactions and to attach all information of sender and recipient at all points in the transaction.12 At the same time, the recommendations call upon those who conduct any transactions – including large cash operations such as gambling – to ‘know their customers’ and to attach this identiﬁcation material to the transfer. The lack of this information is seen as suspicious and is liable to be reported – but the effectiveness of this measure remains to be seen. The FATF is a monitoring and coordinating multilateral body, but its means are limited in comparison with the size of its new mandate. Although OECD member countries are strongly recommended to adopt FATF guidelines, it is not a treaty obligation. After a 30 October 2001 statement of the Special Principles on Terrorism, the FATF released documents to help states identify suspicious activities. The ﬁrst step in the FATF monitoring procedure involved a self-assessment questionnaire to all member states, to track their progress in implementing these new guidelines.13 Subsequently, non-member states were requested to complete this questionnaire as well.

254 Valpy FitzGerald This questionnaire is designed not only to determine if states have been implementing the guidelines but also to provide case studies for other states to follow in their policy design. Drawing upon the results of some of the questionnaires, the FATF produced guidance notes to help ﬁnancial institutions to avoid becoming the unwitting conduit for terrorist funds.14 The G7 itself also took action to increase information sharing and cooperation in broader measures to ﬁght ﬁnancial crime. While the impetus for these actions is relatively new, it draws upon previous anti-money lending and anti-drug smuggling agreements. At the G8 meeting of 13–14 May 2002, measures were taken to increase cooperation across borders, particularly in electronic communications and e-commerce, in an attempt to stop terrorists from abusing the system.15 In November 2001 and in response to Security Council Resolution 1373, the wider G20 group of leading developed and developing countries adopted an action plan to increase cooperation among states and with international institutions to adopt measures necessary to limit international terrorist funding. According to this plan, all states would ratify the existing conventions and resolutions dealing with terrorist ﬁnancing; collectively they would draw up lists of offending groups so that their assets could be frozen and information would be shared to make action possible.16 Liaising with the FATF and other international actors, common standards limiting terrorist ﬁnancing were to be implemented. Also, more developed countries promised ﬁnancial support and assistance to those states that lacked the infrastructure to carry out these missions effectively. Aside from multilateral agreements from groups such as the G7, ASEAN and others, multilateral designations of terrorist suspects remain the most noticeable sign of this cooperation. The United States has used the medium of joint declarations of terrorists and terrorist-supporting groups to promote cooperation in this effort.17 Additionally US ofﬁcials pointed to the willingness of other states to issue designations of terrorist-supporting groups on their own and with US support.18 These declarations and the negotiation required to agree upon them, formed a key part of this cooperation. However, although the United States itself did comply with the ﬁrst FATF recommendation – to ratify the 1999 UN Convention on the Suppression of Terrorist Financing – as early as December 2001, only seven other OECD member states had ratiﬁed the Convention by mid-2002.19 The European Union was quick to support the US-led actions to stop the ﬂow of funds that support terrorist activities and issued three separate lists of designated terrorist supporters whose assets should be blocked by end-2001. However, the legal and ﬁnancial infrastructure appears to be inadequate to implement these decisions, and only the US and the UK froze signiﬁcant sums.20 Meanwhile, informal ﬁnancial transactions systems such as hawala networks have been the subject of increased regulation and anti-money laundering declarations. While states such as the UAE did not ban hawala, it was clearly important that they be seen to participate in the ‘coalition of the willing’. In May 2002 at Abu Dhabi,

Global ﬁnancial information, compliance incentives and terrorist funding

255

representatives from regional banking and law enforcement agencies called for the application of the FATF regulations to the hawala industry, declaring that ‘the international community should continue to work individually and collectively to regulate the hawala system for legitimate commerce and to prevent its exploitation or misuse by criminals and others’, and that to this end, they should ‘designate competent supervisors to monitor and enforce the application of the recommendations of the Financial Action Task Force to hawala dealers and other alternative remittance providers’.21 The proposals to regulate hawala stems from the positive aspects of this system and the fear that any attempt to eliminate informal channels would merely drive more transactions underground. In May 2002, the Central Bank of the UAE announced that hawala operators in the UAE would be licensed and regulated.22 Pakistan has launched a scheme to convince its nationals to avoid unofﬁcial channels, including novel incentives such as baggage allowances and preferential treatment at airports.23 The campaign to interdict terrorist ﬁnancing seems to demonstrate the value of intergovernmental cooperation. However, many measures taken were commitments rather than action. While it is too soon to determine how successful the adopted measures have been, success seems limited mainly to unilateral action within the United States and the United Kingdom. To ensure the wholehearted participation of other member countries of the OECD, and of leading emerging market countries, would require that they have a more effective share in decision making by international bodies. To the extent that the proposal for ﬁscal compliance incentives discussed below were to be implemented, this would also provide a stronger economic motive for governments of small or poor countries to participate more effectively in the suppression of terrorism. The US response to terrorist ﬁnancing was probably the swiftest and most wide ranging. The October 2001 Patriot Act includes several amendments to the Bank Secrecy Act and new anti-money laundering measures.24 Within the Patriot Act, Title III is the International Money Laundering Abatement and Anti-Terrorist Financing Act of 2001. The Financial Crimes Section of the Department and the Ofﬁce of Foreign Assets Control of the Treasury is the government agency responsible for determining policy on terrorist ﬁnancing. The American Banking Association (ABA) has cooperated with this agency to introduce new practices to reduce money laundering. In February 2002, the ABA announced new procedures to reduce the threat of terrorists misusing the US banking structures, including new checks for the opening of bank accounts.25 However, the ABA has also challenged certain aspects of the Patriot Act, claiming that new provisions for cooperation between the government and the banks might impinge upon existing agreements of ongoing information sharing.26 They fear that the creation of new channels might circumvent older ones and even make information sharing more difﬁcult, and that the time restrictions on reporting are unrealistic. The US Treasury has devoted substantial resources to waging the ﬁnancial war on terrorism. The initial step was to identify the terrorists and terror-supporting groups and to freeze their assets. The next was the regulation of charities to ensure

256 Valpy FitzGerald that money intended for legitimate purposes was not used to support terrorism. US ofﬁcials leading the ﬁght against terrorist ﬁnancing thus announced a second phase of attack: informal money transfer systems including hawala and charitable donations. The Secretary of the Treasury announced: ‘We are driving the terrorists out of the modern ﬁnancial infrastructure and into less efﬁcient transfer methods, which are more difﬁcult, slower and more easily intercepted by law enforcement’.27 The designation of Islamic charities abroad as supporting terrorists and blocking of their resources has also been a US-led initiative, and include the Saudi decision to require charities to report to the foreign ministry about charities abroad, more regulation and increased transparency of charities in Pakistan and greater government supervision of charities in Egypt.28 The British Government has a substantial history of anti-money laundering and anti-terrorist ﬁnancing actions, but a new terrorist ﬁnance team was set up in November 2001, bringing together different interested agencies.29 The British Government had begun to take actions against terrorist ﬁnancing since 1992, but after 11 September, the assets of 100 organizations and over 200 individuals were frozen. New anti-terrorist legislation increased the powers of the police to monitor and seize assets of those suspected to be engaged in terrorist activities and increased the requirements of banks to report suspicious activities.30 According to its own progress report in April 2002, the UK was already compliant with most of the FATF recommendations due to provisions in the US Criminal Justice Act of 1993, and was seeking to pass legislation that would satisfy the remaining recommendations. The banks and ﬁnancial services industry has also taken the initiative to introduce stricter controls over anti-money laundering systems. In July 2002, the major UK banks announced that they would identify suspicious accounts established since 1994, before international regulations require.31 It would seem therefore, that unilateral intervention has been more effective than multilateral action. On the one hand, as New York and London are the only truly global ﬁnancial centres, to deny their use to terrorist groups is a major achievement. On the other hand, most of the rest of the world – non-OECD countries and offshore centres in particular – are thereby excluded and arguably become more difﬁcult to monitor.

4.

Creating economic incentives for compliance

The evidence presented above demonstrates that systemic features of the international ﬁnancial system make limiting terrorist ﬁnances very difﬁcult. The reliance on existing international ﬁnancial monitoring arrangements seemed logical, but failed to learn from the lessons of past failure. It is clearly already difﬁcult to ensure compliance by non-OECD states, while it has now become necessary to oblige international private actors to comply as well. While states may face international sanctions (political, economic and even military) there is no direct means of punishing private sector actors, as they have no status in international law.

Global ﬁnancial information, compliance incentives and terrorist funding

257

Above all, the proposed interventions failed to create positive economic incentives for compliance in order to counteract the existing disincentives for disclosure experienced by ﬁnancial intermediaries and regulatory jurisdictions. Moreover, even within highly regulated institutional structures, the nature of bank secrecy makes the tracking of transactions difﬁcult. With the growth of the Internet and the increasing impersonalization of the banking industry, banks ﬁnd it more difﬁcult to detect suspicious transactions. For many offshore banking companies, their very comparative advantage is in high secrecy and minimal records. As a result, cooperation faces obstacles from all sides. Indeed, governments are often unwilling to share information with banks, expecting instead that banks will act as a conduit for information to the government. By deﬁnition, attempts to exercise some form of regulatory control over hawala systems imply that these informal networks will become either increasingly formalized or they will cease to exist. Either way, any curtailment of the operations of informal MTNs is likely to have negative collateral impacts, and in particular upon marginalized users of these networks who have little or no access to formal ﬁnancial services. It would not be desirable to close down these systems – even if it were possible – because they help the poor and the small enterprises that provide the bulk of employment. There are two logically preferable solutions: to oblige the formal banking system to provide these services at low cost to migrants; or to extend existing micro-credit systems – which reach these clients at the local level in developing countries – to the international level by recognizing them as correspondent banks. In sum, there is no strong economic incentive for ﬁnancial intermediaries to comply in reporting dubious transactions as the probability of criminal conviction is low and ﬁnes small in relations to the sums involved. There are in fact strong incentives for non-compliance as the costs of reporting on a large number of smaller clients are high, while high-wealth clients value secrecy very highly and would be lost to the more diligent banker. It is thus necessary to internalize the externalities caused by non-compliant institutions and jurisdictions, and make it unproﬁtable for them to handle these transactions – or at least to make it so expensive that the resources available to criminals and terrorists are signiﬁcantly reduced thereby. This deterrence would be reinforced by governments facilitating civil sanctions that can be imposed on ﬁnancial ﬁrms and professionals who fail to exercise (and report) due diligence background checks on their customers – sanctions that in practice appear to have more impact on banking practice than the remote possibility of criminal prosecution. A logical solution to this problem could be to shift from the present ‘blacklist’ system such as that administered by the US Ofﬁce of Foreign Assets Control to an equivalent ‘whitelist’ system based on persons or ﬁrms registered for tax purposes (and thus monitored) in OECD countries and qualiﬁed emerging markets. Transactions with compliant jurisdictions (that is, one sharing full information with OECD member countries) made by persons or ﬁrms not properly registered

258 Valpy FitzGerald for tax purposes would then be subject to a substantial withholding tax.32 This would create a strong incentive against dealing with unregulated agents. This would have three strong positive effects: ﬁrst, it would make handling unregistered funds unproﬁtable – or at worst cost criminals a great deal to transact, reducing their liquidity and eroding their assets. Second, the process of levying the tax would provide a steady information ﬂow on the pattern of payments not now available. Third, such a system would mean effective closure of tax havens and a large increase in ﬁscal resources both for developed countries to balance increased security expenditure and for developing countries to reduce social disparities. The European Community proposal for the taxation of cross-border interest payments, the ‘withholding tax directive’, is a precedent that shows that a measure of this nature is politically and administratively feasible (EC, 1998). The concern of OECD members to coordinate measures against tax evasion, avoidance and competition between themselves – and the resultant pressure on offshore centres – has until now been based mainly on concern for their own tax bases. To extend such cooperative measures towards the main emerging markets would be a logical and technically feasible step, particularly since there are strong economic arguments for imposing transaction withholding taxes on short-term ﬁnancial ﬂows (Zee, 1998). On more general grounds, the leading scholar of international tax economics argues that the time has come for the establishment of a ‘world tax organization’ that would not impose or collect taxes, but rather support national authorities by exercising surveillance on tax systems worldwide; resolving disputes on tax competition and developing codes of conduct for tax administration (Tanzi, 1996). Supporting the interdiction of criminal and terrorist funds would be a powerful additional justiﬁcation for an overdue international reform.

Acknowledgements I am grateful to Rachel Ziemba of St Antony’s College for diligent and expert research assistance, and to the Carnegie Foundation for funding. The comments of participants at the DIW Workshop ‘The Economic Consequences of Global Terrorism’ on 14–15 June 2001 in Berlin, and of the anonymous referees, were extremely useful in preparing the ﬁnal version.

Endnotes 1 US Treasury Secretary Paul O’Neill, Remarks to Asia Paciﬁc Economic Cooperation Forum: Policy Dialogue on Combating Terrorist Financing, Los Cabos, Mexico, 5 September 2002. Available HTTP: . 2 UN Report on Terrorism, 5 Sept. Available HTTP: <www.un.org/terrorism>. 3 These limitations illuminate the implicit contradiction inherent in this process: intelligence work is generally conducted in secret, but to restrict terrorist ﬁnancing the

Global ﬁnancial information, compliance incentives and terrorist funding

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

259

authorities have to inform the public because interdiction is implemented through the market. While almost all states have accepted this responsibility, they lack the resources or the interest for substantial implementation. Thus, the greatest amount of information can be found on the US Treasury’s Financial Crimes website, followed by the UK Treasury. Conversely, other countries, including other members of the G7 such as France, appear much less involved in dissemination. There are of course other ways of transferring terrorist funds, such as smuggling cash or other high unit value items (gems, gold) and more generally any import-export business. More details about the Wolfsberg Group can be found on HTTP: . On the FATF, see HTTP: . ‘La technique de l’hawala, méthode de ﬁnancement de l’internationale terroriste’, Le Monde, 23 September 2001. FATF-XI Report on Money Laundering Typologies 1999–2000, Available HTTP: . In India as much as half of national economic activity is said to make some use of the hawala system – see ‘Trois grands systèmes parallèles de fonds ont été identiﬁés’ Le Monde, 2 October 2001. ‘Suffering under the weight’ The Banker, December 2001. FATF Special guidelines on Terrorism. Available HTTP: . ibid. Self-Assessment Exercise: FATF Special Recommendations on Terrorist Financing. Available HTTP: . Guidance for Financial Institutions in Detecting Terrorist Financing Activities. Available HTTP:. Christopher Chipello, ‘G-8 Agrees to Step Up Cooperation In Fight Against Terror Financing’, Wall Street Journal, 15 May 2002. Available http://online.wsj.com/article/0„SB1021416029274629600,00.html. Action Plan of the G20 on Terrorist Financing. Available HTTP: . Deputy US Treasury Secretary Kenneth Dam ‘The ﬁnancial front of the war on terrorism-the next phase’ speech at the Council on Foreign Relations, New York, 8 June 2002. ‘US terror fund drive stalls’. Available HTTP: . Austria, Canada, France, Iceland, the Netherlands, Spain and the United Kingdom. See OECD Observer, June 2002. Available HTTP: . O’Neill, op.cit. Declaration from International Conference on Hawala, 15–16 May 2002. Available HTTP: . Available HTTP: . Julia Wheeler, ‘Pakistan money-laundering clampdown’, 6 October, 2001, BBC online. Available HTTP: . On the American Banking Association’s efforts to counter money-laundering, see. Available HTTP: . Available HTTP: . Available HTTP: .

260 Valpy FitzGerald 27 O’Neill op.cit. 28 Testimony of Kenneth Dam, Deputy Secretary of the Treasury before the Senate Committee on Banking, Housing and Urban Affairs, Subcommittee on International Trade and Finance, 1 August, 2002. Available http://www.treas.gov/press/ releases/po3315.htm. 29 UK Treasury, International Fight against Terrorist ﬁnancing-UK progress report, 12 April 2002. Available HTTP: . 30 Anti-Terrorism, Crime and Security Act, 2001. Available HTTP: <www.hmso. gov.uk>. 31 Ruth Kelly, MP Speech to the ﬁnancial services authority (FSA) Money laundering Conference, 11 July 2002. Available HTTP: . 32 At, say, 25 per cent of the gross transaction – which is the retention tax levied presently within the EU for fees paid to non-residents.

References Alexander, K. (2001) ‘The need for efﬁcient international ﬁnancial regulation and the role of a global supervisor’, Journal of Money Laundering Control, 5: 52–65. Bartelsman, E.C. and Beetsma, R.M.W.J. (2000) ‘Why pay more? Corporate tax avoidance through transfer pricing in OECD countries’, CEPR Discussion Paper 2543, Centre for Economic Policy Research, London. Diamond, D.W. (1984) ‘Financial intermediation and delegated monitoring’, Review of Economic Studies, 51: 393–414. EC (1998) ‘Proposal for a Council directive to ensure a minimum of effective taxation of savings income in the form of interest payments within the Community (598PC0295)’, European Commission, Brussels. Errico, L. and Musalem, A. (1998) ‘Offshore banking: an analysis of micro- and macro-prudential issues’, IMF Working Paper 99/05, International Monetary Fund, Washington, DC. FitzGerald, E.V.K. (2002) ‘International tax cooperation and capital mobility’, Oxford Development Studies, 30: 251–266 . FSF (2000) ‘Report of the working group on offshore centres’, Financial Stability Forum, Basle. Gilligan, G. (2001) ‘Going underground – the not so new way to bank’, Journal of Financial Crime, 9: 105–108. Goodhart, C., Hartmann, P., Llewellyn, D., Rojas-Suarez, L. and Weisbrod, S. (1998) Financial Regulation, London: Routledge. Hiller, B. (1997) The Economics of Asymmetric Information, Basingstoke: Macmillan. IBFD (1998) Annual Report 1997-1998, Amsterdam: International Bureau of Fiscal Documentation. JCT (1999) ‘Overview of Present-Law Rules and Economic Issues in International Taxation (JCX-13-99)’, Washington: US Congress Joint Committee on Taxation. Johnson, J. (2001) ‘In pursuit of dirty money: identifying weaknesses in the global ﬁnancial system’, Journal of Money Laundering Control, 5: 126–128. Lee, D.R. (1988) ‘Free riding and paid riding in the ﬁght against terrorism’, American Economic Review, 78: 22–26.

Global ﬁnancial information, compliance incentives and terrorist funding

261

Nawaz, S., McKinnon, R. and Webb, R. (2002) ‘The war on South Asia’s hawala networks: ﬁghting ﬁnancial crime or undermining ﬁnancial services for the poor?’ Institute for Development Policy Management, Manchester University, Manchester: OECD (1998) Harmful Tax Competition: an Emerging Global Issue, Organization for Economic Cooperation and Development, Paris. OECD (2000) Towards Global Tax Cooperation: Progress in Identifying and Eliminating Harmful Tax Practices. Organization for Economic Cooperation and Development, Paris. Philippsohn, S. (2001) ‘The dangers of new technology – laundering on the internet’, Journal of Money Laundering Control, 5: 87–85. Sandler, T., Tschirhart, J.T. and Cauley, J. (1983) ‘A theoretical analysis of transnational terrorism’, American Political Science Review, 77: 36–54. Santomero, A. (1984) ‘Modelling the banking ﬁrm: a survey’, Journal of Money, Credit and Banking, 16: 576–602. Stiglitz, J. (2002) ‘Information and the change in the paradigm in economics’, American Economic Review, 92: 460–501. Tanzi, V. (1996) ‘Globalization, tax competition and the future of tax systems’, IMF Working Paper 96/141, International Monetary Fund, Washington, DC. UN (1998) Financial Havens, Banking Secrecy and Money Laundering, UN Ofﬁce for Drug Control and Crime Prevention, Vienna. UNCTAD (1995) Incentives and Foreign Direct Investment, United Nations Conference on Trade and Development, Geneva. Zee, H.H. (1998) ‘Taxation of ﬁnancial capital in a globalised environment: the role of withholding taxes’, National Tax Journal, 51: 587–99.

16 An economic analysis of security policies Tilman Brück

This chapter analyses public policy choices in the security economy from an economic perspective. It discusses the role of public goods for national and global security and identiﬁes the importance of the ﬁrst- and second-order indirect effects of insecurity on economic activity, which include the behavioural responses of agents and the government to security measures, akin to such effects in insurance economics. Furthermore, key public policy trade-offs are outlined, in particular between security and efﬁciency, globalization, equity and freedom. The analysis identiﬁes suitable policy options for raising security in the national and international contexts and in view of these trade-offs. A suitable balance between market and non-market instruments in achieving security should be aimed for to minimize the adverse effects of aiming for higher security. In addition, the public good nature of security implies that international coordination of security policies is important, despite this process being itself fraught with enforcement problems.

1.

Introduction

The world seems to have become less secure since 11 September 2001. A variety of risks are appearing, are being noticed or are being feared more than before the deadly terror attacks on New York and Washington. These risks include the new global terrorism, large-scale electrical black outs, wars in the Middle East, an increase in computer viruses, worms and spam, attacks by snipers, e-commerce fraud, anthrax attacks, petrol strikes, and international ﬁnancial instability. This chapter analyses public policy choices in such an insecure world. The analysis is based on broad deﬁnitions of risk (both private and social risks), security (both national and global) and security policies (both public and private measures). The three objectives of the chapter are to apply economic concepts suitable for the analysis of public policy choices in the security economy, to identify key policy trade-offs in designing security policies, and to discuss suitable policy instruments for attaining security while minimizing these trade-offs, both nationally and internationally. The chapter will demonstrate how national and global public goods are involved in providing security, how security measures can have strong economic effects

An economic analysis of security policies

263

themselves (in addition to the effects of the insecurity itself), that these effects are composed of both ﬁrst- and second-order effects, and that there exist a number of important trade-offs for policy makers in deciding what level of security should be attained and which means should be adopted in attaining them. These trade-offs include the balancing acts between security and efﬁciency of production and trade, between security and freedom and between security and equity. The chapter is structured as follows. Section 2 deﬁnes and characterizes the security economy, discusses security as national and global public goods, introduces the distinction between direct, ﬁrst-order indirect and second-order indirect effects of security polices, and provides an overview over the empirical scale of these indirect effects. Section 3 discusses the trade-offs between security, on the one hand, and non-security spending, efﬁciency, globalization, equity and freedom, on the other hand. Section 4 derives policy implications, including in the areas of competition, regulation and coordination policies, for both national and international policies. Section 5 concludes.

2.

The security economy

This section will ﬁrst deﬁne the security economy before outlining some of the economic effects of insecurity itself and some of the effects of the responses to insecurity. This distinction is important, as the majority of the costs of insecurity may not stem from the actual risks themselves but from people’s and governments’ strong reactions to such risks. 2.1.

Characterizing the security economy

‘Risk’ can have several economic meanings. First, risk describes the possibility of a harmful event occurring or being induced. Examples may include the likelihood of a cheque bouncing or of a cheque fraudster being detected. Such events may cause substantial damage. Second, risk refers to the variation, variance or volatility of economic indicators such as exchange rates or future investment returns. These movements may induce costs to some economic actors. Third, risk can be deﬁned as an indicator being close to a threshold. Again, there may be some cost or loss involved in being close to a threshold. This is akin to the concept of vulnerability. The cost of a variation then depends on the distance to a given threshold. In this chapter, insecurity is deﬁned as an aggregate and unquantiﬁable form of risk. There are different sources of such risks and hence insecurity in the economy. These are related to the forces of nature, globalization, technological, social, and political developments on the one hand and economic or market forces on the other hand. These risks differ in their probabilities of occurring, the scale of damages they can induce and in the degree of covariance and hence insurability, as argued below. The security economy will be deﬁned for the purpose of this chapter as those activities affected by, preventing, dealing with and mitigating insecurity in the economy. Such broad deﬁnition includes private and public activities in both legal and illegal areas of the economy. Narrower versions of this deﬁnition (such as a

264 Tilman Brück focus on state spending for homeland security or private spending for anti-crime devices) may be adopted by other authors for different purposes. 2.2.

Security as a public good

National security, like a lighthouse, is a prototypical public good. National security is non-rival in consumption; each citizen enjoys the full amount of national security produced, without restricting the consumption of other citizens. Furthermore, it is impossible to exclude citizens from the provision of national security. Hence, the level of national security provided by the private sector would be suboptimal from society’s point of view. This is one justiﬁcation for the public provision or the public regulation of security in a closed economy. In the international context, global – rather than national – security is also a public good (Sandler, 2005). Instances of global insecurity include nuclear wars between the superpowers, dramatic climate change induced by greenhouse gas emissions or transnational terrorism. The nature of the global public good is that individual countries fail to internalize the foreign costs and beneﬁts of their actions and inactions concerning the underlying global risk. National defensive measures aimed at diverting the local harmful consequences of global risks may be over-supplied from a social planning perspective. For example, increasing national precautions against transnational terrorism may reduce such attacks within the country but increase the number of attacks against nationals of that country living abroad or against other countries. This is what appears to have happened with the pattern of transnational terrorist attacks since 9/11, with less protected countries bearing an increasing share of transnational terrorist attacks (Sandler, 2005). National pro-active measures aimed at reducing global risks entail foreign positive externalities; pro-active measures are thus under-provided. For example, unilaterally cutting greenhouse gas emissions reduces global warming but only has negligible positive effects for the pro-active country itself. In the context of transnational terrorism, Sandler (2005) identiﬁes the irony that by focusing the attacks of 9/11 against the United States, the US government was led to exert additional pressure on other states to intensify their anti-terrorist policies. Hence transnational terrorists to some extent are creating conditions which lead to their own activities becoming more difﬁcult to implement in the future.1 As argued by Frey and Luechinger in this volume, such pro-active policies may also involve positive actions, such as bribing terrorists ﬁnancially or politically to become peaceful. However, such actions will be under-provided by national governments with the existence of global threats to security. While the existence of national public goods justiﬁes regulation, global public goods provide the rationale for internationally coordinating these regulatory schemes. This is especially true if the global collective good security is considered to be a weakest-link public good (Hirshleifer, 1983). The prototypical weakest-link public good is a dyke that prevents the rising sea level to ﬂood an island. Each inhabitant can construct a section of the island’s dyke as protection

An economic analysis of security policies

265

from ﬂoods. However, the actual protection equals the height of the lowest section of the dyke. The same concept holds in principle for the general case of international security issues. Even if country A spent a lot on security s (that is it achieves a high sA∗ ) it may be negatively affected by another country B with a lower sB∗ . Through international air transport, country B may export insecurity unwittingly, for example by transporting suitcases with bombs enclosed onboard a plane travelling towards country A – as happened in the Lockerbie bombing. The weakest-link nature of many security-related collective good problems strongly shapes the policy recommendations discussed in Section 4. 2.3.

Direct economic effects of insecurity

Insecurity imposes costs on people who are risk-averse. Standard economic theory demonstrates that economic agents, if they are risk-averse, would prefer a world of less insecurity and are willing to pay a premium to reduce risks. Yet the costs of insecurity are composed of three effects. First, the direct costs resulting from the underlying risky event itself. Second, the indirect ﬁrst-order costs induced by the agent’s reaction to the threat. And ﬁnally, the indirect second-order costs that are caused by the policy responses to the event and to the agents’ reactions. Direct effects of insecurity include losses in property rights, output, utility, health or lives resulting from events such as theft, fraud, computer viruses, power cuts or terrorism. The ﬁrst-order negative effects contain the responses by the directly affected parties, such as precautionary information technology measures taken by a company targeted by computer viruses. The second-order indirect effects include the costs of the measures implemented by government in response to actual or perceived risks. These may include economic policies or more general political reactions to insecurity. The empirical study by Chen and Siems (in this book) investigating the response of capital markets to cataclysmic events concludes that the indirect effects of insecurity may well outweigh the direct effects in terms of economic relevance. The authors also ﬁnd that American capital markets have improved their response times to various cataclysmic events in the last ﬁfty years. It appears that earlier events had not been more dramatic than recent events. Instead, the economy had not been able to deal with crises in an adequate way in earlier years. The authors explain this, inter alia, with improvements in the efﬁciency of the banking and ﬁnancial sectors, providing the necessary post-crisis liquidity to promote market stability. In addition, fast and well-coordinated responses by the monetary authorities helped stabilize the American ﬁnancial system. The study emphasizes that it is not only the nature and the direct effects of a disastrous incident that determines its consequences but also the reactions by agents and policy makers, that is the indirect effects. 2.4.

First-order indirect effects: agent’s reaction

One important ﬁrst-order indirect effect of insecurity is a rise in private sector security spending. Such spending may express an underlying desire to protect

266 Tilman Brück production or to enhance a ﬁrm’s products. As such, security spending may be voluntary or in response to market forces. Security spending could also be obligatory as a result of new security legislation. This distinction has an impact on the competitiveness of ﬁrms. In the ﬁrst case, ﬁrms decide to spend money on security in the short term to minimise long-term costs (for example by spending on building security to avoid or deter ﬁre, thieves or terrorist attacks). Such spending is akin to insurance spending and reﬂects a ﬁrm’s information, perception and preferences. Firms can be said to self-insure against certain risks. It is likely that some ﬁrms have higher costs than others, for instance as a result of their location in high-risk zones. In the second case, ﬁrms respond to market forces for enhanced security measures, for example because employees require employers to have such security measures (for example protecting ex-pat staff on high-risk postings) or because more security has to be embedded into a ﬁrm’s products (for example alarm systems in cars). In these cases, costs rise but potentially revenues rise as well or are prevented from falling. Such measures may involve many ﬁrms in a market although some ﬁrms may opt to provide lower security and hence lower quality products thus occupying a different niche of the market. An intermediate level of cost differentiation may thus be obtained. In the third case, ﬁrms are legally obliged to implement certain security measures (for instance foreign airlines serving the US). In this case, the extra security spending acts akin to an environmental regulation aimed at increasing social welfare. Such regulation typically raises costs but not private, ﬁrm-level beneﬁts.2 Within-sector productivity will fall as a result of such enforced spending. In parallel, new sectors may emerge to service the new security needs, as can be observed in the environmental service sector. For a closed economy, this then implies that costs are borne uniformly by all ﬁrms in a given sector. Internationally, this may not hold and raises important trade policy issues. This third case also contrasts with a new tax imposed on a sector. Taxes also reduce productivity and may affect ﬁrms uniformly. However, taxes have the important implication of raising tax revenue, which can then be used to achieve some other social good or to compensate some other actor. Also, in the case of taxation, the taxable sector and the sector at risk may differ while in the case of state security spending the two are necessarily the same. Hence in the latter case, threatened sectors may be doubly affected by the new insecurity, both through the security risk and the security legalisation. Depending on the circumstances, this suggests the policy recommendation to diverge security measures and their ﬁnancing to reduce the burden of new security measures. Indirect ﬁrst-order effects concerning the behaviour of agents in response to the existence of risky events can also be analysed using standard insurance theory. Insurance can be used by risk-averse agents to overcome the negative effects of risk. By a large set of such agents forming an insurance, they can spread risks and thus improve welfare. Even if the nature of the risk allows such pooling, moral hazard and adverse selection may affect the insurance market.

An economic analysis of security policies

267

In the case of moral hazard, insured agents take higher risks than non-insured agents. For example, the existence of an obligatory terror re-insurance (or even re-re-insurance) may encourage insurers (or re-insurers) to accept more risky customers. This implies that insurance (or risk-sharing mechanisms such as regulation more generally) can actually increase risk levels compared to the absence of insurance (or regulation). In the case of adverse selection, high-risk agents take up insurance but insurance companies cannot identify the nature of these individuals due to asymmetric information. For example, the recipients of credible terror threats are more likely to buy terror insurance thus leaving the insurance company with a risky and potentially uninsurable pool of clients. This implies that insurance (or risk sharing arrangements more generally) may break down entirely if asymmetric information is very prevalent. Furthermore, the insurance market is affected by ﬁrst-order indirect effects on the supply side (Kunreuther et al., 2003). Standard war or terror risks cannot be insured at prices which are acceptable to customers as their occurrence is too strongly correlated across contracts for insurance companies to be able to pool its risks effectively. In order to illustrate how severely the economy can be inﬂuenced by ﬁrst-order indirect effects of insecurity, it is useful to consider some possible instances of such reactions. Greater insecurity (for example due to terrorist attacks) may lead to the re-design of supply chains which in turn reduces the beneﬁts from just-intime production processes. Alternatively, ﬁrms may prefer to source their inputs from local suppliers if these are less affected by certain insecurities. Such local suppliers may be more reliable but also more expensive. In the long term, such cost pressures may induce a variety of changes, such as increases in inventories, investments in new technologies or changes in the balance of horizontal or vertical integration (Shefﬁ, 2001, Hodges and McFarlane, 2003). Other indirect effects of higher insecurity include higher transaction costs of conducting business, including higher transport costs and higher transport insurance rates (Lenain et al., 2002). This will reduce trade ﬂows, the transport and tourism sectors, both domestically and internationally (Nitsch and Schumacher in this volume, Walkenhorst and Dihel, 2002). The decline in trade could reduce the spread of economic activity and boost geographic clustering. But the more clustered an economy is, the more valuable the clustered target is for terrorists, for instance, thus further raising insecurity (Frey and Luechinger in this volume). In fact, large-scale violence impacts cities in three ways. First, the safeharbour-effect encourages people to concentrate to have an advantage in defending themselves from attackers, making cities more appealing in times of rising violence. Second, the target-effect implies that cities are more attractive targets for violence, which creates an incentive to disperse. Third, the transportation-effect suggests that as terrorism often targets means of transportation, violence can increase the effective cost of transportation, which will usually increase the demand for density. However, empirical evidence on war and cities in the 20th century suggests that the effects of wars or terrorism on urbanity are not signiﬁcant.

268 Tilman Brück Having said that, there are notable exceptions in some extreme instances (Glaeser and Shapiro, 2001). Generally, if insecurity thrives on openness, then ﬁrms and households will scale back on openness. For example, less online trade may be conducted in the presence of online fraud and less international outsourcing may be undertaken in the presence of regular riots, roadblocks or strikes abroad. Changes in relative prices as a result of insecurity will lead to a suboptimal re-allocation of resources. Therefore, the insecure economy will have lower GDP growth than would obtain otherwise (Lenain et al., 2002). Higher levels of risk also undermine investor conﬁdence, reducing their willingness to commit to new projects. Over time, higher-risk premiums increase required rates of returns on investments, reducing equity prices and biasing investment decisions against high-risk, high-return, long-term investments towards low-risk, low-return, short-term investments. The cumulative effects of such portfolio adjustments are to change the composition of the portfolio, to reduce overall investment and to retard further economic growth. However, markets will also induce positive feedback effects causing structural shifts. These will occur in favour of products and services which have embedded security as an important characteristic. It is worth noting that reactions to insecurity may not always be justiﬁed, even if they are voluntary, because the degree of insecurity is also a matter of perception. Actual risks are extremely hard to assess, as argued by Brennan in this book. There is strong evidence that people, and by extension policy makers, are poor judges of objective levels of risks. On the one hand, especially when strong emotions such as fear are involved, people tend to focus on the worst-case scenarios rather than on the probability of the outcome occurring. As a result agents over-estimate minor risks or neglect signiﬁcant risks (Sunstein, 2003). In addition, the public representation of insecurity is very skewed (Kunreuther, 2002). Airline crashes, for example, receive more column inches in newspapers than fatal car accidents, although the former cause fewer casualties than the latter. For both reasons it is likely that the private sector and policy makers over-provide security measures and legislation, so that the costs of security may easily exceed its beneﬁts. On the other hand, Kunreuther and Heal (2003) point out that, on the demand side, people tend to under-estimate the risks of natural disasters or terrorist attacks when it comes to making insurance decisions. This ‘it-will-not-happen-inmy-backyard’ mentality represents another obstacle for developing an insurance market for disastrous events. Finally, the events of 9/11 may have revealed to the general public which state of insecurity they actually face. This view is supported by evidence that the likelihood of terrorist attacks has not increased after 9/11 but that agents assess this risk more realistically (Sandler, 2003). This interpretation also implies that structural changes in the economy (for example increasing the share of security-related spending or reductions in demand for airline travel) may not be inefﬁcient but rather are the result of an adjustment process towards a new market equilibrium.

An economic analysis of security policies 2.5.

269

Second-order indirect effects: policy reaction

Government regulation can cause insecurity in two ways. First, there may be an element of regulatory insecurity where an increasing density of regulation, though aimed at raising social welfare, increases uncertainty for ﬁrms operating in an environment of raising legal obligations. Second, certain types of regulation may trigger illegal responses raising insecurity. For example, the period of prohibition in the United States from 1920 to 1933 represents an instance where a policy reaction induced a signiﬁcant but perverse behavioural response, including illegal brewing, smuggling and organized crime. The degree of government security regulation is quite large in many economic sectors. For example, inspections and other security regulations create delays at borders, increase shipping times and reduce border permeability thus reducing trade ﬂows. Such regulation thus enhances the direct effects of insecurity on trade. In addition, standard government regulations in the ﬁelds of national defence, ﬁghting crime and civil rights will impose further costs on businesses (Hobijn, 2003, Phillips, 2001, Lenain et al., 2002, World Bank, 2003). Security regulations imply shifting economic resources between actors, including between sellers and buyers and between private and public agents. The existence of such a burden will reduce the efﬁciency of the market and hence growth. In the insurance market, second-order indirect effects exist as well. For example, the market for terror re-re-insurance has been organized quite differently in various OECD economies (Kunreuther et al., 2003, Wolgast in this volume). While in the United States the government required insurance ﬁrms to offer terror insurance, in Germany for instance the government helped subsidize a monopolist publicprivate partnership re-re-insurer to cover potential terror risks. The US scheme has suffered from insurance ﬁrms offering the obligatory terror insurance but doing so at premiums that are unattractive to most ﬁrms. Thus the insurers fulﬁl their legal obligations without incurring risky and potentially unproﬁtable terror risks. Therefore, this may represent an instance where public intervention and even subsidies are necessary for maintaining some market forces, rather than using regulation (or the lack thereof as was common in most OECD economies before 9/11) to stiﬂe the market for terror insurance. Another second-order indirect effect of insecurity is the increase in public security spending which may have a retarding impact on long-term growth through two channels: ﬁrst, high budgets for defence and homeland security may crowd out more growth-enhancing investments. Second, there is some evidence that public security spending may also crowd out the more efﬁcient private sector attempts to increase security. These points will be addressed in more detail in Section 3. 2.6.

Estimates of the scale of indirect effects

Empirically, it is difﬁcult to directly measure the costs of the indirect effects. The following paragraphs will therefore provide an overview over existing estimates of instances of such costs for the cases of security spending and non-ﬁscal security measures (such as tighter border controls).

270 Tilman Brück In assessing the economic effects of security spending, Hobijn (2003) claims that neither private nor public spending on security will have a major impact on the economy. Private security spending in the United States in his view will reduce labour productivity by 1.12 per cent and multifactor productivity by 0.65 per cent, which in turn results in only small reductions of American GDP. In addition, he predicts that security-related research and development (R&D) will not signiﬁcantly crowd out productivity enhancing R&D. In regard to public security spending, he calculates that homeland security spending will reduce output only by 0.6 per cent over a ﬁve-year period. Judging by the much larger scale of military spending in the 1980s, he believes that to be negligible and to have no effect on the US budget deﬁcit. However, one should interpret Hobijn’s optimistic results with a degree of caution. The analysis contains some important assumptions such that private expenditures for security will only double in the future. Hobijn may also under-estimate the future public spending by the Bush administration, especially when adding homeland and national security spending in the light of the Iraq war and occupation. Nordhaus (2002) in particular contradicts Hobijn’s line of argument. He cautions not to depend too strongly on governmental estimates of future security budgets.3 Nordhaus ﬁnds that the costs of wars, for example, are always grossly under-estimated, which is perfectly rational from the point of view of the warring government. Another analysis partly backing Hobijn’s view is a simulation of the combined growth effects of increased private expenditures for security (up 0.5 per cent of GDP) and increased military spending (up 1 per cent of GDP) ﬁnanced through borrowing (Lenain et al., 2002). This study suggests that real GDP would be reduced by about 0.7 per cent after ﬁve years. The effect is small but permanent and derives from the consequences of undermining ﬁscal consolidation. The postcold war peace dividend is not threatened by such an increase in security-related spending. Moving from the effects of security spending to the effects of security measures, it has been estimated that a one-day delay at the border controls costs 0.5 per cent of the value of the delayed good (Hummels, 2001). Another calculation suggests rising trading costs of 1 per cent to 3 per cent ad valorem after 11 September (Leonard, 2001). Based on such values, it has been estimated that an increase in US inventories of 10 per cent and an increase in US commercial insurance premiums of 20 per cent would cost 0.1 per cent and 0.3 per cent of GDP per year, respectively (Raby, 2003). Another study calculates an elasticity of trade ﬂows (in volume terms) with respect to transport costs (ad valorem) to be in the range of −2 to −3.5 (Limao and Venables, 2001). In international trade, the total global welfare losses resulting from such securityrelated increases in transaction costs after 9/11 have been estimated using a computable general equilibrium model to be relatively low with about 75 billion USD per year (Walkenhorst and Dihel, 2002). Yet that study also suggests that some regions and sectors are hit particularly hard, as goods with a low ratio of value to weight (such as agricultural products, textiles, non-metallic minerals and

An economic analysis of security policies

271

machinery) are vulnerable to an increase in transaction costs. The regions most affected by 9/11 in absolute terms according to Walkenhorst and Dihel are Western Europe, North America and Northern Asia. However, they calculate that Southern Asia, North Africa and the Middle East suffer the most damage in relation to the size of their economies, not least due to the higher import dependence. That means that developing countries may be particularly affected by the ﬁrst- and second-order effects of 9/11. A different methodological approach involves directly estimating the effect of the existence of insecurity on growth or international trade. One such study ﬁnds that international trade ﬂows are signiﬁcantly reduced by the existence of terrorism in a trading partner’s country (Nitsch and Schumacher in this volume). In the short term, this effect is estimated to reduce international trade by 4 per cent if the number of terrorist incidents in one country is doubled. It is open to empirical analysis if and how soon the negative effects of insecurity will wear off in the long term. Increases in efﬁciency may be obtained by better regulation and implementation (Shefﬁ, 2001, Hobijn, 2003, Lenain et al., 2002, Walkenhorst and Dihel, 2002, Raby, 2003, World Bank, 2003). Regulation may be more targeted, thus reducing unnecessary security measures. Markets may respond to existing security measures, ﬁnding new ways to communicate, to produce and to deliver goods. Security measures may deter or identify criminals thus reducing the exposure to risks and hence making the measures superﬂuous in the long term. This may be true. It is, however, not clear if these developments will actually occur. A key policy focus should thus be the monitoring of security spending, the security situation, the security policies and their effects on the economy to adjust measures over time as appropriate. Overall, security spending and security measures do have strong effects. It is less clear that these effects signiﬁcantly restrict growth, trade and other economic activities. However, security policies appear to have a differential impact, depending on the nature of the economy. In the long term, there operate strong forces which will alleviate the negative economic effects of security policies.

3.

Economic trade-offs

Policy makers must assess the beneﬁts and the costs of introducing security measures. This section will consider the trade-offs that are involved in providing security, especially in regard to efﬁciency, equity and liberty. 3.1.

Security spending versus other spending

The ﬁrst trade-off refers to the different types of expenditure by both the private sector and governments. Such butter versus guns decisions include the tradeoffs between different types of public and private sector expenditure between security-related versus other goods and services. The character of public spending on military services is mainly consumptive, with only a small part of military budgets being devoted to R&D (Ram, 1995).

272 Tilman Brück An economic beneﬁt of military spending is the prevalence of peace. Yet this effect is difﬁcult to estimate in practice. In addition, demand effects will increase GDP growth in the short term while in the long term negative effects may prevail: a large defence budget crowds out public investments, thus lowering total factor productivity. Military spending may also increase the budget deﬁcit, the national debt and hence interest rates. Cutting military budgets (or realizing a peace dividend) may thus boost growth through higher capital accumulation, a higher civilian labour force and more productive capital allocation for a given security threat. A similar line of reasoning also holds for the case of private spending on security. Since output is not positively affected by this spending (especially when spending concerns hiring more guard labour) productivity falls. In addition productive investments are likely to be crowded out and hence growth is retarded. Therefore both public and private spending on security may have both expansionary and dampening effects on growth, with the net effect being ambiguous a priori and empirically (Lenain et al., 2002). 3.2.

Security versus efﬁciency

The second trade-off concerns efﬁciency. An economy is efﬁcient when it maximizes production from a given set of resources and technologies. Aiming for security entails both costs and beneﬁts. Such costs may cause frictions in production thus preventing the economy from functioning efﬁciently. This could refer to the equilibrium position on the production possibility frontier (PPF) where the marginal rate of substitution between the good security and the alternative good equals the price ratio between these two goods. If an exogenous shock requires a higher provision of security then this may lead to a reallocation of production from the alternative good to the production of security. The new equilibrium may be efﬁcient if at that point the price ratio also equals the marginal rate of substitution between these two goods. An inefﬁcient level of security production would only be obtained if production of either or both goods was within the PPF or if the marginal rate of substitution did note equal the price ratio, that is if frictions exist which prevented achieving the new equilibrium. Another source of friction may prevent a new equilibrium from establishing itself in response to technological change, which could potentially extend the PPF outwards. The nature of the security markets, of the new technological developments and of government regulation then determines if efﬁciency is obtained. Government policy should thus be geared towards maintaining markets and incentives which make this feasible. Efﬁciency can be visualized as minimal levels of transaction costs, for example when crossing borders or generally in trade. Here there appears to be an obvious trade-off between security and efﬁciency as more border controls increase security but also reduce the speed and ease with which goods and people are moved. In the long run, however, this trade-off may disappear, as argued above. Security-driven improvements may even facilitate trade in the long run. Additional investments in secure facilities and modern technologies can reduce transaction costs.

An economic analysis of security policies

273

Security cost pressures could potentially induce reforms in trade-related institutions and infrastructure with beneﬁcial effects on trade and growth. Better trade facilitation due to deregulation of trade-related sectors, harmonization of customs services and coordination across countries would increase trade among 75 countries by 377 billion USD (World Bank, 2003). Another example concerns the public versus the private provision or regulation of security services and rules. For instance, the United States increased the public sector employment of airport security personnel post-9/11 (Phillips, 2001). It is not clear ex ante if such services must necessarily be provided by the federal or state government or if, with suitable regulation, private ﬁrms may not have provided security more efﬁciently. 3.3.

Security versus globalization and technological change

The third trade-off may occur between security on the one hand and globalization and technological change on the other hand. It is not clear ex ante whether globalization is compounding or extenuating the problems associated with the security economy.4 More generally, one can identify a race between two effects of globalization. On the one hand, the same forces which can bring some countries and sectors such prosperity are highly vulnerable to security threats. It is both the openness and the interdependence that enable various risks to destabilize the international economy (Stevens, 2003). On the other hand, coordination, integration and harmonization associated with globalization reduce the scope for insecurity and make the tracking of the sources of insecurity easier. In addition, globalization is a process which provides ongoing ﬂows of beneﬁts while many forms of insecurity cause one-off, shock-like costs (unlike the ﬁght against insecurity which may cause ongoing costs, too). In an integrated, globalized world economy, building coalitions to ﬁght insecurity by providing public goods may hence be much easier than in a world economy dominated by import-substituting nation states. Accordingly, Chen and Siems, (in this book) conclude that the globalized world has become more stable in face of threats. The policy response to 9/11 especially showed how effectively cooperation can be indeed conducted. The international integration made it both necessary and possible for authorities all over the world to share relevant information and to reconcile policies in order to absorb such tremendous shock. Globalization and technological change induce structural change in open economies. The security economy may especially witness an accelerated structural change (Shefﬁ, 2001, World Bank, 2003). This may be obtained through technological advances induced by investments in security infrastructure, for example through the automation, surveillance and informational exchange in harbours, airports and border crossings. Globalization may thus serve as the very means that makes the trade-off between security and efﬁciency diminish in the long run.

274 Tilman Brück One important policy challenge is the integration of technical security protocols into international organizations, agreements and technical standards (such as the European Union, the World Trade Organisation and the International Organization for Standardization (ISO)). Transparency and harmonization should be sought to reduce transaction costs. In addition, security concerns should not permit the establishment of non-tariff trade barriers. Another policy implication addresses the role of economic winners and losers from structural change induced by new security regulations. This will be discussed further below. At the same time, countries falling behind the evolving international security standards are unable to reap the beneﬁts of globalization if their territory is not seen to be safe or reputable any more (for example by not guaranteeing security, providing smart technologies and protecting supply chains). Those economies will face higher-risk premiums and the cost of protecting assets will rise, reducing foreign direct investments. Competition concerning both the supply of security between countries and the nature of the provision of security within countries could evolve. Some countries may specialize in utilizing their comparative advantage in producing secure or insecure goods (such as the respective examples of the United States and Talibanled Afghanistan in the case of terrorism or Switzerland and some small island states in the case of more or less prudent banking facilities). In addition, countries may choose different models of providing a given standard of security within an international organization. NATO, for example, has contained in its history both democracies and dictatorships as well as armies of recruits and professional armies. For companies, there is geographic choice in their production decisions, both in regard to the desired level of security and the nature in which this level is achieved. As a result each country A would than obtain its (individually) optimal level of security sA∗ , a circumstance which may well conﬂict with the weakest-link nature of international security. 3.4.

Security versus equity

A politically and socially sensitive trade-off concerns the distributional costs of increased security. Analytically, it is not clear ex ante which groups should gain or loose most from higher security. Many security services are provided by the lowskilled (such as guards) but many technology-intensive products will be developed by the highly skilled. If international trade is reduced by higher security-related transaction costs, then this may damage employment in those sectors or countries most affected by such measures. Public sector employment may rise if public security spending focuses on judicial, police, customs and military personnel. However, some of their services can also be subcontracted to private providers, which is an important policy option when considering the efﬁcient provision of security. Governments could consider compensating the losers of security measures within their countries. Internationally, this may be particularly important if losers of the security economy (say groups or entire countries losing from reduced trade

An economic analysis of security policies

275

in developing countries) may themselves be the source of future insecurity. Hence the compensation of losers (and perhaps the taxation of the winners) is strongly related to the causes and the nature of the insecurity. One option may be the accelerated and unilateral reduction of trade barriers for developing countries particularly damaged by the war against terrorism. Another equity issue is related to the access to security services and products. Lower-income groups may, as a result of market forces or due to administrative processes, be excluded from secure products or services. One can also think of social clustering since the poor may only be able to afford property in less safe environments or regions. Policy makers may wish to consider how they can grant egalitarian access to the security economy. Another equity concern has already been mentioned in Section 2. While it is clear that both market-induced and compulsory investments in security measures raise costs for ﬁrms, it is less clear ex ante whether the producer or the consumer bear a larger share of the costs. Using tax-incidence analysis, one can demonstrate that the short-term burden is distributed according to the price elasticities of demand and supply. In the long term, however, the supply elasticity is apt to be perfectly elastic in a globalized world, thus allowing most of the costs to be shifted onto the consumers. 3.5.

Security versus freedom and privacy

The ﬁfth trade-off concerns the balance of civil rights, privacy and individual freedom versus the possible need to curtail these rights in the pursuit of more security. Internet, computing, mobile and wireless technologies are highly vulnerable to security attacks. At the same time, these technologies can be used to monitor movements, usage and proﬁles of individuals or goods – by consumers, regulators and potential perpetrators of crimes. For example, a positive correlation between democracy and the levels of terrorist activities has been observed in various studies (Sandler, 2005). This effect may result from terrorists preferring to act in countries where their attacks will be reported widely.5 This topic raises a number of interesting and relevant points which, albeit, are not all or not exclusively the domain of economic analysis. First, there is a clear trade-off between economic freedom and economic growth, at least in the extreme. The empirical estimation of this trade-off may lead to ambiguous results but analytically it should be clear that a high level of regulation and restriction hampers productivity growth and utility maximization (Paldam and Würtz, 2003). Second, the evolution of the network economy and the evolution of the security economy are closely related. The opportunity to process and link data of low marginal value is growing dramatically. With it rises both the vulnerability of interconnected and interdependent data systems and the opportunities for tracing criminals. Protecting these systems, using their opportunities and maintaining civil liberties requires a ﬁne balancing act. The greater demand for security-induced surveillance and the technological advances in this ﬁeld

276 Tilman Brück facilitate the potential abuse of data mining, social sorting and losses in privacy (Lyon, 2003). In fact, many economic sectors require increasingly complex information chains in production. The monitoring of the origins of food, of industrial chemicals (especially in the European Union), and of dangerous waste products increasingly require source-to-use chains of information. The use of smart tags can thus be expected to rise dramatically as will the use of positioning and navigation systems in combination with mobile technologies (Hodges and McFarlane, 2003). These technological and legal developments demonstrate the rising challenges for the security economy and its policy makers. Resolving civil rights and technological security issues should occur in parallel to the implementation of such information chains. Forcing the resolution of security concerns may actually accelerate the development of source-to-use information systems. In addition, social preferences about the relative value of judicial type I and type II errors may evolve in the security economy. This may be particularly true for very rare but extreme events where the importance of balancing type I errors (where the innocent goes to jail) and type II errors (where the guilty walks free) may be reversed (Kunreuther, 2002). Many societies, when protecting their own citizens from extreme attack, prefer to punish the innocent than to let the guilty escape with committing atrocities. The opportunity cost of inaction weighs particularly heavily in the security economy. This may hence lead to an otherwise excessive level of security regulation and spending. However, there is no obvious, technocratic optimality in assessing the trade-off between security and freedom. In democracies, the voters have to decide what their preferences are in this respect.

4.

Policy implications

Having reviewed the nature of the security economy, the direct and indirect effects of insecurity, and the trade-offs involved in reaching more security, this section will discuss options for designing security policies in the national and international contexts. 4.1.

National policy instruments

Policies aimed at reducing our exposure to risks can be defensive (for example by installing anti-virus software) or pro-active (for example by identifying, arresting and punishing hackers before, as or after they strike). Some of the defensive policies may not be security policies in a narrow sense. Yet it is useful to remember that reducing economic insecurity involves many more policy ﬁelds than law-andorder or economic policies. In addition to defensive and pro-active policies, other policies may aim to reduce the costs of insecurity. This may serve the purpose of reducing the impact of insecurity but also of making deliberate acts of insecurity less attractive to the perpetrators. Frey and Luechinger (in this volume), for example, suggest that raising the marginal costs of terrorists to undertake terrorist attacks by adopting

An economic analysis of security policies

277

deterrence policies may not be the best response to such threats. Instead, terrorism may be fended off more effectively by abating the expected beneﬁts of terrorist acts for the prospective terrorists. Such a policy could be based on strengthening decentralized decision-making since a strike against authority would then have only little effects on the stability of the polity and the economy. The analysis of security policies should also differentiate between different policy instruments. Information and institutions are one group of policies used to achieve deterrence and punishment. Since the probability of malevolence tends to be over-estimated by individuals, information should be used to make risks more transparent. One set of policy instruments includes regulation, supervision and coordination while another involves the provision of ﬁnancial incentives and disincentives, for example through ﬁscal policy. For instance, in addition to the under-provision of security, a market economy may also under-invest in security-related R&D. As R&D generates spillovers for society, the social rate of return is typically higher than the private rate, and hence private investment in R&D typically falls short of the socially desirable level. Hence, even if security was a private good, there would be a case for state subsidies for security-related R&D. The public debate about the ﬁght of terrorism since 9/11, for instance, has focused quite strongly on security spending and the adjustment of civil rights while neglecting some other instruments, such as international coordination, political signals or even, at times, deliberate disinterest (Brennan, in this book). Most importantly, the public debate often failed to ask how market forces may help solve some of the problems that society faces, instead focusing strongly on government intervention, regulation and spending. Even if government regulation of the security economy is necessary, it should be tailored with care. Incentives, expectations and market powers should be emphasized to raise the efﬁciency of the intervention. Crowding out should be avoided to reduce the negative secondary effects of intervention (especially on growth). As noted above, regulation that ‘adds insult to injury’ by placing the incidence of regulation on those already suffering the direct effects of insecurity may be economically unproductive and socially unfair. The post-9/11 insurance sector serves as an example of how public-private cooperation may be feasible in more cases than previously envisaged, for example in providing security services or in implementing new regulatory schemes for re-re-insuring terror (Lenain et al., 2002, Wolgast in this volume). Market forces also provide mechanisms to enhance voluntary security spending. For example, agents investing in security may pay lower insurance premiums or low-risk individuals may accept higher deductibles in insurance contracts as a way of signalling their status and avoiding moral hazards and adverse selection. Such outcomes can be enhanced further by regulation deﬁning liability, which increases the incentives of the guilty party to act responsibly. However, even market forces may not be able to overcome moral hazards and adverse selection entirely, thus implying that insurance markets cannot be relied on to diversify all possible risks through market forces alone.

278 Tilman Brück 4.2.

International policy coordination

As argued above, dealing with global public goods requires international coordination. There are many incidents in the global security economy where uncooperative behaviour falls short of the socially optimal outcome. The aforementioned undersupply of pro-active policies and the over-supply of defensive measures may serve as examples. In the ﬁrst case, nations are stuck in a prisoner’s dilemma with a Nash equilibrium of mutual inaction. The second case resembles the well-known tragedy of the commons (Sandler, 2005). It is the weakest-link public good nature that emphasises the need for international cooperation and intervention. Whenever the overall level of protection is set by the least contributor, competition fails to achieve the efﬁcient level of security both in the private and the public spheres. Kunreuther and Heal (2003) give an example from the private sector by noting that airline A will only install an additional baggage screening system at its checkin counters if other airlines (including airline B) adopt similar systems. This is because the hazard may not only arise from passengers who check in directly with airline A, but also from passengers who check in bags with airline B and then arrange for their luggage to be transferred, without necessarily being checked again by airline A. The security of airline A is hence determined by the weakest link in the chain of airlines. The decision by all airlines to remain unprotected is then a Nash equilibrium of a prisoner’s dilemma game. Only a coordinated approach can break this suboptimal equilibrium by guaranteeing the participation of every airline. International organizations like the International Air Transport Association (IATA) can stipulate rules and regulations for their members. A key policy issue is the sanctions that are implemented locally within member states, which apply in the case of national deviant behaviour. A public sector example is provided by Sandler (2005). He makes the point that even a nation which heavily invests in securing its domestic targets is still vulnerable as it is linked with other less secure countries through trade, travel and migration. It is again the weakest-link character of other countries that cannot (or choose not to) increase domestic and border security which causes the problem. This shortcoming must be overcome by international institutions such as the WTO, OECD, NATO, the UN, and ISO by setting minimum standards in the areas of defence, politics and economics. However, this coordinating task is challenging. International cooperation tends to be myopic due to political (in particular democratic electoral) processes and due to a preference by governments to be autonomous in defence matters. This contrasts with the long-term planning horizons of terrorists (who rarely have credible outside options). In addition, weakest-link countries (which often are developing countries) may do not have the capacity to conform to international standards. As a consequence, the international community must support these states through either cash transfers or in-kind transfers, which itself poses a collective action problem.

An economic analysis of security policies

279

Governments face a standard free-rider situation where every government ‘sits on its hands’ expecting others to provide the transfers. Furthermore, there is a moralhazard problem associated with cash-transfers, since the recipient may divert the money for concerns other than enhancing its security (Sandler, 2005). Despite these obstacles, international security policies should also contain a strong commitment to international development as one of the pillars of global security policy. This view corresponds to the so-called ‘Solana strategy’, the European security strategy proposed by the European Commission (2003). In addition, compensatory or complementary liberalizing policies are important to provide a strong stimulus of world economic growth and to generate the economic surplus to assist vulnerable groups in the global security economy. This applies especially in the area of transport costs and world trade, where additional trade access could be provided to countries at risk from losing out in the security economy (Leibfritz, 2003, World Bank, 2003).

5.

Conclusions

This chapter ﬁnds that in addition to its striking direct effects, insecurity can have important indirect effects via agents’ behavioural and governments’ policy responses. Policy insecurity, the burden and incidence of security regulation, and unintended responses to policies may raise aggregate insecurity further. The emergence of new risks induces structural changes in the economy, both across sectors and across countries. Markets may be inhibited due to insecurity but insightful regulation can also help support the emergence of markets otherwise threatened by the nature of the risks. These issues drive the complex trade-offs which policy makers face in the security economy. For instance, the trade-off between butter and guns does not imply a simple choice between more or less growth, respectively. Likewise, securityrelated transaction costs may rise in the short term but may fall below even previous levels in the long term. There is hence a potential for the emergence of ‘secure growth’, akin to the emergence of ‘green growth’. Policy instruments in the security economy have to account for the complexity of these issues and should therefore not be overtly dependent on regulation alone. Instead market instruments, for example from the insurance markets, can help pool risks and alleviate the costs of remaining risks. Non-markets instruments may be useful for regulating access to the security economy and for coordinating technical standards suitable for the security economy. However, state intervention should not be used to stem necessary structural changes with an economy. Very importantly, a social dialogue is necessary, and will emerge, on the relative merits of civil liberties and extreme forms of insecurity. Concerning domestic policy interventions in the security economy, governments should refrain from executing extreme, hasty or untested policies. Competition alone cannot solve the problems. Neither can state intervention. The best results may be achieved by a portfolio of policies combining political, economic, legal and social means.

280 Tilman Brück Given the public good nature of global security and the resulting externality effects, policy responses to insecurity rely heavily on international coordination efforts, which may be fragile if the public good is of the weakest-link type. Given the diversity of the actors involved and the difﬁculty in monitoring and enforcing cooperative behaviour, policy coordination at the international level is desirable but hard to achieve. Multilateral institutions should play a vital role in this coordination game.

Acknowledgements The author is grateful for helpful comments from Carlos Barros, Reza Lahidji, Patrick Lenain, Todd Sandler, Barrie Stevens, and from conference participants in Paris, Lisbon and Wiesbaden. Till Stowasser once more provided outstanding research assistance. The usual disclaimer applies.

Endnotes 1 It is not entirely clear from an economic point of view why transnational terrorists should pursue such strategy of focusing on their strongest enemy in the ﬁrst place. 2 The distributional long-term effects of such regulation will be addressed in Section 3. 3 Security budgets may even be poorly measured ex post, as argued by Brauer (2004) for the case of the United States. 4 One example of the former effect is that transnational terrorism, itself made possible by globalization, also implies that terrorists have to hit increasingly large targets to make an impact on global current affairs (Sandler, 2005). In other words, there exists a race for nastiness among transnational terrorists. 5 Such negative consequences of press freedom may only be overcome if freedom of the press (or democracy) was a truly universal phenomenon.

References Brauer, J. (2004) ‘United States Military Expenditure’, Mimeo, Augusta State University. European Commission (2003) ‘A Secure Europe in a Better World’, European security strategy by the European Commission, Brussels. Glaeser, E. L. and J. M. Shapiro (2001) ‘Cities and Warfare: The Impact of Terrorism on Urban Form’, NBER Working Paper Series 8696. Hirshleifer, J. (1983) ‘From Weakest-link to Best-shot: The Voluntary Provision of Public Goods’, Public Choice, 41: 371–86. Hobijn, B. (2003) ‘What Will Homeland Security Cost?’, Economic Policy Review, 8(2): 21–33. Hodges, S. and D. McFarlane (2003) ‘What Identiﬁcation, Authentication and Surveillance Technologies for Tomorrow?’, paper presented at the OECD Forum for the Future on ‘The Security Economy: What Trade-Offs in an Open and Mobile Society?’, Paris. Hummels, D. (2001) ‘Time As a Trade Barrier’, Mimeo, Department of Economics, Purdue University, West Lafayette/Indiana. Kunreuther, H. (2002) ‘Risk Assessment and Risk Management in an Uncertain World’, Risk Assessment, 22(4): 655–64.

An economic analysis of security policies

281

Kunreuther, H. and Heal, G. (2003) ‘Interdependent Security’, The Journal of Risk and Uncertainty, 26(2/3): 231–49. Kunreuther, H., Kerjan, E. M. and Porter, B. (2003) ‘Assessing, Managing and Financing Extreme Events: Dealing with Terrorism’, NBER Working Paper Series, 10179. Leibfritz, W. (2003) ‘Auswirkungen des Terrorismus auf die Volkswirtschaften und Implikationen für die Wirtschaftspolitik’, Ifo Schnelldienst 56(1): 14–20. Lenain, P., Bonturi, M. and Koen, V. (2002) ‘The Economic Consequences of Terrorism’, OECD Economics Department Working Papers, 334. Leonard, J. S. (2001) ‘Impact of the September 11, 2001 Terrorist Attacks on North American Trade Flows’, Manufacturers Alliance E-Alert, Arlington Virginia. Limao, N. and Venables, A. J. (2001) ‘Infrastructure, Geographical Disadvantage, Transport Costs and Trade’, World Bank Economic Review, 15: 451–79. Lyon, D. (2003) ‘Mid- to Long-term Implications for Society of the Growing Use of Security Technologies’, Paper presented at the OECD Forum for the Future on ‘The Security Economy: What Trade-Offs in an Open and Mobile Society?’, Paris. Nordhaus, W. D. (2002) ‘The Economic Consequences of War with Iraq’, The New York Review of Books, 49(19). Paldam, M. and Würtz, A. (2003) ‘The Big Bend – Economic Freedom and Growth’, paper presented at the European Public Choice Society annual meeting in Aarhus. Phillips, L. T. (2001) ‘A Crisis of Security and Economics’, Regulation, 24(4): 53–6. Raby, G. (2003) ‘The Cost of Terrorism and the Beneﬁts of Cooperating to Combat Terrorism’, paper presented at the Secure Trade in the APEC Region (STAR) Conference. Ram, R. (1995) ‘Defense Expenditure and Economic Growth’, in Hartley, K. and Sandler, T. (eds) Handbook of Defense Economics, pp. 251–73, Amsterdam: Elsevier. Sandler, T. (2003) ‘Collective Action and Transnational Terrorism’, The World Economy, 26(6): 779–802. ______ (2005) ‘Collective versus Unilateral Responses to Terrorism’, Public Choice, forthcoming. Shefﬁ, Y. (2001) ‘Supply Chain Management under the Threat of International Terrorism’, The International Journal of Logistics Management, 12(2): 1–11. Stevens, B. (2003) ‘Factors Shaping the Demand for Security Goods and Services’, paper presented at the OECD Forum for the Future on ‘The Security Economy: What Trade-Offs in an Open and Mobile Society?’, Paris. Sunstein, C. R. (2003) ‘Terrorism and Probability Neglect’, The Journal of Risk and Uncertainty, 26(2/3): 121–36. Walkenhorst, P. and Dihel, N. (2002) ‘Trade Impacts of the Terrorist Attacks of 11 September 2001: A Quantitative Assessment’, paper presented at the workshop on ‘The Economic Consequences of Global Terrorism’, DIW Berlin. World Bank (2003) ‘Reducing Trading Costs in a New Era of Security’, in Global Economic Prospects 2004: Realizing the Development Promise of the Doha Agenda World Bank: 179–203.

Part V

Epilogue

17 Terror: the ‘ISM’ versus the ‘ISTS’ Geoffrey Brennan

‘No passion so effectually robs the mind of all its powers of acting and reasoning as fear’ Edmund Burke Our Ideas of the Sublime and Beautiful 1756 ‘The only thing we have to fear is fear itself’ Franklin D. Roosevelt Inauguration Address 1933

1.

Introduction

For reasons that are perhaps understandable, much of the discussion of terrorism since late 2001 has been focused on the terrorists – on their motives and the institutions that support them. And in the same way, most of the policy discussion has been directed at the minimization of the terrorist threat. As the title suggests, the focus of this paper will be somewhat different. My primary interest lies in terror itself – and terror as a speciﬁcally social phenomenon. I seek to focus attention on the processes that occur after the initiating ‘terrorist’ acts – the down-stream processes that are called into operation by those acts and which have the effect of ‘terrorizing’ the target population. Although the motives of terrorists are not my prime concern, I am at least going to assume this much – that terrorizing the target population is one of the purposes of terrorists and of terrorist organizations. It is worth noting that this is not selfevidently true. Terrorist acts might have the consequence of sowing terror without that consequence being a particular motive of the perpetrators. Perhaps all that the perpetrators want to do is to inﬂict the immediate harm. However, while this may be true of some individual actors, I do not think that it is in general true of terrorist organizations. Most terrorist organizations have political purposes and securing those political purposes requires that the numbers of people affected by their terrorist activities extends well beyond the persons who are directly harmed in any terrorist episode (or series of them). ‘Terrorism’ as I shall use the term includes the entire process whereby the target population comes to be terrorized – of which process the actions of ‘terrorists’ are but one piece.1 It follows from this view that a critical part of the ‘war on terror’ ought to involve efforts to appropriately manage the response within the domestic population.

286 Geoffrey Brennan Those individuals and institutions within that population that play a role in spreading and amplifying the terror can on this basis be identiﬁed as ‘part of the problem’ – whether they themselves intend the ampliﬁcation or not. Simply put, what terrorists hope for is over-reaction: and the greater the over-reaction, the greater the extent to which one of their prime purposes will be secured. Clearly then, my agenda in this paper connects to other, more conventional discussions of whether the West is ‘over-reacting’ to the terrorist threat. But I hope it will be clear that what I have in mind when I talk of ‘over-reaction’ is rather different from what that term is usually taken to mean. My conception of over-reaction is focussed less on any putatively retaliatory actions that any country might undertake on suspected perpetrators than it is on the psychological state of the general population. Of course, such psychological states have consequences, and any concept of ‘over-reaction’ must include such consequences. But the consequences as such are not my chief concern in this paper. For example, I am not here to be seen as attempting any kind of assessment of the war on Iraq. I am to be seen as trying to identify some of the factors and processes that were politically hospitable to that war and to the widespread retreat from liberal ideals (such as the Patriot Act, Guantanamo Bay conﬁnement, torture of terrorist suspects, etc.) that has characterized the ‘post-9/11’ period. We can usefully think of the response process as involving a sequence of several distinct phases, and several distinct aspects. So in phase 1, occupying the ﬁrst week or so, there is the media reporting of the event itself and its immediate aftermath. There is also the ﬁrst round of policy response – say, in the 11 September case, the halting of all commercial air trafﬁc within and in and out of the US. And there are the messages from senior ﬁgures, not least the politically appointed ones. These latter messages are usually focused on the events themselves rather than interpretation or foreshadowing of more far-ranging policy responses. However, any immediate policy response will require some justiﬁcation, and the associated rhetorical defences will sometimes have implications for longer-term policy action. In phase 2, comprising the couple of months that follow the initiating event, there are the ‘more considered’ press reports, oriented towards interpretation rather than mere description or reliving the moment. In this phase, there are opportunities for public intellectuals and opinion-shapers to offer their comments. There are the liturgical ceremonies – memorial services and the like. There is discussion among policy makers as to what larger actions might be called for: what measures of deterrence or avoidance or prevention are thought appropriate. In this phase, the investigative element plays a major role. Who is to blame? How did it happen? How might it have been avoided? Who blundered? And so on. Then in phase 3, occupying perhaps the next six months, the longer-term policy response emerges. And the formidable rhetorical resources that are embodied within democratic political processes are mobilized. At each point, the media obligingly keeps everyone abreast of the latest developments. And of course, the technological capacity to do this ‘abreast-keeping’ is now sophisticated to a degree unprecedented in human history: Marshall McLuhan’s

Terror: the ‘ISM’ versus the ‘ISTS’

287

global village is well and truly operational (McLuhan and Powes, 1988). Unquestionably, part of the special horriﬁc genius of the 11 September episode lay in its cinematic qualities – more reminiscent perhaps of a Hollywood blockbuster than of real life. These qualities and the salience of the images to which they gave rise undoubtedly played a role not just in mobilizing but also in sustaining public attention. Throughout each of these various phases, individuals at all levels respond and interact. For most ordinary people, the responses will be small-scale and local. People will watch the news coverage and their imaginations will be quickened by the images they see. They will talk to their friends; attend their local churches; say their prayers; reﬂect and ponder. Some people, the natural leaders in each local setting, will have a larger role. They will be expected to have things to say, and will indeed expect this of themselves. So they too will reﬂect and ponder; and search the newspapers and other natural sources for resources to assist them in deciding on the content of their responses. For an initiating act of sufﬁcient salience, like the 11 September incident, virtually everyone is involved. A vast interactive process of response and counterresponse within the ‘observing community’ is brought into play. For most people, it would take some very considerable effort to remain totally isolated from that process. And such effort is one that virtually no one has any incentive to make. On the contrary, to stay aloof would be to demonstrate a remarkable insensitivity to the major events of one’s day and to the spectacle of others’ suffering and thereby to expose oneself to general contempt. Everybody will be expected to have an emotional response to the events and will be expected to share that response with immediate others. Everybody will probably be expected to have views about what should be done – and will expect others to have such views. In that process by which those expectations are met, beliefs form, attitudes solidify, emotional responses are variously supported and ampliﬁed. If they are not supported and ampliﬁed then they will simply erode with time. Clearly, this response process is an intrinsically ‘social’ phenomenon not just in the sense that it involves individuals in interactions with each other but also in the sense that the process lies beyond the control of any individual: the process takes on a life of its own.2 Individuals, operating atomistically, both respond to and, in the process of their response, create the social, psychological, political and economic environment they face. This is a feature familiar to economists: the process emerges, as we might put it, as if ‘by an invisible hand’ – if indeed it is a ‘hand’ rather than a ‘foot’ or a ‘backhand’. The ‘hand/backhand’ distinction is in fact of the essence. Clearly, it has to be an open question as to whether the response that emerges is appropriate to the initiating cause. It is, for example, logically possible that the ripples will be greater the further we move from the epicentre – at least over some signiﬁcant range. And the result in this case is a kind of epidemic of fear. Of course, this is only one possibility in the variety of trajectories that the responsive process can take, but it is an especially interesting one and deserves a special designation. We shall call it the ‘ampliﬁcation case’ – hereafter the A-case.

288 Geoffrey Brennan I should emphasise that it is not an a priori truth that the A-case involves a disproportionate response in any normative sense. It may be that an epidemic of terror is in certain cases entirely appropriate. Still, the circumstances under which this would be so need some explication. There is, I think, a general presumption that epidemics of terror are undesirable – as the quotations cited at the outset suggest. In any event, in what follows I shall attempt to argue that the current situation in the US is an example of the A-case. The ﬁrst step in this argument is to examine the various senses in which terror might be said to be broadly ‘rational’. This occupies us in Section 2. In Section 3, I examine some empirical evidence that I think supports the ‘over-reaction’ hypothesis. In Section 4, I examine some features of the processes of belief formation in situations analogous to the terrorist threat case. Section 5 brieﬂy explores the political element in the whole reaction process. Section 6 offers a summary conclusion.

2.

The rationality of terror

Terror has an epistemic and an emotional aspect. To have a fear of Z includes not just a particular emotional response to Z: it also requires the belief that Z can cause you harm. The belief can be ‘rational’ or ‘irrational’ depending on whether it connects appropriately with relevant evidence. But the emotion can be broadly rational or irrational too: rational if it stimulates you to act in appropriate ways more quickly than rational calculation might allow; irrational if it stimulates you to act inappropriately. And the emotion can be rational in an additional sense – speciﬁcally, if one enjoys experiencing it for its own sake. I shall begin with this last aspect, both because it is relevant normatively and because it is, I think, important in explaining some elements of the reaction process. I shall then turn to the role that terror might play in shaping action. 2.1.

The intrinsic costs of terror

Most of us, most of the time, ﬁnd fear unpleasant. For the most part, we go out of our way to avoid placing ourselves in fearful situations. For good reason, because if our fears are broadly rational, then the things we are afraid of stand to do us real harm. And if fear is going to do this kind of work, it is necessary that fear be unpleasant to us. That said, it is important to acknowledge that some people, some of the time, enjoy fear. They place themselves deliberately in contexts where they will experience fear or some very close cousin, simply for its own sake. Horror movies, roller-coaster rides and a variety of explicitly ‘dangerous’ pastimes exemplify. In some of these cases, the context is constructed so that the emotion is epistemically unjustiﬁed. Even in the most horriﬁc roller-coaster experiences, for example, people presumably believe, at the point when they buy the ticket, that there is little ‘real danger’. The danger is simulated simply for the ‘pleasure’ of enduring the emotional response. People play with their own emotional states.

Terror: the ‘ISM’ versus the ‘ISTS’

289

Horror movies (including all those with graphic displays of violence, pain and the like) are similar. In this respect, genuinely dangerous pastimes – ones speciﬁcally in which the danger is salient to the participants, like hang-gliding or bungee-jumping or parachuting or car-racing – are different. Here, the activity is genuinely dangerous and the belief that this is so is presumably an element in securing the desired ‘thrill’. However, that thrill is presumably an object of desire in its own right, and one that some people are prepared to endure genuine danger to secure. In all such cases, of course, the fear is in a broad sense under the actor’s control – one can have it on one’s own terms, as it were. When fear is thrust upon one from external and involuntary sources, the presumption of pleasure has to be modiﬁed. Nevertheless, the possibility that fear is, at least over some range, rather enjoyable for many people is important. A certain sense of danger can be energizing, invigorating, sense-sharpening: it can give one’s otherwise rather ﬂat and tired life a sense of adventure and signiﬁcance. For related reasons, images that invoke fear often have a particular salience. In contexts where getting attention is more important than giving pleasure, the fearful will dominate the pleasant. This presumably is part of the explanation of why ‘bad press’ drives out ‘good press’ much of the time. The fact that people ﬁnd fearful things arresting and in some cases positively pleasurable plays a role in explaining two aspects of the ‘over-reaction’ process: the appeal that fearful things have for the media; and the fact that exposure to such things is not itself entirely unpleasant for media-consumers. Nevertheless, as far as the normative aspect is concerned, I shall take it that the emergence of fear across the population is in itself an undesirable thing. In this connection, the fact that people can arrange to experience fear through their leisure activities, if they wish to, actually adds some support to the general claim: fear is generally unpleasant and can only be normatively justiﬁed if there are desirable consequences that spring from having it. 2.2.

The dispositional rationality of fear

As the two epigraphs suggest, fear is not in itself a friend of rationality. Intense fear diminishes the capacity for rational thought: instinctual responses take over. Nevertheless, it seems clear that fear has some evolutionary rationale. There are circumstances in which the rush of adrenalin associated with fear enables a more rapid and effective response to external circumstances than would careful calculation. There are presumably also circumstances when the paralysis that is often associated with fear – and to which Roosevelt refers – is an effective strategy, and one that would be less effectively delivered by more reﬂective processes. When you tread on a snake or are ﬂeeing from a hungry bear, instincts may well be a more satisfactory guide to action than cost-beneﬁt calculus. But equally, Burke and Roosevelt are right that fear is not always a good guide. Most of us learn to manage our instincts – to learn, that is, when to suppress them and when to give them full reign. Fear is a rough-and-ready disposition – triggered in circumstances where failure to act quickly may be disastrous. And in lots of

290 Geoffrey Brennan cases where fear is triggered and turns out to have been ill-founded, the damage done is not so great: you may look a little foolish to have started at shadows, but the greater harm you feared has not eventuated. Fear is like an insurance policy. You will act inappropriately under its operation on many, perhaps even most, occasions. It is just that most of those occasions will be low-cost – whereas the beneﬁts in cases fear is called for will often be extremely large. Whether fear is a good disposition to have depends then upon its being triggered as far as possible in appropriate circumstances. Someone for whom it is inappropriately triggered on a regular basis would be described as having a phobia – a psychological ﬂaw that needs treatment. But lots of us will be fearful in cases where the fear is not justiﬁed and still be entirely rational in retaining our disposition to be fearful. To be afraid of snakes in a country where many snakes are extremely venomous; to be afraid of sharks in an environment where sharks are man-eating; to be afraid of heights – such fears have a clear rationale. Reﬂex actions are not irrational just by virtue of being unreﬂective. Nor is it irrational to retain the reﬂex even though it will sometimes (perhaps often) make you act inappropriately. But only if you are acting inappropriately too often and at too great a cost, will it pay you to suppress the reﬂexes. The general message I take from these considerations is that fear in response to the threat of terrorist activity is rational to the extent that: • • •

The probability of terrorist attack is accurately assessed; The cost of terrorist attack is accurately assessed; and The response that the fear stimulates is superior to that made in the absence of fear.

As I have indicated, I do not want to address the third of these aspects head-on. But I do want to say something about the ﬁrst two, because they are matters on which there is at least some evidence. It is to a consideration of that evidence that I now turn.

3.

Some evidence

Consider ﬁrst the question of the probability of terrorist attack. 3.1.

Probabilities

There is a Gallup poll survey that has been conducted every two weeks since the end of September 2001 in which a random sample of US citizens has been asked the following question: ‘How likely is it that there will be further acts of terrorism in the US of the scale of September 11 over the next few weeks?’ The response options included: ‘very likely’ and ‘quite likely’ alongside three other less pessimistic options. Like most polls, the questions are rather vague and the results difﬁcult to render precisely. What does ‘very likely’ mean? A better than even chance? A probability higher than 10 per cent? And what of ‘quite likely’?

Terror: the ‘ISM’ versus the ‘ISTS’

291

A ﬁve per cent chance? It is not possible to say. But what seems clear is that anyone answering ‘very likely’ or ‘somewhat likely’ thought that the chances were far from negligible. Those respondents identiﬁed the prospect of a further terrorist attack as one that seemed at least quite likely to them, in whatever terms likelihood is to be understood. What is striking here is that at no point in the sixteen months between the end of September 2001 and February 2003 (according to a report in a mid-February 2003 issue of the Economist) did the proportion answering ‘very’ or ‘quite’ fall below 50 per cent. That proportion hit a high in mid-November 2001, when it reached 85 per cent, and for almost all of the period it has been above 60 per cent. A little elementary calculation will show that the probability of some repetition of the Twin Towers episode occurring over a period of sixteen months must be very high if the probability of one occurring over the next few weeks is anything other than negligible. There are 23 three-week periods over the period up to February 2003. There are another 27 three-week periods up to the present (more or less) over which no repeat of the Twin Towers event in the US has occurred. So, assuming that the probabilities are independent, the probability, Q, of a similar terrorist event not occurring over 50 periods is: Q = [1 − p]50 where p is the probability of such an event occurring over the next three weeks. Probability of attack in next three weeks

Probability of no attack over entire period

0.02 0.03 0.05 0.1

0.4 0.25 0.09 0.006

The ﬁgures in this table are to be read as follows. If the probability of an attack occurring in the next few weeks were as high as 5 per cent, then the likelihood that an attack would not have occurred over the entire period to the present is less than 10 per cent. So if a 5 per cent chance is interpreted as being ‘quite likely’, then the US seems to have been incredibly lucky to have avoided something that was 90 per cent likely. Even if ‘quite likely’ is interpreted as being as low as 2 per cent, the chance that people were right in their estimate of the likelihood of attack is against the odds. The basic point is that, for plausible values of ‘quite likely’ and ‘very likely’, the fact that a terrorist episode like 11 September has not been repeated suggests that over 60 per cent of people were just plain wrong. They grossly over-estimated the chance of terrorist attack! The fact that 60 per cent of respondents continue to hold the belief that further terrorist attacks in the next few weeks are ‘very likely’ or ‘somewhat likely’ in the face of the evident

292 Geoffrey Brennan fact that none has occurred over an extended period is enough to throw doubt on the rationality of common beliefs about the likelihood of such attacks. The actual record suggests that terrorist attacks are considerably less likely than most Americans believe. At one level, this misperception of relevant probabilities may not be news. It is well-known that many individuals have considerable difﬁculty in making correct inductions in stochastic contexts. And this seems to be so especially when the probabilities in question are small. Equally, it might be argued that respondents are using the probability as a proxy for expected cost – that the reported measure of probability of attack includes in many cases an implicit allowance for how bad the event contemplated would be. Even rather unlikely events might have a salience for potential victims if the outcome is bad enough. For this reason, ‘very likely’ might be proxy for ‘on my mind’. And questions about the size of probabilities might well be re-interpreted as questions about expected losses, or about the reasonableness of being concerned. But if people cannot distinguish between likelihood of events and the costs associated with them, it is difﬁcult to see how they can be making fully rational calculations. The Gallup poll evidence is not sufﬁcient to demonstrate irrationality of response, but it does indicate that people’s estimates of the likelihood of attack are implausibly pessimistic. It suggests people are excessively frightened. 3.2.

Costs

Misperceptions of probabilities might be justiﬁed if the stakes are very large, but the rational actor tradition suggests that there are advantages in dealing with probabilities and costs separately, as we do here. And when probabilities are quite small and costs are large, there is scope for making errors in estimating both – and no necessary presumption that those errors will cancel out. I have no direct evidence on popular perceptions of the size of the losses that might be endured if another terrorist episode occurs. All one can appeal to here is the nature of media rhetoric, and comparison with other potential benchmarks. On the latter front, the point is occasionally made that deaths from terrorism are much smaller than total road deaths – and have been so in almost all places3 and even in the worst of years. Certainly, this was so in the US in 2001 – and a fortiori in the years before and since. I recognize well enough that death from a deliberate act of violence is not the same as death from accident – for all sorts of reasons. For one thing, one can be reasonably sure that deaths on the road will not escalate overnight. 11 September indicated that there are people who bear hatred towards America and are prepared to die in order to express that hatred; and may have the technical ability to kill very large numbers of others – well beyond the numbers involved in the Twin Towers episode. Even so, the road toll does provide some kind of benchmark – one that is readily available and not totally unreasonable. One might plausibly think that until the terrorist threat is of that order there is no call for desperation – at least if the stoical social attitude to road deaths is to be taken at face value.

Terror: the ‘ISM’ versus the ‘ISTS’

293

But other benchmarks – gruesome though they may be – are available; and in this light too, 11 September appears as representing a relatively low level of catastrophe. The death toll from 11 September was, as I recall, around 3,000 at highest estimate. The total death toll in the Second World War was some 60 million – or 12 million a year, or 33,000 per day for each of the ﬁve years of the war. It is no denial of the tragic character of 11 September to register that WWII was a tragedy of a very different order. One media report referred to September 11 as ‘the blackest day in American history!’. In the face of that sort of extravagance, it is worth recalling that 10,000 lives were lost in a single day at the battle of Antietam – at a time when the US population was very much smaller than in 2001. In fact, there were quite a number of days in the civil war when the death toll was in excess of 3,000. The point of these comparisons is not, of course, to encourage some sort of morbid competition between tragedies, to work out which is indeed the worst. It is simply to make the point that catastrophes come in degrees of greater and lesser. And in framing intelligent responses to bad things – and not least in managing one’s fears – assessing just how bad the things in question are is a crucial component. We have good reason for thinking that carefully computing costs when they are large is something we are not very good at doing individually. Whether we have any reason to think that we might be better at doing it collectively is the issue to which we now turn.

4.

Belief, opinion and convergence

Individuals might be prone to make mistakes in assessing various risks and the best responses to them, but those mistakes might be ameliorated by what we might loosely call ‘social factors’. These factors are of two broad kinds. One set relates to the social context in which individuals test and modify their beliefs: perhaps those social forces work to eliminate signiﬁcant errors. The other set relates to incentives that discourage agents from having false beliefs. In illustrating these two kinds of factors, it will be useful to distinguish between cases where knowledge is subject to a division of labour and those where the knowledge in question is ‘common knowledge’. In the ﬁrst case, not everybody needs to know the same things: the domain of different individuals’ beliefs will be rather different. In the second case, everybody is supposed to know the same things (e.g., common language; common use of measures and weights; etc.). I do not want to say much about the ﬁrst case: much of what needs to be said is familiar from Hayek’s famous discussion of ‘the use of knowledge in society’. It may be useful though to emphasize a distinction within the category of ‘private knowledge’ between what the ‘independent’ [IN] and the ‘negatively dependent’ [ND] cases. In the IN case, I beneﬁt privately from the possession of true knowledge – but the beneﬁts I derive are essentially independent of what others choose to do or what their states of knowledge are. I happen to have, let us suppose, knowledge of

294 Geoffrey Brennan good diet that is beneﬁcial to me – but whether others have that knowledge or not is more or less irrelevant to the beneﬁts that I derive from my possession of it. I can provide the knowledge free to those I care about and urge them to eat accordingly – but whether people in general are obese, or eat in ways injurious to their long-term health is probably not a matter of much concern to me. In any event, even if I do care about people at large, providing the relevant information to them does not inhibit my use of it in any way. In the ND case by contrast the beneﬁts I derive are affected: it will be best for me if others hold contrary views. If for example I am good at predicting stock market movements or interest rate changes, the beneﬁts I derive from that special information are larger the greater the errors that others hold. Some knowledge in relation to terrorist risks is of this latter kind. In the wake of the Bali bombings, holidays became extremely cheap in Bali. There was, it seems, a widespread belief that the risk of further terrorist activity in Bali had increased and people adjusted their holiday plans accordingly. Suppose for the purposes of the argument that this belief involved ‘over-reaction’ – that the estimate of the risk was widely exaggerated. Then I can beneﬁt privately from this error – and can beneﬁt more the more widespread and erroneous it is.4 In both IN and ND cases, I have an incentive to acquire correct information; but the incentive in the ND case is greater the greater the error in others’ beliefs. Especially in the ND case therefore there is a self-corrective mechanism in play in relation to error. So-called ‘efﬁcient markets’ depend on these self-corrective mechanisms. That these corrective elements are in play does not of course mean that ‘markets can’t be wrong’: it just means that those who have correct information will do better than those who do not.5 In both IN and ND cases, the information is desired for instrumental reasons: it makes your life go better if you hold correct rather than false views. Not all cases are like this. In many cases, the primary value of true views is just that they are true. And though most of us think that truth has an intrinsic value, the satisfaction we derive from instantiating that value may not make us particularly conscientious in pursuing it. That is, we may all acknowledge that truth is in principle better than falsehood but unless knowing the truth has some additional instrumental value,6 we will not be motivated to spend much to acquire it. And of course, this remains so for those cases where the instrumental value attached to true beliefs is collective in nature. If the value to me of getting things right relies on the effects of my beliefs on common knowledge across the community, then except in special cases I will have little incentive to make sure I have got things right. The case of collective information is basically like that of zero value information: the only incentive I have for making sure that my beliefs are justiﬁed is the intrinsic value I place on having true beliefs. 4.1.

Social forces in belief formation

Beliefs, to be appropriate, ought to be based on relevant empirical evidence. In many cases, ‘relevant empirical evidence’ is taken to be supplied by what

Terror: the ‘ISM’ versus the ‘ISTS’

295

others believe. There are two kinds of reasons for this. One involves the thought that even if you don’t have evidence for a belief, others are likely to have acquired such evidence – so one can have conﬁdence in ‘following the crowd’. These cases are sometimes referred to as ‘informational cascades’ and we shall use that terminology here. The second reason is that if everyone else believes proposition A you will be thought to be a bit of a fool if you don’t believe it too. You will suffer ‘disesteem’ for your eccentric beliefs. Both these considerations encourage convergence of beliefs, but not necessarily convergence around beliefs that are epistemically warranted. I want to say something about each of these. 4.2.

Informational cascades

Informational cascades work directly on beliefs. (Bikhchandani et al., 1992). An agent induces from behaviour (either actual or counterfactual) that others hold particular beliefs and takes it that those others have epistemic warrant for the beliefs they hold. The belief that others have relevant evidence becomes in turn evidence for the belief itself. As you walk along the street, you ﬁnd that everyone is running with an air of desperation in the opposite direction. You induce that there is some reason to ﬂee. You turn and start running too. And you have some reason to do so – even if you don’t know what exactly the substantive cause of their ﬂeeing is. Fedderson and Pesendorfer (1998) give an example where the reasoning is from behaviour counterfactually rather than actually observed. You are on a jury where the requirement for ﬁnding guilt is unanimity. Thus the only case in which you will be decisive is where all others ﬁnd the accused guilty: in all other cases, your judgement is irrelevant. So you reason on the basis that that case applies. But in that case, all the others – who have heard the evidence just like you and have made their own assessments of witness credibility – have decided that the prisoner is guilty. That fact gives you pause. It is a reason for thinking that the accused is guilty. So all jurors think this way. And there are too many convictions. Or so the argument goes.7 Condorcet jury theorems are often taken to rationalize the view that there is epistemic value in majority belief. But it is worth emphasizing that these results depend on the assumptions, ﬁrst that any randomly selected person is more likely to be right than wrong, and second that all individuals make independent judgements. And both these assumptions are questionable. In the ﬁrst place, assuming that on any issue each person is more likely than not to ‘get it right’ is in one sense to assume what is to be shown. The Condorcet jury theorem implies that majority rule in a sufﬁciently large body never gets it wrong. But this is an implausibly strong result. One might think that the majority gets it right more often than not (presumably most democraphiles do) without thinking that this implies that each individual gets it right more often than not across the full set of possible issues. Second, if each uses the jury theorem itself as a consideration in deciding how to vote, then the independence requirement is violated. In that sense the Condorcet theorem cannot represent a reason for following the crowd, because if signiﬁcant numbers did so, the theorem would be imperilled. In well-controlled institutional

296 Geoffrey Brennan contexts, we can perhaps ensure that independence applies (by preventing people from speaking to one another before casting their votes perhaps). But in the formation of ‘common knowledge’ we simply cannot know how many people are following the crowd whether for Condorcet-related reasons or otherwise. In such environments, the ‘informational cascade’ logic explains why beliefs will converge; but it doesn’t give us much reason for thinking that the convergent beliefs are true! In the terrorism context, we can explain how, if a signiﬁcant number of people come to believe that there is a large threat of further terrorist activity, they may carry others along with them in that belief. However, in the circumstances surrounding terrorist threats, it seems highly dubious whether the convergence processes tell us anything other than that there are convergence processes at work. 4.3.

Esteem effects

In my view however many of the forces for such convergence are associated with ‘social pressure’ rather than epistemic considerations (Brennan and Pettit, 2004). The classic exempliﬁcation of this case is the ‘emperor’s new clothes’ fable, in which each observer acts as if he observes the emperor’s clothes in order to avoid the disesteem attached to not holding the relevant belief. A less ‘fairy tale’ example is provided by the case of so-called ‘pluralistic ignorance’ (Prentice and Miller, 1993). The PI case involves three features: • • • •

A belief bi held by each person; A belief b’i held by each about others’ beliefs bj; Behaviour by each in accord with b’i; b’i is false.

Suppose for example that each student in a college dorm believes that drunkenness is disgusting. Each believes that virtually all the other students think that drunkenness is ‘cool’. So each student has bouts of regular drunkenness in order to avoid the (presumed) disesteem of his fellows. And no-one expresses doubts about the desirability of drunkenness for the same reason. So everyone gets drunk regularly, and all of them use that fact to induce that others think that drunkenness is ‘cool’. (There is, incidentally, evidence that supports the view that this is indeed the structure of beliefs around drunkenness among students.) Here, the desire for esteem drives the behaviour. The behaviour drives the belief about others’ values. But the assessment of others’ values is in error. The resultant behavioural/belief equilibrium is however entirely stable to local shocks. The forces of esteem encourage convergence of belief-related behaviour at two possible levels. First, there is the desire to be ‘agreeable’. If A holds a view strongly and advances that view publicly, A is also likely to disesteem people who hold contrary views – to regard them as stupid, or wicked or both. And obversely to esteem those who agree with him as astute and holding the right values.

Terror: the ‘ISM’ versus the ‘ISTS’

297

Others know this. So if they wish to secure A’s esteem, they will be led to express support for A’s views. Or at least keep their objections silent and thereby avoid A’s disesteem. It is, after all, ‘disagreeable’ to be disagreeable; and most of us try to be agreeable much of the time – indeed, whenever the cost of doing so doesn’t become too high. Esteem also operates at a further level, slightly removed. If you go along with the prevailing view – simply, let us say, by holding your peace – you do not require anything positive in the way of argument. The onus typically lies with those who dispute the prevailing wisdom. So if most people in your set believe p and you think ∼p, you are the one who is going to have to produce the arguments. To say staunchly ‘∼p!’ and then offer nothing by way of reasoning to support your claim is just to reveal yourself as an opinionated idiot. But, though you may think ∼ p, you may not feel that you have a good case for what you think or you may doubt your capacity to make it convincing. It is asymmetrically difﬁcult to contest the epistemic status quo. And this fact constitutes an incentive structure to which people predictably respond. Of course, there are cultures in which holding unconventional or unpopular views (or mounting arguments for such views) is a positive source of esteem. The academic culture can be one of them. However, even in academia the approved adventurism is often within fairly limited bounds. Those who dispute the core nostrums of their discipline can expect troubled professional careers – and had better have very good arguments for their case. Safer to restrict one’s clever contentions to the margins of professional concern! In short, there are forces that serve to encourage convergence around majority views and to preserve those views, once in place. Some of these forces may have some epistemic validity; but the association between popularity and truth is in most cases pretty thin. As professional economists, for example, we have good reason for thinking that much that passes for ‘common knowledge’ about the workings of the economy among the population at large is at best highly contestable and at worst downright wrong. Yet there are incentives for individuals to understand how the economy works because there is money to be made from doing so (or money to be lost in not doing so). By contrast, the private advantages from being roughly right about the expected cost of terrorist activities – and the expected costs and beneﬁts of various terrorism-related policies – are for most people quite small. Predictably, what I have said so far about the epistemic authority of common beliefs has focused on the incentives of individuals in various settings to acquire relevant information and/or dispute apparently widely held views. It would be strange not to mention the media in this connection, largely because these are the primary vehicles for the dissemination of information – both directly about the events themselves and indirectly about the judgements in relation to those events that are taken to be the prevailing wisdom. I do not, though, want to focus on the media because too often the media are made the scapegoat for all failings. It will sufﬁce here to make just three brief points. First, by and large, the media give the public what it wants – and if what

298 Geoffrey Brennan it wants does not measure up to the highest ideals of intellectual enquiry, that is more the public’s fault than the media’s. Second, the media always makes a trade-off between getting attention and accuracy of reporting. I do not think that television broadcasters often recycle ﬁlm from other scenes or directly misrepresent events – or that newspaper reporters simply manufacture the quotations they attribute to bystanders or public ﬁgures. But news that is arresting, that grabs attention, that induces a more intense emotional response, gets more attention than news that is carefully balanced and meticulously faithful. Arguably, the trade-off between attention-seeking and faithfulness alters as news is provided jointly with advertising, because circulation/ratings become more signiﬁcant in the advertising setting. If so, then cable TV is something that is to be encouraged; and books and journal articles that are advertisement free are more to be trusted. Third, cures can be worse than the disease they address. Policies that aim to open up the media to competition are one thing. Measures designed to inhibit information ﬂows on the grounds that the information might cause public panic are the stuff of repression. Bruno Frey (2004, ch. 7) has argued that we ought to consider the policy of suppressing the names of terrorist groups who claim responsibility for various terrorist activities on the grounds that the publicity for their cause is a large part of what such groups seek. One can see the logic of this kind of policy; but I cannot say that I have any enthusiasm for it. Once governments start ﬁltering the news the dangers seem to me to be much greater than the beneﬁts. In short, any such policy would seem to me to be just another instance of the kind of over-reaction I deplore.

5.

The politics of terror

Politics, Hume famously remarked, is driven by public opinion. But this way of putting things perhaps understates the role that governments themselves – and political processes more generally – contribute to the formation of public opinion. Certainly within the expressive account of political process that I personally favour, rhetoric plays a signiﬁcant role; political ‘leaders’ are expected by ordinary citizens to have the capacity to comment intelligently on major events, and to interpret their signiﬁcance to the community. And politicians are selected for their capacity to do this well. So even where no policy response as such is called for, political ﬁgures will play some role in the formation of prevailing views. But often policy response will be called for. And such policy responses will have to be ‘sold’ to the electorate – which means that the (often formidable) rhetorical resources of the government will be arrayed in support of policies chosen. By implication, an important feature of the choice among alternative policy responses will be how amenable the various options are to persuasive rhetorical presentation. So while it is correct to depict governments as responding to public opinion, it is also important to see governments as important players in the processes by which public opinion is formed.

Terror: the ‘ISM’ versus the ‘ISTS’

299

It is I think an implication of the expressive account of politics that large, salient events call for correspondingly large symbolic political responses. Those responses will be partly rhetorical and partly substantive policy – but both will have to match in an appropriate way the events to which they respond. I think this requirement often makes for over-reaction of a very particular kind. To illustrate, let me take a natural disaster context. Suppose that ﬁre and ﬂood are about equally likely natural disasters in a particular environment. The effects of both can be ameliorated by certain protective measures, which are expensive and therefore competitive. Suppose a bad ﬁre occurs at time t1. Suppose that there is no reason to think that this fact alters the relative likelihood of a bad ﬁre or a bad ﬂood at t2. But if there were a bad ﬂood, people would be inclined to think that the community had been just very unlucky to sustain two bad events in successive years. On the other hand, if there were another bad ﬁre and there had been no expansion in protective policy in the meantime, policy makers’ heads would roll. That is, the political costs of failure to react to the disaster that is salient are disproportionately high as compared with relative risks. To do nothing (further) about a salient risk would be considered culpable, in a way that doing nothing about an equal but much less salient risk would not be. If this is so, then there are systematic forces supporting ‘over-reaction’ – with over-reaction here measured against the best estimate of objective risks. Now if relative salience is an important feature of this account, it needs to be noted that political ﬁgures contribute to this salience. When natural disasters strike, there is a distinctive opportunity for political leaders to gain exposure: they can wander around the ruins, express the sense of catastrophe the community naturally feels, and generally act the statesman. Well handled, there is on offer a signiﬁcant increase in both general recognition and popularity. So it is to an incumbent’s political advantage not just for the disaster to have happened in the ﬁrst place but also for it to remain salient for a long time. Of course, it would be politically disastrous for a candidate to be causally implicated in the occurrence of the disaster. But he may well do a little (discreetly) to keep the catastrophe on the local agenda: and certainly the longer it stays there, the better for his political fortunes. In the conventional ‘interest-driven’ account of political process that tends to predominate in public choice circles, there is an important presumption of underlying stability in electoral demand. Certainly, the composition of majorities can change and this source of instability has been a signiﬁcant element in the standard account. Still, the interests themselves are broadly stable – and any instability is more the result of internal forces, essentially independent of external events. In the expressive account, the sources of instability are more external, connected to salient events and the emphases that those external events stimulate. And for this reason, there is a greater scepticism about the normative authority of popular preferences. One important element of the public choice apparatus, however, is relevant here – namely, the idea of equilibrium and the forces of convergence around

300 Geoffrey Brennan that equilibrium. After a natural disaster, political leaders of all stripes will try to outdo each other in expressions of sympathy, indignation and determination that something should be done, and so on. If my claims of a tendency towards over-reaction are correct, they are certainly not the monopoly of any particular party or faction. Although Bush is clearly identiﬁed with the Iraqi war and the Patriot Act, the notion of political equilibrium suggests that the policy response to September 11 by a Democrat administration is unlikely to have been very different. Or in the Australian case, it seems unlikely that the opposition Labor party, had it then been in ofﬁce, would not have supported US policy in the way that the Liberal government actually did. The claim that democratic political processes and other features of Western society are especially hospitable to ‘over-reaction’ to salient events like 11 September seems likely to be independent of who happens to be in power when the salient event occurs.8

6.

Summary and conclusion

In this paper, I have argued for three things: • • •

A shift in the focus of attention away from the motivations, incentives and likely behaviour of ‘terrorists’ and towards the mechanisms of response that effectively mobilize the terror; Recognition of the fact that most people’s capacities to make correct judgements of expected costs in the case of small probability/large cost events is generally poor; And recognition that such capabilities are likely to be especially poor in relation to calculations of political relevance and optimal policy response.

I think there is some presumptive evidence that these features have all been present in the US response to 11 September. There is at least presumptive evidence to support the claim that a majority of US citizens overestimate the likelihood of terrorist attack and hold false beliefs about the likely cost of any such attack should it be successful. If this is so, it is a bad thing not just because it is likely to induce excessive policy ‘response’ but also in itself. Terrorism does terrorize; and having a deeply anxious population is only acceptable if that anxiety is necessary to secure proper response. Given my conviction that our current social institutions are excessively hospitable to the creation of ‘terror’ and to associated policy overreaction, I ought no doubt recommend institutional changes that would diminish such hospitality. In fact I make no such recommendations here. That ‘institutional design’ exercise seems to me to be somewhat downstream. At this point, I am more interested in uncovering the processes – in diagnosing the disease rather than in selling my own patent cure. It may even be that there is no institutional cure that is not worse than the disease. Even so, I think that it might be a good thing if more scholars were more sensitive to what I see as a general susceptibility to over-reaction. Recognizing a phobia seems like a good start in overcoming it.

Terror: the ‘ISM’ versus the ‘ISTS’

301

Acknowledgements I am grateful to participants in the 2004 Buchanan Colloquium and to Tilman Brück for useful comments on earlier drafts. Responsibility for remaining inadequacies is entirely my own.

Endnotes 1 And not even a necessary piece since terror can follow in the wake of natural phenomena like earthquakes, bushﬁres, ﬂoods, tsunami, epidemics and the like. In these cases too there can be ‘over-reaction’ – as I believe there probably was in the recent SARS ‘epidemic’ threat. In some ways, these ‘natural disaster’ cases are simpler because there is no perpetrator to be punished or discouraged or accommodated – and a less personalized emotional response. Most of the elements of the response process that are discussed here apply no less in the natural disaster cases. To say this is, however, not to deny that the case where the initiating harm comes about by virtue of intentional human agency evokes special emotional and normative responses that are themselves an important feature of the response process. 2 A few very salient or rhetorically gifted individuals may be able to inﬂuence the process to some extent – but these persons may themselves be as much tools of events as they are genuinely independent players. 3 Israel/Palestine has, I believe, only just become an exception. And even there the difference between road deaths and terrorism deaths is not so very great – at least if I understand the situation accurately. 4 The advantages depend of course on market prices being inversely related to the extent of demand. For my example, upward-sloping supply curves are assumed. 5 In short, I do not take the fact that the price differential for Bali vacations stayed in place for a long time – my sense is that it is still present – is decisive evidence that the probability of further terrorist activity there was sufﬁcient to justify that price differential. The price differential might enable one to calculate the ‘marginal belief’ – but it does not enable one to determine whether that marginal belief is accurate. 6 I include in this extra instrumental value of course the professional beneﬁts among academics of ‘knowing the facts’ and the beneﬁts in esteem of being proven right. 7 Fedderson and Pesendorfer use their reasoning to mount an argument against unanimity as a decision rule. Majority rule would not, for example, be vulnerable to the same distortion. Under simple majority rule, the case in which one is decisive is that where the others are exactly divided: so you can proceed to call the shots exactly as you see them. 8 An implication of this line of reasoning is that popular democracies may be quite hospitable to war. This runs against a famous claim of Kant’s about an alleged association between democracy and ‘perpetual peace’ – but I see little evidence or a priori grounds in support of such an association. I am indebted to Tilman Brück for this point. See also Brennan and Hamlin (2004).

References Bikhchandani, S., Hirshleifer, D. and Welch, I. (1992) ‘A Theory of Fads Fashions, Custom and Cultural Change as Informational Cascades’, Journal of Political Economy, 100: 992–1026. Brennan, G. and Hamlin, A. (2004) ‘The European Constitution and Peace: Taking the Heat out of Politics’, in Blankart, C. and Mueller, D. (eds) A Constitution for the European Union MIT Press, Cambridge, Mass. 1–24.

302 Geoffrey Brennan Brennan, G. and Pettit, P. (2004) The Economy of Esteem, Oxford: Oxford University Press pp. 339 + xii. Fedderson, T. and Pesedorfer, W. (1998) ‘Convicting the Innocent’, American Political Science Review, 92: 23–35. Frey, B. (2004) Dealing with Terrorism: Stick or Carrot?, Cheltenham: Edward Elgar pp. 182 + xvi. McLuhan, M. and Powers, B. (1988) The Global Village, Oxford: Oxford University Press. Prentice, D. and Miller, D. (1993) ‘Pluralistic Ignorance and Alcohol Use on Campus’, Journal of Personality and Social Psychology, 64: 243–56.

Economic Analysis of Terrorism (Routledge Studies in Defence and Peace Economics)

Competing Economic Theories (Routledge Studies in the History of Economics)

Interpreting Classical Economics: Studies in Long-Period Analysis (Routledge Studies in the History of Economics)

In Defence of Objectivity (Routledge Studies in Critical Realism)

Contemporary State Terrorism: Theory and Practice (Routledge Critical Terrorism Studies)

Arms Trade and Economic Development: Theory and Policy in Offsets (Studies in Defence Economics)

Contemporary Economics (Routledge Studies in the History of Economics)

Contemporary Economics (Routledge Studies in the History of Economics)

Targeting Development (Routledge Studies in Development Economics)

Rethinking the Roots of Terrorism (Rethinking Peace and Conflict Studies)

Classical Economics and Modern Theory: Studies in Long-Period Analysis (Routledge Studies in the History of Economics, 63)

Economics and the Mind (Routledge INEM Advances in Economic Methodology)

Historians of Economics and Economic Thought: The Construction of Disciplinary Memory (Routledge Studies in the History of Economics)

Equilibrium and Economic Theory (Routledge Studies in the History of Economics, 11)

The Laws of History (Routledge Studies in History of Economics)

Game Theory and Economic Analysis: A Quiet Revolution in Economics (Routledge Advances in Game Theory, 001)

Entrepreneurship and Economic Development (Studies in Development Economics and Policy)

Economic Careers: Economics and Economists in Britain 1930-1970 (Routledge Studies in the History of Economics, 15)

Understanding 'Classical' Economics: Studies in Long-Period History (Routledge Studies in the History of Economics, 16)

Ideas, Policies and Economic Development in the Americas (Routledge Studies in Development Economics)

Open Economics: Economics in relation to other disciplines (Routledge Studies in the History of Economics)

Routledge Dictionary of Economics

A Bibliography of Female Economic Thought to 1940 (Routledge Studies in the History of Economics)

Europe and Economic Reform in Africa: Structural Adjustment and Economic Diplomacy (Routledge Studies in Development Economics, 9)

Pax Pacifica: Terrorism, the Pacific Hemisphere, Globalization, and Peace Studies

Europe and Economic Reform in Africa: Structural Adjustment and Economic Diplomacy (Routledge Studies in Development Economics, 9)

Measurement, Quantification and Economic Analysis: Numeracy in Economics

Socialism and Marginalism in Economics 1870-1930 (Routledge Studies in the History of Economics)

Game Theory and Economic Analysis: A Quiet Revolution in Economics

Microfinance Perils and Prospects (Routledge Studies in Development Economics)

Social Economics of Jean-Baptiste Say: Markets and Virtue (Routledge Studies in the History of Economics)

Economic Analysis of Terrorism (Routledge Studies in Defence and Peace Economics)

Competing Economic Theories (Routledge Studies in the History of Economics)

Interpreting Classical Economics: Studies in Long-Period Analysis (Routledge Studies in the History of Economics)

In Defence of Objectivity (Routledge Studies in Critical Realism)

Contemporary State Terrorism: Theory and Practice (Routledge Critical Terrorism Studies)

Arms Trade and Economic Development: Theory and Policy in Offsets (Studies in Defence Economics)

Contemporary Economics (Routledge Studies in the History of Economics)

Contemporary Economics (Routledge Studies in the History of Economics)

Targeting Development (Routledge Studies in Development Economics)

Rethinking the Roots of Terrorism (Rethinking Peace and Conflict Studies)

Classical Economics and Modern Theory: Studies in Long-Period Analysis (Routledge Studies in the History of Economics, 63)

Economics and the Mind (Routledge INEM Advances in Economic Methodology)

Historians of Economics and Economic Thought: The Construction of Disciplinary Memory (Routledge Studies in the History of Economics)

Equilibrium and Economic Theory (Routledge Studies in the History of Economics, 11)

The Laws of History (Routledge Studies in History of Economics)

Game Theory and Economic Analysis: A Quiet Revolution in Economics (Routledge Advances in Game Theory, 001)

Entrepreneurship and Economic Development (Studies in Development Economics and Policy)

Economic Careers: Economics and Economists in Britain 1930-1970 (Routledge Studies in the History of Economics, 15)

Understanding 'Classical' Economics: Studies in Long-Period History (Routledge Studies in the History of Economics, 16)

Ideas, Policies and Economic Development in the Americas (Routledge Studies in Development Economics)

Open Economics: Economics in relation to other disciplines (Routledge Studies in the History of Economics)

Routledge Dictionary of Economics

A Bibliography of Female Economic Thought to 1940 (Routledge Studies in the History of Economics)

Europe and Economic Reform in Africa: Structural Adjustment and Economic Diplomacy (Routledge Studies in Development Economics, 9)

Pax Pacifica: Terrorism, the Pacific Hemisphere, Globalization, and Peace Studies

Europe and Economic Reform in Africa: Structural Adjustment and Economic Diplomacy (Routledge Studies in Development Economics, 9)

Measurement, Quantification and Economic Analysis: Numeracy in Economics

Socialism and Marginalism in Economics 1870-1930 (Routledge Studies in the History of Economics)

Game Theory and Economic Analysis: A Quiet Revolution in Economics

Microfinance Perils and Prospects (Routledge Studies in Development Economics)

Social Economics of Jean-Baptiste Say: Markets and Virtue (Routledge Studies in the History of Economics)

Recommend Documents