The Arms Trade: Security and Conflict (Routledge Studies in Defenseeconomics, 5)

Arms Trade, Security and Conﬂict The arms industry and the arms trade are areas of immense importance to both producer...

Author: Paul Levine

86 downloads 506 Views 1MB 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

DOWNLOAD .DOCX DOWNLOAD .PPTX

Arms Trade, Security and Conﬂict

The arms industry and the arms trade are areas of immense importance to both producers and users throughout the world. Yet, at the same time, the economics of this sector is still a relatively unexplored strand of economics. This volume seeks to ﬁll this lacuna by bringing together leading experts from all over the globe, including Todd Sandler and Keith Hartley, and focuses on the important issues surrounding the arms trade such as: • • •

new challenges to arms export controls; alliance formation and expansion; arms races and military expenditure.

This well-rounded, comprehensive book will be of huge interest to students and academics involved in the economics of defence and conﬂict resolution, as well as more general studies of the defence sector. The book will also be an invaluable read for those with an interest in the arms industry such as policy-makers and corporate managers. Paul Levine is Professor of Economics at the University of Surrey, UK. Ron Smith is Professor of Applied Economics at Birkbeck College, University of London, UK.

Studies in defence economics Edited by Keith Hartley University of York

Jurgen Brauer Augusta State University

Volume 1 European Armaments Collaboration Policy, problems and prospects R. Matthews Volume 2 Military Production and Innovation in Spain J. Molas-Gallart Volume 3 Defence Science and Technology Adjusting to change R. Coopey, M. Uttley and G. Spiniardi Volume 4 The Economics of Offsets Defence procurement and countertrade S. Martin Volume 5 Arms Trade, Security and Conﬂict Edited by Paul Levine and Ron Smith Volume 6 Economic Theories of Peace and War F. Coulomb Volume 7 From Defense to Development? International perspectives on realizing the peace divident A. Markusen, S. DiGiovanna and M. Leary

Arms Trade, Security and Conﬂict

Edited by Paul Levine and Ron Smith

First published 2003 by Routledge 11 New Fetter Lane, London EC4P 4EE Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 Routledge is an imprint of the Taylor & Francis Group

This edition published in the Taylor & Francis e-Library, 2005. “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.” © 2003 Paul Levine and Ron Smith for selection and editorial matter; individual contributors for their chapters 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 Arms trade, security and conﬂict / [edited by] Paul Levine and Ron Smith. p. cm. – (Studies in defence economics ; 5) This conference volume contains a set of papers given at the June 1999 conference at Middlesex University Business School on the arms trade security and conﬂict. Includes bibliographical references and index. 1. Arms transfers – Congresses. 2. Defense industries – Congresses. 3. National security – Congresses. I. Levine, Paul, 1944– II. Smith, Ron, 1944– III. Studies in defence economics (Chur, Switzerland) ; v. 5. HD9743.A2 A747 2003 382 .45355–dc21

ISBN 0-203-47716-2 Master e-book ISBN

ISBN 0-203-33941-X (Adobe eReader Format) ISBN 0–415–30648–5 (Print Edition)

2002036939

Contents

List of ﬁgures List of tables List of contributors Acknowledgments

1 Introduction

vii ix xi xiii

1

PAU L LEV I N E A N D RO N S MI TH

PART I

Arms exports and arms production

3

2 The economics of UK arms exports

5

K EI TH H A RTLE Y A N D S TEP H EN MA RTI N

3 Potential and actual arms production: implications for the arms trade debate

21

J U RG EN BRAU E R

4 Export controls, market structure and international coordination

37

MA RÍ A D EL CAR MEN G A RCÍ A - A LO N S O A N D K EI T H HART L E Y

5 Arms export controls and emerging domestic producers

55

PAU L LEV I N E, F OTI S MO U ZA K I S A N D RO N S MI TH

6 The supply-side implications of the arms trade: UK aerospace industry, economic adjustment and the end of the Cold War

78

I A N JACK SO N

7 New challenges to arms export control: whither Wassenaar? RO N SMI TH A N D B ER NA RD U D I S

94

vi

Contents

PART II

Wars, alliances and arms races 8 Rational wars with incomplete information

111 113

PAU L LEV I N E A N D F R A N CI S C O MO RA I Z

9 Alliance formation, expansion and the core

137

TO D D SA N D LE R

10 The determinants of US military expenditures in the context of arms race

158

SY LV I E MATELLY

11 Arms race models and econometric applications

178

J . PAU L D U N N E , EF TY C H I A N I KO LA I D O U A N D RON S M I T H

12 Demand for space expenditure and the space race between NATO and the USSR: an econometric analysis

188

VA SI LI S ZERVO S

PART III

Overview

207

13 Arms trade, arms control and security: collective action issues

209

TO D D SA N D LE R

Index

221

Figures

5.1 5.2 5.3 6.1 6.2 6.3 6.4 6.5 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 8.10 8.11 8.12 8.13 9.1 9.2

The international market for arms The threshold value of military capability The Nash equilibrium in military capability Global defence spend Estimated civil ﬂeet in service – all jets Rising unit cost of military aircraft Unit costs and numbers The Euroﬁghter 2000–Tornado production gap The production function parameter α and the resulting outcome when investment in war is ﬁxed The change in the initial priors υ and the resulting outcome when investment in war is ﬁxed Graphs representing the relation between cost of war and types of countries The effect of changes in the cost of a soft country 2 when investment in war is ﬁxed The conﬂict roller-coaster The payoffs effect of changes in υ0 when investment in war is optimal (small probability of war) The payoffs effect of changes in υ0 when investment in war is optimal (small probability of peace) The payoffs effect of changes in υ0 when investment in war is optimal (increased probability of war) Resources dedicated to ﬁght and prior beliefs Resources dedicated to ﬁght and cost asymmetry Optimal allocations used in a graph representing the probability of war The contradictory effects of the discounting factor is shown After several simulations, we found out that according to our model, strategy 2 is rarely played Alternative alliance conﬁgurations Mutual defense, baseline case

57 61 63 80 81 82 83 85 121 122 123 124 125 127 128 128 129 130 131 132 132 142 143

viii Figures 9.3 9.4 10.1 10.2 10.3 11.1 11.2 13.1 13.2 13.3

Mutual defense, short middle ally Mutual defense, long middle ally US and USSR military expenditures, 1952–97 Observed and estimated military expenditures US economic size and military implication Military expenditure for Greece and Turkey (in constant 1990 million US$) Military expenditure for India and Pakistan (in constant 1995 million US$) Summation and Prisoner’s Dilemma Best shot and coordination game Weakest link and matching

145 146 164 169 170 182 184 212 213 214

Tables

2.1 2.2 2.3 2.4 3.1 3.2 3.3 3.4

UK arms trade (£ million: current prices) Beneﬁts to the United Kingdom Limiting UK arms exports Employment functions for the UK defence industry, 1978–96 Number of arms producers, c.1975–83 vs c.1990–5 ISIC revisions 2 and 3 codes to determine countries’ PDC index PDC index by AP-level, early 1990s Average industrial diversiﬁcation among 283 ISIC revision 2 sub-categories, 1986–95 3.5 School enrollment data 4.1 Two-choice game 4.2 Membership of multilateral military-related export control regimes, as of 1 January 1998 5.1 The emergence of b-producers as price P increases 5.2 Summary of model 5.3 Comparison of arms trade regimes 5.B1 Summary of calibration 6.1 Life cycle costs of ﬁghting units 6.2 The Al-Yamamah project 6.3 The battle of the sexes 6.4 The battle of the sexes in UK aerospace production 6.5 The battle of the sexes in the European aerospace industry 8.1 Three strategies for country 1 8.2 Expected payoffs for the three strategies 8.3 Possible equilibrium outcomes 8.4 Baseline parameter values 8.5 Values for the baseline calibration of optimal allocations 8.A1 Three strategies for country 1 8.A2 Expected payoffs for the three strategies

6 8 10 17 22 26 27 29 32 43 49 64 71 72 75 84 86 88 88 88 120 120 120 121 129 134 135

x

Tables

10.1

Five theoretical examples of action and reaction processes of armaments dynamics between two opposite countries 10.2 Empirical results of the ﬁve theoretical relations (estimation period, 1952–97) 10.3 Correlation coefﬁcients of US and USSR military expenditures and armed forces, 1952–89 and 1952–97 10.4 Autocorrelation and partial correlation for US military expenditures between 1952 and 1989 10.5 Evaluation of the optimal lag for USSR military expenditures 10.6 Model 1 10.7 Determining the lag of model 2 10.8 Model 2 10.9 American defence budget and public deﬁcit (in billions of US current dollars and % of GNP) 10.10 The possible factors of American grievance between 1952 and 1997 12.1 Modelling LSEEU by OLS 12.2 Modelling LSEUS by OLS (sample range 1970–90) 12.3 Modelling LSEUSSR by OLS (sample range 1970–90) 12.4 Unit root tests for LSEUS and LSEUSSR (sample: 1970–90; critical values: 5% = −3.011; 1% = −3.785; constant included) 12.5 Modelling LSEUS by RLS (the present sample is: 1970–90) 12.6 Modelling LSEUS by OLS 12.7 Modelling LSEUS by OLS (the present sample is: 1961–94 less four forecasts, the forecast period is: 1991–4) 12.8 Analysis of 1-step forecasts 12.9 Unit root tests for zt (sample range: 1970–90; critical values: 5% = −3.011; 1% = −3.785; constant included) 12.10 Modelling LSEUS by OLS (sample range 1970–90)

161 161 163 165 166 167 167 168 171 173 194 195 196 197 198 198 199 199 200 201

Contributors

Jurgen Brauer is Professor of Economics in the College of Business Administration at the Augusta State University. J. Paul Dunne is Professor of Economics in the School of Economics, University of the West of England. María del Carmen García-Alonso is Lecturer in the Department of Economics at the University of Kent. Keith Hartley is Professor of Economics in the Centre for Defence Economics at the University of York. Ian Jackson is Senior Lecturer in Economics at the Staffordshire University Business School. Paul Levine is Professor of Economics in the Department of Economics, University of Surrey. Stephen Martin is a Research Ofﬁcer in the Centre for Defence Economics at the University of York. Sylvie Matelly is a Defence Economist at the Institute for International and Strategic Relations (IRIS) in Paris and lectures in International Economy and Trade at the University of Paris XIII. Francisco Moraiz is Lecturer in the Department of Economics at the University of St Andrews. Fotis Mouzakis is Lecturer in Real Estate in Cass Business School at the City University. Eftychia Nikolaidou is Lecturer in Economics at City Liberal Studies, Thessaloniki. Todd Sandler is Professor of Economics in the School of International Relations at the University of Southern California. Ron Smith is Professor of Applied Economics in the Department of Economics, Birkbeck College, University of London.

xii

Contributors

Bernard Udis is Emeritus Professor in the Department of Economics at the University of Colorado at Boulder. Vasilis Zervos is Lecturer in Industrial Economics at the Nottingham University Business School.

Acknowledgments

The editors and publishers would like to thank the following for granting permission to reproduce material in this work: Sage Publications Inc. for permission to reproduce “Alliance Formation, Alliance Expansion and the Core,” Todd Sandler, from the Journal of Conﬂict Resolution 43(6), 727–747. Monterey Institute of International Studies for permission to reproduce “New Challenges to Arms Export Control: Whither Wassenaar?” Ron Smith and Bernard Udis, from The Nonproliferation Review 8(2), 81–92. Taylor & Francis Plc for permission to reproduce chapters 2, 3, 4 and 5. Earlier versions of chapters 2, 3, 4, 5 and 13 were published in a special issue of Defence and Peace Economics, 11(5), 2000, on Arms Exports, Controls and Production edited by Paul Levine, Somnath Sen and Ron Smith. Every effort has been made to contact copyright holders for their permission to reprint material in this book. The publishers would be grateful to hear from any copyright holder who is not acknowledged here and will undertake to rectify any errors or omissions in future editions of this book.

1

Introduction Paul Levine and Ron Smith

The arms trade is interesting because it is where foreign policy concerns such as security, human rights and international order interact most directly with economic concerns such as trade, jobs and proﬁts. The control of arms exports remains a matter of policy concern both within countries and within multilateral organisations. This Conference Volume contains a set of papers presented at the June 1999 Conference at Middlesex University Business School on the Arms Trade Security and Conﬂict. The Conference marked the end of a 4-year ESRC-funded project on the Arms Trade. A full description of the project and other papers produced are available at http://carecon.org.uk/Armsproduction. One can imagine supplier governments determining their levels of arms exports in the light of their economic and security objectives subject to the support of their allies; the constraints of the market; and the nature of the demand for arms by buyer governments. The constraints of the market may include either competition from other suppliers or various forms of cooperation including arms exports controls, agreements between suppliers to restrict sales to particular destinations. The nature of the demand by buyer governments will reﬂect their security objectives, alliances, budget constraints and alternative sources of arms, in particular, the possibility of domestic production. The chapters in this volume examine various aspects of this process at both the empirical and theoretical levels. Part I presents chapters looking at the supply-side and Part II at the demand-side. Todd Sandler was rapporteur at the Conference and his chapter, the ﬁnal one in the volume, pulls together the themes discussed at the Conference as examples of collective action problems. Part I on the supply-side deals with arms exports and production. Keith Hartley and Stephen Martin present the stylised facts of the UK arms trade and examine the costs and beneﬁts of arms exports, the economic impact of limiting arms exports and employment behaviour in the export sector. This opening chapter sets the scene for the rest of the book by demonstrating the opportunities for analytical and empirical work in the emotive area of the arms trade. Jurgen Brauer looks at the evidence on the capability to produce arms and actual production in developing countries. María García-Alonso and Keith Hartley consider controls on dual-use products and the potential for multiple equilibria in the market. Fotis Mouzakis, Paul Levine and Ron Smith examine how the possibility of domestic production by importing nations changes the effectiveness of arms export controls. Ian Jackson

2

Paul Levine and Ron Smith

looks at the supply-side implications of the arms trade, with an emphasis on the United Kingdom and its role in the restructuring of the European industry. The ﬁnal chapter in Part I, by Ron Smith and Bernard Udis, reviews the institutional structure of arms export regulations and the practical difﬁculties such regulation faces. Part II examines the demand-side: wars, alliances and arms races. The ultimate driver of demand for arms is war or the fear of war. Paul Levine and Francisco Moraiz examine possible rational actor models of the decision to go to war based on imperfect information. The demand for arms is often a joint decision in combination with allies. Todd Sandler examines alliance formation, alliance expansion using the concept of the core to explain the process. The next three chapters use arms race models to provide empirical analyses of the demand for military expenditure. Sylvie Matelly looks at the determination of US military expenditures in the context of an arms race with the Soviet Union and considers why US military expenditure did not drop in tandem with Russian military expenditure. Paul Dunne, Eftychia Nikolaidou and Ron Smith examine the interaction of Greek–Turkish and India–Pakistan military expenditures in the context of arms race models. Their results are very different for the two dyads. India and Pakistan show well-determined Richardson-type reaction functions, Greece and Turkey do not. Vasilis Zervos examines the government demand for military and space expenditures in the United States and former Soviet Union, and ﬁnds evidence that there has been a space race similar in nature to arms races examined in previous chapters of the book. This collection indicates the range of interesting issues that defence economists have been analysing and suggests a range of issues that will be on the future research agenda.

Part I

Arms exports and arms production

2

The economics of UK arms exports Keith Hartley and Stephen Martin

Introduction: the policy issues Although the UK arms trade involves both exports and imports of defence equipment and services, the focus is usually on arms exports. This is an area dominated by myths, emotion and special pleading. Critics refer to an ‘ethical foreign policy’, to the immorality of a nation exporting ‘weapons of death’, to the human rights record of some countries to which the United Kingdom exports defence equipment and the costs to the taxpayer of UK arms exports. Supporters of UK arms exports point to its economic beneﬁts in the form of jobs, export earnings and maintenance of the UK defence industrial base (Towle 1998; Webb 1998). Such diversity of views show the scope for independent analytical and empirical work and for critical evaluation of the UK’s arms exports. There are extensive opportunities for applying economic theory, for empirical testing and for the critical evaluation of policy options. Questions arise about what is known, what is not known and what is necessary to know for sensible, informed debates and public choices on the UK’s arms exports. This chapter examines several economic aspects of UK arms exports including their beneﬁts and costs, the economic impacts of limiting arms exports and employment behaviour in the export sector. Three issues are examined in more detail, namely, trade externalities, a case study of the problems and costs of maintaining a UK defence industrial base without exports and a comparative study of employment behaviour in UK export markets compared with the home market. The chapter starts by presenting the ‘stylised facts’ and then considers the importance of government in any analysis of UK arms exports.

The stylised facts of the UK arms trade Evidence on the UK export and import of defence equipment over the period 1975–97 is shown in Table 2.1. The UK aerospace industry has dominated defence equipment exports, accounting for some 70 per cent of arms exports over the years 1975–85, rising to 86 per cent in 1990 and about 91 per cent in 1997 (i.e. military aircraft, missiles and aerospace equipment). Table 2.1 also shows the changing international competitiveness of the rest of the UK defence industrial base.

6

Keith Hartley and Stephen Martin Table 2.1 UK arms trade (£ million: current prices) 1975 Exports of defence equipment Total Split by commodity Military aircraft and parts Guided weapons, missiles, parts Estimates of additional aerospace equipment Armoured ﬁghting vehicles Warships Ammunition Split by destination Middle East and North Africa NATO and other Europe Asia and Far East Latin America and Caribbean Other Africa Imports of defence equipment Total Split by commodity Military aircraft and parts Guided weapons, missiles, parts Armoured ﬁghting vehicles Split by origin NATO and other Europe Asia and Far East Balance of trade

1980

1985

1990

1997

477

1,537

1,753

4,467

6,685

43 13

170 25

174 55

1,206 143

3,532 428

279 42 42 43

1,000 50 59 102

940 127 4 85

2,487 70 3 315

2,987 202 256 20

91 52 20 30 5

158 111 134 13 121

284 247 73 54 155

1,207 396 295 46 36

2,592 1,460 379 145 22

50

147

246

627

1,540

— 25 5

2 65 7

18 119 10

316 190 12

1,290 113 42

46 4 +427

141 5 +1,390

220 16 +1,507

471 12 +3,840

1,216 106 +5,145

Source: MoD (1999). Note: Splits by commodity and country show major items: others account for residuals. Figures are for equipment only.

Ammunition exports declined dramatically in the 1990s; similarly with warship exports in 1985 and 1990, but these rose over the years 1995–7. Most of the UK’s arms exports were to the Middle East (e.g. Saudi Arabia), NATO and Europe. The UK arms imports are also dominated by aerospace equipment, especially during the 1990s, with most of the aerospace equipment imported from the United States of America. Overall, for the period 1975–97, the UK defence industry achieved a substantial balance of trade surplus for defence equipment. Whilst there are economic determinants of both arms exports and imports, the trade in arms is greatly affected by political factors and the role of governments.

The importance of government Governments have both a direct and indirect role in understanding UK arms exports. Most UK arms exports are of equipment that was originally designed,

The economics of UK arms exports

7

developed and produced for the UK’s Armed Forces. As a result, national demands determine the operational requirements, the technology and the unit costs of UK defence equipment (in some instances, possible export prospects might have entered into the criteria for procurement choices). There are few examples of genuine private venture projects designed specially for export markets (examples include the BAe Hawk and some Vickers tanks). Governments have a further role in arms exports through the award of export licences approving the export sale and the associated policing and monitoring of the licence system (Scott 1996). They can also agree not to impose any R&D levies on arms exports and can offer other forms of support, such as favourable loans, overseas aid, military support (e.g. with training) and assistance with offset requirements (Webb 1998). The importance of government means that public choice models and analysis are relevant to understanding arms exports. As a result, UK arms export policy will be inﬂuenced by interest groups that are likely to beneﬁt from such exports. There will be lobbying by major defence contractors seeking export orders, supported by trade unions and professional associations concerned about their future employment and income prospects. Industry will commission management consultants to provide ‘independent’ estimates of the beneﬁts to the UK economy of allowing arms exports. The Ministry of Defence (MoD) and the Department of Trade and Industry, as budget-maximising departments likely to beneﬁt from arms exports, will focus on the beneﬁts for the UK defence industrial base and for the UK economy, tending to ignore or underestimate the costs of arms exports. Opposition to UK arms exports comes from interest groups such as the Campaign Against the Arms Trade, Saferworld and the World Development Movement.

Evaluating UK arms exports: a beneﬁt–cost framework Beneﬁts of arms exports A beneﬁt–cost analysis provides both an analytical and empirical framework for assessing UK arms exports. The approach is illustrative rather than comprehensive. Beneﬁts include employment, a contribution to the balance of payments, support for the UK defence industrial base (e.g. providing security of supply in a crisis), savings on MoD procurement spending, possible trade externalities and other intangible beneﬁts (e.g. world power status; inﬂuencing friends and allies). Table 2.2 provides evidence on the magnitude of some of these beneﬁts, particularly those for which data are available in the public domain. It can be seen that between 1980 and 1998, defence exports have become an increasingly important element of the UK defence industrial base. Employment from defence exports as a share of total UK defence industry employment rose from under 20 per cent in 1980 to 26 per cent in 1990–1 and 37 per cent in 1997–8. The increasing export proportion in the 1990s reﬂected the UK industry’s response to the end of the Cold War and the reduction in UK defence spending (UK defence spending fell by over 20 per cent in real terms during the 1990s: MoD 1999). In fact, between 1994

8

Keith Hartley and Stephen Martin Table 2.2 Beneﬁts to the United Kingdom

Employment (000s) Defence exports Total Balance of payments (£ billion)a Equipment exports Import-saving Support of UK DIBb Defence exports share of UK industry (%) Defence exports share of MoD equipment spend (%)

1980–1

1990–1

1997–8

140 740

145 550

130 355

1.54 4.03

4.47 6.98

6.69 6.34

30.1

50.4

65.8

31.5

50.5

74.3

Source: MoD (1999). Notes a Balance of payments is for equipment exports and import-saving for equipment. Import-saving assumes that imported equipment would be 15% cheaper than UK-produced equipment. The ﬁgures are broad orders of magnitude. b Support for UK DIB. Share of UK industry refers to MoD defence expenditure in the UK by ‘industry group’. Share of MoD equipment spend refers to total MoD spending on equipment.

and 1997, the United Kingdom increased its share of a declining world market for defence exports from 16 to 23 per cent and by 1997 was the world’s second largest defence exporter after the United States of America (HCP 1999). Further beneﬁts from UK defence exports include savings on MoD procurement expenditure embracing both development and production costs. The commercial exploitation levy on export sales is designed to recover MoD’s contribution to development costs and further ﬁnancial beneﬁts accrue to MoD where export sales lead to the spreading of overhead costs on production. Lower unit production costs through learning economies are also possible if the MoD orders additional equipment after export sales (e.g. Hawk trainers). Over the 5-year period 1992–8, the export levy averaged some £50 million per annum and the spreading of company overheads saved MoD some £350 million in 1997–8 (HCP 147 1999; Martin 1999). Exports might also provide the UK defence industry with a competitive stimulus that is then reﬂected in the industry’s efﬁciency and prices paid by MoD; the competition effect might be reinforced if exports enable some contractors to remain in the industry, thus increasing rivalry for future MoD contracts. The remaining beneﬁts include trade externalities where UK arms exports to a speciﬁc country have favourable impacts on the UK exports of civil goods and services to that country; and other intangible beneﬁts, some of which might contribute to the UK’s security (e.g. inﬂuencing friends and allies). Of course, some of these beneﬁts are extremely difﬁcult to measure. Even where measurement is possible, there remains the task of converting diverse performance indicators into monetary valuations, where such valuations allow policy-makers opportunities for ‘discretionary behaviour’ (e.g. vote-maximisation).

The economics of UK arms exports

9

Costs of arms exports Arms exports involve costs as well as beneﬁts; often, the cost side of the equation is ignored. Costs include the alternative-use value of UK resources involved in arms exports, any public subsidies, the costs of the Defence Exports Services Organisation (DESO) and other costs, including any externalities. In 1997–8, UK defence exports employed 130,000 personnel, together with land and capital, giving a total value of resources of £6.7 billion allocated to UK arms exports (based on the value of arms exports: MoD 1999). Questions next arise about the alternative-use value of these resources. Here, there is a scope for further research into the operation of factor markets, distinguishing between short- and long-run adjustment processes and recognising the role of other determinants of wage and salary differences (e.g. human capital). There are, however, some limited studies of redundant defence workers. These show that of those who obtained employment, some 50–60 per cent received lower salaries, almost 20 per cent received the same salary and about 25–30 per cent received higher salaries (Hooper and Butler 1996; Hooper et al. 1996). In other words, most redundant defence workers moved to lower-paid jobs, suggesting that their alternative uses involved lower productivity jobs. On this basis, it can be concluded that workers in defence export industries are likely to earn more than in their next-best alternative, so that there are net beneﬁts to labour from UK defence exports. Governments provide assistance for UK arms exports. These include the costs of DESO, any ﬁnancial assistance for arms exports (via the Exports Credit Guarantee Department or ECGD), military assistance and training, promotional support (e.g. through overseas embassies; military displays, exhibitions and overseas deployments or ‘goodwill’ visits of navy and air force units) and the possible use of overseas aid to support UK arms exports. One of the few studies in the ﬁeld estimated that the UK government provided a net annual subsidy for defence exports of almost £230 million (1995 prices: Martin 1999). However, this estimate is subject to two limitations. First, it is sensitive to the assumptions used in the analysis (e.g. the losses by ECGD, Martin 1999: 34–35); but the assumptions are clearly stated so that critics can identify the sources of disagreement and suggest alternative assumptions. Second, further analysis is needed to estimate the magnitude of government support and assistance that would be provided in the absence of UK defence exports (i.e. governments are likely to provide support for alternative civil exports). Arms exports involve a set of other possible costs. There are possible externalities if the arms exported subsequently represent a security threat to the United Kingdom. Arms exports might also contribute to a regional arms race and an eventual regional conﬂict; and there might be political and ethical costs (e.g. human rights). But these arguments need to be carefully and critically evaluated. There are alternative scenarios. For example, an alternative possibility might be that arms exports contribute to a nation’s self-defence and hence to regional peace and stability. Similarly, it might be claimed that the arms exports to developing countries impose costs on importing countries reﬂected in adverse impacts on their growth

10

Keith Hartley and Stephen Martin

and development; but an alternative scenario of a ban on arms exports might impose even greater costs on a developing nation if it responds by creating an independent defence industry. The task for economists analysing and measuring the economic impacts of the arms trade is to carefully specify the alternative states of the world and the ‘benchmarks’ against which their analysis and empirical work is being undertaken. The next section presents an example of the approach.

The economic impacts of limiting UK arms exports Estimates have been made of the economic impacts of a one-third reduction in the annual value of UK arms exports (Martin et al. 1999). Even such an apparently simple task is fraught with data problems. There is no Standard Industrial Classiﬁcation for UK defence industries (Hartley and Hooper 1995); the value of arms exports differ between different data series (UK, ACDA, SIPRI); confusions arise between orders and deliveries; dual-use items and defence services are excluded from the UK data; and following the creation of the single European market, some UK defence exports were no longer recorded from 1993 (e.g. military radio, radar, simulators). A one-third reduction in UK arms exports will have short- and long-run impacts on jobs, unemployment, regions, industries and the balance of payments. The immediate impact comprises a loss of some 13,000–40,000 jobs, depending on which defence export series is used in the estimation; but not all those who lose their jobs will remain unemployed indeﬁnitely. Estimates suggest that 20 per cent of the redundant defence workers will leave the labour force; of the remaining 80 per cent, over one-half will ﬁnd a job within one year and a further onethird will take over two years to ﬁnd a new job (Hooper and Butler 1996). The geographical impacts of reduced UK arms exports is likely to be concentrated in a few regions with over 80 per cent of export-related employment located in the south-east, south-west and north-west regions of the United Kingdom. Similarly, the industrial impacts are likely to be concentrated on a few industries. Aerospace

Table 2.3 Limiting UK arms exports Job losses in short-run Unemployment Wages and salaries Regional impacts Industrial impacts Balance of payments

Source: Martin et al. (1999).

13,000–40,000 20% leave the labour force; of the remaining 80%, over half will ﬁnd a job within one year Majority moved to lower-paid jobs South-east, south-west and north-west regions of the United Kingdom worst-affected Aerospace industry worst-affected Loss of defence exports (possibly £1.3 billion), equal to 0.5% of total UK exports; over time, a compensating rise in exports from other industries

The economics of UK arms exports

11

is likely to be the worst affected industry, since in 1997, it accounted for some 90 per cent of UK defence equipment exports (MoD 1999). The balance of payments impacts of a one-third reduction in UK arms exports are relatively small. Total UK exports of goods and services would fall initially by 0.5 per cent; but as resources are re-allocated to other industries and regions, there will be compensating gains from additional exports in other industries. For example, an initial one-third cut in UK arms exports equivalent to £1.3 billion might be offset within one year by almost £600 million of additional exports produced by redundant defence workers re-employed in new jobs (Martin et al. 1999). The economic impacts of a one-third cut in UK arms exports are summarised in Table 2.3.

Trade externalities Amongst the potential beneﬁts from UK arms exports are the possible ‘spin-offs’ in the form of UK arms exports having a favourable impact on UK exports of civil goods and services. This can be viewed as one aspect of the claim often made that ‘trade follows the ﬂag’. Such an apparently simple hypothesis is, in reality, much more complex. This can be shown by considering the following general model: T = T (DSP, Z), where T is the UK trade, DSP the UK defence and security policy and Z the other relevant inﬂuences, such as UK macro-economic variables (including exchange rates), world demand and the structure of the UK economy. In the model, UK trade is deﬁned to include visibles, invisibles and foreign direct investment, all or each of which could be part of any ‘trade–ﬂag’ relationship. UK defence and security policy embraces three elements. First, the level of UK defence spending, its allocation between the army, navy and the air force and the geographical distribution of its armed forces (e.g. overseas bases). Second, UK membership of international treaty organisations and military alliances (e.g. NATO); its commitment to being a ‘world power’; overseas visits by UK forces, foreign exercises and overseas military assistance. Third, UK arms exports and its arms export policy. Some of these elements of UK defence and security policy are more easily measured than others. Also, questions arise about the predicted relationship between UK defence and security policy and UK trade. The obvious prediction is of a positive relationship, but the reality is more complex. For example, UK defence spending, especially on defence research and development, might ‘crowd-out’ valuable civil investment with adverse effects on the UK’s international competitiveness (Buck et al. 1993). At the same time, UK defence spending might signal that the United Kingdom is a ‘safe haven’ for foreign investment, so attracting inward investment into the United Kingdom (Ayanian 1994). The economics literature on the arms trade has focussed on the structure of the international arms market, the causes of the arms trade, models of the international arms market, the role of prices and quantities and work-sharing arrangements (Levine et al. 1994, 1998; Anderton 1995; Martin 1996). This section reports the results of a pilot study of one aspect of the ‘trade–ﬂag relationship’, namely, the

12

Keith Hartley and Stephen Martin

hypothesis that UK arms exports have a favourable impact on total UK civil exports. Even this simple hypothesis requires consideration of the underlying causal relationships. One explanation might be related to the role of markets in information and knowledge and trade linkages. Firms in the nations importing UK defence equipment might be impressed by the technology of the equipment and as a result, become aware of the United Kingdom as a major supplier of civil goods and services (including high-technology equipment). There might be associated trade linkage effects as the importing nation establishes contact with UK government ofﬁcials and trade associations, thus promoting contacts and trading opportunities with other ﬁrms in the UK economy. These favourable trade effects might be facilitated by ‘political factors’ reﬂected in inter-governmental agreements supporting trade between the two nations. The pilot study focussed on UK arms exports and UK civil exports to four nations over the period 1970–92 (all data in constant prices and using SIPRI arms export data). The four nations were Brunei, Indonesia, South Korea and Saudi Arabia, each of which have been major markets for UK arms exports. Also, the hypothesis was tested for total UK arms exports and total UK civil exports. Various correlation tests were tried and the main ﬁndings were: 1

2

3

4

For South Korea, there was a statistically signiﬁcant and positive correlation coefﬁcient between UK arms exports and UK civil exports to the country (r = 0.74); and between UK arms exports and UK exports to South Korea as a share of total UK exports (r = 0.82). For Saudi Arabia, there was a positive correlation coefﬁcient between UK arms exports and civil exports, but the relationship was not signiﬁcant at the 5 per cent level (r = 0.33); but, there was a signiﬁcant and positive relationship between UK arms exports to Saudi Arabia and the share of UK exports to Saudi Arabia in total UK exports (r = 0.43). For Brunei and Indonesia, the correlation coefﬁcients were not signiﬁcant (e.g. for Brunei r = 0.04 and 0.17). However, for Indonesia, there was a surprisingly negative correlation coefﬁcient between UK arms exports and UK civil exports to the country, but it was not signiﬁcant at the 5 per cent level (r = −0.35). For the aggregate of UK arms exports and UK civil exports, there was no support for the hypothesis of trade following the ﬂag (r = −0.06). However, the introduction of 1-, 2- and 3-year lags resulted in positive correlation coefﬁcients, but the results were not signiﬁcant (e.g. r = 0.31 for a 3-year lag).

The results of the pilot study are suggestive and exploratory rather than conclusive. There was evidence of trade following the ﬂag for South Korea and possibly for Saudi Arabia, but not for Brunei and Indonesia. Whilst there was a similar lack of support for the United Kingdom as a whole, there were indications that there might be an impact of UK arms exports on UK civil exports after a lag of some years. Overall, the pilot study results identify the scope for a research agenda using a more fully speciﬁed model, alternative arms export data, analysing more nations (e.g. NATO), incorporating more lags into the estimations and extending

The economics of UK arms exports

13

the hypothesis to include variables other than defence equipment (e.g. investment ﬂows and ‘safe havens’).

A UK defence industrial base without exports: a case study Proposals are often made for a complete ban on UK arms exports, but such suggestions rarely address the implications for the future of the UK defence industrial base. This section presents a case study of the problems of maintaining a UK defence industry that has no exports. The industry case study is the UK submarine industry, which specialises in the design, development and construction of nuclear-powered submarines, none of which are exported. Various international obligations prevent the United Kingdom from exporting nuclear-powered submarines, particularly the 1958 licence under which the United States of America transferred nuclear propulsion technology to the United Kingdom. The problems facing the UK submarine industry arise from gaps in development and production work and small production quantities. These are problems that have not been addressed previously in the United Kingdom (Rand 1994). From the late 1950s to the early 1990s, during the Cold War, the UK submarine industry had an overlapping workload on both nuclear and conventional submarines. Typically, during this period, production rates for UK submarines averaged about one per year. Following the end of the Cold War, the United Kingdom reduced its orders for conventional submarines and subsequently withdrew them from service, relying on a ﬂeet of nuclear submarines. Work on the last class of nuclear attack submarines (the Trafalgar class with an order for seven boats) started in the late 1970s and ended in the early 1990s. During the 1980s and until the late 1990s, there was work on four boats of the Vanguard class of ballistic missile submarines (SSBNs). The next generation of SSNs is the Astute class with a contract for three boats awarded in March 1997 at an estimated total cost of £1987 million (1998–9 prices: HCP 519, 1999) and a forecast in-service date of June 2005. After Astute, the next generation of SSN will be the Future Attack Submarine (FASM) with the main contract expected to be awarded around 2008 with possible delays of up to three years. As a result, there will be a gap of some eight years between the completion of design work on Astute and the start of detailed design work on FASM. Such a gap in design and production will have major implications for the future of the UK submarine industry; and unlike other UK defence industries, it cannot rely on exports of nuclear submarines to retain key skills and facilities. In the circumstances, the major issues for the industry and for policy-makers are: 1 2 3

Which are the critical or ‘key’ skills and facilities needed for nuclear submarines? Which skills and facilities are speciﬁc to submarine design, construction and integration? How can these skills and facilities be retained for FASM, including delays, and at what cost? For example, can these skills and facilities be retained by transferring them to other work (e.g. surface warships or civil work),

14

Keith Hartley and Stephen Martin or can they be shut down and restarted in 2008 (and at what cost and how quickly)?

The UK submarine industrial base is a unique UK defence industry. It specialises in one product and depends on one buyer, namely, the UK government (Royal Navy); currently, there is only one ﬁrm in the United Kingdom with the actual experience and facilities for building nuclear-powered submarines (i.e. GEC-Marine at Barrow, previously VSEL and acquired by British Aerospace in 1999). The requirement for nuclear-powered submarines imposes major barriers to new entry, competition and international collaboration. Entry barriers arise from the requirement for a nuclear site licence for construction (GEC-Marine at Barrow), for the manufacture of the nuclear reactor (core factory and recovery plant at Rolls-Royce Marine Power at Derby) and for refuelling (Devonport Management Ltd at Devonport). Finally, unlike some other UK defence industries, the submarine industry provides no obvious wider economic beneﬁts for the UK economy in the form of exports and technology spin-offs to the rest of the economy (cf. military and civil aircraft). Nuclear submarines are highly specialised and technically complex weapons systems that require special skills for aspects of design and construction work. They differ from surface warships in that they are designed to operate for long periods underwater at great depths and quietly with a range of weapons whilst providing a safe environment for the crew living close to a nuclear reactor. As a result, the key skills needed include design staff with submarine expertise; nuclear design staff for the reactor; the skills needed to retain the nuclear site licences for the nuclear reactor, for construction and for re-ﬁt; together with some highly specialised facilities for nuclear submarines that have no alternative use (e.g. core factory; testing facilities). Various solutions are possible to ‘solve’ the problem of gaps in development and production work between Astute and FASM. Some possible options include: 1

2

3

Leave industry to solve the problem. Under this option, those ﬁrms in the UK submarine industry with transferable skills and facilities and alternative work would be most likely to survive. For example, resources might be transferred to other defence projects, including naval surface ships (e.g. new aircraft carriers, new frigates). Nuclear skills might be retained by ‘sharing arrangements’ with other military or civil nuclear establishments (e.g. Devonport on re-ﬁtting work; Sellaﬁeld nuclear site). But there remains the problem of those core skills and facilities that are so highly specialised that they can be used only for submarine work. State intervention to support the industry. Options for the UK Ministry of Defence include ordering another two Astute boats with additional design inputs for modiﬁcations and upgrades; starting the FASM programme earlier than 2008; or awarding technology demonstrator contracts. Shut-down and restart. This would be costly and would take time. US estimates suggest that reconstitution after a 5-year shut-down might exceed $1.7 billion (1992 prices: Rand 1994). Similar estimates for the United Kingdom suggest

The economics of UK arms exports

4

15

restart costs of over £300 million (a guesstimate) and restart time of 10 years for the speciﬁc nuclear capabilities and expertise (Hartley 1999). Consider the alternative of acquiring conventional submarines instead of nuclear submarines: this would ‘remove at a stroke’ the nuclear entry barriers to competition.

Employment behaviour The UK arms export market provides an opportunity for assessing whether defence ﬁrm behaviour differs between competitive and non-competitive markets, with the latter characterised by labour hoarding and inefﬁciency. The UK MoD buys equipment (e.g. aircraft, ships and tanks) and non-equipment or civil items (e.g. food, fuel, clothing). Typically, civil goods and services are purchased in competitive markets where there are large numbers of buyers and alternative suppliers. In contrast, for equipment, MoD is either a monopsonist (e.g. nuclear submarines) or a major buyer and the UK defence industry comprises domestic monopolies for high-technology items. Despite the introduction of a competitive procurement policy after 1983, a substantial proportion of MoD contracts remain non-competitive (e.g. 25 per cent in 1997–8; MoD 1999). As a result, it can be argued that the UK defence equipment market is relatively protected and less competitive than the market for non-equipment or civil items. Similarly, export markets can be regarded as competitive. This section compares employment behaviour in these three markets and tests the following hypotheses: a

b

Hypothesis I compares employment elasticities between MoD’s purchase of defence equipment and its purchase of other civil goods and services (e.g. clothing, food and fuel). It is predicted that the employment elasticity will be lower for defence equipment work (less or non-competitive) compared with other competitive purchases (i.e. MoD equipment vs non-equipment purchases). Hypothesis II compares the employment elasticities for competitive defence export sales and for less-competitive equipment sales to the UK government. It is predicted that the employment elasticity for competitive defence exports will be greater than the corresponding elasticity for non-competitive equipment sales to the UK government.

The model and empirical results Employment data are only available on an annual basis from 1978. Annual data are also available for the type of contracts awarded by MoD ranging between the extremes of contracts priced by competition and those priced on a cost-plus basis. There are, however, data problems that could affect the reliability of the empirical results. Employment data for UK defence industries are estimated from an industrial analysis of both MoD expenditure and export sales, and from estimates of sales per employee for the relevant industries. As a result, any rapid employment

16

Keith Hartley and Stephen Martin

adjustment might reﬂect the method used to estimate defence industry employment (MoD 1988). Nonetheless, these are the only published data available. To test these hypotheses, a standard short-run employment function was estimated of the following form (Hartley and Corcoran 1975; Hartley and Lynk 1983): log Lt = a log Qt + bt + c log Lt−1 where Lt is the employment in the current period, Lt−1 is the employment lagged one period, q the output, t a time-trend representing the capital stock and technology; a = λ/α, b = λr/α and c = 1 − λ. Two employment elasticities can be distinguished. The ﬁrst is the unadjusted elasticity ( ), which is given by the estimated regression coefﬁcient on output ( = a = λ/α). This, of course, incorporates an estimate of the speed with which actual employment adjusts to its desired level (i.e. λ). A second, ‘corrected’ elasticity ( ), which does not include the speed of adjustment (λ), can also be computed and is given by the reciprocal of the short-run returns to labour ( = 1/α). Initially, the model was estimated for employment attributable to MoD defence equipment expenditure (see Eq. (1) in Table 2.4). Output was measured by the value of MoD expenditure on defence equipment in the United Kingdom at constant (1995) prices. The equation was also re-estimated with a dummy variable that took the value of zero from 1978 to 1982, and then the value of one from 1983 to 1996 [Eq. (2a)]. This was designed to test whether the MoD’s move to a more competitive and commercial procurement policy since 1983 had shifted the employment–output relationship (MoD 1983; HCP 189, 1991). A dummy variable that took the value of zero from 1978 to 1987, and then the value of one from 1988 to 1996 was added to see if MoD’s increased willingness to award contracts to overseas suppliers since the late 1980s had had any effect on the employment– output relationship. However, the addition of this variable was unsuccessful. In addition, a dummy variable that took the value of zero from 1978 to 1989, and then the value of one from 1990 to 1996 was added to see if the end of the Cold War had had any effect on the employment–output relationship, but this too was unsuccessful. Equations (3) and (4) in Table 2.4 are identical to Eqs (1) and (2), except that they relate to employment attributable to MoD non-equipment expenditure (e.g. on clothing, food and fuel). Output was measured as the value of MoD non-equipment expenditure at constant (1995) prices. This equation was also re-estimated with a dummy variable [Eq. (4)]. Equations (5) and (6) relate to employment attributable to UK defence equipment exports. Again, Eq. (6) included a dummy variable. Finally, Eqs (7) and (8) relate to all employment in the UK defence industrial base (i.e. employment attributable to MoD total expenditure plus employment attributable to UK defence equipment exports). Most of the estimated equations had signiﬁcant coefﬁcients for output and time, but, in terms of the hypotheses tested, the results were rather mixed.1 For both

0.303 (2.0) 0.285 (1.2) 0.872∗∗ (7.9) 0.877∗∗ (7.5) 0.745∗∗ (6.7) 0.826∗∗ (7.9)

−0.023∗ (3.0) −0.024∗ (2.5) −0.034∗∗ (6.3) −0.035∗∗ (4.8) −0.046∗∗ (7.8) −0.041∗∗ (7.0)

0.878 (0.6) 1.044 (0.5)

−0.790(1.2) −0.793 (1.2)

0.505 (0.6) −0.550 (0.6)

5 6

7 8

3 4

0.561∗∗ (3.7) 0.844∗∗ (4.7) 0.506∗∗ (3.1)

−0.044∗∗ (5.4) −0.034∗∗ (4.0) −0.049∗∗ (5.1)

0.916 (1.0) −1.356 (1.0) 1.007 (1.1)

1 2a 2b

log Qt

Time trend

Constant

Eq. no

Estimated coefﬁcients of

−0.176 (1.2) −0.137 (1.0)

−0.339 (2.1) −0.346 (2.0)

0.349 (1.6) 0.346 (1.5)

0.010 (0.1) −0.050 (0.3) 0.049 (0.2)

log Lt−1

−0.07∗ (2.2)

0.01 (0.1)

0.01 (0.1)

−0.12∗ (2.3) 0.06 (0.95)

DV1983

1.176 1.137

1.339 1.346

0.651 0.654

0.990 1.050 0.951

λ

1.58 1.38

1.53 1.53

2.15 2.29

1.76 1.24 1.88

α

0.63 0.72

0.65 0.65

0.47 0.44

0.57 0.81 0.53

0.975 0.980

0.854 0.843

0.878 0.869

0.980 0.985 0.979

R2

224.0∗∗ 219.0∗∗

34.1∗∗ 23.8∗∗

41.9∗∗ 29.3∗∗

278.6∗∗ 273.0∗∗ 207.8∗∗

F

0.08 1.95

0.06 0.04

0.01 0.00

0.24 0.00 0.17

LM

Notes 1 All equations are log-linear. 2 Figures in parentheses are t-ratios. ∗ Signiﬁcant at the 5% level; ∗∗ Signiﬁcant at the 1% level using a two-tail test. 3 Lt is employment in period t; Q is the real value of MoD equipment expenditure/non-equipment expenditure/defence exports depending on the equation being estimated; DV1983 is a dummy variable that takes the value of zero for 1978–82 and a value of one for 1983–96. In Eq. (2b), DV1983 has been replaced by a variable measuring the percentage of the value of MoD contracts priced by competition. 4 λ is the speed with which actual employment adjusts to its desired level; α is the elasticity of employment with respect to output; is the ‘corrected’ elasticity of employment; R 2 is adjusted for degrees of freedom; F is the statistic testing the null hypothesis that all coefﬁcients in the estimated equation are zero; and LM is the Lagrange multiplier statistic testing for the presence of ﬁrst-order serial correlation.

Sources: MoD (1985, 1988, 1998).

Supported by MoD equipment expenditure Supported by MoD non-equipment expenditure Supported by UK defence equipment exports Supported by MoD expenditure and UK defence exports

Dependent variable: employment (log Lt )

Table 2.4 Employment functions for the UK defence industry, 1978–96

18

Keith Hartley and Stephen Martin

of the MoD equipment expenditure Eqs [(1) and (2a)], output was a signiﬁcant determinant of the level of employment and the time trend was signiﬁcant with the anticipated negative sign. With regard to the lagged dependent variable (Lt−1 ), however, the estimated coefﬁcient (c = 1 − λ) in Eqs (1) and (2a) was not significantly different from zero. This suggests that λ does not differ signiﬁcantly from one and that employment adjusts to its desired level within a year (the periodicity to which the data relate); but these results could reﬂect the method used to construct the data, and hence no further reference will be made to the lagged adjustment process. The dummy variable in Eq. (2a) is signiﬁcant, with the anticipated sign (i.e. the introduction of a more competitive approach to MoD procurement appears to have had a shake-out effect, reducing the level of employment associated with any given volume of equipment purchased). Comparisons can also be made between both the unadjusted and adjusted employment elasticities for the equipment and non-equipment equations. The unadjusted elasticity (the estimated coefﬁcient on output) is greater for defence equipment than for other MoD purchases suggesting that, if anything, defence ﬁrms are more responsive to market signals than their civilian counterparts! Similarly, the true, adjusted employment elasticity ( = 1/α) is also greater for defence equipment than for other MoD purchases. Unlike Eq. (2a), the dummy variable in Eq. (4) is not signiﬁcantly different from zero. This is to be anticipated as the introduction of a more competitive approach to MoD procurement is only likely to have affected those goods and services that had hitherto been purchased via non-competitive contracts. Equations (5) and (6) relate to employment on defence export work. As predicted, the unadjusted employment elasticity is greater for competitive export work than for non-competitive sales to the MoD but this difference is only signiﬁcant if the dummy variable is excluded [i.e. if Eqs (1) and (5) are compared]. Once the dummy is included in the estimating equation [see Eqs (2a) and (6)], the difference between competitive export work and MoD equipment sales almost disappears. Moreover, when the adjusted elasticity ( ) is used to compare the export and MoD sales equations with the dummy included, the employment elasticity for MoD equipment sales (0.81) exceeds that for competitive export sales (0.65)! Finally, Eqs (7) and (8) relate to total employment in the UK defence industrial base. Again, output is a signiﬁcant determinant of the level of employment, the time trend is signiﬁcant, with the anticipated negative sign, and the dummy variable for MoD competition policy is also signiﬁcant and negative. The unadjusted employment elasticity (0.745) is, as anticipated, less than that for competitive export work (0.872) but greater than for equipment sales to the MoD (0.561). However, once the dummy variable is added to the model these differences almost disappear so that all three elasticities fall between 0.826 and 0.877. When the adjusted elasticity ( ) is used to compare the defence industrial base (DIB) equation with the other employment functions, the results were not those anticipated: the aggregate DIB employment elasticity (0.72) is less than that for MoD equipment sales (0.81).

The economics of UK arms exports

19

As an alternative to the dummy, the percentage of the value of MoD contracts priced by competition was also used as an indicator of the degree of MoD’s move to a more competitive procurement policy. Although the inclusion of this variable rather than the dummy had little impact on three of the four equations, it reduced the estimated elasticities for employment supported by MoD equipment expenditure. Equation (2b) shows the regression result with the value of MoD contracts priced by competition variable. A comparison of Eqs (2a) and (2b) reveals that although the competition variable is not signiﬁcantly different from zero [unlike the dummy in Eq. (2a)], the estimated coefﬁcient on output (the unadjusted elasticity) has fallen from 0.8444 to 0.506 and the adjusted elasticity has declined from 0.81 to 0.53. Thus, with this model there is evidence to support Hypothesis II that employment on defence equipment work for MoD adjusts more slowly than employment on similar work for export. Nevertheless, Hypothesis I – that employment on MoD non-equipment work will adjust more quickly than that on MoD equipment work – remains unsupported. Overall, it was surprising to ﬁnd that there was relatively little support for the hypotheses of this study. However, this could reﬂect the MoD’s data construction method.

Conclusion This chapter has demonstrated the opportunities for analytical and empirical work in the emotional area of UK arms exports. Evidence on the beneﬁts and costs of UK arms exports, on trade externalities, on the problems of maintaining a UK defence industrial base without exports and on employment behaviour all show the contribution of economics to sensible, informed debates and public choices on the UK’s arms exports.

Note 1 There was no evidence of ﬁrst-order serial correlation in any of the equations (see the LM test statistic reported in the last column of Table 2.4).

References Anderton, C. (1995) Economics of the arms trade. In: Hartley, K. and Sandler, T. (eds), Handbook of Defense Economics. North Holland, Elsevier, Amsterdam. Ayanian, R. (1994) The real exchange ration enigma: a safe haven solution from the Cold War era. Defence and Peace Economics 5(1), 51–66. Buck, D., Hartley, K. and Hooper, N. (1993) Defence R&D, crowding-out and the peace dividend. Defence Economics 4(2), 161–178. Hartley, K. (1999) FASM Industrial Base Study. Unpublished report for MoD, Centre for Defence Economics, University of York, York. Hartley, K. and Corcoran, W. J. (1975) Short-run employment functions and defence contracts in the UK aircraft industry. Applied Economics 7, 223–233. Hartley, K. and Hooper, N. (1995) Study of the Value of the Defence Industry to the UK Economy. Centre for Defence Economics, University of York, York.

20

Keith Hartley and Stephen Martin

Hartley, K. and Lynk, E. (1983) Labour demand and allocation in the UK engineering industry. Scottish Journal of Political Economy 30(1), 42–53. HCP 189 (1991) Ministry of Defence: Initiatives in Defence Procurement. HMSO, London. HCP 147 (1999) The Appointment of the New Head of Defence Export Services. Defence Committee, House of Commons, The Stationery Ofﬁce, London, March. HCP 519 (1999) Ministry of Defence: Major Projects Report 1998. National Audit Ofﬁce, The Stationery Ofﬁce, London. Hooper, N. and Butler, B. (1996) A case study of redundant defence workers. Defence and Peace Economics 7(2), 149–167. Hooper, N., Butler, B., Hartley, K., Braddon, D. and Dowdall, P. (1996) Defence Industry Redundancies in the South West Region. Report for Department of Employment, Centre for Defence Economics, York. Levine, P., Sen, S. and Smith, R. (1994) A model of the international arms market. Defence and Peace Economics 5(1), 1–18. Levine, P., Mouzakis, F. and Smith, R. (1998) Prices and quantities in the arms trade. Defence and Peace Economics 9(3), 223–236. Martin, S. (1996) The Economics of Offsets. Harwood Academic Publishers, Amsterdam, The Netherlands. Martin, S. (1999) The subsidy savings from reducing UK arms exports. Journal of Economic Studies 26(1), 15–37. Martin, S., Hartley, K. and Stafford, B. (1999) The economic impacts of restricting UK arms exports. International Journal of Social Economics 26(5), 779–801. MoD (1983) Value for Money in Defence Equipment Procurement. Ministry of Defence, London. MoD (1988) Statement on the Defence Estimates 1988, volume 2, HMSO, London. MoD (1999) UK Defence Statistics 1999. Ministry of Defence, The Stationery Ofﬁce, London. Rand (1994) In: Birkler, J. et al. (eds), The US Submarine Production Base: An Analysis of Cost, Schedule and Risk for Selected Force Structures. Rand, Santa Monica. Scott (1996) Inquiry into the Export of Defence Equipment and Dual-Use Goods to Iraq and Related Prosecutions. HMSO, London, February. Towle, P. (1998) Ethics and the Arms Trade. Institute of Economic Affairs, London. Webb, T. (1998) The Armour-Plated Ostrich. Comerford and Miller, London.

3

Potential and actual arms production Implications for the arms trade debate Jurgen Brauer

Introduction Nearly ten years ago, Brauer (1991) published a piece in which he examined developing nations’ potential and actual arms production. He reasoned, as had others before him (e.g. Wulf 1983), that in order to produce armaments indigenously, countries need to possess a set of minimum skills or abilities, especially in regard to industrial structure, industrial diversiﬁcation, and human capital. Classifying developing countries into four arms producing groups, Brauer examined their primary, secondary, and tertiary school enrollment data for 1960 and 1965, and annually from 1975 to 1984 as a measure of human capital formation. As regards industrial structure, he computed an index – the potential defense capacity (PDC) index – that measured, annually for 1976–84, the output value of arms production relevant manufacturing industries as a percentage share of the output value of total manufacturing. With regard to industrial diversiﬁcation, Brauer counted for 1975–84 in how many six-digit sub-categories of the arms production relevant three-digit International Standard Industrial Classiﬁcation (ISIC) code a country produced. For example, Argentina produced in eighteen of the seventy-four six-digit categories within the three-digit category of “manufacture of industrial chemicals” (ISIC code 351) whereas Brazil produced in sixty-two of those seventy-four categories. The four arms production groups were the following. A group of “occasional arms producers” (AP = 1), that is, countries for whom arms production activities were occasionally, but not regularly reported and whose arms production output was minimal. A second group, AP = 2, consisted of countries that consistently produced arms but at limited quantity and sophistication. In contrast, AP = 3 countries were those with sizeable and diversiﬁed arms production activities. As a control group, Brauer also compiled the relevant data for non-arms producers (AP = 0). With respect to the early 1980s, Brauer found a strong correspondence between countries’ potential to produce arms and their actual extent of arms production: those that can, do; and those that cannot, do not. Brauer also noted exceptions: some countries’ potential exceeded their actual arms production (in particular Greece, Turkey, Portugal, Spain, and Mexico); other countries’ actual arms production

22

Jurgen Brauer

appeared to strain their potential (particularly Argentina, Israel, and South Africa). The differences in industrial diversiﬁcation among the three groups of arms producers (AP = 1, AP = 2, and AP = 3) were statistically signiﬁcant.1 The differences among the groups with respect to the school enrollment data were even stronger: between 1960 and 1984, school enrollment differences increased in favor of the AP = 3 countries and in disfavor of the non-arms producers (AP = 0). The present study updates and extends Brauer’s (1991) work in a number of ways. First, instead of examining arms and non-arms producers with reference to the early 1980s, this study examines arms and non-arms producers for the early to mid-1990s. One objective is merely to ﬁnd out what has happened in the ten-year interval. Second, unlike the earlier study, this present chapter includes data on developed, “Western” countries, almost all of which are arms producers (AP = 9). The objective here is not only to compare arms producers among developing nations to non-arms producers but to the arms producers in developed nations and thus to glean whether or not the substantial quantity and sophistication of “Western” arms is within or out of reach of the most engaged arms producers among developing nations. A third objective is to ﬁnd out what happened to the exceptions Brauer noted in 1991, that is, Greece, Turkey, Portugal, Spain, and Mexico on the one hand, and Argentina, Israel, and South Africa on the other. Fourth, to carry out the empirical work, a unique index or ranking of arms producers is constructed.

Deﬁning arms producers Whereas Wulf (1983) and Brauer (1991) dealt with about thirty developing nations producing arms, I now deal with nearly twice as many developing nations: sixtythree countries in all, plus an additional thirty-two developed nations, twenty-nine of whom produce arms (see Table 3.1). These numbers possibly contain a slight over count. The sources used were the following: among other items, the US Arms Control and Disarmament Agency (US ACDA 1997) lists countries’ arms exports. Most arms exporters are also arms producers but exceptions arise when countries import, and then re-export, arms. For instance, the US ACDA (1997) lists the Cape Verde islands as having Table 3.1 Number of arms producers, c.1975–83 vs c.1990–5 AP = 1 AP = 2 AP = 3 Sub-total AP = 0 Sub-total AP = 9 Total Brauer (1991)a 18 This studyb 39

7 12

8 12

33 63

67 60

100 123

N/A 32

100 155

Source: Brauer (1991). Notes a Data coverage in Brauer (1991) varied for the various measures used. The number given in the table pertains to the largest number of countries covered.2 b Data coverage in this study also varies by measure used. The number given pertains to the largest number of countries covered.

Potential and actual arms production

23

exported arms in 1985. To my knowledge, Cape Verde is not an arms producer per se and thus my table over-counts the number of arms producers. Of thirtyeight developing countries listed by the US ACDA (1997), perhaps ﬁve fall into this category (Afghanistan, Cape Verde, Ethiopia, Mali, and Oman). Another source used was Brzoska (1995) who lists thirty-four non-European arms producers among developing nations (counting Turkey as European). Rana (1995), a third source, lists ﬁfty-two nations engaged in the production of “small arms,” including twenty-nine developing nations. Finally, the International Institute for Strategic Studies (IISS 1998, table 32) lists ﬁfty-two arms suppliers (producers and, one presumes, re-exporters) as well as ﬁfty-one producers of “light weapons” (IISS 1998, table 49), developing and developed. Table 3.1 combines the count from these ﬁve sources.3 The drastic increase in the number of arms producers is explained, in part, by the break-up of the Soviet Union and east, southeast, and central European states. For example, instead of the Soviet Union being a single arms producer we now have nine. They are, in alphabetical order, Belarus, Estonia, Kazakstan, Kyrgyzstan, Russia, Tajikistan, Turkmenistan, Ukraine, and Uzbekistan. Similarly, instead of Czechoslovakia there are now two arms producers: the Czech Republic and Slovakia. And instead of Yugoslavia we now have Yugoslavia (i.e. Serbia, Kosovo, and Montenegro), Croatia, and Macedonia.4 The seeming proliferation of arms producers occurs for two main reasons: (a) the break-up of the Soviet realm created many more countries; and (b) many of these new countries are now classiﬁed as “developing” countries but were not so classiﬁed in the Wulf (1983) and Brauer (1991) studies. Wulf (1983) provided a ranking, from one to thirty-two, of arms producers among developing countries. His ranking was based on counts and classiﬁcations of weapon systems reportedly produced by the countries. I do not quarrel with Wulf’s procedure, but for the current study wish to offer a different ranking procedure to the literature. The ranks rely on frequency counts and are constructed as follows: •

• • •

US ACDA (1997) lists arms exports by country annually for 1985–95. For each year in which arms exports are listed, the country receives a score of 1. Chile, for example, is listed with arms exports in eight years and therefore gets a score of 8. It is important to note that I did not add up purported arms export dollar values, ﬁguring that the fact of an arms export counts rather than its reported dollar amount. Brzoska (1995) lists arms producers (all weapon systems) among the developing countries, but speciﬁcally excludes all (east and west) European countries. Each country mentioned gets a score of 1. Rana (1995) lists ﬁfty-two producers of “small arms.” Each country mentioned gets a score of 1.5 IISS (1998, table 32) lists arms suppliers, 1992–7, by dollar value and groups the countries into six supplier categories, based on arms export values. Once again ignoring the purported dollar amounts, a country listed in the

24

•

Jurgen Brauer lowest-valued group gets a score of 1; the United States of America, the only country in the highest-valued group, gets a score of 6. IISS (1998, table 49) also lists producers of “light weapons.” Each country mentioned gets a score of 1.

This procedure produces an arms production raw score of between 0 and 19. To establish an arms production rank from the raw score, the following procedure was used. • • • • • • •

sort countries by arms production raw score (0–19); break ties by sorting countries by the number of years for which arms exports are reported (ACDA data); break further ties by sorting countries by the dollar-arms export category (IISS, table 32); break further ties by sorting countries by light-weapons production only (IISS, table 49); break further ties by sorting countries by small-weapons production only (Rana’s table); break further ties by sorting countries by whether or not they are mentioned in Brzoska’s table (all weapons categories for all developing countries); and split the remaining ties.

Apart from the group of developed nations and non-arms producing developing nations (AP = 9 and AP = 0, the default classiﬁcations of interest), I then looked for obvious, “natural” break points in the AP-ranks for the remaining countries. In all, three additional groups emerge for a total of ﬁve groups. •

• • •

•

AP = 9 the group of developed, industrialized countries (irrespective of AP raw score and rank); this includes thirty-two countries in all, twenty-nine of which produce arms; this deﬁnes the group of countries to whose level and sophistication of arms production developing nations might aspire; AP = 0 the group of non-arms producers (whose AP raw score is zero) among developing nations; AP = 1 the group of occasional arms producers (whose AP raw score ranges from 1 to 6 and whose number of arms export years is ﬁve or less; Jordan and Malaysia represent the upper limit in this group); AP = 2 a group of twelve countries that are consistent arms producers but at limited quantity and sophistication (whose AP raw score ranges from 7 to 13 and whose number of arms export years are, with two exceptions, six or more); and AP = 3 a group of another twelve countries that engage in sizeable and diversiﬁed arms production and whose number of arms export years is ten or eleven.

Potential and actual arms production

25

The production of the raw score and the ranking can be faulted, and suggestions for alternatives are welcome. The intent was to produce an alternative to Wulf’s (1983) ranking, which is much more time-intensive to produce, and to give weight to the frequency of reported incidences of arms production/exports across various sources while avoiding to give weight to dollar values in arms production, be they purported (US ACDA; IISS) or imputed (SIPRI 1999, chapter 11).6

Deﬁning developed and developing nations The World Bank’s World Development Report 1998/99 (World Bank 1999, table 1) deﬁnes high-income countries as those with a 1997 per capita GNP of US$9,656 or more. They rank from number 1 Switzerland (US$44,320) to number 27, Slovenia (US$9,680). These are the countries included in the AP = 9 group, as are a further ﬁve (the Bahamas, Cyprus, Luxembourg, Kuwait, and Qatar) that the World Bank does not rank but whose per capita GNP exceeded US$9,656 in 1997. For convenience, these thirty-two countries are deﬁned as “developed nations”.7 All other countries are deﬁned as “developing nations.” They include the upper-middle income countries Malta (US$8,630) to Turkey (US$3,130), the lower-middle income countries Panama (US$3,080) to Sri Lanka (US$800), and the low-income countries Albania (US$750) to Mozambique (US$90). They also include, importantly, most central and east European nations that emerged from the break-up of the Soviet empire.

Potential arms production: the PDC index To produce arms indigenously requires some industrial structure. In the past, researches examined the industrial classiﬁcation codes listed on the left-hand side of Table 3.2 and measured the local currency value of the output in these categories relative to the local currency value of all manufacturing industries. The reasoning went that the smaller the share of these arms production relevant industries in total manufacturing, the smaller a country’s potential to produce arms. Vice versa, the larger the share of these arms production relevant industries in total manufacturing, the larger a country’s potential to produce arms. This share is known as the PDC index. By the early 1990s, most countries used this classiﬁcation system, known as ISIC revision 2, but a number of countries used the ISIC revision 3 classiﬁcation. To construct a PDC index for those countries, I examined the revision 3 codes listed on the right-hand side of Table 3.2. Also, I computed not only the output value share of PDC industries relative to total manufacturing, but also the employment share of these industries relative to employment in all of manufacturing. The results are displayed in Table 3.3. Start examining Table 3.3 with the results for the seventy-three ISIC revision 2 countries. Among developing nations, there is a clear progression from non-arms producers (AP = 0) to occasional arms producers (AP = 1) to regular arms producers (AP = 2) to substantial arms producers (AP = 3) both with regard to the

26

Jurgen Brauer

Table 3.2 ISIC revisions 2 and 3 codes to determine countries’ PDC index ISIC revision 2

ISIC revision 3

351 Industrial chemicals 352 Other chemicals 371 Iron and steel 372 Non-ferrous metals 381 Metal products

241 Basic chemicals 271 Basic iron and steel 272 Basic precious and non-ferrous metals 273 Castings of metals 281 Structural metal products; tanks, steam generators 289 Other metal products; metal working services 291 General purpose machinery 292 Special purpose machinery 293 Domestic appliances, not elsewhere classiﬁed 311 Electric motors 312 Electricity distribution and control apparatus 313 Insulated cables and wires 314 Accumulators, primary cells, and batteries 315 Lighting equipment, electric lamps 319 Other electric equipment, not elsewhere classiﬁed 321 Electronic valves, tubes, etc. 322 TV/radio transmitters 323 TV/radio receivers 3311 Medical, surgical, testing appliances, etc. 3312 Measuring, testing, navigation appliances 3320 Optical instruments and photographic equipment 341 Motor vehicles 342 Automobile bodies, trailers, semi-trailers 343 Parts and accessories for automobiles 351 Building and repairing of ships and boats 352 Railway locomotives and rolling stock 353 Aircraft and spacecraft 359 Transport equipment, not elsewhere classiﬁed

382 Non-electrical machinery 383 Electrical machinery 384 Transportation equipment 385 Scientiﬁc, measuring, controlling equipment

employment as well as the output share of arms production relevant industries relative to total manufacturing employment and output. The same pattern is obtained for the twenty-nine ISIC revision 3 countries. In both instances, the employment share of PDC industries in total manufacturing is smaller than the output share, suggesting that relatively little employment creates relatively high-valued output. Table 3.3 also presents a weighted average (weighted by the number of countries), combining the information from ISIC revision 2 and 3 countries. Marked differences in arms production potential are evident among the developing nations, as is a clearly identiﬁable trend: the higher a country’s potential, the higher its actual engagement in arms production activities. Another remarkable observation is that the arms production potential of the AP = 3 countries exceeds that of the AP = 9, the industrialized, developed countries. Taken at face value, this suggests that these

Potential and actual arms production

27

Table 3.3 PDC index by AP-level, early 1990s AP-group ISIC revision 2 countries Employment share (%) Output share (%) ISIC revision 3 countries Employment share (%) Output share (%) Sum of n Revision 2: weight-of-countries (in %) Revision 3: weight-of-countries (in %) n-Weighted shares Employment share (%) Output share (%) For comparison Brauer (1991); output share (%)

AP = 0 AP = 1 AP = 2 AP = 3 AP = 9 n 34 17.82 24.32 n 4 16.71 18.08 38 89.47

14 26.85 31.70 5 21.25 26.60 19 73.68

6 34.71 41.08 3 26.31 36.51 9 66.67

8 41.88 45.32 2 39.23 44.35 10 80.00

11 42.18 42.43 15 37.38 42.10 26 42.31

10.53

26.32

33.33

20.00

57.69

17.70 23.66 33 18.69

25.38 30.36 11 31.97

31.91 39.56 5 33.85

41.35 45.13 6 46.28

39.41 42.24 n/a n/a

n 73

29

102

55

Sources: Computed from latest available year reported, usually any year between 1991 and 1995, in United Nations Industrial Development Organization (1998), International Yearbook of Industrial Statistics 1998 (Vienna: UNIDO).

countries (Brazil, Bulgaria, China, the Czech Republic, Mexico, North Korea, Pakistan, Russia, South Africa, and Turkey) could produce weaponry of the sophistication and in the quantities of the average developed country (Australia, Belgium, Canada, Cyprus, Denmark, Finland, France, Germany, Greece, Ireland, Israel, Italy, Japan, Kuwait, the Netherlands, New Zealand, Norway, Portugal, Singapore, Slovenia, South Korea, Spain, Sweden, Switzerland, the United Kingdom, and the United States of America),8 a not altogether inconceivable proposition. Note that even the AP = 2 countries (Hungary, Iran, Romania, Chile, Egypt, India, Indonesia, Ukraine, and Yugoslavia) achieve a PDC index remarkably close to that of the developed countries. From this, it would seem that a large cohort of developing countries has entered the realm of very substantial and sophisticated arms production potential. Even the AP = 1 countries are nearly at the level that the AP = 2 countries achieved in the early 1980s. That and why the more advanced among the developing countries’ arms producers are not yet on par with the likes of Germany, Italy, France, the United Kingdom, and the United States of America is a subject worthy of separate investigation, a task only indirectly attempted in a later section of this chapter. Finally, Table 3.3 compares the new results to those of Brauer (1991). The PDC share for AP = 1 and AP = 3 countries is virtually unchanged between the two time-periods examined, the early 1980s and the early to mid-1990s. The PDC-share for non-arms producers among developing nations (AP = 0) has increased by ﬁve

28

Jurgen Brauer

percentage points. Perhaps in another ten or twenty years, these countries also will enter the realm of arms production potential. The PDC share of the AP = 2 countries has increased by about six percentage points and they are, as pointed out, almost on par with the developed countries, an ominous progression.

Industrial diversiﬁcation The remarkable closeness of AP = 2 and AP = 3 countries’ PDC-share relative to that of developed (AP = 9) countries is perhaps a ﬁgment of the method by which the PDC index is constructed. It measures, recall, the employment and output shares of arms production relevant industries relative to the employment and output value of all of manufacturing. But what if any particular country’s high PDC index obtains solely from employing a lot of labor to produce a lot of output in only one or two, out of the nine, arms production relevant industries (in ISIC revision 2 data)? Brauer (1991) used this reasoning to discover that this was indeed the case, at the time, for Argentina, Israel, and South Africa, and he concluded that these countries, while arms production capable, appeared to strain their available capacities. Clearly, a renewed look at industrial diversiﬁcation within the arms production relevant categories is warranted. Table 3.4 summarizes the results. Data were collected for 155 countries. The nine three-digit ISIC categories consist of a total of 283 sub-categories. For example, ISIC code 351, industrial chemicals, consists of seventy-ﬁve sub-categories, that is, seventy-ﬁve distinct industrial chemicals. On average, the non-arms producing developing nations produced in only two of these seventy-ﬁve categories (e.g. soap), whereas the developed countries produced, on average, in twenty-eight of the seventy-ﬁve categories. Four observations stand out. First, as already observed in Table 3.3, AP = 0 countries clearly are out of the arms production league. They do not score high on the PDC index (Table 3.3), nor are they well diversiﬁed within the ISIC categories (Table 3.4). On average, they produced in only nine of the 283 sub-categories. Second, the AP = 1 countries, while possessing a fairly high share in terms of the PDC index (Table 3.3), are diversiﬁed at relatively low levels in each of the nine ISIC-categories, producing on average in only twenty-nine of the 283 subcategories (Table 3.4). Third, as in Table 3.3, so here in Table 3.4, the AP = 2 and AP = 3 countries are almost indistinguishable from the developed countries (AP = 9). They are as diversiﬁed as are the developed countries and they produce in very nearly as many (AP = 2) or in more (AP = 3) ISIC sub-categories than the developed countries (AP = 9) do. The fourth observation is this: it would appear that the AP = 1 countries in the early 1980s were more diversiﬁed and active in a larger number of ISIC sub-categories (56 or nearly 20 percent of all sub-categories) than in the early to mid-1990s (only 29 or just above 10 percent). It would appear that the AP = 1 countries of the 1990s have fallen further behind in the potential to produce arms. However, this conclusion would be a ﬁgment of the data, for the high score in Brauer (1991) came solely on account of Spain, Portugal, Greece, and Mexico

n 32 12 12 39 60

352

11 4.82 4.67 4.92 2.28 1.42

351

75 27.96 36.58 27.83 8.33 2.15 36 15.86 5.71 7.59

17.29 7.43 7.06

34 14.64 11.00 10.17 2.49 0.63

372

32

31 14.75 17.42 11.92 2.79 0.68

371 57 17.93 18.83 18.08 5.05 1.28

382

59 —N/A— 3.71 21.43 2.43 7.86 2.18 8.24 —N/A—

15

13 3.50 3.33 4.00 1.21 0.52

381

12.29 8.00 6.71

27

34 10.11 12.92 11.33 4.15 1.63

383

10.43 6.14 4.71

21

23 9.18 10.00 9.08 2.77 0.83

384

1.86 0.43 0.24

6

5 1.00 0.58 1.17 0.23 0.12

385

127.43 57.71 55.71

282

283 103.89 115.33 98.50 29.31 9.27

Sum

45.19 20.47 19.75

36.71 40.75 34.81 10.36 3.27

Share (%)

Note The data tables in United Nations (1997) are presented with annual columns for 1986–95. Actual data coverage, by year, varies from country to country. If a country was listed as in any one category in even a single year, the data is represented in the table. The ISIC codes 351–385 are deﬁned in Table 3.2. The number of sub-categories between this study and Brauer (1991) are slightly different.

Source: United Nations (1997). 1995 Industrial Commodity Statistics Yearbook, Production Statistics (1986–95). New York: United Nations Department of Economic and Social Affairs, Statistics Division. This source relies on ISIC revision 2 data.

For comparison (Brauer 1991) n 74 12 9 3 7 38.14 6.43 2 7 15.86 3.86 1 17 15.53 3.47 0

9 3 2 1 0

AP

Table 3.4 Average industrial diversiﬁcation among 283 ISIC revision 2 sub-categories, 1986–95

30

Jurgen Brauer

without which the AP = 1 average industrial diversiﬁcation score would have been, as in this study, just about 10 percent. In contrast, the AP = 2 countries have made astonishing and real advances, from producing in ﬁfty-eight of 282 sub-categories (or 20 percent) in Brauer’s (1991) study to ninety-nine of 283 subcategories (of 35 percent) in this study, putting them nearly on par with the AP = 3 countries (from 127 to 115 sub-categories or from 45 to 41 percent in the early to mid-1990s relative to the early 1980s). I tested whether the differences among the AP groups are statistically signiﬁcant. Listing all developing countries, from AP = 0 to AP = 3, by their industrial diversiﬁcation score, that is, by the number of product categories in which they produce, essentially establishes a rank order of arms production potential. Non-parametric procedures must be used to test for differences in rank-ordered values among independent samples.9 The tests conﬁrm what is visually obvious: • • • •

There are statistically signiﬁcant differences among the four groups, AP = 0, AP = 1, AP = 2, and AP = 3 (Kruskall–Wallis H = 5508; p < 0.0001). When dropping the AP = 0 group from the testing procedure, there still are statistically signiﬁcant differences among the three remaining groups, AP = 1, AP = 2, and AP = 3 (Kruskall–Wallis H = 21.48; p < 0.0001). When dropping the AP = 0 and the AP = 1 groups from the testing procedure, there is no statistically signiﬁcant difference between the remaining AP = 2 and AP = 3 groups (Mann–Whitney U = 65; p > 0.05). Since the industrial diversiﬁcation score for AP = 9 countries lies exactly between the scores for AP = 2 and AP = 3, there is therefore no statistical difference among the industrial diversiﬁcation scores of AP = 2, AP = 3, and AP = 9 countries, that is, between the developed, “Western” countries and the regular and the sophisticated arms producers among the developing nations.

The overall conclusion that emerges from Table 3.4 is that, as in Table 3.3, non-arms producing developing nations do not produce arms because they cannot. Those that only occasionally, even rarely, dabble in arms production (or re-export) are hardly capable of armaments production as regards their industrial structure and diversiﬁcation. But those that produce arms regularly, if at limited quantity and sophistication (AP = 2), would appear just as capable as those that regularly produce fairly large quantities of arms and at high levels of sophistication (AP = 3) and these, in turn, are hardly distinguishable, industrially (and indistinguishable, statistically), from the developed, industrialized, “Western” countries of the world (AP = 9). In a further statistical test, I ranked arms producers (AP = 1, AP = 2, AP = 3) by their industrial diversiﬁcation score and correlated that score with their arms production rank. That is, I correlated their ranked potential for arms production with their ranked actual arms production. The Spearman rank-correlation coefﬁcient is 0.6066 (p < 0.0001). In words, the higher is their potential to produce

Potential and actual arms production

31

arms, the higher is their actual arms production rank. Repeating the Spearman rank-correlation only for AP = 2 and AP = 3 countries, that is, leaving out the AP = 1 countries, results in a coefﬁcient of 0.1151 (p = 0.5923), that is, a statistically insigniﬁcant correlation between potential for and actual arms production among the top twenty-four arms producers among the developing countries. For these countries, the decision to engage in various degrees of arms production therefore depends not on potential (they all have it) but on other circumstances that are not investigated in (nor the subject matter of) this chapter.

Potential in 1980 translates into actual arms production in the 1990s It is instructive to explicitly trace some of the AP = 1 and all of the AP = 2 and AP = 3 of the early 1980s to learn what happened to them with regard to indigenous arms production in the ten or so years to the early to mid-1990s.10 Brauer (1991: 170) wrote: “It is clear that – in the AP = 1 group – Mexico and the three European countries [Greece, Portugal, Spain] could produce more weapons than they do. So could Turkey . . .” By the early to mid-1990s they have – and so has Turkey. In contrast, Brauer (1991: 170) wrote: “. . . in the AP = 3 group, Israel, South Africa, and Argentina apparently strain their industrial resources and abilities.” By the early to mid-1990s, all three continue to be engaged in arms production, but only Argentina appears to have increased its arms production capacity (industrial diversiﬁcation score), whereas South Africa and Israel would appear to continue to strain their industrial abilities. In Asia, the position of India and Indonesia practically reversed. In the early 1980s, India was active in 142 of 282 ISIC sub-categories and heavily engaged in arms production. In contrast, Indonesia produced in only 65 sub-categories and produced arms in limited quantity and sophistication. Ten years later, India produces in only 119 sub-categories and Indonesia in 162, and Indonesia is higher ranked as an actual arms producer than is India.

Human capital formation Brauer (1991: 172, table 4) presented data on primary, secondary, and tertiary school enrollment for 1960, 1970, and annually for 1975–84. He found that primary school enrollment in non-arms producing developing nations (AP = 0) lagged, at 80 percent of their age group, far behind the nearly 100 percent enrollment rates of the other country groups (AP = 1, AP = 2, and AP = 3). As regards secondary enrollment rates, AP = 0 achieved a 31 percent rate by 1984, AP = 1 and AP = 2 were nearly equal at about 48 percent each, and AP = 3 countries achieved a 64 percent secondary school enrollment rate. Incredibly, whereas AP = 0 countries had gained about twenty-two percentage points (from eight to thirty) in secondary school enrollment rates, AP = 1 and AP = 2 countries had gained about thirty points each, and AP = 3 countries gained thirty-six points, so that the enrollment difference increased over time (1960–84) in favor of those countries

32

Jurgen Brauer

that happened to produce many and sophisticated armaments. Exactly the same pattern emerged form tertiary school enrollment. AP = 0 countries advanced from a 2 percent enrollment rate to 6 percent, AP = 1 countries went from 2 to 12 percent, AP = 2 countries from 5 to 13 percent, and AP = 3 countries from 6 to 22 percent.

Table 3.5 School enrollment data Year

AP = 0; n = 58

AP = 1; n = 33

AP = 2; n = 11

Primary school enrollment (as % of age group) 1965 59.67 70.77 81.09 1970 61.38 71.83 84.89 1985 82.57 91.12 100.36 1986 80.55 90.14 98.40 1987 79.03 89.00 98.10 1988 81.33 85.71 98.00 1989 83.02 85.32 97.80 1990 80.00 85.64 97.78 1991 83.45 88.16 97.63 1992 84.79 89.50 97.25 Secondary school enrollment (as % of age group) 1965 10.47 16.64 26.70 1970 14.55 20.14 36.11 1985 27.80 41.04 60.36 1986 23.79 39.55 58.60 1987 25.33 40.32 59.40 1988 27.83 40.00 60.60 1989 27.43 38.50 63.60 1990 25.36 39.43 66.56 1991 29.53 42.84 63.13 1992 32.13 42.04 62.25 Higher education enrollment (as % of age group) 1965 1.76 2.72 5.73 1970 4.86 7.62 12.86 1980 4.82 7.71 10.00 1985 5.56 11.70 12.90 1986 5.30 11.50 13.11 1987 5.45 12.65 13.67 1988 6.17 11.68 14.38 1989 5.98 11.57 14.38 1990 6.00 12.58 14.29 1991 6.31 13.09 14.43 1992 7.72 13.82 15.50

AP = 3; n = 11

AP = 9; n = 27

92.00 92.33 102.13 102.71 104.71 103.38 101.38 96.00 99.50 98.00

103.04a 102.63 102.00 102.92 103.65 103.84 104.40 104.67 103.92 102.27

26.22 35.00 54.50 49.86 50.57 52.25 54.63 54.00 53.00 53.56

54.29 67.13 86.35 87.00 88.81 90.28 89.44 92.65 92.67 90.70

7.22 14.67 11.63 13.75 14.83 14.43 16.25 16.63 16.00 17.75 19.86

12.77 23.24 24.00 27.89 28.85 30.28 31.64 32.88 35.50 38.24 38.85

Source: World Bank, World Development Report (1988–95). Note a School enrollment can exceed 100 percent when students outside the age group enroll.

Potential and actual arms production

33

How has the school enrollment behaved since? Table 3.5 presents data for 1965, 1970, and annually for 1985–92. Examine ﬁrst only the data for developing nations (from AP = 0 to AP = 3). Non-arms producers (AP = 0) have made little improvement. Their primary, secondary, and tertiary school enrollment rates are higher in 1992, but barely so, than they were in 1985, seven years before. The occasional arms producers (AP = 1) hardly fare better than the non-arms producers as regards primary school enrollment (90 vs 85 percent), but much better with regard to secondary school enrollment (42 vs 32 percent) and show a nearly twice as large tertiary enrollment rate by 1992 (14 vs 8 percent). The occasional arms producers, one may say, are signiﬁcantly different from non-arms producers with respect to secondary and tertiary education. A similar break occurs between the occasional (AP = 1) and the regular (AP = 2) arms producers, especially with respect to secondary school enrollment (42 vs 62 percent enrollment rates by 1992). Inasmuch as arms production requires the industrial crafts, a secondary school education is perhaps a prerequisite. Developing nations that regularly produce relatively large quantities of arms at fairly high levels of sophistication (AP = 3) differ from AP = 2 in two ways. Their secondary school enrollment rates are consistently lower (54 vs 62 percent in 1992) but their tertiary enrollment rates are consistently higher by 1992 (20 vs 16 percent). Inasmuch as the sustained production of sophisticated arms requires scientists, engineers, and trade school graduates, the tertiary enrollment rates are perhaps of signiﬁcance. Also note that for the AP = 1, AP = 2, and AP = 3 countries virtually no progress has been made between 1985 and 1992 in secondary school enrollment rates. In contrast, the enrollment rate progress is pronounced and in favor of AP = 3 countries for tertiary school enrollment rates. But the clearest break or difference in the enrollment data comes with respect to the developed countries (AP = 9). Ninety percent of the relevant age group in 1992 enroll at secondary school, a much larger number of children than in developing countries, indeed nearly ﬁfty percent higher than the secondary enrollment rate for the countries with the next highest rates (90 vs 62 percent). This difference is even more pronounced at the tertiary level: developed countries enrolled 39 percent of the age group in tertiary education, whereas the AP = 3 countries managed only half that, at 20 percent. How to interpret these data and noted trends? In light of the ﬁndings regarding industrial structure (Table 3.3) and industrial diversiﬁcation (Table 3.4), where it appeared that AP = 2 and AP = 3 are practically equal to developed (AP = 9) countries, the school enrollment data suggest that, for the time being, highly sophisticated armaments produced and marketed in large variety and numbers (AP = 9) may depend not merely on industrial capacity and diversity but more crucially on the requisite schooling background of countries’ scientists, engineers, and crafts people. In other words, my PDC index as a measure of arms production potential heavily weighs “old,” manufacturing industries rather than “new,” knowledge-intensive industries and is therefore likely to overstate the potential for arms production. Yet the committed arms producers among developing nations are catching up. Relative to the countries studied in Brauer (1991), AP = 2 and

34

Jurgen Brauer

AP = 3 countries show much higher secondary school enrollment rates (in the high ﬁfties rather than high fourties) and have steadily improved their tertiary school enrollment rates as well. Ideally, an improved measure of arms production potential would combine information about manufacturing and knowledge industries into a single augmented PDC index.

Conclusion: implications for the arms trade debate Comparing the ﬁndings of this study with those of Brauer (1991), it would appear that a development dilemma exists: if we want countries to develop economically, that is, in terms of industrial and human capacity, then we must recognize and acknowledge their increasing ability to produce arms. The fact that developing countries’ industrial and human capital has increased and that it has gone handin-hand, after the end of the cold war, with nearly a doubling of known arms producers among them (from thirty-three to sixty-three, see Table 3.1) should give pause. At present, the arms trade debate is dominated by discussion about arms supply restrictions, such as the voluntary UN arms trade registry, and is aimed especially at the big world-market arms suppliers such as the United States of America, the United Kingdom, and France. A recent review and assessment resulted in a “bleak” outlook for arms control success (Cornish 1996). But suppose that these efforts were completely successful. Would that be the end of arms manufacturing and arms trade? Of course not, and there are two primary reasons for that: ﬁrst, there is continued demand for arms, especially among developing nations, and second, as this paper empirically demonstrates they increasingly possess the capacity to manufacture arms themselves. (This is especially true in the realm of so-called small arms which do not require very much industrial and human capital background to produce, yet do nearly all of the killing in the wars within and among developing nations.) Since we cannot and do not want to restrict countries’ ability to improve their manufacturing capacity, it seems to me that we need to supplement studies on arms supply restrictions with those focusing on arms demand reductions, however difﬁcult that may turn out to be. If it is naive to believe that one can inﬂuence countries’ demand for armaments, it is equally naive to believe that restricting arms supplies will somehow prevent developing countries from supplying their own needs in this regard on account of their growing industrial and human capacity to do so.

Acknowledgments I thank Jonathan Brauer for assistance with data entry and statisticians Dr Thiruvaiyaru and Dr Sethuraman for assistance with running the Spearman rank correlations. In addition, I thank an anonymous reviewer, seminar participants at the University of Cambridge (Centre for International Studies, June 1999)

Potential and actual arms production

35

and “Arms Trade, Security and Conﬂict” conference participants at Middlesex University Business School (June 1999) for useful comments.

Notes 1 The non-arms producers (AP = 0) were not included in the statistical test on differences in industrial diversiﬁcation because an earlier table showed (Brauer 1991: 167, table 2), relative to the arms producers, a drastically lower level of industrialization to begin with. 2 Wulf (1983) lists thirty-two arms producing developing nations; Brauer (1991) uses thirty-three, adding Spain to Wulf’s list since economic indicators of Spain such as GNP and average life expectancy did not, then, differ markedly from Israel, Greece, and Singapore, all of which Wulf’s list included. 3 Complete listings are available in Appendix A of the draft version of this paper posted at the author’s website: http://www.aug.edu/∼sbajmb/paper-london2.PDF. 4 Wulf (1983) lists an additional seven countries that either produced “only ammunition” and another ﬁve engaged in the construction of “simple ship hulls” (p. 313). These he excluded from his study, whereas I include them. 5 Oddly, Rana’s listing leaves out the United States of America as a producer of “small arms”. I added a 1 for the United States of America anyway. Rana’s listing also includes Taiwan, but Taiwan is excluded from this study because statistics on Taiwan are not available from international organizations such as UNIDO and the World Bank on which I rely elsewhere in this study. 6 Of course if the procedure to create the raw score and tie-breaking rules were different, the rankings would change. My intuition is that changes would occur at the margin but not in substance. Comparing alternative rankings may be the subject of a separate research effort. For example, running a variety of simple correlations between the ranking reported here and those reported by SIPRI (1999, table 11.1) results in coefﬁcients of about 0.7. 7 A further developed country is Iceland, but data availability was so spotty that it is excluded from analysis. Iceland, incidentally, does not maintain any armed forces. 8 When Cyprus, Kuwait, New Zealand, and Australia are excluded from the computations in Table 3.3 (on the basis that they are particularly natural-resource based), the PDC employment share for the remaining developed countries is 51.22 percent and the PDCoutput share only 40.72 percent. 9 See Appendices B, C, and D in the online source cited in note 3. 10 For details, see table 5 of the paper listed in note 3.

References Brauer, J. (1991) Arms production in developing nations: the relation to industrial structure, industrial diversiﬁcation, and human capital formation. Defence Economics 2(2), 165–175. Brzoska, M. (1995) Spread of conventional weapons production technology. In: Lodgaard, S. and Pfaltzgraff, Jr., R. L. (eds), Arms and Technology Transfers: Security and Economic Considerations Among Importing and Exporting States. United Nations Institute for Disarmament Research (UNIDIR), New York, Geneva, pp. 19–47. Cornish, P. (1996) Controlling the Arms Trade. Bowerdean, London. International Institute for Strategic Studies. (1998). The Military Balance 1998/99. Oxford University Press, Oxford.

36

Jurgen Brauer

Rana, S. (1995) Small arms and intra-state conﬂicts. Research paper #34. United Nations Institute for Disarmament Research (UNIDIR), Geneva. SIPRI (1999) SIPRI Yearbook 1999: World Armaments and Disarmament. Oxford University Press, Oxford. United Nations. (1997) 1995 Industrial Commodity Statistics Yearbook, Production Statistics (1986–1995). United Nations Department of Economic and Social Affairs, Statistics Division, New York. United Nations Industrial Development Organization. (1998) International Yearbook of Industrial Statistics 1998. UNIDO, Vienna. United States Arms Control and Disarmament Agency. (1997) World Military Expenditures and Arms Transfers 1996. US ACDA, Washington, DC. World Bank (1988–95; 1999) World Development Report. Oxford University Press, New York. Wulf, H. (1983) Developing countries. In: Ball, N. and Leitenberg, M. (eds), The Structure of the Defense Industry: An International Survey. St Martin’s Press, New York, pp. 310–343.

4

Export controls, market structure and international coordination María del Carmen García-Alonso and Keith Hartley

Introduction In the post-Cold War era, the perceived security requirements of the main producers of weapons have changed signiﬁcantly.1 Among these concerns, we have civil strife and ethnic conﬂict (e.g. Kosovo), terrorism, regional threats, the proliferation of weapons of mass destruction and even the environmental consequences of improper storage or disposal of military related material, especially nuclear technologies. In 1992, the Security Council of the United Nations (UNSC) determined that the “proliferation of all weapons of mass destruction constitutes a threat to international peace and security” and the member countries committed “to working to prevent the spread of technology related to the research for or production of such weapons and to take appropriate action to that end.” The fact that among the members of the UNSC we ﬁnd the most important exporters of weapons highlights the importance of this commitment. Indeed, prior to 1992, Western industrialized countries had already started to organize supply-side controls on transfer of weapons-related technologies. At present, there are a number of multilateral export control regimes that include the nuclear suppliers group (NSG) and the Zangger Committee (ZC), which address nuclear technology; the Australia Group (AG), which deals with chemical and biological military-related material; the Missile Technology Control Regime (MTCR) and the Wassenaar Agreement (WA), which refer to conventional arms and dual-use goods and technologies. Also, extreme export controls in the form of embargoes have been monitored by the UN and NATO (e.g. Iraq). As the spin-offs between the civil and the military sector now ﬂow in both directions and the speed of technological innovation is very high in these sectors, the scope of the above-mentioned export controls extends to goods or technologies that have a direct or potential military applications, the so-called dual-use products. In this chapter, we present different ways of aggregating the exports of dual-use products to give the security perception of exporter countries and their consistency with the relevant export control regimes. Also, we analyze different models of export controls highlighting the role of the perception of security, market structure and competition between exporting ﬁrms in determining the existence of multiple

38

María del Carmen García-Alonso and Keith Hartley

equilibria and, therefore, the need for coordination between countries in setting export controls. There is some literature that analyzes the issue of arms control (Hartley and Sandler 1995; Sandler and Hartley 1995). This has been, in part, motivated by the arms control treaties between the United States of America and the former Soviet Union in the 1980s. At that time, the security perception of those countries was highly inﬂuenced by their involvement in an arms race. This situation resulted in an extensive literature that focused on the analysis of defence expenditure in defence alliances with defence expenditure considered as a public good to which the members of an alliance contribute (Hartley and Sandler 1999). In addition to arms races, the trade in arms affects perceptions of security and raises the related issue of export controls. A number of recent papers have introduced the trade in dual-use products as an important factor in the study of security among the main exporters. The common feature is that these models consider exports of arms and dual-use products as having a negative effect on the suppliers’ security, that is, these exports generate negative externalities on exporter countries. This idea approaches the problem of export controls for dual-use products in the same way as some environmental problems like global warming and nuclear waste anti-dumping (Moody-O’Grady 1995; Rotillon et al. 1996). Both issues have generated a need for agreements at the international level and discussions over the concept of security: military security or environmental security. The key difference between them is the importance that the market structure and competition between exporter ﬁrms has on the ability to control the exports of dual-use products: this creates a double interaction between countries at the security level and at the exports proﬁts level, which sometimes generates a need for the design of speciﬁc mechanisms that help multilateral agreements to be implemented. In the arms trade literature, Levine et al. (1994) and Levine and Smith (1995) present a model that speciﬁes not only the economic features but also political characteristics of the arms trade. They provide a formal model of trade that allows for competing forward-looking suppliers whose welfare depends on both the economic beneﬁts from the sales and the security repercussions of the recipient’s behavior. In García-Alonso and Levine (2002) and García-Alonso (1999), the consideration of military ﬁrms and exporting governments as different decision makers is introduced; strategic effects involved in the decision on domestic weapons procurement and arms exports are analyzed. Also, Levine and Smith (1997) and García-Alonso (2000) introduce, the issue of exports controls for dual-use technologies in two different ways. Levine and Smith (1997) studied different forms of cooperation between arms exporters deciding on domestic procurement and arms exports and compares, among other things, the equilibrium quantities in each case. García-Alonso (2000) examines the relationship between the optimal subsidization policies of R&D investments of exporters of dual-use technologies with the existence of a unilateral or multilateral concern for security in exporter countries.

Export controls, market structure

39

A key difference between the various papers is the way each of them models the effect of exports on security. While they all contemplate the negative effect of exports on the suppliers security, the way in which these exports are aggregated differs depending on whether or not product quality is considered. The issue of aggregation technologies has also been raised in the context of public goods. Conybeare et al. (1994) discussed different ways of aggregating military capabilities of the members of a military alliance. Also, it has been applied to international environmental agreements in Sandler and Sargent (1995). Cornes and Sandler (1996) provide a detailed review of the relevant literature. However, export control agreements have not yet been analyzed under this perspective. The chapter is organized as follows. First, it presents some stylized facts of arms exports. Second, it examines the different ways of aggregating exports of dual-use products to give the security perception of exporter countries and their consistency with the relevant export control regimes. Third, it analyzes different models of export controls highlighting the role of the perception of security, market structure and competition between exporting ﬁrms in determining the existence of multiple equilibria and, therefore, the need for coordination between countries in setting export controls. Finally, it describes the most signiﬁcant characteristics of the existing dual-use export controls.

Some stylized facts This section describes some of the main features of the international arms trade that our model seeks to address. Arms exports are big business in a market dominated by relatively small numbers of supplying and buying nations. Over the period 1993–7, the United States of America was the world’s leading supplier accounting for 47 percent of total weapons deliveries; the top ﬁve suppliers, including Russia, United Kingdom, France and Germany, accounted for 82 percent of total deliveries. Over the same period, the top ten recipients or defence equipment accounted for over 50 percent of total imports (ranked by value of imports, these were Saudi Arabia, Taiwan, Turkey, Egypt, South Korea, China, Japan, Greece, Kuwait; SIPRI 1998). It might be expected that small numbers of nations, especially as suppliers, are more likely to reach an international agreement on arms export controls. Participation in the world arms market will partly reﬂect the different comparative advantages of both buyers and sellers with the market reﬂecting vertical differentiation. Some nations will specialize in supplying costly, high-technology arms exports (e.g. France, United Kingdom, United States of America); others will specialize in supplying cheaper, low-technology arms exports (e.g. Brazil, China). Similarly, buying nations will have different income levels affecting their ability to pay with poorer nations demanding cheaper and, hence, lower-technology arms, with such demands reﬂecting the comparative advantage of their national armed forces (e.g. labor-intensive conscript forces with limited human capital will require “simple technology” weapons). Arms exports are determined by the usual price and non-price variables, but they are “different” in their dependence on political variables. Governments dominate

40

María del Carmen García-Alonso and Keith Hartley

the market through support for their national defence industries and their role in allowing arms exports and determining the terms of trade (e.g. R&D levies, favorable ﬁnancial terms). Governments might also favor certain nations (e.g. allies, former colonies, friends), which might then be viewed as “captive markets.” Defence industries have the characteristics of both a military and economically strategic industry. They are economically strategic in terms of R&D intensity, spinoffs and decreasing cost reﬂecting both economies of scale and learning (Sandler and Hartley 1999). With high ﬁxed R&D costs and decreasing unit production costs, output is a major determinant of unit total costs. Disarmament following the end of the Cold War has resulted in fewer new projects and smaller national orders, leading to pressures for defence companies to seek export markets. For national governments, arms exports are a means of maintaining their defence industries in an era of disarmament with exports resulting in “wider economic beneﬁts” (Hartley 2000). But there are trade-offs with governments having to choose between support for their domestic defence industries and their concern with the possible impacts of arms exports on national security (e.g. importing nation might be a future threat to the exporting nations either directly or through regional conﬂicts that involve the exporting nation). This concern with both national defence industries and national security might affect the exporting nation’s attitude to the form of export controls where these vary between quantitative and qualitative. At the limit, there might be a complete ban on arms exports, but such a policy is not costless and the nation has to estimate the economic impacts of a ban, including the costs of retaining any defence industrial capability for national defence needs. Obvious costs from an export ban include losses of jobs and exports, higher prices for national defence equipment and the costs of either retaining capacity or shutting it down and re-starting between gaps in orders (Martin et al. 1999; Hartley 2000). An alternative to a complete ban might be to control the quantity of arms exports to certain nations or to control their quality. Restrictions on quality mean that the exporting nation would refuse to supply the latest high-technology equipment to foreign buyers. For example, combat aircraft and helicopters might be exported but without the latest radar, avionics and stealth features; conventional submarines might be exported but not nuclear-powered submarines. Of course, importing nations might respond to these various forms of export controls by buying from other nations or developing their own national defence industry (cf. Israel and South Africa). Policies on arms exports can be either national or international, with international agreements offering the prospects of a public good in the form of peace. But international collective agreements involve substantial transaction costs in identifying participating countries, negotiating an agreement and then monitoring and enforcing it. Collective decisions are needed on the deﬁnition and range of weapons to be included in the agreement, the target importing nations to which the agreement applies, the policing and monitoring arrangements, the penalties for non-compliance and the incentives created for illegal trading. Even the deﬁnition of weapons causes problems, especially for dual-use equipment (e.g. helicopters can be used as civil passenger transports or as military transports and

Export controls, market structure

41

observation posts). And, in the ﬁnal analysis, actual arms export controls are implemented by nation states. These stylized facts form the background for our formulation of models of export controls.

Technologies for aggregation of international negative externalities This section examines different ways in which negative externalities caused by exports of dual-use products can be aggregated to give a perception of the security consequences for the suppliers of this trade. Also, we discuss which ﬁts better with each of the current export control arrangements. The discussion ignores the effect on security of domestic procurement in each of the exporter countries.2 Let qi denote the military capability exported by country i and Si represent the overall effect of exports on supplier i’s security: Si = Si (q1 , . . . , qn ),

(4.1)

where n is the number of exporters and Si is decreasing in all arguments. The speciﬁc aggregations we now propose resemble those used in the public good literature for aggregating individual contributions in the supply of public goods. However, the interpretation is somewhat different; we obtain the aggregate negative externality on suppliers produced by exports of dual-use equipment. The constant elasticity of substitution (CES) function is extensively analyzed in Cornes and Sandler (1996) as a generalized technology of public good supply aggregation. The arms trade literature has also made use of the CES function3 for the aggregation of arms exports. García-Alonso (1999) used it as part of a general security function in which the effect on security of domestic procurement was also considered: Si = −

n

1/σ σ

(qi )

,

(4.2)

i=1

where 1/(1 − σ ) is the elasticity of substitution between the exports of different countries in the exports security function. In our context, the elasticity of substitution determines the impact any country’s exports has on security. When σ = 1, all countries’ exports are perfect substitutes on the ﬁnal negative impact on security: Si = −

n

qi .

(4.3)

i=1

This aggregation ﬁts best with military capability being related to quantity of homogenized dual-use products exported and provides a rationale for quantitative export controls. Also, it could be applied to global warming where products are homogenized in terms of the units of CO2 they produce.

42

María del Carmen García-Alonso and Keith Hartley

When σ approaches ∞, we have a technology of aggregation in which what determines the global negative externality is the maximum of the exported military capabilities: Si = − max(q1 , . . . , qn ).

(4.4)

An interesting property of this aggregation is that increases in any of the countries’ exported military capability do not affect security unless it goes beyond the level exported by any of the other countries. In this case, we get close to the bestshot technology for aggregating individuals’ contributions to public goods. This was used ﬁrst in Hirshleifer (1983) for public goods. García-Alonso (2000) used it for the aggregation of dual-use goods exports to give a perception of security on exporter countries. This method ﬁts well when the quality and not the quantity of the product exported is what really affects security. However, note that with this type of aggregation, only the highest exported quality affects security. One could also consider the case in which other lower exported qualities also affect security;4 however, the best-shot aggregation provides a good index of security perception of countries that set qualitative export controls specifying lists of products that cannot be exported, based on their applicability to military purposes (e.g. AG). Let us now discuss which deﬁnition of military capability seems to correspond best with the existing dual-use export controls. In general, military capability is a function of both the quality and quantity of exports. Export controls could be either quantitative or qualitative. The ﬁrst term would embrace controls that set a numerical ceiling on permitted classes of the amount of exports. The second term would imply limits on the performance or quality of exports of dual-use products (this classiﬁcation is similar to the one proposed in Panofsky 1990). In practice, the existing export control agreements refer to the quality or powerfulness of dualuse technologies exported. It does not seem that the countries taking part in these agreements perceive quality and quantity as being substitutes for each other. For instance, in the United States of America, The National Defense Authorization Act on High Performance Computers (HPC)5 controls and restricts the exports of powerful computers to a group of ﬁfty countries including Russia, China and Israel. It is obvious that in this case quality and quantity are not substitutes for each other. In other words, having a computer that is able to simulate a nuclear experiment is not at all the same as having many computers that do not have enough power to do it. It could be argued that, in terms of security, it should not be the same one restricted country having access to a higher quality than several restricted countries managing to acquire it. For instance, if India had access to HPC and Pakistan did not, would it be better or worse for the exporter’s security than both getting it? There is no clear answer. In principle, one of the two countries having HPC should be worse, but Pakistan and India (as most of the importers of restricted dual-use goods) are involved in an arms race. Under this consideration, exporting a high-technology good to India only could dangerously destabilize the arms race; besides, the willingness of Pakistan to pay for the high technology would increase making it even more tempting for producer countries to export it

Export controls, market structure

43

to Pakistan too. We can then conclude that quality alone gives a good index of the deﬁnition of military capability for countries who are interested in implementing export controls. In the following section, we use the best-shot technology in presenting different non-cooperative games that attempt to explain the strategic interaction between countries involved in export control arrangements and how their decisions are affected by the exports market structure. We ﬁrst use a discrete choice model in which governments are faced with the decision of whether or not to restrict a domestic ﬁrm’s export quality. We then extend the discussion to a continuous choice and general security function situation.

The restriction game Discrete choice model One of the distinguishing characteristics of the exports control problem is the twolevel interaction between governments deciding their optimal export controls and still caring about their domestic ﬁrms that compete in the international market for dual-use products. The decision faced by exporting countries can be presented in the following way: W i = π i + θ i Si ,

(4.5)

where Wi is social welfare, πi the domestic ﬁrm’s proﬁts, Si = − max(q1 , . . . , qn ) and θi the degree of security concern of the government; so, if θi = 0 the government’s and the ﬁrm’s problem would be equivalent. Consider an export market composed of two countries with one ﬁrm each exporting dual-use products to the rest of the world. Firms compete in qualities and governments face the decision on setting export controls on the quality that the home ﬁrm exports or not. It is very helpful at this point to start with a discrete choice model and a best-shot technology for aggregation. In this case, the problem faced by governments is indeed a two-choice game and it is presented in Table 4.1. As can be seen, the proﬁt of each ﬁrm depends on the quality that both export, πi (q1 , q2 ), the ﬁrst quality is the quality exported by the ﬁrm in country 1 and second term is the quality exported by the ﬁrm in country 2. The game we present in this section is a version of García-Alonso (2000). That paper presented a three-stage game in which governments commit to an R&D subsidy for the domestic ﬁrm before ﬁrms choose their R&D investment. Once governments know the outcome of the R&D process, they choose the optimal Table 4.1 Two-choice game 1, 2 R

NR

R π1 (qL , qL ) − θ1 qL , π2 (qL , qL ) − θ2 qL π1 (qL , qH ) − θ1 qH , π2 (qL , qH ) − θ2 qH NR π1 (qH , qL ) − θ1 qH , π2 (qH , qL ) − θ2 qH π1 (qH , qH ) − θ1 qH , π2 (qH , qH ) − θ2 qH

44

María del Carmen García-Alonso and Keith Hartley

technology security policy that takes the form of a maximum exportable quality. This policy becomes relevant when the domestic ﬁrm is successful in R&D and it consists of allowing or not the domestic ﬁrm to export the innovation depending on the quality available to the competitor. Given this policy, ﬁrms make their quality choices and, ﬁnally, compete in prices. Our intention is to focus on the game theoretical interaction between exporter countries whose ﬁrms have been successful in developing a higher quality. The fact that governments are not able to commit on export controls before ﬁrms invest in R&D makes our simpliﬁed game valid for our purposes. All R&D costs are sunk and governments know the quality that competitors are able to sell. It is then left with the exports control decision. Our objective is to analyze the problem involved in implementing a multilateral export control regime given that member countries already have access to the technologies whose exports they intend to restrict. There are two qualities that ﬁrms are able to produce: qH and qL , with qH > qL . The dilemma faced by governments consists of deciding whether or not to allow the higher quality to be exported. The result of this game will depend crucially on the kind of competition between exporting ﬁrms and on the degree of concern for security, θi , of each government. Consider a situation in which importer countries care about quality but they also perceive the product offered by each ﬁrm as different: in other words, we consider a model of horizontal differentiation in which quality matters (see Economides 1989 for an example). In this case, the proﬁt functions of each ﬁrm would be increasing in its own quality so that π1 (qL , qH ) < π1 (qH , qH ), π2 (qH , qL ) < π2 (qH , qH ). Also, if both ﬁrms export the same higher level of quality, their proﬁt is not smaller than they would obtain if none of them could export the innovation, that is, πi (qL , qL ) ≤ πi (qH , qH ), i = 1, 2. Finally, proﬁt is a decreasing function of the competitor’s quality, that is, π1 (qL , qH ) < π1 (qL , qL ), π2 (qH , qL ) < π2 (qL , qL ), π1 (qH , qH ) < π1 (qH , qL ) and π2 (qH , qH ) < π2 (qL , qH ). Under these conditions, given that government 2, for instance, restricts the export quality of its ﬁrm to qL , government 1 will restrict its ﬁrm too if and only if when it restricts, welfare is higher, W1 (qH , qL ) < W1 (qL , qL ). For this to be the case the degree of security concern of government 1 should be higher than π1 (qH , qL ) − π1 (qL , qL ) ≡ θ1∗ . q H − qL

(4.6)

This provides a cut-off line between security concerned and unconcerned governments. In this context, a security concerned government would be that which would restrict the exported quality if it were the only one to have the high quality. If both governments are security concerned (i.e. θ1 > θ1∗ and θ2 > θ2∗ ), we have two Nash equilibria in pure strategies6 in the game presented in Table 4.1: either both governments restrict (R) or neither of them restricts (NR) the domestic ﬁrm. This is because restricting one’s own ﬁrm only affects security if it lowers the maximum exported quality. If the high-quality level is available from the other country’s company then it is better for the government not to restrict its home ﬁrm

Export controls, market structure

45

so that it can compete effectively. These two possible equilibria can be ranked by comparing the welfare that the governments achieve in each of them. It is easily shown that governments prefer the equilibrium in which both restrict if their concern for security is higher than πi (qH , qH ) − πi (qL , qL ) ≡ θi∗∗ . qH − qL

(4.7)

Note that since the proﬁt functions are decreasing in the competitor’s quality, the degree of security concern for which governments prefer a global restriction is smaller than the degree of security concern for which one government decides to restrict when it is the only country to have the high quality, that is, θi∗ > θi∗∗ . In general, the discrete choice games with two symmetric Nash equilibria, one of them being Pareto dominant, was ﬁrst analyzed in Harsanyi and Selten (1988) and considered as a possible example of negative externalities in Dybvig and Spatt (1983). The problem with this game is that it is difﬁcult to give a clear prediction of which of the two equilibria will result. Indeed, even if both governments get together before deciding whether to restrict or not and assure they plan to restrict we should not expect that they believe each other (Aumann 1990). The reason is that, independently of its own choice, government 2 gains if government 1 restricts; hence, government 2 would always say it is going to restrict, even when it is not planning to restrict. This characteristic of the game is due to the fact that a ﬁrm’s proﬁts depend on both exporters’ quality, so that ﬁrms always prefer to have a leadership in quality with respect to the other ﬁrm. If ﬁrms had their own captive markets the situation would be different. In the extreme case, having a captive exports market means that the ﬁrm’s proﬁt function would no longer be dependent on the other ﬁrm’s exported quality. In this situation, the game would have the features of the Stag Hunt game and pre-play communication would help countries to implement a multilateral restriction. It has been discussed (Sandler and Sargent 1995) that, especially in coordination games in which players do not trust each other, a mixed strategies equilibrium makes sense. For the game in Table 4.1, the mixed strategies equilibrium is given by the probabilities of the other country cooperating that makes them indifferent between restricting or not. For government 2, this probability is ρ2 =

π1 (qH , qH ) − π1 (qL , qH ) . π1 (qL , qL ) − π1 (qH , qL ) + θ1 (qH − qL ) + π1 (qH , qH ) − π1 (qL , qH ) (4.8)

Therefore, if both governments implement the restriction with a probability bigger than ρi , then both governments’ restricting equilibrium would be the predicted outcome. Note that an increase in the degree of security concern clearly favors the chances of the restricting equilibrium. As predicted for these types of games in the literature, an increase in the number of countries that have access to high quality and take part in a multilateral restriction game would diminish the chances of observing the restriction on quality

46

María del Carmen García-Alonso and Keith Hartley

implemented. The ﬁrst reason is that, according to our deﬁnition of security, one country not restricting exported quality has the same effect on security as all countries defecting. As a consequence, ρi would be bigger with more countries and the chances of a multilateral restriction smaller (see Fudenberg and Tirole 1995: chapter 1; Sandler and Sargent 1995). Besides, an increase in the number of exporting countries will also increase competition and affect negatively the per country proﬁts and, therefore, the payoff matrix. Consider now the effect of the existence of uncertainty over the degree of security concern of governments. For instance, imagine a situation in which government 1 has some uncertainty over the degree of security concern of government 2 such that P2 is the probability that θ2 > θ2∗ and 1 − P2 is the probability that θ2 < θ2∗ . Rewriting the game in Table 4.1 in terms of expected welfare for government 2, we conclude that, although none of the governments’ restriction remains a Nash equilibrium in pure strategies, both governments’ restricting equilibrium might disappear if it is sufﬁciently unlikely that government 2 cares about security. It is also worth noting that if government 2 does care about security, it is in its self-interest to reveal it to the security conscious country. Continuous choice model The previous section assumed that, when restricting, countries were faced with a discrete choice. This makes sense in cases in which innovation processes are discrete and the qualitative differences between two different generations of products are signiﬁcant in terms of security. In other instances, like a missile’s speed or precision, it is very difﬁcult to justify a discrete choice framework. Also, if countries were attempting to restrict quantity a continuous choice would be more suitable. This section analyzes the restriction game in a continuous choice environment. Governments will be confronted with a continuous choice of restriction. Our objective is to see under which conditions we obtain multiple equilibria in this framework.7 For this purpose, we will use a model in which governments must decide the amount of dual-use products they allow their ﬁrms to export taking into account both the domestic ﬁrms’ proﬁts and the security consequences of such exports. We use the general form of security described in Eq. (4.1) and the welfare function stated in Eq. (4.5). Therefore, the problem for government 1 is presented as follows: Max π1 (q1 , q2 ) + θ1 S1 (q1 , q2 ). {q1 }

(4.9)

The ﬁrst-order condition for this problem is ∂π1 (q1 , q2 ) ∂S1 (q1 , q2 ) + θ1 = 0. ∂q1 ∂q1

(4.10)

Export controls, market structure

47

This is also the reaction function of government 1 to the quantity allowed to be exported by government 2. A necessary but not sufﬁcient condition for having multiplicity of symmetric Nash equilibria is that both reaction functions have a positive slope greater than 1 at a Nash equilibrium:8 ∂q1 (∂ 2 π1 (q1 , q2 )/∂q2 ∂q1 ) + θ1 (∂ 2 S1 (q1 , q2 )/∂q2 ∂q1 ) =− > 1. ∂q2 (∂ 2 π1 (q1 , q2 )/∂q12 ) + θ1 (∂ 2 S1 (q1 , q2 )/∂q12 )

(4.11)

The denominator of this expression must be negative for this is the slope measured at a Nash equilibrium. Therefore, a sufﬁcient condition for the slope to be positive is that the numerator is positive. In other words, we need the government’s marginal welfare of increasing the exported military capability to rise with increases in the other government’s export capability. Note that this is a property present in the discrete choice games studied in the previous section. When the other country does not restrict, it becomes welfare improving for the government not to restrict as opposed to the case in which the other country restricts exported quality. As by exporting military capability countries are generating a negative externality not only to themselves but also to the other exporters, we can say, in general, that equilibria in which the exports level is higher will imply lower welfare.9 If there is a unique Nash equilibrium, the level of exported military capability is higher than the Pareto optimal one: in this case we would have a Prisoner’s Dilemma problem. In other words, we have a unique Nash equilibrium that is Pareto inferior. Introducing vertical differentiation So far, we have analyzed the nature of the game played by countries that, having access to the “state-of-the-art” technology and, based on security grounds, aim to prevent ﬁrms from exporting it to other countries. However, there are also countries that do not have access to the high-technology products but, instead, they export low-technology products to importers who cannot afford to buy the highest quality. Using the literature on vertical differentiation (e.g. Shaked and Sutton 1982), when importers have different incomes, lowtechnology ﬁrms would still have some positive proﬁts by selling a lower quality at lower prices to low-income countries. An interesting feature of the vertical differentiation model is that the optimal quality for the low-technology ﬁrm is an increasing function of the quality exported by the high-technology ﬁrms. Therefore, in a continuous choice environment, if countries involved in exporting high qualities manage to agree to diminish the quality of their exports, low-technology exporters would also diminish their qualities in order to keep the quality gap between them. This would hold even if their governments did not care about security (as long as the restricted quality is above the technological frontier of low-technology countries).

48

María del Carmen García-Alonso and Keith Hartley

Evaluation of multilateral military-related export control regimes This section identiﬁes the characteristics of the multilateral military-related export control regimes that reveal the coordination problems involved in these regimes and the coordination mechanisms designed by some of them in order to achieve an effective multilateral restriction regime.10 Table 4.2 presents the membership of the different export control regimes and the year in which each was created. First, note that most of the regimes were created by the end of the Cold War and, therefore, the perception of military-related exports as having a negative effect on security is a characteristic of the post-Cold War era. Interestingly, the last regimes to be created are those that target speciﬁcally dual-use products and technologies. This can be seen as the consequence of the spin-offs between the civil and the military sector increasingly ﬂowing in both directions. It is a general characteristic of these export control regimes that they are implemented through national export control mechanisms. Therefore, they are all voluntary arrangements. The conditions for membership differ between them, but they have in common that the prospective member must develop or have national export controls on the relevant items. Besides, the overall approach of the country to non-proliferation issues is also taken into account. So far, we have seen that the military-related export controls are multilateral in the sense that they are implemented by a group of countries whose common characteristic is having national export controls, which signals an individual concern for security. But, these agreements do not have direct enforcement power: membership is voluntary and there is no penalty system. What then is the role they play in facilitating multilateral restrictions? The same question has been posed for other multilateral institutions that attempt to promote international trade cooperation (e.g. Maggi 1999 for the World Trade Organization). In that case, we have a Prisoner’s Dilemma, which, as is well known, when played repeatedly leads to a multiplicity of equilibria. This chapter has used a static game structure where a multiplicity of equilibria arises and, with that, the need for coordination on the good equilibria. Some of the suggested roles of a multilateral institution, like export control arrangements, would be to act as an information gathering agency, generate a sense of “international obligation” and promote multilateral rule making. In doing so, these multilateral arrangements add an important feature to the individual national export controls of the security concerned countries, which is the joint implementation of coordination mechanisms. Information gathering is made in most of the export control arrangements through agreed lists of controlled items. These lists describe the products and technologies that are considered relevant for each speciﬁc export control arrangement. In building up these lists member countries make explicit the items over which member countries are security concerned. This plays an important role in the national implementation of the export control because, as was seen in the previous section, uncertainty about the degree of security concern of other potential exporter

Table 4.2 Membership of multilateral military-related export control regimes, as of 1 January 1998 State

Zangger Committee,a 1974

NSG,b 1978

Australia Group, 1985

MTCR,c 1987

EU dual-use Regulation, 1995

Wassenaar Arrangement, 1996

Argentina Australia Austria Belgium Brazil Bulgaria Canada China Czech Republic Denmark Finland France Germany Greece Hungary Iceland Ireland Italy Japan South Korea Luxembourg The Netherlands New Zealand Norway Poland Portugal Romania Russia Slovakia South Africa Spain Sweden Switzerland Turkey UK Ukraine USA Total

× × × ×

× × × × × × ×

× × × ×

× × × × ×

× × × ×

×

×

×

×

n.a. n.a. × × n.a. n.a. n.a. n.a. n.a.

× × × × × ×

× × × × × ×

× × × ×d × × × × × × × × × × × ×

× × × × × × × × × × × × × × × × ×

× × × × × × × × × × × × × × × × × ×

× ×d × 33

× × × 34

× × ×d ×

×

× × × × × × × × × × × × × × × ×

×

× × × × ×d ×

× 30

× 29

× × ×

× × × × × n.a. n.a. × × n.a. n.a. × × n.a. n.a. n.a. × n.a. n.a. n.a. n.a. × × n.a. n.a. × n.a. n.a. 15

× × × × × × × × × × × × × × × × × × × × × × × × × × × × × 33

Source: SIPRI Yearbook (1998). Notes The years in the column headings indicate when the export control regime was formally established, although the groups may have met on informal basis before then. n.a. not applicable. a The European Commission is represented in this regime as an observer. b The Nuclear Suppliers Group. The European Commission is represented in this regime as an observer. c The Missile Technology Control Regime. d This state became a member of the regime in 1997.

50

María del Carmen García-Alonso and Keith Hartley

countries makes it less likely that an equilibrium in which countries multilaterally decide to restrict is achieved. Also, especially for dual-use goods, it is sometimes technically difﬁcult to deﬁne which goods can have potential military application. Therefore, deciding over lists is a way for some countries to realize the military application of some dual-use products about which they might not be aware. In the case of the ZC and the NSG, the list of controlled items is called the Trigger List, the reason being that any potential export of a listed item to a nonnuclear weapon state triggers the need for International Atomic Energy Agency (IAEA) safeguards. The safeguards aim to verify that the potential importer is not using nuclear material or equipment to develop or produce nuclear weapons. In May 1997, the IAEA safeguards were strengthened and renamed as full-scope safeguards. A key element of the new safeguards is an enhanced information system managed by the IAEA based on expanded declarations on nuclear transfers. The members of the NSG have agreed on implementing full-scope IAEA safeguards over potential transfers of items in the Trigger List. Again, this makes clear the way in which countries make explicit their perception of security and, therefore, helps in achieving the multilateral implementation of export controls. We have seen the positive aspects of the existence of agreed lists of controlled items. Nonetheless, there might be some risks involved in relying too much on lists when dealing with dual-use goods. First, there is the updating problem: for technologies in which innovation speed is high it can be difﬁcult to have lists that react quickly to changes in the technological patterns of military-related products. Encryption technology is one of the best current examples. Besides, ﬁrms competing in export markets might advance a potential restriction over the items it attempts to export and, therefore, try, artiﬁcially, to change their characteristics so that they are not subject to export controls: this does not necessarily mean that their goods are not military-related anymore. Maybe, this is the reason why in the WA the list of controlled items has not yet been released. Another way of avoiding the lack of ﬂexibility of the lists system is to add discretionary mechanisms for speciﬁc cases. Here, we could include the catch-all or know rule used in the MTCR. The idea is that, if an exporter is aware that an item will contribute to the proliferation of weapons of mass destruction, the export should be prevented whether or not it conforms to technical parameters of the commodity control list. There is one ﬁnal coordinating mechanism that can help security concerned countries to implement a multilateral export controls and which can be found in the regulations of the Australia Group. In June 1993, the AG adopted a so-called no-undercut policy.11 The policy seeks to avoid a situation in which an AG member competing for a lucrative business deal tendered by a potential proliferator would grant an export licence under the presumption that otherwise another AG state would do so. The AG countries honour the decisions of other AG states to deny a particular export. If an AG country does not grant an export licence it notiﬁes the other AG states of its decision and provides them with information regarding the goods, their destination and the end-user. If, however, a second

Export controls, market structure

51

AG member has doubts about or disagrees with the proliferation risk assessment on which the original denial was based, it is obliged to consult with the country that denied the export licence before proceeding with a sale, which otherwise would undercut the original denial. The outcome of this consultation mechanism can be either that the state which has issued the denial notiﬁcation revokes it, and thus allows the export to proceed, or that both countries agree on the soundness of the denial and, consequently, refuse the licence. (SIPRI Yearbook 1998) The no-undercut policy is a tool that recognizes the role played by competition between different exporter ﬁrms in the implementation of export controls. As seen in the previous section, designing credible mechanisms by which governments make explicit their decision not to take competitive advantage from other countries’ restrictions is especially important in export industries where competition is ﬁerce and exporting ﬁrms have no captive markets. It is not by chance then that the need for an undercut policy was intensively discussed among the members of the WA. However, the members of the WA have not yet agreed to refuse a licence for a transfer of the same product to the same destination where another member has denied it. Most of the export control agreements also have regulations that prohibit unauthorized re-transfers. These regulations become especially important in the case of dual-use products. A country might decide to export a dual-use product to a country that does not care about security as long as it is sure that it is not going to be used for a military application. But, this country could re-transfer the product to a country that would use it for military purposes as it does not care about security. The prohibition of transfers attempts to avoid this event. Altogether, we have seen that multilateral export controls have some features that make them different from other international institutions. The need for coordination can arise in a static game environment. The reason for this is the irreversibility of the decision over exporting some of the most dangerous technologies. A decrease in CO2 emissions can always ameliorate the pollution problems and eliminating barriers to trade can improve welfare, but once a (potentially) massive destruction technology has been transferred to an aggressive country, the exporter countries will have to bear the security costs in perpetuity.

Summary and conclusions This chapter has analyzed military-related export control regimes using a game theoretic analysis of the interaction between countries setting export controls under different market structures. We have explained that the military-related export ﬁrms are involved in oligopolistic competition, but they still have strong links with their governments, a fact that encourages the existence of captive markets. Also, high R&D investments have quickly increased the quality level that make these products more dangerous and the deﬁnition of dual use more difﬁcult. In this context, exporter countries face a trade-off between security and proﬁts. After discussion

52

María del Carmen García-Alonso and Keith Hartley

of different methods of aggregating the effects of exports on producer countries, we have concluded that exported quality provides a good index of the exporters’ perception of security. In a discrete choice environment, the exports control game was shown to be a coordination game in which it cannot be predicted with certainly whether or not the Pareto superior restricting equilibrium is obtained. The existence of captive markets increases the chances of coordination, but an increase in the number of players or the existence of uncertainty over the degree of security concern of the other country decreases these chances seriously. This emphasizes the important role played by the interaction between competing exporter ﬁrms in the prediction of the viability of multilateral exports control. We have also identiﬁed the properties of the perception of security and exporter ﬁrms’ proﬁts that result in the export control game becoming a coordination game in a continuous choice environment. Finally, we have studied the effect that high-technology exporters’ restrictions can have on low-technology producers using results from the vertical differentiation literature. The analysis was then supported with different examples of actual international agreements. Throughout the chapter we have focused on multilateral export controls in which all countries have a concern for security. However, there are other situations in which a security concerned country can enter into a bargaining process with a security unconcerned country in order to persuade it through some compensation not to export a speciﬁc security sensitive technology (Jehiel et al. 1996 and Rotillon et al. 1996) or try to implement a penalty system (Smith and Udis 1998).

Acknowledgments This chapter has been written in the context of the participation of the author in a group working on arms trade, ﬁnancially supported by the ESRC under the grant R000235685. We would like to thank Javier Coto, Alan Carruth, Andy Dickerson, Todd Sandler and Ron Smith for helpful comments.

Notes 1 For an exhaustive analysis of these changes, especially in Europe, see Mollas-Gallart and Robinson (1998). 2 See, e.g. García-Alonso (2000) and Levine and Smith (1997) for an analysis of the strategic interactions between domestic procurement and arms exports decisions. 3 For a full explanation of the CES production function, see Varian (1992). 4 For instance, we could consider Si = −αmax qmax − j =max αj qj (where α are some, quality adjusted weights). 5 HPC are those of speeds above 2,000 theoretical operations per second (MTOPS). The most recent regulations on the exports of HPC have been implemented in February 1998 (Federal Register, Vol. 63, No. 22, Tuesday, February 3, 1998, Rules and Regulations). 6 If only one of the two countries cares about security the game would clearly have a unique Nash equilibrium in pure strategies in which none of the governments restricts. However, countries could still engage in a bargaining process that could lead them to a Pareto

Export controls, market structure

53

superior equilibrium. Such a possibility has been analyzed in Jehiel et al. (1996) and Rotillon et al. (1996) for nuclear technology exports and global warming, respectively. 7 The issue of multiplicity of equilibria and coordination failures has also been analyzed in the macroeconomics literature. See Cooper and John (1988) as an example. 8 This comes from the sufﬁcient conditions required for the existence of a Nash equilibrium: ∂π1 (q1 , q2 ) ∂S1 (q1 , q2 ) lim + θ 1 > 0, q→0 ∂q1 ∂q1 q q ∂π1 (q1 , q2 ) ∂S1 (q1 , q2 ) lim + θ1 < 0. q→∞ ∂q1 ∂q1 q q 9 With positive externalities the implication would be the opposite: see Cooper and John (1988). 10 The information we analyzed here is based on SIPRI Yearbook (1997, 1998) and the web pages they refer to. 11 See US Arms Control and Disarmament Agency, “Australia Group” (28 October 1997), URL .

References Aumann, R. (1990) Communication need not lead to Nash Equilibrium. Mimeo, Hebrew University of Jerusalem. Conybeare, J. A. C., Murdoch, J. and Sandler, T. (1994) Alternative collective-goods models of military-alliances: theory and empirics. Economic Inquiry 32, 525–542. Cooper, R. and John, A. (1988) Coordinating coordination failures in Keynesian models. The Quarterly Journal of Economics 103, 441–463. Cornes, R. and Sandler, T. (1996) The Theory of Externalities, Public Goods and Club Goods. Cambridge University Press, Cambridge. Dybvig, P. H. and Spatt, C. S. (1983) Adoption externalities as public goods. Journal of Public Economics 20, 231–247. Economides, N. (1989) Quality variations and maximal variety differentiation. Regional Science and Urban Economics 19, 21–29. Fundenberg, D. and Tirole, J. (1995) Game Theory. MIT Press, Cambridge, MA. García-Alonso, M. C. (1999) Price competition in a model of arms trade. Defence and Peace Economics 10(3), 273–303. García-Alonso, M. C. (2000) The role of technology security in a model of horizontal differentiation. International Journal of Industrial Economics 18(5), 747–773. García-Alonso, M. C. and Levine, P. (2002) Domestic procurement, subsidies and the arms trade. In: Brauer, J. and Dunne, P. (ed.), Arming the South: The Economics of Military Expenditure, Arms Production and Arms Trade in Developing Countries, New York, Palgrave. Harsanyi, J. and Selten, R. (1988) A General Theory of Equilibrium Selection in Games. MIT Press, Cambridge, MA. Hartley, K. (2000) The beneﬁts and costs of the UK arms trade. Defence and Peace Economics 11(5), 445–460. Hartley, K. and Sandler, T. (1995) In: Arrow, K. and Intriligator, M. (eds), The Handbook of Defence Economics. North-Holland, Amsterdam. Hartley, K. and Sandler, T. (1999) NATO burden-sharing: past and future. Journal of Peace Research 36(6), 665–680.

54

María del Carmen García-Alonso and Keith Hartley

Hirshleifer, J. (1983) From weakest-link to best-shot: the voluntary provision of public goods. Public Choice 41, 371–386. Jehiel, P., Moldovanu, B. and Stacchetti, E. (1996) How (not) to sell nuclear weapons. The American Economic Review 86, 814–829. Levine, P. and Smith, R. (1995) The arms trade and arms control. The Economic Journal 105, 471–484. Levine, P. and Smith, R. (1997) The arms trade, Economic Policy 12(25), 336–370. Levine, P., Sen, S. and Smith, R. (1994) A model of the international arms market. Defence and Peace Economics 5, 1–18. Maggi, G. (1999) The role of multilateral institutions in international trade cooperation. The American Economic Review 89, 190–214. Martin, S., Hartley, K. and Stafford, B. (1999) The economic impacts of restricting UK arms exports. International Journal of Social Economics 26 (5), 779–801. Mollas-Gallart, J. and Robinson, J. P. (1998) Assessment of Dual-use Technologies in the Context of European Security and Defence. Final Report for the Scientiﬁc and Technological Options Assessment (STOA), European Parliament. Moody-O’Grady, K. (1995) Nuclear waste dumping in the oceans: has the Cold War taught us anything? Natural Resources 35, 695–709. Panofsky, W. (1990) Barriers to negotiated arms control. In: Arrow et al. (eds), Barriers to Conﬂict Resolution. Norton. Rotillon, G., Tazdaït, T. and Zeghni, S. (1996) Bilateral or multilateral bargaining in the face of global environmental change? Ecological Economics 18, 177–187. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge. Sandler, T. and Hartley, K. (1999) The Political Economy of NATO. Cambridge University Press, Cambridge. Sandler, T. and Sargent, K. (1995) Management of transnational commons: coordination, publicness, and treaty formation. Land Economics 71, 145–162. Shaked, A. and Sutton, J. (1982) Relaxing price competition through product differentiation. Review of Economic Studies 49, 3–13. SIPRI (1997 and 1998) Yearbook: Armaments, Disarmament and International Security. Oxford University Press, Oxford. Smith, R. and Udis, B. (1998) New challenges to arms export control: whither Wassenaar? Discussion Papers in Economics 98–4. University of Colorado, Boulder. Varian, H. R. (1992) Microeconomic Analysis. Third edition. London: Norton.

5

Arms export controls and emerging domestic producers Paul Levine, Fotis Mouzakis and Ron Smith

Introduction In the conﬂicts with Croatia, Bosnia and Kosovo, Serbia had the advantage of a domestic capability to produce the type of arms most useful in such conﬂicts. This advantage was a product of two factors. Given the ambiguous position of former Yugoslavia between Soviet communism and Western capitalism the effective price, political and economic, of importing arms was high creating incentives for import-substituting industrialization. Second, for internal geo-political reasons those production facilities were mainly located in Serbia, rather than the other republics. In many other conﬂicts, the domestic arms production capabilities of the combatants plays a central role. This can be the ability to produce either conventional or unconventional weapons as with Iraq’s ability to produce chemical and biological weapons; India, Pakistan, Israel and White South Africa’s ability to produce nuclear weapons. The possibility of domestic production raises a difﬁculty for arms export control measures, since embargoes, by raising the effective price of imports, increase the incentive for domestic production. The Carter embargo on arms sales to a number of Latin American countries, adopted for human rights reasons, spurred many of them particularly Argentina and Brazil into developing their own arms industries. The empirical background to the international arms market and the policy issues are discussed in Levine and Smith (1997b) on which the stylized facts used in this chapter are based.1 The international institutions that regulate arms exports are discussed in Smith and Udis (1998). The literature on alliances, collaborative production and the arms trade are extensively discussed in Hartley and Sandler (1995) and Sandler and Hartley (1995). The model used here draws on the arms trade models of Levine et al. (1994), which uses a Cold War structure, and Levine and Smith (1995, 1997a), which have antagonistic recipients similar to those below but do not allow for military cooperation and collaborative production. The models in those papers are dynamic in that military stocks accumulate over time. Dynamics is important for issues of stability and credibility, an important focus of those papers, but is not central to our concerns. In all our earlier papers, we treated the division between buyers and producers as exogenous. This is a useful approximation for many questions, particularly about

56

Levine et al.

top of the line major weapons systems where the barriers to entry are massive. However, for other questions, the choice between importing and domestic production is central. This is particularly the case for light weapons, military vehicles and artillery. In this chapter, we provide a formal model of the domestic production decision. We generalize the model of Levine and Smith (1999) in particular to allow for the possibility that restrictions on the exports of arms by the main producers will drive up the price and encourage importing countries to set up their own arms production sector. The essence of the model is that there is a switch point at which countries become large enough to make it worthwhile to establish a domestic arms industry. This requires large ﬁxed costs so the average cost of realizing a given level of military capability with domestic production has to be below its marginal cost with only imported arms, to cover the initial investment. Most arms control regimes are designed to raise the marginal economic or political cost of importing thus increasing the incentives for domestic production. While we will work with a single price, it should be noted that it will contain economic and political elements. For instance, the marginal cost to Pakistan of taking delivery of the F16s it had already paid for would have been giving up its nuclear weapons programme, something it was unwilling to do. Domestic and imported arms are not perfect substitutes, they have different characteristics in combat and we allow for this. In addition, demand for both is a derived demand originating from the countries’ strategic interaction with its neighbours. We model this demand as deriving from an arms race in which both sides base their weapons acquisition decisions on a hypothetical conﬂict subject to a budget constraint. This model then allows us to determine the conditions under which a combatant switches from being an importer to a producer. The rest of the chapter is organized as follows. The next section sets out the model of buyers of high-tech major weapons systems who can also emerge as domestic producers of low-tech arms. Of the following two sections, the ﬁrst models the suppliers of high-tech arms and the second brings the demand and supply sides together in a partial equilibrium model and compares three arms trade regimes: laissez-faire (LF) trade, the uncoordinated regulation of exports and a producer cartel consisting of coordinated regulation. The ﬁnal section concludes the chapter.

Buyers of high-technology arms The world is divided up into three groups: a small number of suppliers of hightech weapons (group s) and a large number of buyers engaged in regional conﬂicts (group b). The latter group in turn consists of non-producers of arms, who are therefore completely reliant on imports for their defence needs, and producers of low-tech weapons which can be used for domestic defence and exported (Figure 5.1). The two types of military goods high-tech goods produced by the s group and low-tech goods by the b group are inputs into a military capability production

Arms export controls and domestic producers

57

s-group

b – group (producers)

b – group (non-producers)

Figure 5.1 The international market for arms.

function together with labour. Labour and arms combine in ﬁxed proportions in produce military capability and by a suitable choice of units we assume that one unit of low- or high-tech military goods combines with one unit of labour. Low-tech goods are also produced in the rest of the world in a competitive sector regarded as exogenous in our model. The military capability production function has three properties: ﬁrst, constant returns to scale; second, the marginal product of the hightech input tends to inﬁnity as the input tends to zero (so some input of the high-tech good is essential for the production of military capability) but the marginal product of the low-tech input remains ﬁnite as that input tends to zero. The latter has the implication that this input is not essential for military capability. The third property concerns the increase in military capability with respect to an increase in a unit of imported high-tech good, or a unit of domestically produced low-tech good, or a unit of imported low-tech good. We require these three ‘marginal products’ to be ranked in precisely this order. In other words the ranking of effectiveness of military goods is: high-tech, domestically produced low-tech and imported low-tech. Let ds and db be government procurement of domestically produced arms in group s and b, respectively. Let ms and mb be government procurement of imports of arms from the s and b group, respectively. Then military capability functions with these three properties are: s-group:

γ

Ks = ds 1 (ds + mb )(1−γ1 ) γ ms 2 (ms

+ db )

(5.1)

b-group with domestic production:

Kb =

b-group without domestic production:

Kb = ms 3 (ms + mb )(1−γ3 ) , (5.3)

(1−γ2 )

(5.2)

γ

where γi ∈ [0, 1]; i = 1, 3. The ﬁrst two properties, constant returns to scale and the inﬁnite marginal product of the high-tech good, but ﬁnite marginal product of the low-tech good, are clearly satisﬁed. The third property, the ranking of the

58

Levine et al.

marginal products, is also satisﬁed if ∂Ks ∂Ks > ∂ms ∂ds

and

∂Kb ∂Kb ∂Kb > > . ∂ms ∂ds ∂mb

Differentiating Eq. (5.1) we have that 1 ∂Ks 1 − γ1 1 ∂Ks = γ1 + = γ1 + , Ks ∂ms ms + d s Ks ∂ds 1 ∂Kb 1 ∂Kb = γ2 + , Kb ∂ms Kb ∂db

1 ∂Kb 1 ∂Kb − = γ3 − γ 2 . Kb ∂ds Kb ∂mb

Hence, the third property holds if γ1 > 0 and γ3 > γ2 > 0. The rest of this section concentrates on the b-group. We assume nb identical pairs of buyers each engaged in an antagonistic relationship with its neighbour. The total number of buyers 2nb ns so that we can assume that the buyers of the high-tech good from the s-countries are price-takers who ignore the strategic effect of their purchasing decision and production decisions on the price of that good. Each unit of arms combines with one unit of labour at a price wb (i.e. there is no substitutability between personnel and weapons). The national resource (balanced trade) constraint and government military spending for a typical b-country with a domestic arms industry and with national income Yb is given by Yb = Cb + Gb ,

(5.4)

where Cb is consumption and military expenditure is given by Gb = (wb + P )ms + (wb + pb )db .

(5.5)

In Eq. (5.5) P is the price of the imported high-tech good, pb is the procurement price of the domestically produced good and the price of an internationally traded consumption good is normalized at unity. In a partial equilibrium setting, the military sector is assumed to be small relative to the economy as a whole. Then the government decisions regarding procurement and exports do not have a signiﬁcant effect on factor prices and national income. Thus, Yb and wb may be regarded as exogenous and the use of a cost function C(yb ) in Eq. (5.6), which ignores possible changes to factor prices, is justiﬁed. The domestic procurement price pb for domestically produced arms, db , must be sufﬁcient to ensure the participation constraint: b = pb db + Pb xb − C(yb ) ≥ 0,

(5.6)

where yb = db + xb is total output, xb are exports at the world price for lowtech arms given by Pb , and C(yb ) is the cost function. Assuming constant returns in production, we put C(yb ) = Fb + cb yb where Fb are the set-up ﬁxed costs expressed as costs per period. For an optimizing government seeking to minimize

Arms export controls and domestic producers

59

the cost of providing a given military capability, Eq. (5.6) must bind. Then Eq. (5.5) becomes Gb = (wb + P )ms + (wb + cb )db + Fb − (Pb − cb )xb .

(5.7)

The government’s objectives for each buyer country are deﬁned by the following Cobb–Douglas utility function of consumption (Cb ) and security (Sb ) for the ﬁrst country: Ub = Cbωb Sb1−ωb ;

ωb ∈ (0, 1)

(5.8)

and an analogous function of Cb∗ and Sb∗ for its rival. An important aspect of the set-up is the choice of security function, Sb in Eq. (5.8). We follow Levine and Smith (1995, 1997a, 1999) and assume that each b-country’s security function is an increasing linear function of its own military capability (equal to imported arms by choice of units) and a linear decreasing function of its adversary’s capability. Thus, for the ﬁrst of our typical pair we can write: Sb = αb Yb + βb Kb − Kb∗ .

(5.9)

In Eq. (5.9), the ﬁrst term represents the ﬁxed beneﬁts of defence and βb the variable beneﬁts of defence or the ‘Lanchester coefﬁcient’ measuring how many units of military capability are required to defeat one unit of your opponent’s. We assume αb > 0 and βb ≥ 1 and ruling out ﬁxed or variable beneﬁts from attack.2 Consider the case of symmetric buyers. The b group is large in number and each country in the group can therefore be assumed to be price-takers with respect to the price P of imported high-tech goods. We assume that the market for low-tech goods is competitive so that the price is driven down to the marginal cost. Hence, in Eq. (5.7) Pb = cb and the last term disappears. From Eq. (5.6), the procurement price is given by pb = cb + Fb /db and the government in effect ﬁnances the set-up costs through the procurement process. Consider a typical pair of rivals choosing military capability Kb and Kb∗ . Given price-taking, the ﬁrst government’s optimization problem decomposes as follows: 1 2

Minimize with respect to inputs ms , mb , or ms , db and Fb the cost Gb of producing a given military capability Kb given prices P , wb and Pb = cb . Maximize Eq. (5.8) with respect to military capability Kb given Kb∗ (both produced at minimum cost).

This then gives rise to reaction functions for each country and a Nash equilibrium at their intersection. At this Nash equilibrium both countries are choosing their optimal cost-minimizing level of military capability given the military capability of the other country. We now carry out stages (1) and (2) of this problem. In stage (1) ﬁrst assume that the choice of whether to be a domestic producer of low-tech arms has been made. Then for a typical pair of buyers in the b-group with

60

Levine et al.

domestic production, function (5.2) applies and the optimization problem (1) is then: minimize Gb = (wb + P )ms + (wb + cb )db + Fb with respect to ms , db γ given Kb = ms 2 (ms + db )1−γ2 . The result is

1 − γ2 db = γ2

wb + P − 1 − 1 ms = φ2 (P )ms wb + c b

(5.10)

say and ms =

Kb . [1 + φ2 (P )]1−γ2

(5.11)

Hence, given that there exists a military production sector, domestic production is worthwhile iff the parameter capturing the contribution of the mixed high- and low-tech good satisﬁes: γ2 ≤

P − cb . P + wb

(5.12)

Otherwise, we impose φ2 (P ) = 0. The minimum cost of delivering a particular military capability is then given by Gb = (wb + P )ms + (wb + cb )db + Fb (wb + cb )φ2 (P ) + wb + P Kb + Fb = 2 (P )Kb + Fb = [1 + φ2 (P )]1−γ2

(5.13)

say. For a b-group with without domestic production, Fb = 0 and production function (5.3) applies. The optimization problem is now to minimize Gb = (wb + γ P )ms + (wb + cb )mb with respect to ms , mb given Kb = ms 3 (ms + mb )1−γ3 . The result is

1 − γ3 mb = γ3

wb + P − 1 − 1 ms = φ3 (P )ms wb + c b

(5.14)

say, and ms =

Kb . [1 + φ3 (P )]1−γ3

(5.15)

Hence, b-group import low-tech goods if and only if the parameter capturing the contribution of the mixed high- and low-tech good satisﬁes: γ3 ≤

P − cb . P + wb

(5.16)

Arms export controls and domestic producers

61

Otherwise, we impose φ3 (P ) = 0. The minimum cost of delivering a particular military capability is then given by Gb = (wb + P )ms + (wb + cb )db (wb + cb )φ3 (P ) + wb + P = Kb = 3 (P )Kb [1 + φ3 (P )]1−γ3

(5.17)

say. From the fact that γ3 > γ2 , it is clear that φ3 < φ2 . Furthermore, Appendix A shows that 3 > 2 . Choice of domestic production by b-group b-Group countries will choose to be producers at a threshold military capability K = Kˆ b at which: ˆ 2 (P )Kˆ b + Fb = 3 (P )K;

i.e. at Kˆ b = Fb /[3 (P ) − 2 (P )].

(5.18)

Then for Kb > Kˆ b the average cost of realizing a given level of military capability with domestic production drops below its marginal cost with only imported arms. Figure 5.2 illustrates these results by plotting the minimum cost Gb (Kb ) of providing a given military capability Kb given the price of high-tech imports P . For Kb < Kˆ b , the country is a non-producer importing high-tech arms from the s-group and low-tech arms from other b-group countries. For Kb > Kˆ b it becomes optimal for the country to invest in a domestic arms production sector. Since 3 > 2 , for Kb > Kˆ b where a domestic production sector now exists, the marginal cost of producing an extra unit of military capability is less than the case without the domestic sector. As the next sub-section shows, this has an important bearing on the inefﬁciency of the Nash equilibrium for the two countries.

Gb Slope Θ2

Producers Fb

Slope Θ3

Nonproducers Kb

Kb

Figure 5.2 The threshold value of military capability.

62

Levine et al.

The Nash equilibrium in military capability Now turn to stage (2) of the optimization problem of the b-countries: the maximization of Eq. (5.8) with respect to military capability Kb and given Kb∗ . Consider ﬁrst b-countries who are domestic producers. The ﬁrst-order condition for the ﬁrst country is, −ωb 2 (P ) (1 − ωb )βb + = 0, Yb − 2 (P )Kb − Db αb Yb + βb Kb − Kb∗

(5.19)

from which the reaction function of the ﬁrst country is obtained as Kb =

ωb Kb∗ (1 − ωb )(Yb − Fb ) αb ωb Yb − + . 2 (P ) βb βb

(5.20)

Then, in an symmetric Nash equilibrium we have Kb = Kb∗ =

[(1 − ωb )βb − ωb 2 (P )αb ]Yb − (1 − ωb )βb Fb . 2 (P )(βb − ωb )

(5.21)

Similarly, for b-countries with no domestic military production the reaction function corresponding to Eq. (5.20) is given by Kb =

ωb Kb∗ αb ωb Yb (1 − ωb )Yb − + βb 3 (P ) βb

(5.22)

leading to a symmetric Nash equilibrium: Kb = Kb∗ =

[(1 − ωb )βb − ωb 3 (P )αb ]Yb . 3 (P )(βb − ωb )

(5.23)

Hence given the price P , there exists a threshold size of county Yˆb (P ): Yˆb (P ) =

3 (P )(βb − ωb )Kˆ b , [(1 − ωb )βb − ωb 3 (P )αb ]

(5.24)

where Kˆ b is given by Eq. (5.18), for which if Yb > Yˆb (P ) then the b-countries produce. Figure 5.3 shows this shift in the Nash equilibrium. AB given by K ∗ = αb Yb + βb K is an attack function which expresses the stock of imports that the second country needs to successfully attack its neighbour. If security Sb = αb Yb + βb K − K ∗ > 0, then K ∗ is below this level and the ﬁrst country is secure. Similarly CD given by K = αb Yb + βb K ∗ is the attack function for the ﬁrst country and Sb∗ = αb Yb + βb K ∗ − K > 0 ensures security from attack for the second country. Together, the bankruptcy boundary and the attack functions deﬁne a feasible mutual security (FMSR) which is shown as the shaded regions in Figure 5.3 for the case considered in this chapter, of countries of identical size and structure. [The case of asymmetric countries is considered in Levine and Smith (1997a).] AB and CD

Arms export controls and domestic producers

63

K∗ B (K ∗)max N⬘

A

FMSR

E⬘

N

E D

F

F⬘ C

K max

K

Figure 5.3 The Nash equilibrium in military capability.

are the attack functions for countries 2 and 1, respectively. The degree of military security in their relationship is described by the size of the FMSR. If the ﬁxed cost of attack, αb , increases then AB moves up and CD moves to the right increasing the size of the FMSR. If the Lanchester coefﬁcient, βb , the combat exchange rate, increases thus moving in favour of the defender, AB swivels in an anti-clockwise direction and CD in the opposite direction and the FMSR increases. Given the other country’s military capability, each country prefers to be as far into the interior of the FMSR as possible. The Cournot–Nash equilibrium without domestic production as given by Eq. (5.23) is shown at intersection of the reaction functions EN and FN at point P. With domestic production the reaction functions (5.20) and (5.22) shift outwards to E N and F N resulting in a new Nash equilibrium with a higher military capability if (Yb − Fb )/2 > Yb /3 ; that is, if Yb > 3 Fb /(3 − 2 ) = 3 Kˆ b using Eq. (5.18). This will hold if military expenditure without domestic military production Gb = 3 Kb at the threshold Kb = Kˆ b lies within the bankruptcy boundaries of the two countries. Investing in military capability produces a negative externality, since one country’s security is its rival’s insecurity. The Nash equilibrium therefore results in an inefﬁcient allocation of resources between military expenditure and consumption. A cooperative arms control agreement between the countries would eliminate this inefﬁciency, but we rule out this possibility in this chapter. A switch to domestic production brings about a switch to a Nash equilibrium with higher levels of military capability and military expenditure and therefore involving greater inefﬁciency. Thus, we have the following proposition. Proposition 1 For two identical buyers in an antagonistic relationship of sufﬁcient size so that Kˆ b lies within their bankruptcy boundaries, a switch into domestic military production is possible and causes switch to a Nash equilibrium with higher levels of military capability, military expenditure and inefﬁciency.

64

Levine et al.

From this analysis, given the price P of the high-tech arms import from the s-country, we obtain two crucial switch or threshold points, (P − cb )/(p + wb ) = γ ∗ (P ) say, and Yˆb (P ) that determine the pattern of domestic production and trade in the b-countries. If γ ∗ ≤ γ2 then b-countries never produce even if they have an established domestic arms industry. Since γ ∗ (P ) < 0 this scenario occurs if the price P is sufﬁciently low and/or the parameter γ2 capturing the effectiveness of the high-tech good in contributing to military capability is sufﬁciently high. Formally, in this case φ2 (P ) = φ3 (P ) = 0 and 3 (P ) = 2 (P ) = wb + P and the threshold size Yˆb (P ) becomes inﬁnite. Next consider γ2 < γ ∗ < γ3 . Then if Y b ≤ Yˆb < Yb ≤ Y b the b-country of that size will establish a domestic industry and the smaller b-countries of size Yb < Yˆb will choose only to import from the s-countries. Starting from this case, if we let the price P fall sufﬁciently until γ2 < γ3 < γ ∗ then the non-producing b-countries will then choose to substitute some low-tech for the high-tech import. When the threshold size is such that Yˆb ≤ Y b , all b-countries become producers. Then since imported low-tech arms contribute less to military capability than domestically produced low-tech arms, but have the same marginal cost, b-countries cease to import them. Table 5.1 summarizes these results. The arms control regimes studied in the next section affect the decisions of the b-countries through an increases in the price P . But the effectiveness of the regimes to actually reduce the build-up of arms in the region depends on the incentive to substitute imports for domestic production, albeit of inferior weapons, and this incentive grows as the price P increases. To examine to effect on aggregate military capability we need to examine the changes with respect to price in the thresholds ˆ ) and Yˆ (P ) separately. Differentiating Eq. (5.18), we have K(P Fb [3 (P ) − 2 (P )] dKˆ b =− . dP [3 (P ) − 2 (P )]2

(5.25)

From Appendix A we have the results that i /i = γi , i = 2, 3. Hence Fb [3 (P )γ3 − 2 (P )γ2 ] dKˆ b =− . dP [3 (P ) − 2 (P )]2

(5.26)

Table 5.1 The emergence of b-producers as price P increases

γ2 ≤ γ ∗ < γ3 (φ2 ≥ 0, φ3 = 0) γ2 < γ3 ≤ γ ∗ (φ2 ≥ 0, φ3 ≥ 0)

Yˆb ≤ Y b

Y b < Yˆb < Y b

Y b ≤ Yˆb

All b-producers No b-importers All b-producers No b-importers

Some b-producers No b-importers Some b-producers All b-importers

No b-producers No b-importers No b-producers All b-importers

Note γ ∗ = (P −cb )/(P +wb ). b-Importers are b-countries who import low-tech arms. All b-countries are s-importers.

Arms export controls and domestic producers

65

Since 3 > 2 and γ3 > γ2 , it follows from Eq. (5.26) that dKˆ b /dP < 0; that is, the threshold level of military capability at which b-countries switch to domestic military production decreases with the price P of the high-tech good. Now consider changes in the threshold Yˆb (P ). Differentiating Eq. (5.24), we ﬁnd that dYˆb (P ) 3 (βb − ωb ) (1 − ωb )(1 − ωb )βb γ3 3 Kˆ b dKˆ b = + . dP (1 − ωb )βb − ωb 3 αb dP [(1 − ωb )βb − ωb αb 3 ]2 (5.27) From Eq. (5.26), the ﬁrst term in Eq. (5.27) is negative, but the second term is positive. The effect of price on the threshold size of country is therefore ambiguous. An increase in price brings about reduction in the military capability threshold at which it beneﬁts countries to switch to military production for a given military capability (the ﬁrst term), but at the same time encourages substitution of consumption for security which reduces that level of military capability (the second term). If the substitution effect is sufﬁciently small then the threshold size Yˆb (P ) decreases with price. The decrease in Kˆ b (P ) and the possible decrease in Yˆb (P ) as price increases implies that aggregate military capability in the region increases. These results are summarized as follows. Proposition 2 As the price P of high-tech arms imports increases then the threshold military capability Kˆ b (P ) decreases. If the substitution of consumption for security is sufﬁciently weak, then threshold size Yˆb (P ) also decreases, more b-countries become producers of low-tech arms and aggregate military capability increases. Can the effect highlighted in this proposition be mitigated by other changes that reduce the incentive to set up a domestic industry? To answer this question, we need to examine ways in which the threshold size can be increased. Examining Eq. (5.23) one obvious way is to increase the cost of a partial technology transfer by increasing the set-up costs Fb , for example, by strengthening IPR or imposing controls on investment goods involved. A less obvious route is to improve the sophistication of the high-tech good making substitution by the low-tech good less effective. This happens when the parameter γ2 reﬂecting the contribution of the high-tech import to military capability in Eqs (5.2) and (5.3) increases. These results we summarize as follows. Proposition 3 As set-up costs Fb and/or the contribution of high-tech arms imports to military capability increases (i.e. as γ2 and γ3 > γ2 increase) then the thresholds Kˆ b (P ) and Yˆb (P ) increase and aggregate military capability in the b-countries falls for a given price P.

Suppliers of high-tech arms On the supply side, there are ns identical countries. We assume that 1 − γ1 is sufﬁciently small to ensure that the s-group of countries only use their domestically

66

Levine et al.

produced weapons and that there is one regulated producer per supplier country (in fact, the market structure in the major supplier countries is often characterized by one national champion for each major weapons system). Let P be the price of the military good in the international market and psi be the procurement price paid by domestic government i.3 In each period producer i supplies dsi goods to its own government and exports xsi . Since there is self-sufﬁciency in weapons for producer countries, dsi will constitute the total domestic procurement of country i. Unlike the large number of b-countries, the small number of s-countries are not price-takers and set price strategically. Let ysi = dsi +xsi be total output of producer i and assume an identical constant returns to scale cost function C(ysi ) = Fs + cs ysi for each producer. In this cost function Fs represents the cost per period of ﬁnancing the initial R&D and physical investment. Proﬁts per period are then given by si = psi dsi + P xsi − C(ysi ) = πsi (psi , P ) − Fs = πsi (P , P ) − (psi − P )gi − Fs ,

(5.28)

where πsi (psi , P ) denotes operating proﬁts per period at procurement and market prices psi and P , respectively. The government then pays the minimum procurement price that will satisfy the participation constraint si ≥ 0 for its sole producer. From Eq. (5.28) this implies: psi =

C(ysi ) − P xsi Fs − πsi (P , P ) =P + . dsi dsi

(5.29)

Thus, if operating proﬁts at procurement price psi = P are less than the singleperiod equivalent of ﬁxed costs, the domestic government must subsidize its producer. On the other hand, if the world market power and scale of production is such that operating proﬁts are greater than ﬁxed costs, then the procurement price is less than the market price. The government then, in effect, taxes some of the producer’s oligopolistic proﬁts. Now consider the government’s optimization problem. As for s-countries, each unit of arms combines with one unit of labour. Then assuming balanced trade in consumption and military goods altogether, the supplier government i faces a national resource constraint: Ysi = Csi + Gsi = Csi + (wsi + psi )dsi ,

(5.30)

where Ysi is national income, Csi is consumption and Gsi is military expenditure. In Eq. (5.30), the price of the consumption good is normalized at unity and, as for b-countries, the military sector is assumed to be small relative to the economy as a whole. Then Ysi and wsi may be regarded as exogenous and the use of a cost function C(ys ) in Eq. (5.28), which ignores possible changes to factor prices, is justiﬁed.

Arms export controls and domestic producers

67

The government’s objectives in the supplier country are deﬁned by a utility function that depends on consumption (Csi ) and security (Ssi ). We assume a Cobb– Douglas functional form: Usi = Csiωs Ssi1−ωs ;

ωs ∈ (0, 1).

(5.31)

As before, an important aspect of the set-up is the choice of producers’ security function, Ss in Eq. (5.31). We adopt the following formulation for producer i: µ

Ssi = (αs Ys + βs dsi )1−µ Sri ;

µ ∈ [0, 1).

(5.32)

The ﬁrst term in Eq. (5.32) is the domestic component of security in a world where there is no obvious source of immediate threat as was the case in the Cold War. Countries merely seek domestic security by building up domestic arms as an insurance in a environment of uncertainty. The domestic component of security consists of two terms: a ﬁxed beneﬁt of defence, which is dependent upon economic size and a variable beneﬁt dependent upon military expenditure. The parameter βs can be interpreted as a ‘Lanchester coefﬁcient’ developed in the demand side of the model above. The second term is regional global security and this will depend on the nature of alliances in the event of a regional conﬂict. Consider the following general form of global security function: Ssi = f

ns

dsj , KbL (P ) + KbH (P )

f1 > 0; f2 < 0

(5.33a)

j =1

where KbL and KbH are the aggregate military capabilities of the smallest and largest b-countries (deﬁned in full in the next section), respectively. Equation (5.33a) says that regional security from the viewpoint of the suppliers depends positively on their total military capacity (as potential allies in the post-Cold War era) but each supplier is also concerned with the total build up of major weapons systems in each region. This may reﬂect the use of these arms for repressive purposes. It may also reﬂect the ability to intervene collectively in regional disputes.4 As one would expect, results are particularly sensitive to the speciﬁcation of these security functions. Levine et al. (1994) took a taxonomic approach, making the security of suppliers a quadratic function of the stock of arms sold to the recipient. This covered four cases: security increases with sales, typical of sales to allies; security falls with sales, typical of sales to adversaries; security increases then falls as they acquire arms beyond their legitimate security needs and start to become potential threats in the regions; security falls then increases, as by selling them arms you link initially hostile countries into your international security regime. These assumptions are appropriate for the Cold War context whilst the model of this chapter is more applicable to post-Cold War scenarios like the Gulf or former Yugoslavia, or the need to evacuate civilians from a war zone.

68

Levine et al.

The speciﬁc formulation used is linear in arguments and assumes that all s-countries together intervene in regional conﬂicts: Sri = dsi + d˜si −

rβ [KbL (P ) + KbH (P )], nb

(5.33b)

where ns and 2nb are the numbers of producers and buyers, respectively, and we have written total procurement = dsi + d˜si . Equation (5.33b) then says that regional security of each supplier depends on the total military capability of supplier countries to intervene in r number of regional conﬂicts at any one time. Sr > 0 is necessary to intervene effectively and βr has the same interpretation as βs , for sellers.5 In a non-cooperative game each producer country chooses dsi and xsi taking d˜si and x˜si (deﬁned as for d˜si ) as given. There is then a free-rider problem that leads to gi , being too low and a beggar-thy-neighbour problem of xsi being too high, in the absence of cooperation. Moreover, when subsidies are present this is exacerbated because a low dsi puts pressure on the government to allow more exports. Below we explore these interconnections and study the consequent beneﬁts of a cooperative arms trade regime in the form of a producer cartel. Now consider the supplier governments’ optimization problem. Substituting for the procurement price from the participation constraint (5.29) and for consumption from the national resource constraint (5.30), the governments’ utility function can be expressed in terms of procurement dsi and exports xsi . For the case of regulated trade, these are the s-governments’ decision variables. For the case of unregulated trade the ﬁrm chooses exports given government procurement and price. Regimes We now develop the details of the three regimes: LF, non-cooperative regulation (NCR), and a producer cartel. It is convenient to consider ﬁrst the case of noncooperative regulation. Non-cooperative regulation In a Nash game, total world government procurement is dsi + d˜si and each country maximizes its utility taking d˜si , the combined procurement of all other countries, as given. Similarly total world exports is X = xsi + x˜si and each country maximizes its utility taking x˜si , the combined exports of all other countries, as given. Substituting for Gsi and Ssi from Eqs (5.30) and (5.33), respectively, we ﬁrst rewrite the utility function (5.31) in the form: log Usi = ωs log(Ysi − (ws + psi )dsi ) + (1 − ωs ) × [(1 − µ) log[αs Ys + βs dsi ] + µ log Sri ].

(5.31 )

Arms export controls and domestic producers

69

Then the ﬁrst-order conditions are given by ∂Usi ∂psi ωs (1 − ωs )µ ∂Sri =− dsi + = 0; ∂xsi Ysi − (wsi + psi )dsi ∂xsi Sri ∂xsi

(5.34a)

∂Usi ∂psi −ωs ws + psi + dsi = ∂dsi Ysi − (ws + psi )dsi ∂dsi (1 − µ)βs µ ∂Sri + + (1 − ωs ) = 0; αs Ys + βs dsi Sri ∂dsi

i = 1, ns

i = 1, ns

(5.34b)

where from Eqs (5.29), (5.33b) and the Nash assumptions we have: cs − P − xsi (∂P /∂xsi ) ∂P 1 ∂pi ; = ; = P (X) = ∂xsi dsi ∂xsi X (P )

i = 1, ns (5.34c)

and ∂Sri rβ [KbL (P ) + KbH (P )] ∂Sri =− = 1; ∂xi nb X (P ) ∂dsi

i = 1, ns .

(5.34d)

Since all countries are identical the equilibrium for NCR and all other regimes has the property that all allies choose identical levels of procurement and exports, gi = ga , xi = xa say, and non-allies similarly choose identical levels dsi = ds , xsi = xs say. Laissez-faire Now go back to the ﬁrst arms trade regime, LF. For this case, the government ﬁrst chooses procurement price, psi according to Eq. (5.29) and procurement volume dsi . The ﬁrm then maximizes single-period proﬁts given by Eq. (5.28) given these decisions. This leads to the ﬁrst order condition cs − P − xsi

∂P = 0; ∂xsi

i = 1, ns

(5.34e)

which replaces Eq. (5.34a). The ﬁrst-order condition for procurement, Eq. (5.34b), is as before. We can now see from Eqs (5.34a) and (5.34e) that unregulated trade coincides with the LF regime µ = 0. Producer cartel In this regime, all ns producer countries enter into a cartel of jointly regulated trade (i.e. the cooperative determination of determination of xi , i = 1, ns ) which takes into account the regional security implication of their decision. Government

70

Levine et al.

procurement remains uncoordinated as in the ﬁrst two regimes. The ﬁrst-order conditions are as for unregulated trade except in Eq. (5.34e) we now have ∂P dX ns ; = P (X) = ∂xsi dxsi X (P )

i = 1, ns .

(5.34f)

To summarize, suppliers’ behaviour is described by three equations: a participation constraint (5.29), which gives the procurement price, and Eqs (5.34a) or (5.34e) for regulated and unregulated (LF) trade respectively, and (3.34b), which determine the decision variables, procurement volume and exports. To close the model we need to determine the world demand for exports function X(P ) to which we now turn in the next section.

Market equilibrium and comparison of regimes To determine the price P of high-tech arms, we equate world supply and world demand. To calculate the latter, suppose that the size of b-countries in identical pairs is uniformly distributed with support Yb ∈ [Y b , Y b ]. Then using Eqs (5.20) and (5.21) the demand by the lower and upper tails of b-countries for the exports of s-countries is given by

ˆ Yb (P ) [(1 − ω )β − α ω (P )]Y 2nb b b b b 3 b MsL (P ) = dYb [(wb + cb )φ3 (P ) + wb + P ] (βb − ωb )(Y b − Y b ) Y b (5.35a) MsH (P ) =

2nb (βb − ωb )(Y b − Y b )

¯ Yb [(1 − ω )β − α ω (P )]Y − (1 − ω )β F b b b b 2 b b b b dYb . × [(wb + cb )φ2 (P ) + wb + P ] Yˆ (P ) (5.35b)

To complete the model, we also need the total military capacity by b-countries: KbL (P ) + KbH (P ) = [1 + φ3 (P )]1−γ3 MsL (P ) + [1 + φ2 (P )]1−γ2 MsH (P ). (5.36) Then the world market equilibrium for high-tech arms is given by X(P ) = ns xs = MsL (P ) + MsH (P ),

(5.37)

from which (P ) + MsH (P ). X (P ) = MsL

This closes the model, which we summarize in Table 5.2.

(5.38)

Arms export controls and domestic producers

71

Table 5.2 Summary of model Participation constraint Fs − (P − cs )ys ps = P + ; ds

ys = ds + xs

Regulated exports ωs X(P ) cs − P − Ys − (ws + ps )ds ns X (P ) +

(P ) + KbH (P )] (1 − ωs )µ rβb [KbL = 0 ns ds − (r/nb )βb [KbL (P ) + KbH (P )] nb X (P )

Procurement µ −ωs (1 − µ)βs (ws +c)+(1−ωs ) + Ys − (ws + ps )ds αs Ys + βs ds ns ds − (r/nb )βb [KbL (P ) + KbH (P )] Demand and market equilibrium X(P ) = ns xs = MsL (P ) + MsH (P ) where MbL and MbH are given by Eq. (5.35). Threshold size Yˆb (P ) =

3 (P )(βb − ωb )Kˆ b [(1 − ωb )βb − ωb 3 (P )αb ]

Domestic production low-tech imports by b-countries Mb = φ3 (P )[MsL (P ) + MsH (P )] Db = φ2 (P )MsH (P ) if Yˆb ≤ Y b . Otherwise D b = 0. Domestic production by s-countries Ds (P ) = ns ds Notes x = exports; p = procurement price; P = world market price; d = domestic procurement; y = x + d = total output; C(y) = cy + F = cost function; ns = number of suppliers; 2nb = number of recipients; = 0; = 1 for LF regime, = = 1 for NCR regime and = 1, = ns for the producer cartel.

Comparison of regimes The full solution of the market equilibrium requires numerical simulations to which we now turn. We use the calibration adopted in Levine and Smith (1999) which assumed the b-group did not produce. (Full details are provided in Appendix B.) Table 5.3 compares the three arms control regimes: LF, NCR and a producer cartel. Two scenarios are presented. On the left-hand side of the table, ﬁxed costs Fb are set sufﬁciently high that the b-countries never produce. On the right-hand side, Fb is set at the calibrated value that sees b-countries producing. The crucial

72

Levine et al.

Table 5.3 Comparison of arms trade regimes Regime

P d˜s x˜s log Us m ˜s m ˜s ˜ db m ˜b kb kb log U b log U b Kb

High Fb

Low Fb

LF

NCR

Cartel

LF

NCR

Cartel

1.00 1.88 0.119 −1.652 0.827 0.827 0 1.681 4.987 3.324 −0.0689 −0.4743 0.997

1.27 1.85 0.098 −1.651 0.679 0.679 0 1.711 4.637 3.091 −0.0689 −0.4744 0.927

–

0.944 1.901 0.095 −1.652 0.425 0.882 2.209 1.660 6.132 3.395 −0.0721 −0.4743 1.120

0.883 1.906 0.112 −1.652 0.454 0.952 2.194 1.628 6.188 3.475 −0.0720 −0.4743 1.111

0.904 1.904 0.105 −1.652 0.443 0.926 2.200 1.641 6.167 3.446 −0.0720 −0.4743 1.115

0.65 0 −0.285 0 0 0 0 0 0 −0.0576 −0.4630 0

Notes d˜s = ps ds /Ys ; x˜s = P xs /Ys d˜ b = pb d¯b /Y b where pb = pb (Yb ); p¯ b = pb (Y b ) etc; d˜ b = p b d b /Y b ; m ˜ b = Pb mb /Y b where Pb = cb . All values are percentages except the welfare and aggregate military capability Kb = KbL + KbH .

feature provided by these simulations of the full equilibria is the endogenous determination of the price P , the transmission mechanism for the arms control regimes. If domestic production by b-producers is ruled out by very high ﬁxed costs, progressing from the LF regime to the NCR regime sees the price increasing in equilibrium, resulting in lower exports, lower military capability of b-countries and higher regional security of s-countries. Then proceeding to a producer cartel and still ruling out production by b-countries, supplier countries have the option of increasing the price still further and enjoying the beneﬁt of higher proﬁts, or implementing a complete embargo which will increase their regional security. For our parameter values supplier countries choose the latter and welfare of both supplier and buyer countries increases signiﬁcantly in this disarmament (for b-countries only) scenario. If domestic production by b-countries is allowed (low Fb in Table 5.3), the nature of these regimes change drastically. Independently of the regime LF, NCR or the cartel both buyers and sellers are worse off in the market equilibrium with domestic production by b-countries. Since an increase in P would have the effect of encouraging domestic production and raising military capability in b-countries, optimizing s-countries now design regimes in which the price is by far lower, and in fact in NCR or the cartel it is actually less than under LF. As in Levine and Smith (1999), the welfare of b-countries is extremely ﬂat, the welfare beneﬁt from a more efﬁcient Nash equilibrium under LF just about cancelling the loss from the income effect. s-Countries must beneﬁts from a cartel (by design) so the optimal regime for all parties turns out to be the producer cartel, but the gains are small if b-countries produce.

Arms export controls and domestic producers

73

Conclusions This chapter has presented a partial equilibrium model of the international market for arms which highlights the importance of incentives for countries to set up their own domestic arms sector when faced with restrictions on arms imports. In Levine and Smith (1999) where buyers could not be producers we found that an increase in the price of imported arms could actually beneﬁt buyers (i.e. increase their welfare). Simulations suggested the following Pareto-ranking of regimes: Cartel NCR LF where denotes a Pareto improvement. The mechanism through which this happens is the change in the price of high-tech arms. Without domestic production in b-countries, proceeding from regime LF to the cartel causes the world price to increase, thus discouraging the build-up of imported arms in those countries. In the more general model of this chapter which allows for an exogenous competitive world sector for low-tech arms, a cartel results in a complete arms embargo by s-countries and very signiﬁcant welfare beneﬁts for all parties. In this chapter we also allow for b-countries to enter domestic production as the price increases. This encourages higher levels of military expenditure and a less efﬁcient Nash equilibrium in b-countries. It also results in lower regional security on the part of s-countries. The latter is taken into account by s-countries in their optimal design of regimes, and the increase in price in going from regime LF to a producer cartel no longer occurs. The main result is that the possibility of domestic production by b-countries matters for arms control regimes. Irrespective of the arms control regime all parties – small sellers and large buyers – lose out when b-countries are producers. Measures aimed at reducing exports to securitysensitive areas of the world need to be accompanied by policies to discourage the emergence of domestic producers in these regions.6

Acknowledgements An earlier version of this chapter was presented at the conference: ‘The Arms Trade, Security and Conﬂict’ on 11–12 June 1999 at Middlesex University Business School. We are grateful to the participants for their comments. We are also grateful to the ESRC for ﬁnance under grant R000235685.

Appendix A Properties of the function i (P ), i = 2, 3 (i) 3 > 2 Deﬁne φ(γ ) =

(1 − γ ) wb + P − 1 − 1. γ wb + cb

(5.A1)

74

Levine et al.

Then φ3 = φ(γ3 ) and φ2 = φ(γ2 ). Similarly deﬁne (γ ) =

P − cb . γ [1 + φ(γ )]1−γ

(5.A2)

Since γ3 > γ2 then 3 > 2 iff (d/dγ ) > 0. This we can show by taking logarithms of Eq. (5.A2) and differentiating with respect to γ to obtain: 1 (1 − γ ) 1 d =− + + log[1 + φ(γ )]. dγ γ γ2

(5.A3)

Hence, in the region of interest for which φ(γ ) ≥ 0 we have that (d/dγ ) > 0 and hence 3 > 2 . (ii) 3 (P ) > 2 (P ) Taking logarithms Eq. (5.A2) and differentiating with respect to P we obtain 1 d(γ ) 1 (1 − γ ) = . − (γ ) dP P − cb wb + P

(5.A4)

Hence (d3 /dP ) = 3 γ3 > (d2 /dP ) = 2 γ2 using result (i).

Appendix B Calibration Our approach is to choose calibrated parameters so as to give plausible outcomes, broadly consistent with the stylized facts regarding military expenditure/GDP ratios and supplier exports/military expenditure, for the equilibrium solution of a baseline model. For the latter the market is dominated by four identical producers (e.g. the United States of America, the European union, Russia and China),7 there are twelve regional conﬂicts involving twenty-four identical pairs of recipients and regional security is given a very low priority in comparison with domestic security by supplier countries (in fact we assume that regional security is ignored). Other data used are the ratios of expenditure on personnel and equipment, a rough guestimate of the level of subsidies provided for the UK military production industries and the folk wisdom regarding conventional war that successful attackers require a military capability superiority of 3 : 1. These data along with the assumption of Cobb–Douglas preferences are sufﬁcient to reveal all the underlying parameters deﬁning preferences, security and technology. First, consider the parameter values: {βi , ωi , wi , Yi }, i = s, b; βr ; c, d and D; ns , nb , r and µ. Our approach is to choose these parameter values to give plausible outcomes for military expenditure/GDP ratios for a baseline model in which the market is dominated by ns = 4 identical producers, there are nb = 12 regional conﬂicts, regional security is ignored by the suppliers (i.e. µ = 0 in which

Arms export controls and domestic producers

75

case r is irrelevant), b-countries do not produce nor do they import low-tech arms. With these assumptions, putting µ = 0 we have the military expenditure ratios: Gs (ws + ps )ds (ws + p)[(1 − ωs )βs − ωs αs (ws + MR(P ))] = = Ys Ys βs [ws + ωs MR(P ) + (1 − ωs )ps ] (5.B1) Gb (wb + P )ms [(1 − ωb )βb − ωb (wb + P )αb ] = = Yb Yb (βb − ωb )

(5.B2)

In these two expressions, write the ratios of expenditure on personnel and equipment as Rs = ws Ls /ps ds = ws /ps and Rb = wb Lb /P ms = wb /P for suppliers and recipients, respectively. We normalize P = 1 for the baseline calibration. The folk wisdom is that attackers require a military capability superiority of 3 : 1; that is, β = 3. If we put βs = βb = βr = 3 and similarly assume that consumption– security preferences and the ﬁxed beneﬁts of defence are equal for suppliers and buyers, then using data for {Gi /Yi , Ri }, i = s, b we can solve Eqs (5.B1) and (5.B2) to reveal the underlying parameters, ωs = ωb and αs = αb . Table 5.B1 Summary of calibration Data and imposed parameters Gs /Y s Gb /Y b Rs Rb βb = βs d τ ns nb xs /Gs c = cb /cs f = Fb /Fs k = Y b /Y s γ1 γ2 γ3 r/nb

0.03 0.05 1 1 3 0 0.3 4 12 0.12 0.5 0.02 1.5 1 0.03 0.065 1/6

Derived parameters αb = αs ωb = ωs cs Fs /Ys = 2nb Ym /(ns Ys )

0.20 0.85 0.88 0.005 0.14

76

Levine et al. From the market clearing condition, we have that =

2nb Ym ms /Ym Gb /Yb (wb + P ) = = , ns Ys xs xs /Ys

(5.B3)

where Ym = (Y b + Y b )/2 is the average size of the b-countries and we have used Eq. (5.B2). The expression on the left-hand side of Eq. (5.B3) is the relative GDP of buyers to suppliers. This is chosen to be consistent with the military expenditure/GDP ratio chosen above and stylized facts regarding data on exports of arms. To calibrate the cost function C(ys ), we have from Eq. (5.34e) that cs = MP (P ) is obtained given the normalization P = 1. Guestimates of the subsidy τs = (ps − P )/P suggests τs ≈ 0.3 (Dunne and Smith 1996). Then using Eq. (5.A2) we have: Fs = [(1 + τs )P − cs ]ds + (P − cs )xs

(5.B4)

which determines Fs . The calibration applies to an oligopolistic market with four producers who ignore the regional security consequences of their procurement and export policies by setting µ = 0 in their security function (5.32). When regional security matters we require a numerical value for r/nb , the proportion of regional conﬂicts with which each supplier government expects to be involved. We put r/nb = 1/6 (the United States of America is said to prepare for intervention in up to two regional conﬂicts at any one time). The calibration is summarized in Table 5.B1.

Notes 1 See also Anderton (1995) for a recent survey. The scale and pattern of domestic production of arms in the developing world is discussed in Brzoska (1999). 2 But see Levine and Smith (1997a) for a discussion of the latter case. 3 We assume a common international price in this chapter. A model where different recipients pay different price depending on whether they are allies or potential adversaries is set out in Levine et al. (1994). 4 Notice that we do not distinguish between the military contribution of allies (e.g. NATO modelled as an alliance between equal producers, the United States of America and the European Union) and the rest (e.g. Russia and China). This distinction is made in Levine and Smith (1999). 5 Our formulation of this security function is similar to that used in the alliance literature, see Sandler (1993) and Sandler and Hartley (1995: chapter 2), for a recent treatment. The difference is that it has the additional feature that allies may help arm the potential enemy through weapons exports. As for βs , parameter βr can be interpreted as a Lanchester coefﬁcient. Alternative approaches are discussed in the concluding section. 6 Paradoxically, Levine and Smith (2000) show that uncertainty about the future supply of imported weapons or their prices can reduce the probability of proliferation. 7 In reality there are of course substantial differences between the four main suppliers – the United States of America, the European Union countries, Russia and China. We ignore these here and model a hypothetical market of four identical blocs.

Arms export controls and domestic producers

77

References Anderton, C. H. (1995) Economics of arms trade. In: Hartley, K. and Sandler, T. (eds), Handbook of Defense Economics, Vol. 1, North-Holland, Amsterdam. Brzoska, M. (1999) The scale and pattern of domestic production of arms in the developing world. Defence and Peace Economics 10, 139–170. Dunne, P. and Smith, R. P. (1996) The arms trade and employment. Mimeo, University of Middlesex. Hartley, K. and Sandler, T. (eds), (1995) Handbook of Defense Economics. North-Holland, Amsterdam. Levine, P., Sen, S., and Smith, R. P. (1994) A model of the international arms market. Defence and Peace Economics 5, 1–18. Levine, P. and Smith, R. P. (1995) The arms trade and arms control. Economic Journal 98, 471–484. Levine, P. and Smith, R. P. (1997a) The arms trade and the stability of regional arms races. Journal of Economic Dynamics and Control 21, 631–654. Levine, P. and Smith, R. P. (1997b) The arms trade. Economic Policy October, 336–370. Levine, P. and Smith, R. P. (1999) The arms trade game: from laissez-faire to a common defence policy. Oxford Economic Papers 52, 357–380. Levine, P. and Smith, R. P. (2000) Arms export controls and proliferation. Journal of Conﬂict Resolution 44, 885–895. Sandler, T. (1993) The economic theory of alliances. Journal of Conﬂict Resolution 37, 446–483. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge. Smith, R. and Udis, B. (1998) New challenges to arms export control: whither Wassenaar? Mimeo. University of Colarado at Boulder.

6

The supply-side implications of the arms trade UK aerospace industry, economic adjustment and the end of the Cold War Ian Jackson

Introduction Given the past and present trends, if national and Alliance security requirements are to be met at reasonable cost, the MoD can either procure weapons or maintain the DIB, but it cannot afford to do both. (Hartley et al. 1987: 72)

The international trade in defence equipment and military hardware is a highly complex market due to the procurement procedures and the longevity of the equipment life cycle (Gummett and Stein 1997). Domestic and global defence manufacturers create the supply of arms, whereas the demand from the armed forces is derived from national and alliance security requirements (Sandler and Hartley 1999). However, both supply and demand conditions are constrained by the size of national and international defence budgets, which are allocated by a state-controlled defence procurement agency.1 In order to make the security requirements operational, the defence procurement agency has to decide upon a budget allocation that is an efﬁcient use of resources. However, the lowest cost source of equipment (relative to the world price) may not guarantee long-term supply or quality and if imported from overseas could result in domestic job losses (Lovering 1995). Therefore, the decision to make weapons within a country or buy from international suppliers is crucial in terms of cost, quality and the reliability of suppliers. In the United Kingdom, this dilemma affects the Ministry of Defence (MoD), the domestic Defence Industrial Base (DIB) and Her Majesty’s Armed Forces as identiﬁed above by Hartley et al. (1987) and can be assessed in terms of a make-or-buy decision under the transaction costs paradigm (Williamson and Masten 1995). However, if there is a gap between the end of one defence equipment programme and the start of the next, then military exports can be used to supplement declining domestic demand and preserve defence employment. This forms the supply-side of the arms trade and is vital in periods of declining demand, as witnessed during the end of the Cold War (Hartley and Sandler 1995). The aim of this chapter is to focus on the UK aerospace industry as an illustration of the economic adjustment experienced by defence ﬁrms. Whilst some of

The supply-side implications of the arms trade

79

the change is unique to aerospace, a large proportion of the adjustment process is generic and relevant to other major areas of manufacturing, such as telecommunications, pharmaceuticals and food processing. The main areas of concern include the supply-side issues of rising unit costs and overcapacity, plus the demand-side issues of decreasing order size and limited defence budgets. In particular, there is support for the notion that excess capacity in the aerospace industry has forced defence companies to engage in a programme of exports, which has augmented the trade in military goods and services. The next section will provide an overview of UK aerospace industry. In the subsequent sections, we will discuss the economic adjustment, with a particular reference to transaction cost problems; the production and procurement dilemma with respect to the Euroﬁghter 2000 and Tornado production gap; and the international arms trade using a game theoretic approach. The ﬁnal section will present some conclusions and suggest further research strategies. Overall, operational issues of arms trade will be assessed using a transaction cost approach and public procurement policy will be analysed using game theory.

The UK aerospace industry The UK aerospace industry is a pivotal sector in the domestic DIB and can be deﬁned as those companies that manufacture components and provide services for the United Kingdom and global production of air and space vehicles. In essence, the industry generates complex engineering solutions in aircraft, missile and satellite production, via hundreds of ﬁrms, not all of who would necessarily consider aerospace activity as their core business. There have been many signiﬁcant technical advances that have been vital to the industry, namely, the jet engine, the rocket, radar, smart materials, as well as market changes, including the increased demand for civilian aircraft travel and variable demand for military aircraft. Overall, aerospace is an important sector, because it supplies ﬁxed-wing and rotary-wing aircraft to the three main services: ﬁghter, strike and reconnaissance aircraft to the air force, troop transport and ground support aircraft to the army and aircraft operational at sea to the navy (Postan et al. 1964). In more recent years, the UK aerospace industry has experienced a signiﬁcant amount of market uncertainty, as well as technical uncertainty. Defence markets in aerospace have witnessed massive changes since the end of the Cold War, culminating in greater pressure for defence production to consolidate through merger activity.2 Figure 6.1 shows the dramatic downturn in global defence spending, in constant prices, following the end of the Cold War period in the early 1990s. Whilst global defence spending is likely to stabilise, nevertheless, it has been forecast to remain at approximately 50 per cent of the 1985 value until 2010. This has important implications for defence equipment production, especially in aerospace, which has historically been a defence-dependent industry. The government, in addition to supplying export credit guarantees to potential defence exporters, funds a large proportion of military Research and Development (R&D). It is interesting to note the potential change in UK military aircraft holdings forecast until 2010 in Appendix A. The total number of UK active combat aircraft

80

Ian Jackson 1,400 1,200

Sbn

1,000 800 600 400 Forecast 200 0 1980

1985

1990

1995

2000

2005

2010

Figure 6.1 Global defence spend. Source: British Aerospace Annual Report and Accounts (1997).

is forecast to reach a peak of 509 in 2007, as Euroﬁghter 2000 comes on-stream and decreases to the year 2003 ﬁgure by 2010 (IDDS Almanac 1994). The annual average number over this period is 484 combat aircraft, which indicates a strong commitment to the previous Cold War status quo. This is because all the UK combat aircraft (including Euroﬁghter 2000) are a product of Cold War technology and Cold War doctrines. As a result, it is unlikely that these technologies will be easily surrendered as the justiﬁcation of NATO is based on the Cold War (Sandler and Hartley 1999). An important test of this philosophy will be the proposed Joint Strike Fighter (JSF), which will be one of the ﬁrst aircraft designed after the Cold War. This will also test the strength of the pan-European alliance in aerospace. The decline in global defence spending since 1989 (except during the 1991 Gulf War3 ) has reduced effective demand for military equipment, including aircraft, which translates into a highly competitive market for defence ﬁrms. The civilian aerospace markets continue to be a highly competitive environment for all incumbent ﬁrms.4 However, whilst base ﬂeet numbers and committed deliveries are forecast to decrease slightly, the potential for growth in the open market sales is forecast to increase enormously by 2016, as shown in Figure 6.2. On the supply-side, considerations of the industry offer an important insight into how a large-scale industry attempts to design and manufacture a highly complex product. Whilst there are a few very large aerospace ﬁrms, for example, British Aerospace plc (BAe), GKN-Westland Limited, Hunting plc and Rolls-Royce plc,5 there are literally hundreds of other smaller suppliers, who nevertheless remain important to the long-term proﬁtability of the entire industry. Indeed, Lamming (1993) showed this to be the case within the UK motor car industry. In particular, sub-contracting work to the smaller car component manufacturers had strengthened the position of the main car producers and not weakened the market position. Whilst

The supply-side implications of the arms trade

81

Number of aircraft

20,000 15,000

Open market

10,000

Committed deliveries

5,000 0

Base fleet 1997 2000

2004

2008

2012

2016

Figure 6.2 Estimated civil ﬂeet in service – all jets. Source: British Aerospace Annual Report and Accounts (1997).

the Society of British Aerospace Companies (SBAC) Equipment Sector Research Project (SBAC 1998) has initiated a mapping of the supply chain in aerospace, nevertheless, there is a clear parallel between aerospace and car production, which needs to be explored in much greater depth (Hartley and Hooper 1995). On the demand-side, the considerations for UK aerospace suppliers are equally important. For example, in the defence markets the prime contractor, who has the main contractual responsibility for a project, usually faces a bi-lateral monopoly. This is where monopoly and monopsony coincide in the same market arena, which leads to derived demand for the defence equipment from the armed forces. Even though there is often greater scope for competition at the sub-contractor level, negotiation between the MoD and the large defence ﬁrms remains an important feature of the industry together with direct competition. A more intense level of competition exists in civilian markets, but there are also many cooperations between airlines and civilian aircraft manufacturers in order to satisfy effective demand, such as leasing and maintenance agreements. Furthermore, both the range and scope of technology (as well as the rate of technical change in the industry) remain a signiﬁcant market issue. Civilian markets expect high safety standards and defence markets dictate high-performance levels, both of which can only be achieved through a considerable amount of investment in R&D (Rolls-Royce 1997). Finally, there are institutional factors that affect the aerospace market, including UK industrial policies, EU competition rules and the trading formulae of GATT.6 This will alter the proﬁtability of ﬁrms and their rate of survival, due to such disparate factors as privatisation, government subsidies and artiﬁcial barriers to entry and concentration ratios. Indeed, the extent to which horizontal and vertical merger activity in global aerospace markets is permitted (and also encouraged) by political institutions can ultimately determine success and failure in speciﬁc markets by individual ﬁrms. This is due, in part, to the strategic importance of the aerospace supply chain and the interaction between ﬁrms that operate within it. The aerospace industry forms an important part of the UK manufacturing base along with similar industries such as motor vehicles, shipbuilding and electronics. It is also dependent upon government contracts together with other industries

82

Ian Jackson

such as pharmaceuticals and medical equipment. According to SBAC (1998), the aerospace industry had an annual turnover of £15.02 billion in 1997, employing 121,000 people and contributing £3.1 billion to the balance of payments.7 As previously stated, there is a well-established aerospace supply chain, which includes companies that view aerospace as only one part of a wider business portfolio; for example, British Steel Engineering Steels (Pugh 1993). In sum, the size, complexity and the length of product life cycles in the aerospace industry allows it to be assessed in a number of strategic ways by recognising the heterogeneous nature of aerospace companies. An obvious bifurcation of the population of aerospace companies is between those suppliers who are primarily involved with defence contracts (GKN-Westland Limited) and others who mainly supply civilian markets (British Airways Engineering Limited). The next section will begin to consider these issues as the main area of analysis and attempt to develop the research in a transaction cost framework (Williamson 1975, 1985), before considering game theory applied to defence procurement.

Economic adjustment in the UK aerospace industry The UK aerospace industry is currently experiencing a period of profound economic change, due to a combination of market and technological uncertainties in both defence and civilian markets. This chapter will address the issues of economic adjustment experienced by UK aerospace ﬁrms. The main areas of concern include the supply-side issues of rising unit costs and overcapacity, plus the demand-side issues of decreasing order size and limited defence budgets. Figure 6.3 shows the supply-side problem of rising unit cost in the global military aerospace industry. In the United Kingdom, BAe claim that the comparative unit costs of front-line combat aircraft have declined relative to those in the United States of America.

Comparative unit cost

Price at entry into service of front-line combat aircraft F22

F15E F18 Eurofighter Tornado

Phantom Lightning

1950

1960

1970

1980

1990

2000

2010

Date of first production delivery

Figure 6.3 Rising unit cost of military aircraft. Source: British Aerospace Annual Report and Accounts (1997).

The supply-side implications of the arms trade

Unit cost

Number of aircraft ordered for national forces

Unit costs (constant prices)

Numbers

83

1950

1990 Year

Figure 6.4 Unit costs and numbers. Source: Sandler and Hartley (1995).

However, Euroﬁghter 2000 and Tornado remain more expensive than the previous generation of UK military aircraft, namely the Lightning and the Phantom, indicating the supply-side pressure of rising unit costs. It is unknown where the JSF will ﬁt into the picture (Hartley 1985). There have been at least two main effects on the capabilities of the aerospace industry, namely, decreasing order size8 (demand), in addition to rising unit costs (supply). Indeed, demand and supply are interdependent, if learning curves and economies of scale are to be reached. Figure 6.4 shows an indication of the effects of unit costs increasing over time, whilst order size is decreasing over time. Whilst this is not necessarily a straightforward linear function it does demonstrate that these two issues are interrelated. Whilst quality of output is important, there is also a minimum efﬁcient scale of production. According to British Aerospace: The trade-off between cost and capability is universally identiﬁed as a key differentiator in defence equipment procurement. (British Aerospace Annual Report and Accounts 1997: 7) This is a process of derived demand known as the Iron Triangle or the ‘Savoy Maﬁa’ (Lovering 1995) and illustrates a problem of contested terrain. The defence manufacturers who supply the armed forces attempt to maximise proﬁts and, in turn, the armed forces who supply security attempt to maximise capability, whereas the defence procurement agency attempt to minimise the costs of the entire operation. In terms of the Iron Triangle, the trade-off is a circular problem, between the domestic DIB, the armed forces and the MoD. The danger is that confusion and secrecy in the defence procurement process can lead to waste and fraud, in addition to other related ‘pork-barrel’ problems.9

84

Ian Jackson

An important feature of this work is horizontal and vertical integration and how it affects the economic, contractual and organisational arrangements of the ﬁrm and its operating size (Bittleston 1990). Overall, this is an issue relating to the boundaries of the ﬁrm and the make-or-buy10 strategy of defence procurement. This is the transaction cost approach, which analyses the make-or-buy issue on two levels. First, the extent to which a defence procurement agency acquires from domestic sources (make), as opposed from international markets (buy). Second, the extent to which the successful prime contractor manufactures the equipment in-house (make) as opposed to purchasing from sub-contractors and suppliers (buy). This has important implications for the arms trade, because it will establish the size of domestic production and net exports (or net imports). The process of procurement and adjustment to change will be examined in the next section on transaction costs (Dugger 1993).

The EF-2000 and Tornado production gap There are massive life cycle costs associated with military aircraft. There are costs in research, design and development (ex ante costs) and there are costs in operational service and support (ex post costs) as shown in Table 6.1. In transaction cost terms, these elements all generate a potential for hold-up in the overall life cycle of the equipment. See Masten (1984) for a transaction cost approach to aerospace production in the United States of America and Crocker and Reynolds (1993) for a transaction cost approach to aero-engine procurement in the US Air Force. These studies reveal partial evidence in support of the transaction cost hypothesis that aerospace suppliers do attempt to minimise the costs of a transaction, using the make-or-buy approach to the procurement–production mix. As a consequence of the ex ante and ex post costs, there is considerable uncertainty in the life cycle of military aircraft production. The best-case scenario for the aerospace industry is the peak production of one aircraft programme to be immediately replaced by the next. This will allow the preservation of jobs motive to be met (and in some cases hidden) and thereby reduce the turnover costs of employment. However, this runs contrary to the implied proﬁt motive of transaction costs, where the lowest cost option is predicted, whether it is make or buy. Nonetheless, delays and controversy over the content and timing of the Euroﬁghter 2000 has

Table 6.1 Life cycle costs of ﬁghting units Ex ante (procurement) hold-up

Ex post (in service) hold-up

Selection Development Production investment Production Operational investment

Service operational personnel Operational consumables Service support personnel Contractor support plus spares Disposal

Source: Kirkpatrick (1995) with addition by the author.

The supply-side implications of the arms trade

85

140 120 100 80 60 40 20 98

97

19

96

19

95

19

94

19

93

19

92

19

91

19

90

19

19

19

89

0

Personnel employed in military aircraft (thousands)

Figure 6.5 The Euroﬁghter 2000–Tornado production gap. Source: Estimates by the author.

caused a potentially disastrous production gap for BAe at the end of the Tornado programme, even accounting for the mid-life update (MLU) work on existing and older Tornado aircraft. By the late 1980s, the employment proﬁle of exiting BAe work appeared to have a gap effective from the end of Tornado production and before the beginning of Euroﬁghter 2000 production. This production gap is shown in Figure 6.5. There is a clear cycle of activity, as the Tornado production (and Euroﬁghter 2000 development) reaches a peak and before the Euroﬁghter production fully begins. As suggested above, the ﬁrst best solution for the UK aerospace industry would have been a seamless transition from one aircraft project to another, especially since there is already a time lag between the Harrier-AV-8 programme and the JSF. Another lag in the industry between the Tornado and Euroﬁghter 2000 would have been a serious ﬁnancial and logistical problem. This is because a military aerospace workforce is very expensive to recruit, train and retain in terms of time and resources. A series of disjointed demand for military aircraft can create problems for all parties concerned, which can lead to periodic bouts of overcapacity or uneven patterns of employment for the aerospace workforce, although this does depend on the relative transferability of resources. As shown in Figure 6.5, the Tornado production began to end in the early 1990s, whereas the Euroﬁghter 2000 production did not start until the late 1990s. As far as BAe is concerned it had at least three options. First, to ﬁll the production gap with exports and hence preserve jobs and skills till Euroﬁghter 2000 production was ready to sustain the workforce. Second, exit military markets in the long term and replace defence jobs in aerospace with civil jobs in aerospace. Finally, it could attempt a combination of the two. A decision on the most appropriate strategies to adopt is dependent upon the export potential of the equipment. Consider the two extreme cases, where X is deﬁned as the level of net export potential for an industry.

86

Ian Jackson

Case 1: X = 0. If domestic suppliers in a given industry are faced with a zero potential for net exports, then in the face of declining demand, it should either exit the market or diversify into other areas. For example, X = 0 in the UK submarine industry, where a bi-lateral monopoly exists between the prime contractor VSEL and MoD. For security reasons and a lack of international demand, exports are not a viable option (Hartley 1999). Although Germany has recently exported submarines to Israel, the really sensitive sector of the market is nuclear-powered submarines, which VSEL produce at Barrow-in-Furness, Cumbria. Case 2: X > 0. Where X has the potential to be a positive value, then domestic suppliers will have the additional option of voice, which is an effective lobby of government. This can take the form of funds for export initiatives and international political pressure on partner nations and allies. If X > 0, but not signiﬁcantly large, then a combination of exit and voice is required (Hirschman 1970). In the case of defence ﬁrms following the Cold War this has meant merger, selling off defence interests and even more lobby activity by the Defence Manufacturers Association (DMA) and SBAC, on behalf of such manufacturing sites as Warton, Lancashire. The proﬁtability of a global industry such as aerospace is related, in part, to how well companies can compete in international markets. UK aerospace exports as a percentage of turnover have risen from less than 50 per cent in 1980 to 75 per cent in 1997 (SBAC 1998). Indeed, the biggest single export order received by the United Kingdom is the Al-Yamamah programme to supply the Royal Saudi Air Force with Tornado and Hawk military aircraft and related products.11 This export order alone is estimated at £8 billion in 1991 prices (Martin 1996). SBAC (1998) claim that the UK aerospace industry has contributed an average of £2 billion per year to the UK balance of payments from 1988 to 1997 inclusive and a very large proportion of this is due to the Al-Yamamah project. The full details of AlYamamah are speciﬁed in two parts, where phases 1 and 2 ran between 1991 and 1999. An indicative breakdown of the project is given in Table 6.2. Given the crucial timing of Al-Yamamah from 1991 onwards and of course the sheer size of the project, estimated at £8 billion, it is then possible to argue that

Table 6.2 The Al-Yamamah project Phase 1 72 Tornado aircraft 30 Hawk aircraft 30 PC-9 3 mine-hunters Construction of support facilities Phase 2 48–100 Tornado aircraft 88 Black Hawk Helicopters 60–120 Hawk aircraft Construction of a new air base Source: Estimates by the author.

The supply-side implications of the arms trade

87

this export order allowed BAe to partially bridge the Euroﬁghter 2000–Tornado production gap. The dual motive to preserve jobs and make proﬁt for shareholders was actually achieved simultaneously,12 which partially solves the major problem of overcapacity. The trend for UK aerospace companies to form consortia with other European companies has been well established over the past thirty years. With the exceptions of the Harrier Jump Jet13 and the Hawk Trainer,14 UK aerospace companies have only built military aircraft with other partner European companies, since the cancellation of the TSR-215 in 1966. As shown previously, this is due, in part, to the rising unit costs of aircraft (supply-side) and declining order size needed by the RAF (demand-side). There is a similar trend in civilian aerospace, which led to the creation of Airbus Industries, a pan-European consortia, aimed at breaking the long-standing civilian market dominance of Boeing. The next trend requires European aerospace companies to go one stage further than forming a consortium. Increased merger activity is evident in the United States of America and cannot be escaped by European companies, if there is to be a longterm future for the industry. Merger activity as a means of economic survival is a question of when rather than if, especially for the larger, aerospace-dependent companies. As a result, this has important implications for defence procurement policy. In 1997, the UK aerospace received a total order intake of £13.93 billion, but only 26 per cent came from other EU member states. Whilst 24 per cent came from UK sources, 51 per cent came from outside the European Union, SBAC (1998). This has had the effect of further complicating the European context as well as the supply-side of the UK aerospace industry. In order to capture these issues in terms of public policy, the next section applies a game theoretic approach to European aerospace production (Owen 1995).

A game theoretic model of European aerospace production16 A game theoretic approach to arms production must appreciate the mutually dependent features of the Iron Triangle. The prisoners’ dilemma game of non-cooperative behaviour does not fully capture this interdependence between the players, because it contains conﬂict between the players. In the battle of the sexes game shown in Table 6.3, there is non-cooperative behaviour without conﬂict. Whilst the players are selﬁsh they are in love with each other and as a result would sacriﬁce their own ﬁrst preference in favour of the other (Rasmusen 1994). This would generate not one, but two Nash equilibria, as shown in bold. The way to solve this game is via the principle of ﬁrst mover advantage. If the man wants to go to the price ﬁght, then he must buy the boxing tickets in advance. This recognises the interdependence in decision-making, whilst acknowledging there is no conﬂict, as in the prisoners’ dilemma model. The same is true for the woman and hence where there is more than one Nash equilibrium the model is solved through ﬁrst mover advantage. In the case of arms production, the MoD and the UK DIB do have alternative objective functions. However, it is possible that this would be sacriﬁced in the

88

Ian Jackson Table 6.3 The battle of the sexes Woman

Man

Prize ﬁght Ballet

Prize ﬁght

Ballet

2, 1 0, 0

0, 0 1, 2

Table 6.4 The battle of the sexes in UK aerospace production UK DIB

MoD

Buy Make

Buy

Make

2, 1 0, 0

0, 0 1, 2

Table 6.5 The battle of the sexes in the European aerospace industry European DIB

UK DIB

Merge Autonomy

Merge

Autonomy

4, 15 0, 0

0, 0 15, 4

interests of the other, replicating the alleged cosy relationship between the players. That is, the MoD through the procurement agency fully recognise that the lowestcost option of aircraft procurement is to import (buy); whilst the UK DIB fully recognise that the option to maximise proﬁt is to manufacture aircraft themselves (make). The outcomes buy–buy and make–make are both Nash equilibria. As described above, the solution to the game is the ﬁrst mover advantage. For example, if the UK DIB lobby long and hard enough then the make–make option will be adopted (Bellany 1994), as shown in Table 6.4. This approach could be extended further to incorporate a multi-directional battle of the sexes, as detailed below. In this case, the players are the UK aerospace DIB and the European aerospace DIB, where the main nations would be France and Germany. The game would be to merge or remain autonomous. Once again, to merge would have its advantages in terms of scale economies and sharing the ﬁxed cost of development; however, to remain autonomous would similarly be beneﬁcial in terms of security of supply and the preservation of domestic industry. Therefore, this is a politically sensitive situation, especially if consideration is taken of the long history of UK collaboration with the US aerospace DIB on AV-8 and T-45. In Table 6.5, the ﬁrst mover advantage would win the game for either the UK DIB or the European DIB as predicted by the battle of the sexes model and in this example the stakes are increased, represented by the greater inequality between the

The supply-side implications of the arms trade

89

two equilibria (Kreps 1990). Therefore, it is predicted that the battle over the size and inﬂuence in the European aerospace DIB would be a ﬁerce and protracted affair, as currently being witnessed across France, Germany and the United Kingdom.

Conclusions Overcapacity in the UK aerospace defence industry has been partially caused by the insistence of the MoD to procure from largely domestic sources, in spite of the signiﬁcant length of the project life cycle. Although the UK aerospace industry itself has lobbied long and hard to create this situation, overcapacity is a major problem in defence and aerospace equipment, especially at the end of one programme and before the replacement programme has come on-stream. In order to bridge the gap and preserve employment, defence ﬁrms will resort to an export drive that will fuel the global arms trade. As a result of this action, there is ever-increasing offsets and counter-trade, which runs counter to the predictions of the transaction costs approach, discussed earlier. By 1997, BAe claimed to support over 3,000 military aircraft in service with twenty-three different air forces. Rather than this being a success story of a dynamic aerospace company, it could be the result of overcapacity in the domestic aerospace markets and the production gap between aircraft programmes. This, in turn, makes it imperative for UK aerospace companies to search for and nurture new and existing international arms markets, respectively. The truly successful aerospace picture is found on the civilian side, where in 1997 British Aerospace supported 1,630 Airbus aircraft in service with 146 different airline operators. Although there have been alleged payments of illegal state subsidies by the European Union, that could have unfairly improved the situation for Airbus Industries. The current arms trade in defence aerospace could create long-term problems in the future for the UK aerospace industry. If the reduction in global defence spending is maintained, then there will be less scope for international sales in military aircraft. If there are to be fewer opportunities for international arms sales in military aircraft, will the civilian sector be enough to bridge the production gap between Euroﬁghter 2000 and the next generation of military aircraft, such as the JSF and the Future Unmanned Aircraft (FUMA)? Or will the MoD and the UK aerospace industry be forced to recognise that it can either make (at a premium to the taxpayer) or buy (to the cost of domestic aerospace employment)? However, it cannot afford to do both as Hartley et al. (1987) pointed out in the quote that introduced this work. That is why game theory and, in particular, the battle of the sexes can show interdependency in decision-making for aerospace production. Also, it can reveal the issues of public policy and industry response to a decline in defence spending. Finally, game theory predicts a stronger role for government and trade associations such as SBAC and DMA in lobby activity, because ﬁrst mover advantage is crucial (Sandler 1997). In sum, arms trade in aerospace may not create new employment, it may merely act to preserve existing employment. If it is desirable to reduce the trade in military aerospace, then the root cause of overcapacity in domestic markets needs to be addressed. However, if the ‘buy’ option is pursued then this will be difﬁcult in terms

90

Ian Jackson

of job losses in a highly skilled and technical industry, such as aerospace. In other words, the arms trade in aerospace should be partially understood as a problem of excess supply, rather than as a result of a demand from foreign countries. This will lead to merger activity and consolidation across the European Union, as modelled by game theory. Further research in the timing and duration of military aircraft contracts is required to fully substantial these claims. More work is also required on the impact of decisions on the civil aerospace and projects such as Airbus, as well as the impact on relations with the US aerospace industry.

Acknowledgements I am grateful for the many comments on an earlier draft of this chapter given by delegates at the ‘Arms Trade, Security and Conﬂict’ Conference, organised at Middlesex University Business School on 11 June 1999. In particular, I would like to express my thanks to Ron Smith for many constructive suggestions. Also, the helpful comments of Nick Adnett, Alistair Dawson, Paul Downward, Keith Hartley, Todd Sandler and Tim Webb are also acknowledged. All remaining errors are my own.

Appendix A Projected UK holdings of combat aircraft, 1994–2010 Year

Jaguar

Harrier

Tornado F-3

Tornado GR-1

EF 2000

Sea Harrier

Total active

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

53 53 53 53 53 53 53 43 43 29 29 0 0 0 0 0 0

77 85 85 85 85 85 85 84 84 84 84 84 84 84 80 80 80

102 102 102 102 102 102 102 102 102 102 102 102 102 102 102 102 102

179 179 179 179 179 179 179 179 179 179 179 179 179 172 154 136 120

0 0 0 0 0 0 0 0 18 36 54 72 90 108 126 144 162

42 44 52 60 60 60 60 59 59 59 58 53 48 43 35 30 25

453 463 471 479 479 479 479 467 485 489 506 490 503 509 497 492 489

Source: IDDS Almanac (1994) in Forsberg (1994). Note Hartley in correspondence with the author has indicated that the MoD plan to begin the phase-out of the Tornado F-3 as early as 2004, which will reduce the total number of active aircraft.

The supply-side implications of the arms trade

91

Appendix B The leading UK aerospace suppliers in 1997 Company name

Turnover 1997 (£ M)

Employees, 1997

British Aerospace plc Cobham plc Dowty Aerospace Limited GEC-Marconi Avionics Limited GKN Westland Limited Hunting plc European Gas Turbines Limited Matra-Marconi Space Limited Meggitt plc Penny & Giles plc Rolls-Royce plc Siemens Plessey Limited Short Brothers Limited Smiths Industries plc

7,267 323 661 283 94 1,316 280 400 265 35 4,334 216 353 1,076

43,400 4,260 614 3,656 1,524 12,592 2,761 2,694 3,767 446 42,600 2,405 8,158 13,582

Source: Various companies’ reports and accounts.

Notes 1 For a clear and accurate assessment of UK defence procurement and the hidden costs of the arms trade, see The Armour-Plated Ostrich by Webb (1998). 2 For example, the merger between BAe and GEC-Marconi announced in 1999 to create BAE-Systems, as well as others throughout Europe involving German, French and Italian defence ﬁrms. This will be developed in the section on ‘A game theoretic model of European aerospace production’. 3 The Gulf War in 1991 was the ﬁrst occasion that the Tornado aircraft had been ﬂown under combat conditions, which is remarkable since it had been designed as long ago the late 1960s. 4 For example, the on-going competition between Boeing and Airbus Industries took an unexpected turn in 1998 when Boeing announced approximately 45,000 job losses world-wide, as a response to Airbus winning a greater market share in Europe, North America and the Far East. 5 See Appendix B for a more comprehensive list of the top UK aerospace companies in 1997. 6 The intense competition and rivalry between Boeing and Airbus Industries was an integral feature of the Uruguay Round of GATT, with claim and counter-claim of illegal government subsidies levelled at both the United States of America and the European Union. 7 This has implications for intra- and inter-industry trade, as well as offsets and countertrade. 8 For example, Tornado (n = 398) and Euroﬁghter 2000 (n = 250). 9 The term ‘pork-barrel’ is an expression referring to the political implications of defence procurement and dates back to the American Civil War. 10 The nature of make-or-buy decision is seen as central to the transaction costs approach. 11 The related products include hardened aircraft shelters and runways, plus the training of aircrew and groundcrew as well as spares.

92

Ian Jackson

12 A game theoretic approach to this issue is described in the section on ‘A game theoretic model of European aerospace production’. 13 The Harrier was developed into the Sea Harrier for naval duties and also developed into the AV-8 with American collaboration for the US Air Force. 14 The Hawk was developed into the Goshawk for naval duties and also developed into the T-45 with American collaboration for the US Air Force. 15 The multi-purpose TSR-2 (tactical, strike, reconnaissance) was cancelled part way through the production phase, as a result of spiralling costs and project delays. 16 Todd Sandler initially suggested this approach to me. Ron Smith has also provided much help in developing the ideas.

References Bellany, I. (1994) Reviewing Britain’s Defence. Gower, Aldershot. Bittleston, M. (1990) Co-operation or Competition? Defence Procurement Options for the 1990’s. Adelphi Papers 250 IISS, London. British Aerospace plc (1997) Annual Report and Accounts. BAe, Farnborough. Crocker, K. and Reynolds, K. (1993) The efﬁciency of incomplete contracts: an empirical analysis of air force engine procurement. Rand Journal of Economics 24(1), 126–146. Dugger, W. (1993) Transaction costs and the state. In: Pitelis, C. (ed.), Transaction Costs, Markets and Hierarchies. Blackwell, Oxford. Forsberg, R. (ed.) (1994) The Arms Production Dilemma: Contraction and Restraint in the World Combat Aircraft Industry. MIT Press, Cambridge, MA. Gummett, P. and Stein, J. (eds) (1997) European Defence Technology in Transition. Harwood Academic Publishers, London. Hartley, K. (1985) Defence procurement and industrial policy. In: Roper, J. (ed.), The Future of British Defence Policy. NIESRI/Gower, Aldershot. Hartley, K. (1999) The costs and beneﬁts of the Arms Trade. Keynote address at ‘The Arms Trade, Security and Conﬂict’, Middlesex University, 11 June. Hartley, K. Hussain, F. and Smith, R. (1987) The UK defence industrial base. Political Quarterly 58(1), 62–72. Hartley, K. and Hooper, N. (1995) Study of the Value of the Defence Industry to the UK Economy. Centre for Defence Economics, University of York, York. Hartley, K. and Sandler, T. (eds) (1995) Handbook of Defence Economics. Edward Elgar, Aldershot. Hirschman, A. (1970) Exit, Voice and Loyalty. Harvard University Press, Harvard. IDDS Almanac (1994) In: Forsberg, R. (ed.), The Arms Production Dilemma: Contraction and Restraint in the World Combat Aircraft Industry. MIT Press, Cambridge, MA, appendix table 3. Kirkpatrick, D. (1995) The rising unit cost of defence equipment – the reasons and the results. Defence and Peace Economics 6, 263–287. Kreps, D. (1990) Game Theory and Economic Modelling. Oxford University Press, Oxford. Lamming, R. (1993) Beyond Partnership: Strategies for Innovation and Lean Supply. Prentice Hall, Hemel Hempstead. Lovering, J. (1995) Opportunity or crisis? The remaking of the British arms industry. In: Turner, R. (ed.), The British Economy in Transition. Routledge, London, chapter 5. Martin, S. (ed.) (1996) The Economics of Offsets: Defence Procurement and Counter-trade. OPA, Amsterdam.

The supply-side implications of the arms trade

93

Masten, S.E. (1984) The organisation of production: evidence from the aerospace industry. Journal of Law and Economics 27, 403–417. Owen, G. (1995) Game Theory. Academic Press, London. Postan, M. M. et al. (1964) Design and Development of Weapons. HMSO, London. Pugh, P. G. (1993) The procurement nexus. Defence Economics 4, 179–194. Rasmusen, E. (1994) Games and Information, 2nd edn. Blackwell, Cambridge, MA and Oxford. Rolls-Royce plc (1997) Annual Report and Accounts. Rolls-Royce, London. Sandler, T. (1997) Global Challenges. Cambridge University Press, Cambridge. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge. Sandler, T. and Hartley, K. (1999) The Political Economy of NATO. Cambridge University Press, Cambridge. SBAC (1998) UK Aerospace Industry Statistics 1997 (www.sbac.co.uk). London. Webb, T. (1998) The Armour-plated Ostrich: The Hidden Costs of Britain’s Addiction to the Arms Business. Comerford and Miller, West Wickham. Williamson, O. E. (1975) Markets and Hierarchies. Free Press, New York. Williamson, O. E. (1985) The Economic Institutions of Capitalism. Free Press, New York. Williamson, O. E. and Masten, S. E. (1995) Transaction Cost Economics. Edward Elgar, Aldershot.

7

New challenges to arms export control Whither Wassenaar? Ron Smith and Bernard Udis

Introduction It is widely recognized that arms exports can have negative externalities on national security. As a result there are almost universal national controls on arms exports. There are also a variety of international organizations that attempt to regulate the international trade in arms through the coordination of suppliers. The theory of such coordination is developed in Levine and Smith (2000a). That paper examines the interaction between coordination of the supply of weapons through international export controls, coordination of the production of weapons through collaboration and coordination of the levels of military expenditure through alliances. In this chapter, we wish to examine the institutional features of arms export controls: the practical issues in arms export control, the international organizations that are trying to regulate exports and the factors that determine the effectiveness of the various arrangements. Of the large number of organizations involved we will focus on the Wassenaar Arrangement (WA). The end of the Cold War allowed the formation of the WA, the ﬁrst global multilateral agreement covering both conventional weapons and sensitive dual-use goods and technologies. The arrangement includes Russia, but not China, and is designed to prevent destabilizing acquisitions of weapons and technologies through a formal process of transparency and consultation. The WA is the successor to the now defunct Coordinating Committee for Multilateral Export Controls (CoCom) which was designed to restrict the transfer of arms and dual-use technologies to the Communist countries. Mastanduno (1992) provides a history of CoCom. Pierre (1997: 398) calls the WA ‘The best hope for creating a supplier-based multilateral regime’. Keller and Nolan (1997: 123) say the WA ‘has received scant attention from the policy community and ridicule from the arms lobby, and it is presently languishing with no high level involvement’. In surveying these challenges, we ﬁrst set out the background of what is being controlled, why and how. Second, we discuss some general theoretical issues that arise in international agreements of this sort and in particular the ways in which arms export control differs from classic arms control. Third, we discuss the disruptive potential of technical change. Then we examine the role of the WA in

New challenges to arms export control

95

the light of this analysis. Many of these issues have been extensively discussed in the literature, so we can summarize them brieﬂy.1

Background Although they overlap, it is useful to distinguish ﬁve main categories of militarily useful goods and services that may be subject to control. These are: weapons of mass destruction, major weapons systems, light weapons, dual-use technologies, and services (e.g. training and maintenance). They raise very different issues and involve very different market structures and control mechanisms. There is a widespread, though far from unanimous, consensus that proliferation of Weapons of Mass Destruction, WMD, (nuclear weapons, biological and chemical weapons (BCW), and their delivery systems) should be controlled. This is reﬂected in a set of Treaties and Supplier groups discussed below. Whatever their limitations, illustrated by the 1998 Indian and Pakistani nuclear tests, these arrangements have inhibited proliferation of WMD. In contrast, there is no such consensus that there should be general controls on the sales of conventional weapons, since a state’s right to self-defence, embodied in the UN Charter, gives it the right to buy arms from abroad. However, there may be speciﬁc reasons to control particular transfers and such reasons are set out in the 1991 declaration by the ﬁve Permanent Members of the Security Council (P5). Pierre (1997: 377–380) discusses the P5 guidelines. Typically, transfers should be restricted if the recipient’s acquisition of arms is excessive or destabilizing. Not only is it difﬁcult to deﬁne destabilizing acquisitions of weapons, but some like Gray (1992) argue that the concept of destabilizing weapons is itself incoherent: a category mistake, since governments not weapons cause war. Nonetheless, attempts are being made within various export control arrangements to deﬁne destabilizing acquisition either in terms of the capabilities of the weapons or the characteristics of the recipients. Within conventional weapons, it is useful to distinguish major systems from light weapons. The latter include small arms, land mines, small mortars, and manportable missiles. While light weapons have probably caused 90 per cent of the casualties in recent wars (Pierre 1997: 4), major weapons systems have the potential to change regional balances of power. Light weapons are also more difﬁcult to control than major systems partly because whereas major systems are supplied by an oligopoly, light weapons are competitively supplied by large numbers of producers. The theory of the decision as to when a country becomes a producer of arms is examined in Levine et al. (2000) and Levine and Smith (2000b). With the exception of the anti-personnel land-mine agreement there has been little action on light weapon control. The trade in light weapons is reviewed in Boutwell et al. (1995), Greene (1997) discusses control. Dual-use systems raise difﬁculties partly because the technologies required to produce weapons of mass destruction and their delivery systems all have important industrial applications (nuclear, biological, chemical, and space) and partly because, as CoCom found, it is difﬁcult to deﬁne militarily relevant technologies. Services are not generally subject to

96

Ron Smith and Bernard Udis

control but are increasingly important in the transfer of military knowledge and technology. Without training and support services many states have difﬁculty using the advanced weapons they have acquired. There are a variety of reasons for states wishing to control arms exports. The simplest is strategic embargo, preventing the sale of weapons to a potential enemy to limit its military capability. This was the main objective of CoCom, though as Mastanduno (1992) points out, a recurrent source of conﬂict between the United States and its allies was whether the objective of CoCom was strategic embargo or more general economic warfare. Strategic embargo can overlap with the prevention of the proliferation of particular types of arms, for example, inhumane weapons or WMD, either in general or to particular pariah states like North Korea, Iraq, Iran, Libya, and previously, South Africa. It may be feared that the external supply of weapons may prolong a war, hence the embargo of supplies to former Yugoslavia. It may be feared that the introduction of particular arms, for example, high-technology weapons, in a region may destabilize international relations either through an expensive arms race or through encouraging pre-emptive aggression. So for instance, until late 1997 the United States had a policy of not selling F16s to Latin America (the exception was Venezuela where there were particular US security concerns over Cuba). Supplier states may prohibit arms sales to governments with poor human rights records; either as a general indication of disapproval or from a concern that they may be used in internal repression. The embargo on China after Tiananmen Square and the objections to selling to Indonesia, during its illegal occupation of East Timor, are examples. Suppliers may fear that the cost of the arms purchases will damage the recipient’s economy, through the excessive accumulation of debt or the diversion of resouces from development needs. These objectives are wider and vaguer than those of classic arms control, which were to: reduce the probability of war; reduce the adverse consequences of war should it happen; and reduce the cost of military preparations.2 It was always recognized that there were trade-offs between these objectives: thus measures which reduce the probability of war might be more expensive. However, the tradeoffs in traditional arms control are much simpler than those in arms export control. For instance, if arms control stops you from acquiring a system, for example, ABM, you save money; if arms export control stops you from making a sale to another country, it costs you money.3 We return to the differences between traditional arms control and arms export control below. There are a variety of international mechanisms for controlling arms exports, but most rely on national arms export regulations. These usually involve: lists of products that require a license/notiﬁcation; lists of countries to which exports can or cannot be made; a list of criteria used to determine uncertain cases; customs procedures to stop unlicensed exports; some system to guarantee the end use of the equipment to stop it being re-exported; and in multilateral systems, notiﬁcation of exports that have been prohibited and perhaps a no-undercutting rule (countries agree not to approve a sale that has been refused a license by another country). Not all countries have well-functioning export control systems and, as Iraq demonstrated, even reasonably well-functioning systems can be evaded. The

New challenges to arms export control

97

developments in the European Union (EU) system of arms export controls, should be seen as an extension of national arms exports controls since the Single Market, which removes restrictions on cross-border trade within the EU, has reduced the effectiveness of national controls. The early development of EU policy is reviewed in Cornish (1995). There are a range of UN embargoes or sanctions that prohibit the supply of weapons to particular states. There are also a range of treaties, by which participants agree to restrict their use of certain sorts of weapons. Examples include the Non-Proliferation Treaty, NPT; the Landmine Treaty; the Chemical Weapons Convention, CWC; the Biological and Toxin Weapons Convention, BTWC. SIPRI (1999: Annexe A) describes these agreements. Such treaties are open to anybody who wishes to sign, and very large numbers of countries do sign: often over ninety. By signing, countries gain reputation beneﬁts (and perhaps economic beneﬁts as with the NPT) from being members of the club. The costs of signing are not high. Most of the signatories did not want these weapons and, for those that do, enforcement is not very strict with punishment for violation rare. For instance, the 1972 BTWC has no mechanism to check compliance, though talks have been going on since 1994 to develop a veriﬁcation protocol to the Treaty. There are also a range of transparency measures of which the UN arms transfer register is the most important. The UN requests states to report transfers of seven types of arms: battle tanks, armoured combat vehicles, large calibre artillery systems, combat aircraft, attack helicopters, warships, and missiles/missile launchers. Over ninety countries have reported transfers. These categories are also used by the WA and the ﬁrst ﬁve were used in the Conventional Forces in Europe (CFE) Treaty. Finally, there is a patchwork quilt of Supplier Agreements, sometimes called regimes.4 Such restricted clubs include: the Nuclear Suppliers Group or London Club, established in 1975; the Missile Technology Control Regime established in 1987; the Australia Group formed in 1985 to monitor CBW; and Wassenaar, established in 1996, which covers conventional weapons. The same areas are often covered by both treaties and supplier groups but, whereas the former are formal and open to all, the latter are informal with restricted membership. The supplier groups tended to begin by cooperation between a small number of countries and now typically include about thirty industrialized countries. Other countries may adhere to the guidelines of these arrangements without being members. CoCom was a supplier group by one side, aimed at the other side. The neutral states, though not members, tended to broadly abide by CoCom controls; albeit often after US bilateral pressure had been applied. Again, there may be economic incentives to join, for example, the promise of space cooperation with the United States encouraged Russia to comply with MTCR restrictions, see Piyakev et al. (1998). These are arrangements or agreements not treaties; they are enforced, like CoCom, by national legislation of members. Notice that in this context it is usual to think of suppliers as states, whereas it is companies, often transnational, that do the actual supplying and Reinicke (1994) discusses some implications of this. The natural supplier group – small, dominating a large part of the market, and with no ambiguities about membership – is the P5. They did start discussions,

98

Ron Smith and Bernard Udis

and drew up a set of guidelines; but after President Bush’s decision to sell F16s to Taiwan, China stopped cooperating and has not participated since. Direct arms export controls are not the only way to restrict transfers. The United States bought twenty-one Moldovan Mig 29 ﬁghters under the DOD Cooperative Threat Reduction (Nunn-Lugar) Programme, to prevent them being acquired by Iran (Boese 1997). This programme was also used to help persuade Belarus, Ukraine, and Kazhakstan to become ex-nuclear powers. A formal analysis of such measures is given in Jehiel et al. (1996).

General issues If arms exporters can agree to coordinate their actions, the impact of controls will be increased. Thus, collective action, cooperation, is more effective in meeting common arms export control objectives than individual action. However, individual states have an incentive to free-ride on or defect from the agreement, as they trade-off the longer-term security beneﬁts of control with the short-term economic incentives to sell.5 Thus, arms export controls have both public good aspects (nonrivalry, non-excludability) and prisoners’ dilemma cartel stability aspects (there is a joint interest to restrict supply, but an individual interest to cheat if you can do it without detection). These are repeated games, individual suppliers in the club decide whether to abide by the agreement or defect over a sequence of possible sales. Such collective action problems are discussed in Sandler (1992, 2000). While the game theory literature has not produced robust predictions, it does suggest that ﬁve factors are important for the outcome of such games: the structure of payoffs; the discount rate or myopia of the agents, that is, how they weight longand short-term factors; how well agents can monitor each other’s behaviour; the credibility of any threats, in particular whether it would be in the interests of the agent ex post to carry out the threat; and the sequence of moves, for example, whether there is a leader. The credibility of punishment threats can be increased by forms of pre-commitment, for example, ‘burning your bridges’ to make credible the threat that you will not retreat. Leadership plays a central role in creating a coalition which can reach a consensus on a particular issue, deﬁned during the negotiations. While leadership may be provided by small countries which act as facilitators, it is more often provided by large countries particularly the United States. One general question in such games is when does a centralized organization do better than decentralized trigger strategies, which punish defection. The most famous of these trigger strategies is the ‘tit for tat’ strategy in the Prisoners’ Dilemma: begin by cooperating and then do what the other agent did in the last period (Axelrod 1985). The main problem with ‘tit for tat’ is that it is vulnerable to imperfect monitoring of whether the other agents are cooperating or defecting (Axelrod 1997). In general, trigger strategies do not necessarily enforce the ﬁrst best cooperative and states do not rely on them; instead they create international organizations to try and solve the collective action problem. The reasons why states choose to act through international organizations is discussed by Abbott and Snidal (1998: 9) who note that international organizations provide centralization

New challenges to arms export control

99

(a concrete and stable administrative apparatus managing collective activities) and independence (the authority to act with a degree of autonomy and often with neutrality in deﬁned spheres). The autonomy allows them to ‘launder’ activities which would be unacceptable if done by a single nation. International organizations also reduce transactions costs, internalize externalities, gain economies of scale, and pool costs and risks. They can act as trustee, allocator, information provider, and arbiter. Certainly, arms export control organizations help share intelligence and operating procedures, usually through a ﬂow from large to small country members. The dynamics of the international negotiations involved are discussed in Hampson and Hart (1995). However, the WA has almost no autonomy compared to international organizations like NATO or the IMF. In this area, the principals, the nation states, are unwilling to delegate authority to their agent, the WA. Given that there is a strong case for an international organization, the obvious question is why are there so many supplier groups doing such closely related things? Why is there a patchwork quilt rather than an overarching organization?6 One possible analogy is with difﬁcult numerical problems with many linked variables and many conﬂicting constraints. In such cases, good solutions to such complicated conﬂict-laden calculations can best be found by breaking the entire problem into nonoverlapping domains, patches, with some coupling between the patches. Kauffman (1995: chapter 11) provides a non-technical discussion. Having competing organizations with different forms means that solutions or practices which work at one can be transmitted to the others as weapon lists were transferred from CFE to the UN Register to the WA. Pierre (1997: 428) emphasizes the importance of a composite approach: ‘a multifaceted regime with overlapping and complementary institutions and initiatives, none of which may be fully adequate by themselves but which are mutually reinforcing’. There are two dimensions to the optimal size of an arms export contol patch: the range of activities that it controls and the range of members that it includes. In principle, each margin would be determined by a comparison of costs and beneﬁts of extension. The marginal member or activity would be the one at which costs of extension just equalled the beneﬁts. It is worth returning to the distinction between classic arms control and arms export controls.7 Sandler and Hartley (1995) emphasize two features of classic arms control. First, the greater its success in achieving security and stability, the less important it is. The 1817 agreement between the United States and Britain to limit naval vessels on the Great Lakes and Lake Champlain is an example; compliance with this agreement has not been an issue in US–UK relations. Second, arms control is only needed in a hostile adversarial environment, but in such an environment the level of trust and conﬁdence required to negotiate an agreement does not exist: when achievable, arms control is not needed; and when needed it is not achievable. Gray (1992) also emphasizes this paradox. Arms control lives on this ﬁne line between being needed and being achievable. It exists on the edge of chaos, a characteristic of many adaptive systems (e.g. Kauffman 1995). Although they now seem to overlap, the origins of arms control and arms export control are

100

Ron Smith and Bernard Udis

quite distinct. Arms control involved agreements between enemies, arms export control involved agreements between allies to embargo an enemy to reduce its military capability. The overlap occurs because the nominal allies in arms export control (the United States, Russia, France in the WA, for instance) behave like enemies in arms control agreements. The classic theories of arms control asked; what were the incentives for states to come to an agreement and the incentives for the states subject to the agreement to comply with it? These incentives will be a function of the technologies involved; institutional structure and the veriﬁcation possibilities. For classic arms control to work all parties to the agreement must beneﬁt, you must be able to write a contract about something you can monitor and verify and there must be penalties for noncompliance. Arms control works easiest where there are a small number of parties involved (easier to negotiate and more likely for all to beneﬁt); where there is an identiﬁable object (piece of equipment or technology which can be monitored); that object has an unambiguous military purpose; and where control of that object restricts destabilizing military capability. Even in the classic framework, there are problems of distinguishing between avoidance and evasion and between what the parties know and what they can prove. Arms export control, in contrast, tends to involve fairly large numbers of parties, diffuse incentives and punishments, severe monitoring and punishment problems, and poorly identiﬁed objects of control which makes veriﬁcation difﬁcult. The problems of CoCom are illustrative in each of these respects. The classic arms control framework does not apply to arms export control in that the targets of control, potential recipients, are not willing parties to the contract. They are either excluded from agreements, which are between suppliers, or coerced to agree like Iraq at the end of the Gulf War. From the demand side, as Moodie in Harkavy and Neuman (1994: 136) points out, arms export controls appear to be hypocritical, selective, and discriminatory. Recipients may recognize the joint interest of supplier groups in restricting the ﬂow of destabilizing weapons, but they are more aware of the suppliers joint interest as a cartel to restrict the ﬂow of goods or technology in order to raise prices and maintain the suppliers’ military dominance and joint monopoly. Countries, like ﬁrms, have to choose how to source essential inputs; whether by buying on the open market, buying within established relationships or producing for themselves. Restricting supply to recipients changes those choices. Supply restrictions make it more attractive for countries to set up their own arms industry, to develop unconventional weapons and to use unorthodox methods of transfer. Karp (1994) and Klare (1997) discuss the use of black and grey markets to acquire arms. Domestic production, unconventional weapons, and unorthodox transfers reduce the transparency of arms acquisition. Because arms export control involves the regulation of others rather than agreement, it tends to show the classic forms of regulatory dialectic characteristic of ﬁnancial markets. Regulations are avoided and evaded and new regulations are required. Reinicke (1994) discusses this process. Any control system involves compliance costs, and the costs of arms export control fall heavily on the commercial ﬁrms that are potential suppliers of the

New challenges to arms export control

101

equipment. At some stage, the compliance costs can outweigh the regulatory beneﬁts, to the supplying nations. However, since it is the ﬁrms that pay the cost and the military that reaps the beneﬁt, supplying nations have to make a distributional judgement. In the United States this judgement is expressed in terms of whether the Department of Commerce or Department of State should have control of arms exports. Compliance costs were always an issue within CoCom and when they became too high relative to the regulatory beneﬁts, the system worked poorly. Changing technology, which we discuss below, changes the number of ﬁrms which may be able to supply militarily relevant goods and services, changing both regulatory beneﬁts and compliance costs. Traditional approaches to export control may now be infeasible. Moodie in Harkavy and Neuman (1994: 132) argues that nontraditional approaches to control will be needed ‘broadening its scope, increasing its ﬂexibility, simplifying its methods, and enhancing the speed of its accomplishments’. This may involve a shift from controlling supply as such to monitoring the end use and application of the technology in order to stop diversion to military purposes. Supplier groups usually require unanimity to operate (e.g. to change lists, admit new members). This is almost certainly unavoidable, but raises problems of maintaining speed and ﬂexibility. States may also have incentives to join clubs in order to use the unanimity rule and thus their veto power to stop the club working. For instance, it has been suggested that this was the UK’s main motivation to join the EU, EEC as it then was known. One cannot assume that desire to join a club signals agreement with its objectives. The slow progress of the WA has raised similar suspicions about the motives of some of its members. Unanimity requirements also mean that change has to be incremental, establishing procedures and precedents in relatively uncontroversial areas such as inhumane weapons and then extending the scope as conﬁdence in the arrangement grows. If such organizations are effective, they establish a set of norms that maintain compliance. Axelrod (1997: chapter 3) and Cooter (2000) discuss the evolution of such norms. Existing arms export control organizations are a long way from establishing strong norms of this sort.

Technology The impact of technology on arms control is not a new problem. The 1922 Washington Naval treaty restricted the range of Japanese battleships, giving Japan incentives to avoid the constraint through the development of aircraft carriers. Craft (2000) discusses the effect of the treaty, Carus (1994) discusses military technology and the arms trade. CoCom, while not watertight did raise the cost of gaining technology and weapons to the Soviet Union. But CoCom had problems handling dual-use technologies and keeping lists updated. Like CoCom, the WA has to face changing military technology and Keller and Nolan (1997: 123–124) say that if the WA is to have any effect ‘it must at a minimum, identify a set of speciﬁc technologies and particular weapons that must not be proliferated’. Adjusting to changing technologies is made more difﬁcult by changing relationships

102

Ron Smith and Bernard Udis

between civil and military technologies, spin-in rather than spin-off, and increasing international economic integration, globalization. There have been long cycles in the relationship between military and civil technology. After Second World War, military technology was in advance of civilian technology; thus, the military were familiar with the military applications of a technology before it was widely adopted by civilians and in certain cases they could restrict that adoption in ways that reduced the potential threat. For example, the military could put restrictions on the GPS resolution available to civilians because they controlled the satellites, see Lachow (1995). Now that civilian technology is often in the lead, the military may ﬁnd it better to adopt an operating civilian technology as they did with the Iridium global satellite-communications system. Col. Robert Weber, programme manager for the Defense Information Systems network, a unit of the DOD, is quoted as saying ‘The DOD used to be a leader in satellite communications. Now everyone has caught on to space . . . . We’ll take advantage of [the private sector’s] economies of scale’ (Wall Street Journal, 26 January, 1998: B4). They also have to take the consequences when the private sector ﬁnds it unproﬁtable to provide, as was the case with Iridium. There are cases where traditional military technologies (e.g. encryption and simulation) have such important commercial applications that private ﬁrms are willing to invest heavily in R&D to duplicate the effective knowledge that the military will not release. When there is a large and obvious proﬁt potential, as with cryptography, private R&D will dwarf military R&D. Unhampered by bureaucratic and security restrictions private R&D may also be more ﬂexible, more innovative and better organized. In addition, attempts by the military to control the technology will run into major difﬁculties. Traditionally, dual-use technologies were process technologies like the machine tools Toshiba sold to the Soviet Union, used to improve submarine propellers. Now they are often product technologies directly useful in combat. The product may be types of information technology (IT) such as communications systems like ﬁbre-optic networks; civilian sub-components that can be assembled into weapons; or civilian products, like remotely piloted vehicles, designed for crop-spraying, but capable of being used as delivery or reconnaissance vehicles. The military have always thought about future technologies and how they may change combat; but usually they could focus on a few critical military technologies. Even then it was difﬁcult to forecast the tortuous path that must be followed in moving from a technology, to a weapons system, to an appropriate doctrine for its use and ﬁnally to a change in military organization to deploy the new system effectively. Now it is more difﬁcult because there are a large range of generic technologies that have the capability to revolutionize military affairs. Some of these are familiar but changing rapidly, like bio-technology or IT; others are relatively new like smart materials, nano-technology, or high-temperature superconductors. These technologies are being developed with private rather than public funds; often within transnational companies where technology ﬂows globally; and are perceived as being central to a country’s economic competitiveness and export sucess. By the time the military utility of such technologies has been

New challenges to arms export control

103

identiﬁed, signiﬁcant commercial interest has already developed, making controls more difﬁcult to apply. Even where there are clear technological trajectories, like Moore’s Law,8 the applications of the technology are unpredictable. What are often called the ‘killer applications’ made possible by Moore’s Law which were crucial to the diffusion of the technologies – word-processing and spreadsheets on PCs or e-mail on the Internet – were not predicted in advance by those who understood the technology. These general changes in technology give rise to three particular sources of military concern. The ﬁrst is with respect to traditional military products (e.g. advanced ﬁghters or precision guided weapons). Competitive pressure in the arms export market means that ﬁrms are increasingly selling state of the art equipment. In some cases, the recipients may not have the skills in organization or military doctrine to use these complex systems effectively; in other cases, the systems may be relatively autonomous and simple to use. Argentina was unable to adjust its air-force tactics (matching altitude of attack with bomb-fusing) or its army tactics rapidly against the United Kingdom during the Falklands/Malvinas war; but it could use an Exocet to destroy a British warship. Gormley (1998) discusses cruise missiles and the inadequacy of the MTCR to control them. The second source of concern is that competitive pressures are forcing arms ﬁrms to transfer technologies as part of the offset requirement for arms deals, thus diffusing military technologies. Unfortunately for the recipients, they often do not have the broader scientiﬁc and industrial infrastructure to absorb the technology in a way that allows them to develop their own systems; Martin (1996) and Spear (1997). Traditional military systems are highly specialized and complex and even advanced industrial countries have difﬁculty with the systems integration problems involved. In many cases, technology transfer creates dependence, rather than independence. The third source of concern is about innovative military applications of generic civilian technologies; the military equivalent of the ‘killer applications’, which diffused PCs and the Internet. This is the most speculative, but potentially the most dangerous concern. Silverberg (1994: 126) gives an example. ‘One has visions of an Indian software engineer developing a new algorithm that can jam command and control broadcasts, selling it to Iran, which then uses it to bring the US Navy to a halt in the Strait of Hormuz’. The great shortage of software engineers, partly because of the efforts made to cope with the feared Y2K problem, means that state of the art skills are being widely diffused. It is now common for US software engineers to work on the program during the day, then beam it by satellite to India, Israel, or Russia, where another team works on it during the US night; passing it back the next morning. Controlling such technology transfers raises obvious difﬁculties for traditional export control mechanisms. Predicting such ‘killer applications’ poses even greater problems for traditional intelligence communities and the threat from these new technologies will probably only be taken seriously, when they are unexpectedly used by a small or revisionist state to change the balance of military power. Having observed the Gulf War and Kosovo, revisionists states will realize that they should not compete on types of battleﬁeld that the United States

104

Ron Smith and Bernard Udis

dominates. Instead, they should use the tactic of asymmetric warfare, using generic capabilities widely diffused in the South, not traditional military technologies monopolized by the North to achieve their objectives.

Wassenaar With the end of the Cold War, it was no longer possible to retain CoCom and at a high-level meeting in the Hague on 16 November 1993 it was decided that the organization would be phased out by 31 March, 1994. After considerable discussion, it was succeeded by the WA on Export Controls for Conventional Arms and Dual-Use Goods and Technologies, named after the town in the Netherlands where it was provisionally established in 1995. It was formally established in Vienna in July 1996. It differs from CoCom by including Russia and other members of the former Warsaw Pact, though not China, and being a transparency rather than a control regime. The principal goal of the Arrangement is to gather information which will reveal potentially destabilizing accumulation of weapons or militarily useful dual-use technologies. Participating states agreed to maintain national controls on transfers of items speciﬁed in the dual-use and munitions lists; notify other members of aggregate transfers of munitions to non-participating states and of individual cases where licenses to transfer dual-use items have been denied. In principle, unlike CoCom, the WA is not directed at any state or group of states; in practice, it is directed at countries whose behaviour is a cause for serious concern, particularly Iran, Iraq, Libya, and North Korea. The ﬁrst operational plenary session was held in Vienna in December 1997. The WA is still young and while this immaturity leaves it currently ineffective it may make it sufﬁciently pliable to be shaped in ways that would allow it to evolve into a more effective organization. Within WA, the Big 6 (France, Germany, Italy, Russia, United Kingdom, and the United States) have agreed informally to meet for more intensive consultation and more intrusive information sharing. Thus, a subset of WA may operate. The issue of the optimal composition is important. In principle, the WA is open to any country that is a producer or exporter of arms or items subject to control; adheres to the NPT, BTWC, and CWC; and maintains and applies effective national export controls. The WA, with thirty-three members, is a fairly big group (smaller than the NPT but bigger than the P5), which lacks the coherence that was provided to CoCom by the shared perception of a Soviet threat. Building a coherent vision of shared interests is important. The WA is a supplier group and will inevitably lack legitimacy on the demand side. Supplier groups are cartels with a common interest not merely in preventing the spread of destabilizing arms and technologies, but in restricting supply to raise price. Such groups will be seen by the ‘have-nots’ as an attempt by the ‘haves’ to maintain their monopoly position in arms and technology. To give WA greater legitimacy, various ‘outreach measures’ to integrate recipient states into the agreement are important. The WA already conducts outreach to potential members but this could be extended to potential recipients subject to control. The outreach measures could include participation in regional restraint

New challenges to arms export control

105

discussions aimed at establishing demand-side agreements to stop the diffusion of particular systems or technologies to a region. At the June 1997 meeting of the General Working Group of the WA, the United States tabled a paper entitled ‘Emerging Weapons of the 21st centrury’. It identiﬁed ﬁve areas where the United States expects the development and improvement of advanced weapons capabilities within the next ﬁfteen to twenty years. These areas include WMD; long-range precision weapons; reconnaissance, surveillance, and target acquisition systems; counters to precision strike capabilities; and information warfare. The United States argued that these issues were of concern to WA because of the threat of proliferation of such capabilities through exports or diversion or upgrading existing weapons and the danger of increasing the combat capability of states that might deploy them in an irresponsible or destabilizing way. The traditional way of handling technology changes is through a process of list review through which old items are removed and new items added. This is a process which tends to lag behind very rapid changes in technology. Within CoCom, list review was fraught with difﬁculties; and lists became out of date: including technologies which were widely available and not including new technologies. Of particular danger are what might be called precursor technologies which open the possibility of vaulting over the traditional stages in the development of militarily useful products and technologies. Presently, WA lacks the means to identify such developments. The WA munitions list, the seven categories of the UN arms transfer register,9 excludes many munitions, particularly retroﬁt equipment, which can substantially upgrade the military capability of old tanks, combat aircraft, or warships. Extending the list has been the subject of discussion within the WA. At the December 1998 meeting, the control lists were changed to reduce coverage of civil telecommunications and update controls on encryption. Different states are subject to different domestic political and bureaucratic constraints and different international perceptions. Several states have expressed fears that the United States, traditionally the dominant supplier, will try to use controls in its own interests: to maintain its monopoly. This was a recurring problem with CoCom. Despite the fears of US exploitation of its position, the United States is the natural leader, if only because it exports the most arms. However, the United States, partly because of the constitutionally imposed separation of government powers, has difﬁculty providing consistent, coherent leadership over long periods; instead there is a process of sequential attention to different goals as different issues become salient in Washington. This myopia, a focus on current Washington concerns, has weakened US leadership. Myopia is not just a US problem, France has often appeared myopic for a different reason: too focused on short-term economic self-interests at the expense of longer-term common interests. There is one other institutional feature. For bureaucratic reasons, the United States, unlike other nations, tends to separate its delegation membership by regime; rarely do the same persons participate in the various arrangements. Thus, if items arise at MTCR or Australia group which are relevant to WA, the United States representative has difﬁculty in responding on the spot in a meaningful way.10 Conversely, the French are quite centralized; in consequence if the crucial decision-maker has not made

106

Ron Smith and Bernard Udis

a decision the representatives at all the arrangements have to stall. Technology does not respect these artiﬁcial boundaries between regimes. Trying to create a comprehensive regime that combines the existing control and transparency systems would not be sensible, but systematic cross-fertilization and coherent decision-making at a national level would help to transmit insights and speed decisions in the various arrangements. At an institutional level, WA needs to develop an effective monitoring system, including its own IT system to track movements of weapons and sensitive technologies based on information provided by national participants. This would also give the WA more autonomy. There is an alternative approach. Reinicke (1994) proposes a very innovative arms export control system working on a self regulatory basis at the level of the ﬁrm; modelled on ﬁnancial Self-Regulatory Organizations (SROs). This would be very different from the traditional WA framework, but it is not clear whether a multilateral system of this sort would be feasible. Relative to ﬁnance, far more ﬁrms would be involved; there is no natural deﬁnition of the market since so many technologies can be used for military purposes; there is no experience with international SROs even in ﬁnance; the activities subject to control are not as well deﬁned as in ﬁnance; arms sales are more politically sensitive than ﬁnancial transactions; and the monitoring requirements seem to exceed the capability of current IT and reporting capabilities. As an indication of the possibility of learning between arrangements, it is possible that WA could consider if there are any lessons in the progress of the Anglo-French proposal on a Code of Conduct for Arms Exports agreed by EU ministers in May 1998. The code outlines eight guidelines to govern the process by which states give licenses, establishes consultation mechanisms and requires all EU countries to submit annual reports on arms exports. The Anglo-French proposal faced opposition from other EU members: smaller countries were concerned that the controls are not strict enough and that consultations were only bilateral (between the state which denied a license and that which was considering supplying) rather than multilateral. The code is not legally binding but could increase transparency. In the negotiations the conﬂict of interests between the main exporters, Britain and France, and the other EU countries was apparent.

Conclusions While we cannot make ﬁrm predictions about whether Wassennaar will develop into an effective organization; we can identify certain factors that might determine its effectiveness. These factors come from the theory of repeated games and are very closely linked. The ﬁrst is the structure of payoffs or incentives. The long-term payoff is the security beneﬁt of a reduction in proliferation. To the extent that the participants share a perception of a serious threat, they will work together and solve the technical problems of control. This is a lesson from CoCom, the Europeans cooperated with the United States best when they perceived a serious Soviet threat. However, it may take more than the India–Pakistan nuclear tests to persuade the participants that new

New challenges to arms export control

107

technologies pose a serious threat; it may require the use of a ‘killer application’ by a revisionist power. The short-term payoffs are the economic consequences of a sale. At present, the economics rewards non-compliance, through the proﬁts from potentially destabilizing sales. There may be scope to design rewards for compliance as was done with the NPT, Russian adherence to the MTCR and the denuclearization of former Soviet Republics. The second is the degree of myopia or the discount rate; we saw myopia could arise for a variety of reasons, different in France and the United States. For the WA to work requires the participants to accept short-run economic and political costs in return for long-run security beneﬁts. The relative weight they give to the present and future is thus crucial. Of course, an imminent threat reduces myopia about possible security consequences. The Iraqi invasion of Kuwait made countries more concerned about other possible threats and more concerned to enhance control. The effect wore off, however. Nolan (1988: 522), reviewing the lessons from the CAT negotiations, comments on the danger of ‘a permanent ascendance of short-term over long-term interests’. The third is the monitoring capability of the system. While, the main source of monitoring will be national export control systems and national intelligence sources, the WA by pooling and tracking this information could increase the degree of transparency, and increase shared perceptions of potentially threatening developments in technology, intentions, or capabilities. To do this effectively will require better IT capability and more ﬂexible procedures to identify what should be monitored. Traditional list review procedures to determine what will be monitored are likely to be too slow. This requires the suppliers being willing to delegate authority to the WA, making it a much more autonomous body. The fourth is the degree of credibility in the system: the continuity of policies and certainty of consequences. Everyone needs to know that defection will be detected (monitoring) and punished (credibility) while compliance will be recognized and rewarded. At present, the WA has little credibility. Finally is the degree of leadership. It would require concerted commitment by the major arms producers, particularly the United States, United Kingdom, and France, to change the structure of payoffs, to put more weight on longer-term security consequences, develop the monitoring capability of the WA and to allow it to build credibility and autonomy. Without this commitment, which currently does not exist, it is unlikely that Russia and the smaller producers will cooperate or that Wassennaar will work.

Acknowledgements During calendar 1997, Bernard Udis was a Foster Fellow at the US Arms Control and Disarmament Agency, where his responsibilities included the Wassenaar Arrangement. The opinions in this chapter are those of the authors and should not be attributed to ACDA or the US government. Ron Smith is grateful to the Department of Economics, University of Colorado at Boulder for their hospitality during 1997–98, when most of the work on this chapter was done and to the ESRC for

108

Ron Smith and Bernard Udis

ﬁnancial support under research grant R00235685 on models of the arms trade. We are also grateful to Mary Ann Casey, Yongmin Chen, Saadet Deger, Paul Levine, and Somnath Sen for comments on an earlier draft. An earlier version of this paper was published in The Nonproliferation Review, Summer 2001, 8(2), 81–92, copyright Monterey Institute of International Studies.

Notes 1 Useful references are van Ham (1993), Harkavy and Neuman (1994), Mussington (1994), RAND (1996), and Pierre (1997). 2 Schelling and Halperin (1962) is a classic statement of arms control theory. Gray (1992) is a fundamental critique of the whole activity. 3 The economics–security trade-offs and their implications for control are analysed in the context of formal models in Levine and Smith (1995, 1997). 4 Regime has a specialized meaning in International Relations; for instance, Mastanduno (1992: 5) says ‘By “regime,” I mean a set of rules, norms and expectations around which actors’ expectations converge in a given issue area’. It is a subject of dispute whether these agreements are regimes in this sense, so we have avoided the term. Pierre (1997: chapter 15), discusses what an arms control regime might involve. In Levine and Smith (2000a), regime is used to denote particular combinations of forms of coordination. 5 The size of these economic beneﬁts to the supplier nations is a matter of dispute. Commercial conﬁdence means that little of the detail is disclosed and the contracts are very complex involving subsidized R&D, soft credit, and offsets. The selling ﬁrms clearly make proﬁts, they would not take the contracts otherwise; but these proﬁts may come from supplier subsidies not recipient payments. Johnson (1994) discusses ﬁnancial aspects of the arms trade. Recent trends are reviewed in Anayiotis and Happe (1997). The costs and beneﬁts of offsets are discussed in Udis and Maskus (1991) and Udis (1994). 6 A framework for examining the relationship between the various patches is given in Harkavy (1980). 7 There are, of course, cases where they overlap and some arms transfer negotiations with the Soviet Union or Russia involve very traditional elements. Two examples are: the US–Soviet negotiations on conventional arms transfers (CAT) control discussed in Nolan (1988) and the US–Russian negotiations on the MTCR described in Pikayev et al. (1998). 8 Moore’s Law – that the number of components on a chip, or computer power per dollar, doubles every eighteen months – was proposed in 1965 and still seems to hold, though it must hit physical limits eventually. 9 Battle tanks, armoured combat vehicles, large-calibre artillery systems, combat aircraft, attack helicopters, warships, and missiles/missile launchers. 10 RAND (1996: chapter 7), discusses the limitations of the US arms export control process and Nolan (1988) discusses the impact of bureaucratic disputes on the CAT.

References Abbott, K. W. and Snidal, D. (1998) Why states act through formal international organizations. Journal of Conﬂict Resolution 42(1). Anayiotis, G. and Happe, N. (1997) Recent trends and the ﬁnancing of the arms trade. Peace Economics, Peace Science and Public Policy 4(4), 1–30. Axelrod, R. (1985) The Evolution of Cooperation. Basic Books, New York.

New challenges to arms export control

109

Axelrod, R. (1997) The Complexity of Cooperation, Princeton Studies in Complexity. Princeton University Press, Princeton, NJ. Boese, W. (1997) US buys Moldovan aircraft to prevent acquisition by Iran. Arms Control Today 7(7), 28. Boutwell, J., Klare, M. T., and Reed, L. W. (eds), (1995) Lethal Commerce: The Global Trade in Small Arms and Light Weapons. American Academy of Arts and Sciences, Cambridge, MA. Carus, W. S. (1994) Military technology and the arms trade: changes and their impact. In: Harkavy, R. E. and Neuman, S. G. (eds), The Arms Trade: Problems and Prospects in the Post-Cold War World, The Annals of the American Academy of Political and Social Science, September (535), pp. 163–174. Cooter, R. D. (2000) Law from order: economic development and the jurisprudence of social norms. In: Olson, M. and Kahkonen, S. (eds), A Not So Dismal Science, Oxford University Press, Oxford, chapter 9. Cornish, P. (1995) The Arms Trade and Europe, Chatham House Papers. The Royal Institute of International Affairs, London. Craft, C. B. (2000) An analysis of the Washington Naval Agreements and the Economic Provisions of Arms Control Theory. Defence and Peace Economics 11(2), 127–148. Gormley, D. M. (1998) Hedging against the cruise missile threat. Survival 40(1), 92–110. Gray, C. S. (1992) House of Cards: Why Arms Control Must Fail. Cornell University Press, Ithaca, NY. Greene, O. (1997) Tackling Light Weapon Proliferation: Issues and Priorities for the EU. Saferworld, London. Hampson, F. O. and Hart, M. (1995) Multilateral Negotiations: Lessons from Arms Control, Trade and the Environment. The Johns Hopkins University Press, Baltimore, MD. Harkavy, R. E. (1980) Harmonizing policies across arms control domains: dilemmas and contradictions. In: Harkavy and Kolodziej (eds), American Security Policy and Policy Making: The Dilemmas of Using and Controlling Military Force, Lexington Books, Lexington, MA, pp. 129–147. Harkavy, R. E. and Neuman, S. G. (eds) (1994) The Arms Trade: Problems and Prospects in the Post-Cold War World. The Annals of the American Academy of Political and Social Science September (535). Jehiel, P., Moldovanu, B., and Stacchetti, E. (1996) How (not) to sell nuclear weapons. American Economic Review 86(4), 814–829. Johnson, J. L. (1994) Financing the arms trade. In: Harkavy, R. E. and Neuman, S. G. (eds), The Arms Trade: Problems and Prospects in the Post-Cold War World, The Annals of the American Academy of Political and Social Science September (535), 110–121. Karp, A. (1994) The rise of black and gray markets. In: Harkavy, R. E. and Neuman, S. G. (eds), The Arms Trade: Problems and Prospects in the Post-Cold War World, The Annals of the American Academy of Political and Social Science September (535), 175–189. Kauffman, S. (1995) At Home in the Universe. Oxford University Press, Oxford. Keller, W. and Nolan, J. E. (1997) The arms trade: business as usual. Foreign Policy Winter (109), 113–125. Klare, M. T. (1997) The Subterranean arms trade: black-market sales, covert operations and ethnic warfare. In: Pierre, A. J. (ed.), Cascade of Arms, Brookings for the World Peace Foundation, Washington, DC, pp. 43–71. Lachow, I. (1995) The GPS dilemma: balancing military risks and economic beneﬁts. International Security 20(1), 126–148.

110

Ron Smith and Bernard Udis

Levine, P. and Smith, R. (1995) The arms trade and arms control. Economic Journal 105(429), 471–484. Levine, P. and Smith, R. (1997) The arms trade. Economic Policy October (25), 335–370. Levine, P. and Smith, R. (2000a) The arms trade game: from laissez-faire to a common defence policy. Oxford Economic Papers 52, 357–380. Levine, P. and Smith, R. (2000b) Arms export controls and proliferation. Journal of Conﬂict Resolution, 44(6), 885–895. Levine, P., Mouzakis, F., and Smith, R. (2000) Arms export controls and emerging domestic producers. Defence and Peace Economics 11(5), 505–531. Martin, S., ed. (1996) The Economics of Offsets: Defense Procurement and Countertrade, Harwood, London. Mastanduno, M. (1992) Economic Containment: CoCom and the Politics of East West Trade. Cornell University Press, Ithaca, NY. Mussington, D. (1994) Understanding Contemporary International Arms Transfers, Adelphi Paper 291. International Institute of Strategic Studies, London. Nolan, J. E. (1988) The U.S.–Soviet conventional arms transfer negotiations. In: George, A. L., Farley, P. J., and Dallin, A. (eds), U.S.–Soviet Security Cooperation: Achievements, Failures, Lessons, Oxford University Press, Oxford, pp. 510–523. Pierre, A. J. (ed.), (1997) Cascade of Arms. Brookings for the World Peace Foundation, Washington, DC. Pikayev, A. A., Spector, L. S., Kirichenko, E. V., and Gibson, R. (1998) Russia, the US and the Missile Technology Control Regime, Aldephi Paper 317. International Institute for Strategic Studies, London. RAND. (1996) Arms Proliferation Policy: Support to the Presidential Advisory Board. RAND, Santa Monica, CA. Reinicke, W. H. (1994) Cooperative security and the political economy of nonproliferation. In: Nolan, J. (ed.), Global Engagement: Cooperation in the 21st Century, Brookings Institution, Washington, DC, chapter 5, 175–234. Sandler, T. (1992) Collective Action. University of Michigan Press, Ann Arber, MI. Sandler, T. (2000) Arms trade, arms control and security: collective action issues. Defence and Peace Economics 11(5), 533–548. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge, UK. Schelling, T. C. and Halperin, M. H. (1962) Strategy and Arms Control. Twentieth Century Fund, New York. Silverberg, D. (1994) Global trends in military production and conversion. In: Harkavy, R. E. and Neuman, S. G. (eds), The Arms Trade: Problems and Prospects in the PostCold War World, The Annals of the American Academy of Political and Social Science September (535), pp. 122–130. SIPRI. (1999) SIPRI Yearbook, Armaments, Disarmament and International Security. Oxford University Press for Stockholm International Peace Research Institute, Oxford. Spear, J. (1997) The role of offsets in the international arms trade. Paper presented at the ISA Annual Meeting, April 1997, Toronto. Udis, B. (1994) Offsets in defense trade: costs and beneﬁts, Unpublished NATO Research Report. Udis, B. and Maskus, K. E. (1991) Offsets as industrial policy: lessons from aerospace. Defense Economics 2, 151–164. van Ham, P. (1993) Managing Non-Proliferation Regimes in the 1990s, Chatham House papers, Pinter Publishers for the Royal Institute of International Affairs, London.

Part II

Wars, alliances and arms races

8

Rational wars with incomplete information Paul Levine and Francisco Moraiz

Introduction Political theory has approached the study of war from many different perspectives. There is a group of theories called ‘realist’ explanations of war, based on the assumption that countries act rationally and, therefore, follow a predictable behaviour derived from utility maximization. These theories use a variety of approaches including game theory, which has produced a new way of looking at conﬂict. For instance, animal contests, litigation, wage negotiation and international conﬂict are clear examples of how game theory has been successfully applied. There are two different kinds of explanations of war offered by political science rational models. This ﬁrst one studies a scenario in which leaders are rational but overestimate their chances of military victory. The second argues that states may lack information about their adversary’s willingness to ﬁght. This is directly linked to possible informational asymmetries in the cost of ﬁghting. Our approach is closest to this second strand. We confront the problem from an economic perspective. The literature from which we draw includes public choice models of economics and conﬂict that focus on directly unproductive activities. The economic models of proﬁt-seeking have concentrated largely in the theoretical analysis of lobbying for monopoly protection, import licences and tariffs in commerce. Krueger (1974) generated the term rent-seeking and Bhagwati (1982) makes a taxonomy of rent-seeking activities that are directly related to governmental policies. Tullock (1980) analyses optimal strategies from players who participate in a lottery for a ﬁxed prize. It can be taken as one of the ﬁrst examples of conﬂict with endogenous allocation of resources and has generated developments in many different directions. The concept of the contest-success functions (CSFs) and the technology of appropriation has been studied by Dixit (1987), Hirshleifer (1989) and Stergios (1992). Intriligator (1982) reviewed the formal research in conﬂict theory and ﬁnally Brito and Intriligator (1985) produced one of the ﬁrst formal models that incorporates redistribution of resources by negotiation and imperfect information. In previous models (Moraiz 1998), we set up a framework that endogenizes the decision to ﬁght by comparing the outcomes of war and agreement in terms of

114

Paul Levine and Francisco Moraiz

the utility level that each decision produces. We showed that, in an environment with perfect information, there is always a Pareto-superior bargaining outcome due to the cost that war and conﬂict produce. The next step is obvious. We must introduce incomplete information. However, one of the problems of these games is that the equilibrium outcome can be a result of the chosen assumptions regarding the bargaining procedure and informational asymmetries. The equilibrium concept changes whether there is a single country that makes offers or both. The equilibrium will change if the bargaining process is extended to more than two periods. The valuation of the cost of ﬁghting can be introduced in many different ways. Countries can be represented either as tough with low externality cost or soft with high cost. These factors feature prominently in our model. Fudenberg and Tirole (1983) draw the following conclusions about the effects of valuations and discount factors on the strategic behaviour of buyers and sellers in a model of bargaining with incomplete information: First, a decrease in the buyer’s discount factor may make him better of in spite of the fact that, being more impatient, he becomes more vulnerable to a high demand. We explained this phenomenon by the impossibility of commitment to take given actions (here to accept a compromising offer) which is required by the concept of perfect equilibrium. Second, increasing the contract zone (e.g. by making the seller more eager to sell) may increase the possibility of disagreement. Third, if the buyer has complete information about the seller, the seller may charge a higher price in the second period than in the ﬁrst; the buyer may nevertheless refuse such a ﬁrst-period offer since there is the possibility that the seller is soft and will charge less in the second period. Fourth, increasing the number of periods may have surprising welfare effects: it can decrease efﬁciency even when the one-period game has an inefﬁcient solution. The message from Fudenberg and Tirole (henceforth F&T) is clear. We have to be very careful about general assertions about parameter changes and the effect on conﬂict. The solutions have to be treated within a speciﬁc context and the conclusions of our work should be interpreted carefully. Despite these caveats, models of game theory and conﬂict can give us important insights into the decision-making process leading to the outbreak of war. They can also give us some indication of what would be considered optimal policies if we were facing an ideal world of rational players, even if rational behaviour is accompanied by problems of incomplete information. The rest of the chapter is organized as follows. The section ‘The model’ describes the model and the set up for the game. Then we discuss the Perfect Bayesian Equilibrium (PBE) for a given total income and a given end-of-bargaining settlement. This section is a generalization of F&T to allow for any priors on the part of the less-informed country. The section ‘The full equilibrium’ close the model by making output dependent on the allocation of resources to production and the settlement after bargaining (i.e. following a conﬂict) dependent on the allocation of

Rational wars with incomplete information

115

resources to ﬁghting (as in Hirshleifer 1995b). Analysis in this section relies on numerical simulations. Finally, conclusions and suggestions for future work are provided.

The model Our model paper integrates a model of efﬁcient allocation of resources of Hirshleifer (1995b) into a model of bargaining with incomplete information by Fudenberg and Tirole (1983). It will set out to explain the foundations of conﬂict behaviour assuming rational income-maximizing agents. We consider two abstract players who have to divide a common scarce resource. We call those players countries 1 and 2, but it could be any other abstract players. We assume some change in the environment that breaks the previous equilibrium of forces. For example, assume that a new resource is found and it is arbitrarily distributed. Country 1 has found the resource and takes possession, country 2 thinks that there is some potential to make some economic gain and makes a claim for a new redistribution of resources. Since country 1 would be happy maintaining the status quo it has some strategic advantage. Therefore, we give country 1 the advantage of having the initiative (making the offers). Both countries are rational and before ﬁghting takes place they will try to ﬁnd a settlement to avoid the cost of war. Prior to the bargaining process, they have to decide in advance how to allocate their respective initial resources into production effort and ﬁghting effort. This allocation is irreversible until the discovery of new resources. After the allocation of resources, a new equilibrium takes place either by war or by settlement. Countries share the resources obtained in a common production process and the output is distributed according to this new situation.1 Each participant balances between productive exploitation of the current resource base and acquisition of the results of production. Correspondingly, there are two separate technologies: a technology of production and a technology of appropriation, conﬂict and struggle. Consider ﬁrst the technology of production. We adopt a common CES function and the technology of conﬂict follow exactly Hirshleifer production and CSFs: I = A(E1α + E2α )1/α = f (E1 , E2 ),

(8.1)

where I is the jointly generated output, A is a technological constant and E1 and E2 are the inputs of production. This CES production function has an elasticity of substitution = 1/(1 − α), and takes a variety of shapes depending on its value.2 In the case where α = 1, the production function is a simple linear function. As α approaches zero, tends to unity and we have a Cobb–Douglas production function. As α approaches −∞, it gradually transforms into a Leontief technology.

116

Paul Levine and Francisco Moraiz

The participants have an initial resource constraint given by R i = Fi + Ei ,

i = 1, 2,

(8.2)

for country i where Fi is devoted to ﬁghting and Ei to production. A distinctive feature of our set-up is the inclusion of costs of war that are quantiﬁed differently by different countries. For country 1, the cost of war in monetary terms is a proportion γ of output I and γ is common knowledge. Country 2 may be one of two types, a ‘tough’ country for whom the utility loss γ is relatively low, and a ‘soft’ country for whom the utility loss is γ where γ > γ . Country 1 does not know which type it faces and adopts priors υ0 and 1 − υ0 that country 2 is soft and tough, respectively. In a PBE, these probabilities are revised as bargaining proceeds in accordance with Bayes rule. The two types of country 2 will, in general, choose different allocations of total resources between production and ﬁghting effort. Let the soft and tough countries choose E 2 and E 2 , respectively. Total output is then given by the production functions I = f (E1 , E 2 ) and I = f (E1 , E 2 ) if country 2 turns out to be soft or tough, respectively, where f (E1 , E2 ) is given by Eq. (8.1). Now let us turn to the technology of appropriation, which takes the form of Hirshleifer’s CSFs. Let the probability of country 1 winning the war be denoted by p = g(F1 , F 2 ) and p = g(F1 , F 2 ), respectively, for the two types of country 2. Country 1 does not know for certain which type it is confronting and adopts an initial probability p = υ0 p + (1 − υ0 )p. The functional form g(F1 , F2 ) of the appropriation technology follows the ratio variant: p1 =

F1M

; M

F1M + F2

p2 =

F2M F1M + F2M

.

(8.3)

The parameter M is called the decisiveness coefﬁcient and affects the degree to which greater ﬁghting effort translates into battle success. The function captures how expenditure in ﬁghting effort (F1 , F2 ) translates into a certain probability of victory. For country 2, the probability of victory is given by 1 − p and 1 − p for the soft and tough, respectively. In sequencing of events, countries ﬁrst commit to the allocation of their factors of production to output or ﬁghting. This then determines the probabilities of winning a war for the two types of country 2 and for country 1 conditional on the type of country 2 it faces. Second, bargaining takes place. As we have mentioned, we assume that country 1 is the less-informed player and makes the offers. After two offers either agreement is reached or a war ensues which country 1 can win with an updated probability p. In detail, the game in extensive form is given by the following sequence of events: 1

Country 1 has prior υ0 that country 2 is soft. Countries simultaneously commit themselves to the resource allocation (Fi , Ei ), i = 1, 2, which determines expected output for country 1, E[I |υ0 ] = υ0 I + (1 − υ0 )I , actual output I and I for the two types of country 2, and probabilities p, p and p.3

Rational wars with incomplete information 2

3

4 5 6

7

117

It is convenient to consider offers in the form of shares of output given by I = I if country 2 is soft and I = I if country 2 is tough. At the ﬁrst stage of the bargaining process, country 1 claims an amount s1 I , which corresponds to an offer of (1−s1 )I to the soft type of country 2 and of I −s1 I to the hard type. Alternatively, we can write the claim as an amount s1 I , which corresponds to an offer of I − s1 I to the soft type of country 2 and of (1 − s1 )I to the hard type. (We conﬁne ourselves to pure strategies for country 1.) Country 2 accepts the offer with probabilities r 1 (s1 ) and r 1 (s1 ) for types γ (tough) and γ (soft), respectively. Acceptance ends the game with singleperiod payoffs s1 I for country 1 and (1 − s1 )I or I − s1 I for country 2 of types soft and tough, respectively. Otherwise we proceed to: Country 1 updates its probability from υ0 to υ1 . Country 1 claims an amount s2 I , which corresponds to an offer of (1 − s2 )I to the soft type of country 2 and of I − s2 I to the hard type. Country 2 accepts the offer with probabilities r 2 (s2 ) and r 2 (s2 ) for types γ and γ , respectively. Acceptance ends the game with single-period payoffs s2 I for country 1 and (1 − s2 )I or I − s2 I for country 2 of types soft and tough, respectively. Otherwise we proceed to: War ensues which type of country 1 wins with probability p = υ1 p + (1 − υ1 )p. Soft country 2 expects to win with probability 1 − p and the tough country with probability 1 − p. Expected single period payoffs are: (1 − γ )[υ1 pI + (1 − υ1 )p I ] for country 1, and (1 − p)(1 − γ )I and (1 − p)(1 − γ )I for tough and soft country 2, respectively.

The equilibrium for a given allocation between output and ﬁghting effort The full equilibrium is described by (F1 , F 2 , F 2 , s1 , r1 , s2 , r2 ) and depends upon cost of war parameters γ , γ , γ , the prior υ0 , discount factors δ1 , δ 2 , δ 2 and functional forms f (·) and g(·). In this section, we solve for the equilibrium of events 3–6 described above for a given allocation (F1 , F 2 , F 2 ), which then determines E[I |υ0 ] and the probabilities of winning for a given υ0 . This corresponds closely to the game with one-sided incomplete information described in F&T, except that we generalize the game to one with any prior υ0 . The PBE of this game is based on the following ﬁve elements: Range of possible offers. This is based on the solution to the complete information game. In bargaining period 2, a soft country is guaranteed of a payoff (1 − γ )(1 − p)I = (1 − s 2 )I say. It follows that any claim by country 1 s2 I ≤ s 2 I will be accepted by the soft country. Similarly, any claim s2 I ≤ s 2 I will be accepted by the tough country where s 2 = 1 − (1 − γ )(1 − p). Therefore, only claims in the interval [s 2 I , s 2 I ] or [s 2 I , s 2 I ], depending on whether s 2 I < s 2 I or s 2 I > s 2 I ,

118

Paul Levine and Francisco Moraiz

will be accepted by country 2 in bargaining period 2. In the rest of this section we assume that s 2 I < s 2 I ; that is, if the claim by country 1 is less if the country 2 is tough. This may, in fact, not be the case if F 2 > F 2 ; that is, if the soft country devotes more resources to ﬁghting activity. However, the details of the equilibrium are very similar in both cases. Appendix A considers the case where s 2 I > s 2 I . Under complete information, country 1 facing a tough adversary will only claim s 2 I and agree to a settlement without war if (1 − γ )pI < s 2 I .

(8.4)

Using s 2 = 1 − (1 − γ )(1 − p), condition (8.4) holds if p < 1, which, of course, is always the case. Similarly, country 1 facing a soft adversary will only claim s 2 I and agree to a settlement without war if p < 1, which again always holds. In bargaining period 1, soft country 2 will accept any offer: (1 − s 1 )I ≥ δ 2 (1 − s 2 )I . Thus, country 1 can claim up to s 1 I where s 1 = 1 − δ 2 (1 − s 2 ) if it confronts the soft country. Similarly, country 1 can claim up to s 1 I where s 1 = 1 − δ 2 (1 − s 2 ) if it confronts the tough country. As for period 1, we assume that s 1 I < s 1 I . Then, only claims in the interval [s 1 I , s 1 I ] will be considered by country 2 in bargaining period 1. By similar reasoning as before, under complete information country 1 will offer (1 − s 1 )I to the tough country and (1 − s 1 )I to the soft country, and the offer will be accepted. Tough, soft country 1 distinction. Consider a one-shot game where country 1 has priors υ0 that country 2 is soft. Deﬁne a ‘soft’ country 1 that prefers to make a low claim s 2 I with certain agreement to making an offer s 2 I that is only accepted by a soft country 2. The condition for this is s 2 I > υ0 s 2 I + (1 − υ0 )(1 − γ )pI .

(8.5)

Otherwise the country is ‘tough’. A threshold intermediate share s˜ . Deﬁne (1 − s˜ )I as the lowest offer that soft country 2 will accept in the ﬁrst period of bargaining when it expects a generous low claim s 2 I in the second period. If δ 2 is the discount factor of soft country 2 then s˜ is given by (1 − s˜ )I = δ 2 (I − s 2 I ).

(8.6)

Bayesian updating of υ0 to υ1 . By Bayes rule: υ1 = prob(γ = γ | s1 refused) = =

prob(s1 refused | γ = γ ) prob(γ = γ ) prob(s1 refused) (1 − r 1 (s1 ))υ0 . (1 − r 1 (s1 ))υ0 + (1 − r 1 (s1 ))(1 − υ0 ) (8.7)

Rational wars with incomplete information

119

A threshold probability r˜ . Deﬁne r˜ as the value of r 1 that makes a tough country 1 just indifferent between playing soft and tough in the second period when r 1 (s1 ) = 0. Then: υ1 = prob(γ = γ | s1 refused) =

(1 − r 1 (s1 ))υ0 (1 − r 1 (s1 ))υ0 + (1 − υ0 )

(8.8)

and s 2 I = υ1 s 2 I + (1 − υ1 )(1 − γ )p I .

(8.9)

Hence, from Eq. (8.5): r˜ =

υ0 − υ1 , υ0 (1 − υ1 )

(8.10)

where υ1 = Lemma 1 Proof

s 2 I − (1 − γ )pI (s 2 I − (1 − γ )pI )

(8.11)

.

If country 1 is tough then 0 ≤ r˜ ≤ 1.

Using Eqs (8.10) and (8.11) we have that:

r˜ =

υ0 s 1 I + (1 − υ0 )(1 − γ )pI − s 2 I υ0 (s 2 I − s 2 I )

.

(8.12)

Since s 2 I > s 2 I the denominator of Eq. (8.10) is positive. From the deﬁnition of a tough country 1 the numerator is also positive. Hence, r˜ > 0. After some algebraic manipulation of Eq. (8.10), we ﬁnd that r˜ < 1 iff: (1 − γ )pI < s 2 I ,

(8.13)

which is precisely the condition (8.4) described above. Proposition 1 Consider the strategies in Table 8.1. In a PBE, a soft country 1 chooses from strategies 1 or 2 depending on which strategy maximizes the expected payoff given in Table 8.2; a tough country 1 chooses from all three. If strategy 3 is chosen, then, if country 2 turns out to be tough, war ensues. Proof When Eq. (8.5) holds, country 1 is soft, and will play s 2 I in the last period. Recall that country 2 (tough) will refuse any offer in the ﬁrst period smaller than (1 − s 1 )I . If the ﬁrst period offer is rejected, country 1 will update its beliefs according to event 7. Since r 1 (s1 ) = 0, it implies that υ0 > υ1 . Knowing what country 1 will offer in the second period, country 2 will accept any offer in period 1 if s˜ is lower than s˜ = 1 −

δ 2 (I − s 2 I )

. (8.14) I Now, we assume that country 1 is tough. It will play any strategy in Table 8.1. A soft country will accept any offer bigger than (1 − s˜ ). If any offer smaller than

120

Paul Levine and Francisco Moraiz Table 8.1 Three strategies for country 1 Strategy

Period

Country 1

Country 2 (soft)

Country 2 (tough)

1

1 2

s1I —

r=1 —

r=1 —

2

1 2

s˜ I s2I

r=1 —

r=0 r=1

3

1 2

s1I s2I

r = r˜ r=1

r=0 r=0

Table 8.2 Expected payoffs for the three strategies Strategy

Country 1

Country 2 (soft)

Country 2 (tough)

1

s1I

I − s1I

(1 − s 1 )I

2

υ0 s˜ I + (1 − υ0 )δ1 s 2 I

(1 − s˜ )I

δ 2 (1 − s 2 )I

3

υ0 [˜r s 1 I + (1 − r˜ )δ1 rs 2 I ] +(1 − υ0 )(1 − γ )δ1 pI

r˜ (1 − s 1 )I + (1 − r˜ )δ 2 (1 − s 2 )I

δ 2 (1 − γ )(1 − p)I

Table 8.3 Possible equilibrium outcomes No war

Possible war

Strategy 1 Strategy 2

Acceptance in ﬁrst period Soft country accepts in the ﬁrst period and tough in the second Strategy 3 Soft country accepts sometimes in period 1, always in period 2 Tough country never accepts

(1 − s˜ ) is made by country 1 in period 1 and country soft would accept it with probability r 1 (s1 ) > r˜ , then – using Eqs (8.10) and (8.11) – it can be shown easily that s 2 I > υ2 s 2 I + (1 − υ2 )(1 − γ )pI . That is, country 1 will be soft in period 2, which is a contradiction, since country 2 (soft) will be better off refusing ﬁrst-period offer r 1 (s1 ) = 0. Simultaneously, if country 2 (soft) will accept an offer in period 1 with probability r 1 (s1 ) < r˜ , then country 1 will be tough in the second period and country 2 will be better off accepting offer in period 1. Thus, the equilibrium strategy for country 2 is to play r 1 (s1 ) = r˜ .4 Consequently, there is a unique perfect equilibrium in this game with three possible strategies that depend on the particular values of the parameters of this model. According to these parameters, country 1 can play any of the strategies in Table 8.2. The result of the choice of strategies by country 1 is summarized in Table 8.3.5

Rational wars with incomplete information

121

Table 8.4 Baseline parameter values υ0 = 0.5, M = 1, α = 1 Country 1

Country 2 (soft)

Country 2 (tough)

δ1 = 0.9 γ = 0.2 R1 = 100

δ 2 = 0.9 γ = 0.15 R 2 = 100

δ 2 = 0.9 γ = 0.1 R 2 = 100

Income distribution under different values of in CES function 24 22

Country 1 Country 2 (tough) Country 2 (soft)

Expected income

20 18 16 14 12 10 –1

–0.8 –0.6 –0.4 –0.2

0

0.2

0.4

0.6

0.8

1

Figure 8.1 The production function parameter α and the resulting outcome when investment in war is ﬁxed.

Numerical computation of the three strategies In order to analyse the equilibrium properties of the model, we ﬁrst carry out a numerical computation allowing for some variation of the key factors, whilst holding others constant. In this section, we hold (F1 , F 2 , F 2 ) constant thus keeping ﬁxed the probability of winning a potential war after disagreement in period 2. We choose F1 = F 2 = F 2 , thus setting the probability of either country winning at p = 1/2. Other parameters are as given in Table 8.4 unless stated otherwise. Figures 8.1 and 8.2 show graphs of expected payoffs given in Table 8.2 associated with changes in parameters α and υ0 from their baseline values. All the results are intuitive. First, a war equilibrium in our model might happen when country 1 ﬁnds optimal to play strategy 3. This is more likely to happen when country 1 thinks it is facing a soft country; that is, when υ0 is high. In Figure 8.1, we choose υ0 = 0.5

122

Paul Levine and Francisco Moraiz Income under different values of υ0 80

Expected income

70

60

50 Country 1 Country 2 (tough) Country 2 (soft)

40

30

20 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Initial priors υ0

Figure 8.2 The change in the initial priors υ and the resulting outcome when investment in war is ﬁxed.

and vary α in the range [−1, 1]. Then, the parameter values shown in strategy 1 are chosen and settlement always occurs in period 1. The parameter α is a measure of interdependence and, consequently, as it increases so does the payoffs for both countries. A more interesting result is obtained by varying the initial priors, see Figure 8.2. When country 1 believes that there is a probability of facing a soft country, υ0 > 0.8, then it uses strategy 3. Although, there is a general loss of income, given the high probabilities of facing a soft country, the expected income is increasing towards what would be the perfect information optimal offer. There is an increase in the income of country 1, since E[I |υ0 ] increases with the probability of facing a soft country only when it plays strategy 3. For country 2 (soft), there is a loss of income and a gain for country 2 (tough). Under imperfect information the resulting outcomes are more realistic than in some models of costly conﬂict with perfect information. Country 2 enjoys some strategic advantage from its private information and this diminishes the strategic advantage that country 1 has by making the offers. Another interesting result is that a high prior can produce a war outcome independently of the resource allocation process. Given certain values of υ0 , country 1 will ﬁnd optimal to play strategy 3. However, this threshold value depends also on the magnitudes of γ and γ as the next result shows.

Rational wars with incomplete information

123

In Figure 8.3, there are several graphs, all of them, with ﬁxed allocations to ﬁghting efforts, representing the relation between the informational problem measured as the difference perceived by country 1 about the cost of war of country 2, γ = γ − γ , and the prior υ0 . It produces two regions: the region ‘Peace’ corresponds to the use by country 1 of strategies 1 or 2 in Table 8.3. This gives a settlement without war. The region called ‘War’ is the result of country 1 playing strategy 3 in Table 8.3. The probability of war is conditional on country 1 choosing

0

0 Peace

Peace

War War

∆

Graph A F1 = 5, 50 F2 = 5, 50

∆

Graph D F1 = 50, 95 F2 = 5, 50

F2 = 50, 95

F2 = 5, 50

0

0 Peace

War War

∆

Graph B F1 = 5, 50 F2 = 50, 95

∆

Graph E F1 = 50, 95 F2 = 50, 95

F2 = 50, 95

F2 = 5, 50

0

0 Peace War War

Graph C F1 = 5, 50 F2 = 50, 95 F2 = 5, 50

∆

War

Graph F F1 = 50, 95 F2 = 5, 50

∆

F2 = 50, 95

Figure 8.3 In graphs A–F, the difference between γ and γ is located in the horizontal axis (cost of war γ ) and the prior belief about facing one of the two types of country 2, υ, in the vertical axis (probability of facing soft). Different combinations of υ and γ produce two scenarios (Peace and possible War). The investment in ﬁghting (F1 , F 2 , F 2 ) is ﬁxed.

124

Paul Levine and Francisco Moraiz

strategy 3. It means that country 1 prefers to face a situation of war with some probability rather than using strategy 1 that is always accepted in the ﬁrst round. In graphs A–C of Figure 8.3, country 1 has less ﬁghting power that country 2. In graphs D–F, country 1 has a higher ﬁghting power. In order to test the effect of arms races, the same simulation was carried out for low and high levels of arms expenditure. In each graph there should be two loci that plot the frontier between peace and war strategies. One for low ﬁghting effort and one for high. In graph A, there are two war areas. The one on the left corresponds to the scenario where s 2 I > s 2 I and the one at the bottom corner corresponds with the inverse situation. This occurs since we exogenously ﬁxed F 2 > F 2 . We also contrasted cases where country 1 is less likely to win the war than country 2, represented by the graphs on the left, and simulations where country 1 has a greater probability of winning the war, on the right-hand side of Figure 8.3. An overall increase in arms expenditure has a negative impact on peace for those cases where country 1 had a disadvantage in military power with respect of country 2. Alternatively, increases in ﬁghting expenditure has a positive effect when country 1 had a positive advantage in ﬁghting capability. These results are consistent with the empirical results about arms races and war. This seems to support the conventional conclusions about the effect of deterrence, risk uncertainty and ambiguous effect of arms races in political models of conﬂict.6 In Figures 8.4 and 8.5, the allocation of resources is ﬁxed as in Figures 8.1 and 8.2. In the ﬁrst case, there is a large difference in the perception of the

Income for country 1 and both types of country 2

Different cost for country 1, υ0 = 0.7 80 Country 1 Country 2 (tough) Country 2 (soft)

70 60 50 40 30 20 10 0 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

of country 2 (soft)

Figure 8.4 The effect of changes in the cost of a soft country 2 when investment in war is ﬁxed.

Rational wars with incomplete information

125

F1 = 50, F2H = 50, F2L= 25, = 0.2, H = 0.15, L = 0.1 80 Country 1 Country 2 (tough) Country 2 (soft)

75 70 Expected income

65 60 55 50 45 40 35 30 0

0.1

0.2

0.3

0.4 0.5 0.6 Initial priors υ0

0.7

0.8

0.9

1

Figure 8.5 The conﬂict roller-coaster. When the tough country invests signiﬁcantly less than the soft, the condition (1 − γ )(1 − p) > (1 − γ )(1 − p) does not hold. Consequently, country 1 plays strategies given in Appendix A. when 0 < υ0 < 0.4 country 1 uses strategy 2. From 0.4 < υ0 < 0.5 it uses strategy 1. From 0.5 < υ0 < 0.85 it uses strategy 3 and for υ0 > 0.85 it uses strategy 1 again. Under these circumstances, the probabilities of war increases when country 1 is not sure about the type of adversary it is facing.

externality cost that country 1 has about country 2 (γ = soft and γ = tough); in the second case, the externality cost of war is small for all types (γ = 0.2, γ = 0.15, γ = 0.1). In Figure 8.4, the probability of facing a soft country is relatively small for all types (υ0 = 0.3). Given some equal probabilities of winning the war, the effect of asymmetric information is quite signiﬁcant in this ﬁgure. For values of γ higher that 0.5, country 1 is better off playing strategy 3, which is reﬂected by the jump in the payoffs curves. Figure 8.5 is a special case. This ﬁgure shows how unstable the equilibrium can be when investment is ﬁxed and the soft country has a higher expenditure in arms than the tough one. Fixing allocations to ﬁghting capabilities can have a roller-coaster effect in conﬂict. When countries can allocate resources optimally, the soft country 2 never spends more resources in ﬁghting capabilities, producing a greater system stability. The war probabilities in Figure 8.5 are more complex. Although it has the same probabilities of winning the war than soft country 2, the fact that it faces a tough country (low cost of war) with a military disadvantage creates many opportunities for conﬂict.

126

Paul Levine and Francisco Moraiz

The full equilibrium The previous section studied out-of-equilibrium strategies with respect to the initial allocation of resources. The next step is to decide what would be the optimal allocation of resources given the nature of the game. One important remark is that in our model we assume a tie-breaking rule in strategy 3 by which country 2 (soft) plays r(s1 ) = r˜ . Otherwise, we will have scenarios where participants update their beliefs in an arbitrary fashion that causes multiple equilibria. Therefore, the possibility of ﬁnding the optimal allocation of resources in this framework should be ruled out. This could be tackled by introducing some equilibrium reﬁnements, which would be more interesting in a two-sided asymmetric information framework. In order to ﬁnd out the optimal strategy, we form a three-player Nash game for any of the three strategies. The Nash equilibrium in initial allocations Consider ﬁrst the case where country 1 adopts strategy 1. Then, country 1 chooses F1 according to max s 1 I

wrt F1

s.t. R1 = F1 + E1 .

(8.15)

Country 2 (soft) chooses F 2 according to max (I − s 1 I )

s.t. R2 = F 2 + E 2

wrt F 2

(8.16)

and country 2 (tough) chooses F 2 according to max (1 − s 1 )I

wrt F 2

s.t. R2 = F 2 + E 2 .

(8.17)

We substitute the linear constraint into the objective function and take the derivatives with respect to F1 , F 2 and F 2 . This maximization problem produces three reaction curves that can be solved simultaneously in order to ﬁnd the Nash equilibrium: RC1:

∂I ∂s 1 I+ s = 0, ∂F1 ∂F1 1 ∂I

∂s 1

I−

∂I

s 1 = 0, ∂F 2 ∂F 2 ∂F 2 ∂(1 − s 1 ) ∂I RC1 (tough): I+ (1 − s 1 ) = 0. ∂F 2 ∂F 2

RC2 (soft):

−

(8.18) (8.19) (8.20)

For the other strategies, every player maximizes its payoff according to Table 8.2. It also produces three reaction curves in each strategy that every country solves simultaneously. Then, country 1 compares the outcomes of these three strategies and chooses the one that has the highest payoff.

Rational wars with incomplete information

127

Numerical results Once country 1 knows what the equilibrium allocation of resources will be, it can play the game in the same manner that we presented in the section on the equilibrium between output and ﬁghting effort. We developed this game in a Matlab algorithm that ﬁnds the unique equilibrium for any parameter in the game. In Figures 8.6–8.8, we present the outcomes of this simulation. In Figure 8.6, the externality cost for country 1 was γ = 0.7 and for country 2 (soft and tough) γ = 0.3 and γ = 0.2. Strategy 3 was played for υ1 > 0.9, which gives a very low probability of war. When the values of the cost of ﬁghting were low (γ = 0.2, γ = 0.15, γ = 0.1), the probability of war nearly disappears and tends to the perfect information solution (see Figure 8.7). The difference between the costs of countries 1 and 2 affects the expected income, but not the probability of war. The highest probability of war occurred when we introduced high difference between γ and γ . The rest of the parameters in the model took the values as in the Table 8.5.

The effect of changes in prior beliefs 85 Country 1 Country 2 (tough) Country 2 (soft)

80

Expected income

75 70 65 60 55 50 45 40

0

0.1

0.2

0.3

0.4 0.5 0.6 Priors 0

0.7

0.8

0.9

1

Figure 8.6 The payoffs effect of changes in υ0 when investment in war is optimal. In this ﬁgure, the difference between γ and γ is not high (γ = 0.7, γ = 0.3, γ = 0.2) and the cost of conﬂict for all players is very low. This produces war with a small probability represented by the jump in the three curves for values of υ0 > 0.9. Naturally, the three curves are horizontal as long as country 1 does not choose strategy 3, which is the only one that depends on υ0 .

Income for country 1 and both types of country 2

= 0.7, L = 0.3, H = 0.2 and different values of 0 in CSF 64 62

Country 1 Country 2 (tough) Country 2 (soft)

60 58 56 54 52 50 48 46 44

0

0.1

0.2

0.3

0.4 0.5 0.6 Priors υ0

0.7

0.8

0.9

1

Figure 8.7 The payoffs effect of changes in υ0 when investment in war is optimal. In this ﬁgure, although the cost of war was very low for all players (γ = 0.2, γ = 0.15, γ = 0.1), country 1 always chooses strategy 1, which produces peace. Compared to Figure 8.6, we can see that asymmetric cost between countries 1 and 2 has little effect in the choice of strategy (but affects the expected income). The effect of changes in prior beliefs 110 Country 1 Country 2 (tough) Country 2 (soft)

100

Expected income

90 80 70 60 50 40 30

0

0.1

0.2

0.3

0.4 0.5 Priors υ0

0.6

0.7

0.8

0.9

1

Figure 8.8 The payoffs effect of changes in υ0 when investment in war is optimal. When the difference between γ and γ is very large (γ = 0.7, γ = 0.6, γ = 0.1), the probability of war increases drastically. Unless we have strong beliefs that a country is soft, we cannot get a war scenario which in this model is represented by the decision of country 1 to use strategy 3 in Table 8.2.

Rational wars with incomplete information

129

Table 8.5 Values for the baseline calibration of optimal allocations α = 1, M = 1 Country 1

Country 2 (soft)

Country 2 (tough)

δ = 0.9 γ = 0.7 R1 = 100

δ 2 = 0.9 γ = 0.3 R 2 = 100

δ 2 = 0.9 γ = 0.2 R 2 = 100

The allocation of resources to arms

Fighting effort percentage of resources

0.7 Country 1 Country 2 (tough) Country 2 (soft)

0.6 0.5 0.4 0.3 0.2 0.1 0

0

0.1

0.2

0.3

0.4

0.5 0.6 Priors υ

0.7

0.8

0.9

1

Figure 8.9 Resources dedicated to ﬁght and prior beliefs.

These results indicate that priors and cost of war work in a complementary manner. In a model of one-sided incomplete information war, outcomes are the result of a combination of high probabilities of facing a soft country with large differences between the cost faced by soft and tough participants. Finally, unless country 1 decides to follow strategy 3, the expected income is constant in relation the priors υ0 . The point where the income curve jumps for all the countries in Figures 8.7 and 8.8 corresponds to a change from strategy 1 to 3.7 Finally we show in Figures 8.9 and 8.10 the ratio of resources dedicated to ﬁghting effort under optimal allocations. In Figure 8.9, we plot the ratio of resources against the probability of facing a soft country υ0 . When this probability is larger than 0.7 country 1 plays tough.

130

Paul Levine and Francisco Moraiz The allocation of resources to arms with cost asymmetry 55

Allocations to fighting effort

50 45 40 35 30 25 20

Country 1 Country 2 (tough) Country 2 (soft)

15 10

0

0.1

0.2 0.3 0.4 0.5 Difference in the cost of war, ∆

0.6

0.7

Figure 8.10 Resources dedicated to ﬁght and cost asymmetry.

Paradoxically, despite facing some probability of war, it reduces its military expenditure. This is due to the change of strategy. In this scenario, the probability of facing a strong adversary is so low that country 1 decides to take a gamble, plays tough and reduces its expenditure in arms at the same time. Thus, both types of country 2 are forced to increase the expenditure in arms to compensate for the risky behaviour of country 1. Figure 8.10 is also interesting. In most of our simulations, the expenditure dedicated to ﬁghting effort experiences little changes with changes in other parameters. However, differences in the cost of war have a great impact on the optimal allocation of resources. This is also the case in previous models with perfect information where the cost of ﬁghting is one of the main factors in determining optimal allocations. In this graph, the probability of facing a soft country is υ0 = 0.5. The cost of ﬁghting increases with the asymetric cost. This is due to some constrains that we put previously in the model.8 Therefore, Figure 8.10 shows that the tougher the adversary, the more one should spend in ﬁghting effort. Figures 8.11–8.13 represent the relation between war (or peace) and the informational problem – the difference perceived by country 1 about country 2’s cost of war, γ = γ − γ – and the initial probability (υ0 ) of facing one type or the other. It produces two regions. As in Figure 8.3, the region Peace corresponds with the use of strategy 1 or 2 in Table 8.3 by country 1 and War is conditional on country 1 playing strategy 3 in Table 8.3. For instance, in Figure 8.11, when γ is between 0.3 and 0.4, the conditional probability of war equals 0.2. (The probability of

Rational wars with incomplete information

131

The war frontier Probability of facing a tough country (1 – 0)

1 0.8 Peace

0.6 0.4 0.2

War

0 0.1

0.5 Difference in the cost of war, ∆ = –

0.9

Figure 8.11 This graph is similar to the graphs in Figure 8.3. But we have used optimal allocations. The probability of using strategy 3 is positively related with increases in γ and υ. We can see that the probability of war is positively related to the probabilities of facing a soft country and to the magnitude of the information asymmetry. M = 1, α = 1, R1 = R2 = 100 Country 1

Country 2 (soft)

Country 2 (tough)

δ = 0.9

δ = 0.9

δ = 0.9

war was never bigger than 0.5 per cent when participants allocate their resources optimally.) The peace area in the graph always produces a peaceful outcome. However, the war area should be read carefully. For any combination of γ and υ0 that falls in the area, there is a probability of war, which will take place only if country 2’s true type is tough p = (1 − υ). In Figure 8.12, the simulation was carried out with very low values of the discounting factor for tough and soft participants. After an initial increase in the probability of war, it decreases for values of γ > 0.5 in the horizontal axe. The reason is that since the discounting factor is very low, facing the probability of war makes strategy 1 more attractive in relative terms because both countries are very eager to reach an agreement in period 1. Remember that in order to face a war, we must wait for period 2. So, with high probabilities of facing a soft country 2, υ0 > 0.4, waiting becomes very expensive. But there is a point where the beneﬁts obtained from the discounting factor outweigh the beneﬁts of exploiting the difference in the cost of war, producing a substitution effect in strategies. In Figure 8.13, we have three areas. In this occasion, the three strategies in Table 8.2 are played by country 1. When country 1 believes that there is less than 0.6 probability of facing a soft country, it plays strategy 2. This only happens

132

Paul Levine and Francisco Moraiz

Probability of facing a tough country

(1 – υ0) The war frontier 1 0.8 Peace

0.6 0.4 0.2

War

0 0.1

0.9

0.5 Difference in the cost of war, ∆ = –

Figure 8.12 The contradictory effects of the discounting factor is shown. The values of the discounting factors were set to an extremely low level, δ ≈ δ ≈ δ ≈ 0.2. There is a point where the effect of impatience is more important than the possible gains of war. The more country 1 is likely to face a tough country, the more likely it is to play strategy 1.

Probability of facing a tough country

(1 – υ0)

The war frontier 1

0.8

Peace

0.6 0.4

Peace (with agreement in second period by tough type)

0.2 War 0 0.1

0.5

0.9

Difference in the cost of war, ∆ = –

Figure 8.13 After several simulations, we found out that according to our model, strategy 2 is rarely played. For that to take place, we had to give low values to δ = 0.05 and high values to δ = 0.95.

when the difference between δ and δ is considerably large. For situations where a tough and soft country have the same degree of impatience, country 1 never plays strategy 2. Although we cannot generalize, we can conclude from our simulations that: ﬁrst, in those situations where country 1 has a strategic advantage in both negotiation and

Rational wars with incomplete information

133

military terms, a peaceful settlement is more likely to occur – that is to say, when the status quo is not challenged we have more probabilities of seeing a peaceful outcome. Second, an optimal allocation of resources is also optimal in producing peaceful outcomes. There is an abundance of evidence in the literature of conﬂict showing that leaders are subject to limited information, which produces misconception and bias in the choice of policy. Uncertainty about the cost of war can bring together some aspects from the rationalist explanations of war and alternative theories that argue that leaders are sometimes irrational. Asymmetric information about the cost of war could produce a pattern of behaviour from leaders that, although being rational, could act as if they would neglect the costs of war or enjoy the beneﬁts without paying the costs. Optimal allocation of resources does not guarantee a more peaceful situation. However, since changes in some of the factors can produce ambiguous results, it seems more obvious, from the perspective of the public choice, that interventions directed to reduce the informational gap are likely to have a better result than those directed to control the allocation of resources into ﬁghting activities.

Conclusions and developments This chapter presents a model of conﬂict that differs from other models of war in that we combine incomplete information with an endogenous probability of winning the war, which for most part of the literature of bargaining over conﬂict is exogenous. We allow participants to establish a negotiation process over the possible consequences of endogenous allocations to production and ﬁghting efforts, absent in the old literature of conﬂict. Participants always try to avoid costly conﬂict that takes place in some extreme circumstances. The main conclusions of the model are: •

•

The likelihood of war with endogenous optimal allocation of resources were signiﬁcantly lower than the probability of a war under a ﬁxed initial allocation of resources. This indicates that the optimal prior allocation of ﬁghting and productive activity on the negotiation process has a more signiﬁcant impact in the outcome of war than other factors traditionally considered more important such as the military strategic advantage or the probabilities of victory in contest. We have experienced variations in all exogenous parameters. The parameters that have the most inﬂuence are (a) prior beliefs, υ0 and (b) the difference between the externality cost of the soft and tough countries (γ − γ ). This difference moves the break point of strategy 1 to 3 for country 1 to a lower value of υ0 , increasing the difference between the ﬁnal payoffs of countries 1 and 2 (soft and tough). The war decisiveness parameter M and the degree of integration α affect the expected income for both countries but have a small role in determining the outbreak of war.

134 •

Paul Levine and Francisco Moraiz In this model, the probabilities of winning the war and the size of the ﬁnal outcomes are endogenous. Although war is not an efﬁcient outcome, it is sometimes a PBE due to the effect of asymmetric beliefs about the cost of ﬁghting. We conclude that if the ﬁrst casualty of war is truth we should add that the ultimate victim of asymmetric information is peace.

A number of directions for future research is suggested by our results. The obvious question is what happens when both countries have incomplete information. F&T found that not strictly separating equilibrium can exist when the two participants are uniformed. Bayes rule places no restriction on the posteriors of country 2. The choice of conjectures and additional restrictions may lead to pooling or separating equilibria. Separating equilibria work as a restriction to the optimal allocation of resources, which will provide a higher probability of war outcomes. It would be also interesting to study the effect of long-term war. Our model assumes a once-for-all war. There are some models where participants can decides in each period if they want to take part in war or not. This would allow us to design new conjectures and signalling mechanisms, increasing the probability of war in the short run but reducing the cost of war in the long run.

Acknowledgement The authors are grateful to the ESRC for ﬁnancial support under grant R0002335685.

Appendix A Alternative payoffs Payoffs when country 2 soft plays tough Table 8.A1 Three strategies for country 1 Strategy

Period

Country 1

Country 2 (soft)

Country 2 (tough)

1

1 2 1

s1I — s˜ I

r=1 — r=0

r=1 — r=1

2

s2I

r=1

—

1 2

s1I s2I

r=0 r=0

r = r˜ r=1

2 3

Rational wars with incomplete information

135

Table 8.A2 Expected payoffs for the three strategies Strategy Country 1

Country 2 (soft)

Country 2 (tough)

1

s1I

(1 − s 1 )I

I − s1I

2

(1 − υ0 )˜s I + υ0 δ1 s 2 I

δ 2 (1 − s 2 )I

(1 − s˜ )I

3

(1 − υ0 )[˜r s 1 I + (1 − r˜ )δ1 s 2 I ] + υ0 (1 − γ )δ1 p)I

δ 2 (1 − γ )(1 − p)I

r˜ (1 − s 1 )I + (1−˜r )δ 2 (1−s 2 )I

Notes 1 Countries at war do not normally trade with each other. Some authors have studied the relation between trade and war. They found that countries with very high levels of trade do not normally make war, but once the war takes place it is normally longer and more devastating. For a review of those issues see Gowa (1994) and Mansﬁeld (1994). 2 See Varian (1992). 3 F 2 and F 2 are not observed by country 1. 4 For country 1, in order to be indifferent between playing tough or soft in second period, country 2 must play a mixed strategy. It must also be indifferent between accepting and refusing the ﬁrst-period offer. Therefore, the ﬁrst-period offer of country 1 must satisfy this condition. The probability of making a tough offer in the second period must be σ2 (1) =

5 6

7 8

(1 − s 2 )I − (1 − s 2 )I s2I

,

which makes country 2 (soft) indifferent between the payoff in period 1 and the expected outcome of period 2. According to Rasmunsen (1989), strategy 3 in this model has one equilibrium offer but multiple equilibria of acceptance probabilities. This is due to the fact that country 2 must mix between accepting and rejecting and country 1 must also mix between playing tough and soft in the second period. We assume that country soft will accept when r 1 (s1) = r˜ without ruling out the multiple equilibria because at this value is the largest probability of country 2 accepting immediately and avoiding the loss of utility in delay. In our simulations we considered the case s 2 I > s 2 I and/or s 1 I > s 1 I , which occurs when the soft country 2 allocates much higher resources to ﬁghting than the tough country 2. The table of payoffs for this case is given in Appendix A. However, we must be careful with the fact that our model chooses country 1 and the private information of country 2 in an arbitrary manner. So far there is no consistent empirical study of the inﬂuence of asymmetric information about the cost of war. Regarding the effects of arms races and war that provide similar stylized facts to the ones found by our simulation, Huth et al. (1992, 1993) and Sample (1997). High values of δ ruled out strategy 2. γ ≥ γ ≥ γ . We have already shown that for the case where γ ≤ γ ≤ γ country 1 does not need to care about asymmetric information and plays as if it had perfect information regarding only its own cost.

136

Paul Levine and Francisco Moraiz

References Bhagwati, J. N. (1982) Directly unproductive, proﬁt-seeking (dup) activities. Journal of Political Economy 90(5), 989–1002. Brito, D. L. and Intriligator, M. D. (1985) Conﬂict, war, and redistribution. The American Political Science Review 79, 943–957. Dixit, A. (1987) Strategic behaviour in contests. American Economic Review 77, 891–898. Fudenberg, D. and Tirole, J. (1983) Sequential bargaining with incomplete information. Review of Economic Studies L, 221–247. Gowa, J. (1994) Allies, Adversaries, and International Trade. Princeton University Press, Princeton, NJ. Hirshleifer, J. (1989) Conﬂict and rent-seeking success functions: ratio vs. difference models of relative success. Public Choice 63, 101–112. Hirshleifer, J. (1995a) Anarchy and its breakdown. Journal of Political Economy 103, 26–52. Hirshleifer, J. (1995b) Theorizing about conﬂict. In: Hartley, K. and Sandler, T. (eds), The Handbook of Defense Economics. Elsevier, Amsterdam, pp. 165–190. Huth, P., Gelpi, C. and Scott Bennett, D. (1992) System uncertainty, risk propensity, and international conﬂict among the great powers. Journal of Conﬂict Resolution 3, 478–517. Huth, P., Gelpi, C. and Scott Bennett, D. (1993) The escalation of great power militarized disputes: testing rational deterrence theory and structural realism. American Political Science Review 87, 609–623. Intriligator, M. D. (1982) Research on conﬂict theory: analytic approaches and areas of application. Journal of Conﬂict Resolution 26, 307–327. Krueger, A. O. (1974) The political economy of the rent-seeking society. American Economic Review 64, 291–303. Mansﬁeld, E. D. (1994) Power, Trade, and War. Princeton University Press, Princeton, NJ. Moraiz, F. (1998) A model of bargaining and conﬂict. PhD thesis, Department of Economics, University of Surrey. Rasmusen, E. (1989) Games and Information. Blackwell, Oxford, Chapter 10.5, pp. 239–241. Sample, S. (1997) Arms races and dispute escalation: resolving the debate. Journal of Peace Research 34, 7–22. Stergios, S. (1992) Cooperation, conﬂict, and power in the absence of porperty rights. American Economic Review 82, 720–739. Tullock, G. (1980) Efﬁcient rent-seeking. In: Tollison, R. D., Buchanan, J. M. and Tullock, G. (eds), Toward a Theory of the Rent-seeking Society. Texas A&M University Press, College Station. Varian, H. R. (1992) Microeconomic Analysis. Norton.

9

Alliance formation, expansion and the core Todd Sandler

The recent admittance of the Czech Republic, Hungary, and Poland into NATO in March 1999 prior to its ﬁftieth anniversary raises questions as to how many other Partnership for Peace (PFP) nations will eventually become members. In particular, will future entrants include the Baltic states (Estonia, Latvia, and Lithuania), Slovakia, Slovenia, Austria, Switzerland, or Romania? Thus far, the debate about expansion has centered on the associated costs, with estimates ranging from $2 billion (NATO 1995) to $125 billion, depending on alternative assumptions, time horizons, and expansion costs deﬁnitions.1 Although beneﬁts from expansion are brieﬂy mentioned, the current focus on costs implicitly assumes that expansion beneﬁts always outweigh expansion costs and that these beneﬁts do not differ among alternative expansion scenarios. Neither of these implicit assumptions is defendable. The expansion issue is related to alliance formation (i.e. an expansion from zero allies), which hinges on whether prospective allies view their membership as providing a net gain after associated costs are covered. These net gains are dependent on alliance size and composition, which can affect beneﬁts (e.g. cost reductions, enhanced deterrence) and costs (e.g. decision making, joint maneuvers). The reigning model in the alliance literature is the public good model, originally formulated by Olson and Zeckhauser (1966) and later extended by van Ypersele de Strihou (1967), Sandler and Cauley (1975), Sandler (1977), Murdoch and Sandler (1982), McGuire (1990), and others to impure public goods, joint products, and other considerations. Although this non-cooperative game-theoretic model is suited to examining burden sharing, defense demands, and allocative efﬁciency, the model is ill-suited for investigating the distribution of net beneﬁts, alliance formation, expansions, or stability (i.e. is alliance membership stable). In fact, the pure public representation of alliances implies that all willing nations should be included, because beneﬁts per existing ally remains unchanged while cost per ally falls owing to cost sharing when new allies are admitted (Sandler and Hartley 1999). A club theory formulation of alliances can address membership size based on the thinning of forces and cost-sharing considerations when conventional weapons protect borders (Sandler 1977; Sandler and Forbes 1980; Murdoch and Sandler 1982). Club theory cannot, however, readily investigate alliance stability and

138

Todd Sandler

expansion while accounting for locational considerations. To address these issues along with alliance formation, we put forward an alternative paradigm based on cooperative game theory. In particular, a mutual defense game is presented for which the motivation for forming an alliance is cost savings as borders become interior and no longer need protecting (Gardner 1995). These interior borders then provide a beneﬁt-based representation of alliance formation and expansion where spatial considerations – allies’ location and size – become crucial. Other factors (e.g. transaction costs, natural defenses, deterrence) are appended to the basic model to increase realism and foster further insights. The absence of a need to guard interior borders implicitly assumes that nations ally themselves with friendly nations, from which an attack will not be threatened. While there are some notably exceptions to this rule (e.g. Germany and Russia in 1941, Greece and Turkey over Cyprus in 1974), modern-day allies are often drawn from nations sharing economic linkages, democratic principles, and membership in other international organizations (e.g. ANZUS, Japan–US alliance). These factors have been shown in a recent study to inhibit military conﬂict (Russett et al. 1998). Furthermore, these authors established that alliance membership, economic interdependence, and democratic values help explain membership in international organizations, which serve a problem-solving function. Economic interdependence, for example, implies that conﬂict among dependent nations would result in signiﬁcant mutual economic costs. For the Cyprus dispute, the membership of Greece and Turkey in both NATO and the United Nations provided a mechanism for ending the hostilities and separating the forces. Except for the Cyprus incident, NATO has included friendly nations sharing democratic values, economic interests, and membership in a nexus of international organizations (e.g. the European Union, PFP, Organization of Security and Cooperation in Europe). A number of NATO’s missions are to promote this economic interdependence and the network of overlapping organizations (Sandler and Hartley 1999: chapter 7). The requirements for PFP membership, a precursor to NATO membership, mandate that entrants possess democratic principles and do not have territorial disputes with neighbors (Partnership for Peace 1996). The primary purpose of the chapter is to present a simple cooperative game theory representation of alliance formation and expansion when allies seek mutual defense from a conventional threat in all directions along their perimeters. This theory represents an alternative paradigm for viewing alliances. In particular, the core solutions, if they exist, are identiﬁed for alternative scenarios of mutual defense involving diverse spatial conﬁgurations, warfare scenarios, and alliance sizes. Second, the paradigm is applied to illuminate the NATO alliance and the likely outcome of its expansion. Third, the analysis is extended to include numerous additional considerations. The study demonstrates the importance of spatial and locational considerations when determining the bargaining strength of the allies and the resulting distribution of the net gains from allying. Geographical conﬁgurations matters: the same number of allies can have vastly different cooperative equilibria (i.e. cores) depending on how they are arranged. In fact, the classic prediction that the core

Alliance formation, expansion and the core

139

shrinks when the number of participants increases may not apply for some spatial conﬁgurations of allies. The type of threat is also an important factor; nations facing insurgencies offer little if any cost saving from mutual defense and make for bad allies. Transaction costs considerations can have a profound effect on the distribution of alliance gains. The remainder of the chapter contains seven primary sections. The ﬁrst puts forward theoretical preliminaries and deﬁnitions associated with a cooperative game representation. In the second section, a number of three-country mutual defense scenarios are analyzed; in the third section, four- and ﬁve-country alliances are investigated. The effects of deterrence, as derived from a strategic arsenal, are brieﬂy presented in the fourth section. Insights gleaned from the analysis are then applied to NATO in the ﬁfth section. Other considerations are examined in the next section, followed by concluding remarks.

Preliminaries A few preliminary deﬁnitions and concepts are required to understand the bargaining game of mutual defense.2 Agreements in bargaining games must be unanimous so that each participant has a veto. This means that an agreement must provide all participants with as much or more than they can obtain on their own. If an agreement is not consummated among bargaining parties, then any dissatisﬁed party can leave the talks with its disagreement payoff, equal to what a potential participant(s) can earn on its (their) own. When parties in a negotiation have unequal disagreement payoffs, the party with the largest such payoff or option is at a bargaining advantage and can be anticipated to gain relatively more from an agreement, if reached. Thus, if a potential ally’s strategic location is responsible for much of the gains from a proposed alliance, then that ally is anticipated to get a larger share of these gains from such an alliance. A coalition is any subset of a population N = {1, 2, . . . , n} of players or countries. Three kinds of coalitions are germane to the model: (a) single-player or singleton coalitions, {i}, of just player i; (b) a grand coalition of all n players, denoted by N; and (c) intermediate coalitions, S, of 2 to n − 1 players. An intermediate coalition is a proper subset of the population, so that S ⊆ N. The power set denotes all coalitions or subsets of the relevant population, including the null set. In the case of population N, the power set minus the null set contains 2n − 1 potential coalitions, which can form and are relevant for determining the stability of an agreement among a coalition of players. The characteristic or coalition function indicates the sum of net payoffs to the set of players in any coalition of the population. For coalition S, the characteristic function, denoted by v(S), represents the “security” payoff to the group if it were to leave the negotiations and strike out on its own. Thus, any approved agreement must guarantee a sum of payoffs to participants for all potential coalitions that is at least as good as their coalition values, v(S). Based on the characteristic function, the grand coalition has value v(N), while the singleton sets have value v({i}) for i = 1, 2, . . . , n.

140

Todd Sandler

The net payoffs to any set of s individuals in coalition S from accepting a proposal in a negotiation is vector u = (u1 , u2 , . . . , us ), where ui is the agreement payoff for individual i. The utility levels in this vector are assumed to be transferable among people. Transferable utility means that monetary side payments between players are permissible and that the players’ utilities (i.e. payoffs) are essentially linear for the range of possible payoffs (Luce and Raiffa 1957). Allies often engage in side payments to one another (e.g. German and Japanese contributions to Desert Storm, or US transfers of nuclear technology to the United Kingdom and France), which support the transferable utility assumption underlying the analysis. Any positive afﬁne transformation, such that ui = aui + b for a > 0 for all i, does not affect the strategic features of an underlying bargaining game owing to linear invariance (Luce and Raiffa 1957: 186–187; Binmore 1992; Gardner 1995). Such transformations merely change the origin and the unit of measurement of the payoffs. Suppose that a proposal is presented to the grand coalition. If such a proposal is individually rational to accept, then it must satisfy the following inequalities: v({i}) ≤ ui

(9.1)

for all i in N. That is, each individual must be better off or at least no worse off by its payoff from the proposal relative to going it alone as a singleton set. Individual rationality ensures that there are no regrets from the agreement. Payoffs associated with an acceptable proposal must fulﬁll individual rationality and Pareto optimality: v(N) =

n

ui ,

(9.2)

i=1

so that the grand coalition offers the population the same payoff as the aggregate payoff of any proposal involving everyone. This Pareto optimality requirement implies group rationality where the entire group is able to match any payoff of the coalition of the whole. Payoff vectors satisfying Eqs (9.1) and (9.2) are called an imputation. For purposes here, only those imputations in the core are of interest. A proposal is in the core if the following 2n − 1 inequalities are satisﬁed: ui for S ⊆ N. (9.3) v(S) ≤ i∈S

Since S can be the singleton sets as well as the grand coalition, the inequalities in relation (9.3) satisﬁes Eqs (9.1) and (9.2) and coalition rationality, for which no intermediate coalition in the power set of non-empty sets can block the proposal by doing better when leaving the negotiations. Quite simply, the core ensures that no coalition can form and do better than what it gets in aggregate payoffs from the agreement. Consequently, no coalition will block the core agreement when the inequalities in relation (9.3) are satisﬁed. A core to a game may be empty as is the

Alliance formation, expansion and the core

141

case of the majority rule game for which voters try to divide a ﬁxed prize among themselves. Every majority coalition can be blocked, so that a stable coalition never forms (Gardner 1995: 400–401). The mutual defense game will, however, often have a set of solutions in the core.

Mutual defense: three countries Nations are initially assumed to confront a threat to their perimeters or borders that can come from any direction. The focus is on threats involving conventional weapons, because cost saving from sequestrating interior borders applies to these weapons and not to strategic weapons. The impact of strategic weapons on the analysis is taken up after the basic models are presented. A baseline case, taken from Gardner (1995), serves as a benchmark for comparison with myriad novel cases put forward in this chapter. The cost of defending a country from an external threat is assumed to depend on the length of its perimeter. Even if a country stations its mobile troops in the interior and deploys them to its borders only when an attack is imminent, the cost of defending the country is still proportional to the perimeter of the area guarded. This follows because the greater this perimeter, the more forces that must be deployed to its border to provide adequate protection during crises. A country the size of France needs more forces for security than, say, Belgium no matter where the forces are stationed during peaceful times. Thus, the assumption that defense costs against conventional attacks are related to perimeter length is appropriate. Baseline case Suppose that three contiguous countries are contemplating a mutual defense pact where each country is a unit square with four 1-in. borders to protect from an external threat in all directions. Further suppose that the three countries are conﬁgured as pact a in Figure 9.1, so as to form a 3 in. × 1 in. rectangle. The cost of protecting a country or alliance against a conventional attack from abroad is assumed proportional to the defended perimeter. In particular, each side of the countries depicted is assumed to cost 1 to defend, so that v({i}) = −4 is the cost of protecting nation i on its own. Similarly, v(S) represents the cost of protecting an alliance of countries in S. For the three-country region, there are seven nonempty possible coalitions, including the three singleton sets, that could form. To be in the core, the payoff u = (u1 , u2 , u3 ) for the three-country mutual defense pact must satisfy the following inequalities: v({i}) = −4 ≤ ui

i = 1, 2, 3

v({1, 2}) = −6 ≤ u1 + u2 v({2, 3}) = −6 ≤ u2 + u3 v({1, 3}) = −8 ≤ u1 + u3 v({1, 2, 3}) = −8 = u1 + u2 + u3 .

(9.4)

142

Todd Sandler

1

2

3

1

(a) Baseline model

1

2

3

2

3

(b) Shortened interior ally

1

3

2

3

(d) Non-contiguous ally

(c) Lengthened exterior ally

1

2

4

1

2

4

3

(e) Four-ally rectangle

1

(f) Four-ally square

2 3

5 4

1

(g) Five-ally pentagon

4

3

2

5 (h) Five-ally cross

Figure 9.1 Alternative alliance conﬁgurations.

It is easier to interpret the requirements for the core if the ui payoffs are transformed so that the disagreement point has payoffs of (0, 0, 0) as all nations go their separate ways. The positive afﬁne transformation that accomplishes this task, while maintaining the maximum cost savings at 4, is the transformation ui = ui + 4, i = 1, 2, 3. When this transformation is applied to the righthand inequalities in system (9.4), there is no change to the strategic structure. As a consequence, the core now must satisfy the following requirements: v({i}) = 0 ≤ ui

i = 1, 2, 3

v({1, 2}) = 2 ≤ u1 + u2 v({2, 3}) = 2 ≤ u2 + u3 v({1, 3}) = 0 ≤ v({1, 2, 3}) = 4 =

u1 u1

+ u3 + u2

(9.5) + u3 .

Alliance formation, expansion and the core

143

(0, 0, 4)

u⬘2 + u⬘3 = 2

(0, 2, 2)

(2, 0, 2)

u⬘1 + u⬘2 = 2

Core (1, 2, 1) Centroid (0, 4, 0)

(2, 2, 0)

(4, 0, 0)

Figure 9.2 Mutual defense, baseline case.

The coalition function in system (9.5) indicates that alliances containing one or more interior borders (e.g. {1, 2}, {2, 3}) achieves cost savings from borders that no longer require protection. Each interior border saves 2 in costs, because an attack can come from either side. It is these cost savings which motivates alliance formation. Alliances containing a middle country has something to bargain over in contrast to an alliance of the two noncontiguous end countries where there are no interior borders. As interior borders grow in number or length, the motivation for allying increases in the mutual defense game. The core of the mutual defense game can be found graphically in Figure 9.2 in an equilateral triangle whose three vertices correspond to a single nation getting the entire gain from cost sharing in the three-country alliance – that is, ui = 4 for i = 1, 2, 3. Moreover, the three sides of the triangle satisfy u1 + u2 = 4, u2 + u3 = 4, and u1 + u3 = 4, where each of the two-country alliances gets the entire gain. The perpendicular distance of a point in the triangle from any side indicates the payoff of the unlisted ally on that side of the triangle – for example, the payoff to nation 3 is the height from the base where payoffs to nation 1 and 2 sum to 4. Every interior point of the triangle has a non-zero payoff for all three nations, but not every point in the triangle in Figure 9.2 is in the core. To be in the core, equation system (9.5) must be satisﬁed, so that 2 ≤ u1 + u2 and 2 ≤ u2 + u3 must hold. The line u1 + u2 = 2, parallel to the triangle’s base, denotes the payoffs where nation 1 and 2 get an aggregate beneﬁt of 2, while nation 3 receives 2. The ﬁrst inequality rules out all of the cross-hatched area above this line in Figure 9.2, while the second inequality rules out the shaded area to the right of u2 + u3 = 2. The unshaded rhombus with vertices (0, 2, 2), (2, 0, 2), (0, 4, 0), and (2, 2, 0) represents the proposals in the core.

144

Todd Sandler

Since the region representing the mutual defense core is convex, any line segment joining two core proposals is itself in the core. The center of mass at (1, 2, 1) in Figure 9.2 is located at the intersection of the two diagonals of the rhombus, each of which links two of the four “extreme” core proposals where one or two nations receive no beneﬁts. The center of mass gives a centroidal or median outcome of the core, where the middle ally gains half of the cost savings and the two end allies evenly divide the remaining savings of 2. This centroid has much to recommend it, since it accounts for the relative bargaining strength as reﬂected in a coalition’s disagreement payoffs. Certainly, the middle country is essential for cost savings, because it is this ally’s relative location vis-à-vis the other two allies that creates interior borders from which savings arise. This simple game representation ignores many complications that add realism, but these are now introduced. Alternative border lengths: the interior ally We shall ﬁrst consider cases where the interior or middle country’s width is smaller than the widths of the end countries. Suppose that the three potential allies are conﬁgured along a line as before, but the dimension of the middle country is now 1 2 in. × 1 in., while the two outer countries are still 1-in. squares (see pact b in Figure 9.1). Before any transformation of the utility payoffs, the core must satisfy the following equations. v({1}) = −4 ≤ u1 v({2}) = −3 ≤ u2 v({3}) = −4 ≤ u3 v({1, 2}) = −5 ≤ u1 + u2

(9.6)

v({2, 3}) = −5 ≤ u2 + u3 v({1, 3}) = −8 ≤ u1 + u3 v({1, 2, 3}) = −7 = u1 + u2 + u3 . A positive afﬁne transformation of ui is then found that maintains the maximum cost saving at 4, while keeping the three disagreement payoffs at zero or above. The following transformation has this property: ui = 4(ui + 4)/5. When this transformation is applied to the right-hand inequalities and the single equality of system (9.6), the core must then satisfy, v({1}) = 0 ≤ u1 v({2}) = 4/5 ≤ u2 v({3}) = 0 ≤ u3 v({1, 2}) = 12/5 ≤ u1 + u2 v({2, 3}) = v({1, 3}) = v({1, 2, 3}) =

u2

12/5 ≤ + u3 0 ≤ u1 + u3 4 = u1 + u2 + u3 .

(9.7)

Alliance formation, expansion and the core

145

(0, 0, 4) u⬘2 + u⬘3 = 12 5

( (0,

12 , 8 5 5

8 5

, 0, 12 ) 5

, 0, 8 ( 12 5 5

) (

8 5

,

4 5

,

8 5

)

u⬘1 + u⬘2 = 12 5

)

Core

(

4 5

,

12 , 4 5 5

)

Centroid

(0, 4, 0)

(

8 5

,

12 , 5

0)

(4, 0, 0)

Figure 9.3 Mutual defense, short middle ally.

The core is solved graphically in the equilateral triangle in Figure 9.3, based on the payoff requirements in system (9.7). The rhombus, representing the core of this mutual defense game, is formed by the two lines where u1 + u2 = 12/5, u2 + u3 = 12/5, and the sides of the triangle where u1 = 0 and u3 = 0. The center of mass of the unshaded core is at (4/5, 12/5, 4/5) where the advantage of the interior country has grown at the expense of the end countries as compared with the baseline case, owing to the interior country’s smaller size vis-à-vis its neighbors. Next consider the scenario in which the interior ally’s width is just a quarter of an inch. Based on the procedure just used, the center of mass of the core is (8/11, 28/11, 8/11).3 Once again, the interior country’s advantage at the median position has increased. In the limit, the interior country can shrink to a 1-in. length of border with no width and two directions to defend. By applying the same procedure, we identify the center of mass of the core in this limiting case to be (2/3, 8/3, 2/3), where the interior country’s share is now two-thirds of the entire cost savings. While this last case seems extreme, it is analogous to an interior country in an alliance surrounded by allies so that in effect there are no external exposed borders to guard. This location places it in an extremely strong bargaining position and the ﬁnal imputation should reﬂect this advantage. When the interior country has a longer width than the two end countries, its bargaining advantage diminishes, insofar as it has a longer perimeter to defend on its own than its two potential allies. The interior nation’s position, if it were to walk away from the negotiations, is less than the other two countries. At the same time, the end countries need the interior country if any cost savings are to arise, and this means that the interior country still maintains some leverage. At what point does the interior country’s disadvantage of a longer width just offset its locational advantage so that all three allies share equally in the cost savings

146

Todd Sandler

at the centroid of the core? Again consider a three-country conﬁguration, but let the middle country’s dimension be 1.5 in. × 1 in. while the two end countries are 1-in. squares. The following equations must then be satisﬁed at a core solution: v({1}) = −4 ≤ u1 v({2}) = −5 ≤ u2 v({3}) = −4 ≤ u3 v({1, 2}) = −7 ≤ u1 + u2

(9.8)

v({1, 3}) = −8 ≤ u1 + u3 v({2, 3}) = −7 ≤ u2 + u3 v({1, 2, 3}) = −9 = u1 + u2 + u3 . We transform ui for i = 1, 2, 3 so that ui = 2(ui + 5)/3 in order to get a nonnegative disagreement point and a maximum cost savings of 4 for the three-ally mutual defense pact. It is these transformed equations (not displayed) that are used to identify the core graphically. In Figure 9.4, the unshaded rhombus determining the core has vertices at (2/3, 4/3, 2), (2, 0, 2), (2, 4/3, 2/3), and (2/3, 8/3, 2/3) with sides corresponding to the relevant segments of the following binding four equations: u1 + u2 = 2; u3 = 2/3; u2 + u3 = 2; and u1 = 2/3. The center of mass of this rhombus is at (4/3, 4/3, 4/3) for which all allies split the cost savings equally among them. If the interior country’s width grew beyond 50 percent longer than that of the end countries, then the median core position favors the two end

(0, 0, 4)

u⬘1=

2 3

u⬘2 + u⬘3 = 2 ( 23 ,

(0, 2, 2)

4 3

, 2)

(2, 0, 2)

u⬘1 + u⬘2 = 2

Core

(0, 10 , 3

2 3

(

) (

(0, 4, 0)

4 3

,

4 3

,

4 3

)

Centroid

(

2 3

,

2 3

10 , 3

,

8 3

0)

,

2 3

)

(2,

4 3

,

2 3

u⬘3 =

)

(2, 2, 0)

Figure 9.4 Mutual defense, long middle ally.

(4, 0, 0)

2 3

Alliance formation, expansion and the core

147

countries. When, for example its width is twice that of the end countries, the centroid of the core is at (3/2, 1, 3/2), as the reader can verify.4 Next suppose that the three countries have dimensions of 1 in. × 2 in. so that the interior borders are now longer in the baseline case. This change enhances the possibility of cost savings, but leaves unchanged the middle country’s relative bargaining strength. The median position of the core is now (2, 4, 2) where cost savings are 8 as two interior borders of length 2 do not need defending in 2 directions apiece. Clearly, the spatial conﬁguration of the allies vis-à-vis internal and external border lengths play a crucial role for alliance formation and the distribution of a mutual defense alliance’s gain. The pure public good theory of alliances could not identify this insight owing to an absence force thinning and the resulting savings from interior borders. Moreover, the public good theory focused on allocation and indicated little about the distribution of gains. Although club theory representations acknowledged spatial considerations, the models in the literature never quantiﬁed these cost savings or the unique position of interior countries.

Alternative border lengths: the end allies In a three-country alliance, the perimeter of an end country may differ from both the other end country and the interior country. Consider pact c drawn in Figure 9.1, where ally 1 and 2 are unit squares, but end country 3 is 2 in. × 1 in. The core to this game is deﬁned, with our procedure, by a rhombus with vertices at (1, 2, 1), (2, 1, 1), (1, 3, 0), and (2, 2, 0) and a center of mass of (3/2, 2, 1/2). In this example, the interior ally maintains its share of 2 owing to its unchanged relative position as compared with the baseline case, but the remaining gain of 2 is now unevenly divided between the two outer countries. The short outer ally gets three times the gain of its longer counterpart at the core’s median position, and pulls nearer to the bargaining strength of the interior country. Eventually, an outer country’s width can become so long that the alliance only confers cost savings on country 1 and 2, so that this outer country will be excluded from the alliance. When, however, an outer country is shorter than the other two allies, it gains a larger share of savings at the core’s median. Suppose that country 3 has a width of 21 in. and a length of 1 in., while the other outer ally and middle ally are unit squares. The center of mass is now (4/5, 8/5, 8/5), where the shorter exterior ally has pulled even with the interior ally, leaving just a ﬁfth of the savings for the other exterior ally. As another example of a three-country mutual defense pacts, consider pact d in Figure 9.1, where country 3 is not contiguous to country 1 and 2, and all countries are unit squares. There is now a single interior border between country 1 and 2. In the associated equilateral triangle (not shown),5 the core is the base of the triangle consisting of the line u1 + u2 = 2, with a center of mass at (1, 1, 0). There is clearly no reason to include a noncontiguous ally and, if included, this ally receives nothing at any core position. This case is germane when considering the addition of allies, such as Estonia, to NATO (see section on “NATO alliance”).

148

Todd Sandler

Directional threat Obviously, mutual defense pacts may confront a threat from some directions and not from others. For the baseline, case of pact a in Figure 9.1, there is little substantial change to the results if the threat is in the direction affecting interior borders. Consider pact a where the attack can only come from either the east or west. The cost of defending each country in isolation is 2. Additionally, the cost of defending the two contiguous allies is just 2, while the cost of defending the two non-contiguous allies is 4. For the three-ally pact, the cost of defense remains at 2. By transforming the ith country’s utility by u1 = ui + 2, we ﬁnd that the conditions deﬁning the core are unchanged from the baseline line case, so that equation system (9.5) and Figure 9.2 apply, with the median of the core still at (1, 2, 1). If, however, only the east needs guarding in the baseline situation, then the three-country alliance offers a maximum cost savings of 2. The reader can conﬁrm that the median of the core is (1/2, 1, 1/2), where the relative advantage of the interior country remains the same. This last case is most germane to NATO during the ﬂexible-response years of the Cold War when conventional weapons were important. It is important to note that the results in terms of relative payoffs are really unaffected when the threat is in the direction of the interior borders. When, instead, the threat is from the north or south, there are no cost savings from interior borders and the core is empty so that there is no advantage to allying when confronted with a conventional war threat.

Mutual defense pacts: four and ﬁve allies When the analysis is extended beyond three countries, the spatial conﬁguration has more degrees of freedom and can play an even greater role. This follows because the same number of countries can have different interior borders and, thus, diverse cores depending on their conﬁguration, even when their perimeters are unchanged. In Figure 9.1, both defense pact e and f involve four unit-square allies. For pact e, there are two interior countries leading to a cost savings from mutual defense of 6 from three interior borders, no longer needing protection. For pact f , the four countries form a 2-in. square alliance with four interior borders giving cost savings of 8. The dimension of the four-country problem means that a graphical solution is too cumbersome and the inequalities and the equality associated with the core must be solved mathematically. For the rectangular pact in e, the median of the core is at (1, 2, 2, 1), where the middle two countries maintain their twofor-one advantage over the exterior countries, analogous to the baseline case. The 2-in. square pact f means that there are no interior countries, as every ally would have two exterior and two interior sides. The centroid of the core is now at (2, 2, 2, 2) with all allies gaining equally. We consider two alternative ﬁve-country mutual defense alliances to identify some essential considerations. For pact g, the ﬁve potential allies are one-inch equilateral triangles that ﬁt together to form a symmetric pentagon with 1-in. sides.

Alliance formation, expansion and the core

149

This case highlights how the actual conﬁguration of the potential allies can promote alliance expansion. There are now ﬁve interior sides with a cost savings of 10 to be shared among the allies. Because there is no country with a positional advantage – each country is adjacent to two allies and has a single exterior side – all countries share equally at the median position of the core located at (2, 2, 2, 2, 2). Next Suppose that 5 in.-square countries are conﬁgured as pact h in Figure 9.1, where country 3 has four interior sides and the other four countries have one interior side and three exterior sides. The advantage of country 3 is equal to the other four nations put together, so that the core’s median is at (1, 1, 4, 1, 1). As the number of countries in a mutual defense alliance grows, not only can the maximum cost savings increase, but also the manner for distributing this gain increases in its variety. Pact h is likely to result in a two-bloc negotiation between the interior country and the four exterior countries. The center of mass for the core of pact h illustrates that some countries can be in such a pivotal location that it can demand a relatively large gain even at the core’s median. When an attack can come from any direction, a ﬁve-country alliance of unit-square allies is the smallest where a four-sided country can be fully interior. Alliances with more than six members can certainly contain fully interior allies with advantages similar to that of country 3. The analysis can be extended beyond ﬁve allies, but yield little in the way of new insights. Typically, the core is anticipated to shrink when the number of participants increases; however, this need not characterize a mutual defense pact. Consider pacts d and f, consisting of three and four nations, respectively. The core for pact d is the straight line connecting (2, 0, 0) and (0, 2, 0), while the core for pact f is some three-dimensional rhombus with (2, 2, 2, 2) at its center of mass. Moreover, the core associated with ﬁve-nation pact g is larger than that associated with fournation pact h. This realization that the core can grow with the addition of some strategically located allies is yet another reason that spatial considerations matter.

Strategic weapons and deterrence Thus far, the model only applies to alliances set up to protect borders from a conventional attack. The NATO alliance also relies on strategic weapons that deter an attack – conventional or strategic – through a threat of retaliation. If strategic weapons possess sufﬁcient range that their deployment position is immaterial, then new allies can be protected with no additional costs incurred by the ally with such weapons (Olson and Zeckhauser 1966). The sequestration of interior borders loses any importance for strategic weapons: if there are no retaliatory risk factors, then the core consists of all friendly nations allying within an alliance, solely dependent on strategic weapons. The United States could extend its strategic deterrence to additional allies, leading to no real size restrictions. However, the United States would demand political concessions from its allies to obtain some beneﬁt from taking other countries under its nuclear umbrella (Morrow 1991), especially because retaliatory risks arise. The presence of strategic weapons can, therefore, justify why a non-contiguous nuclear ally, like the United States or Britain, would join

150

Todd Sandler

a mutual defense pact even though its own borders are not sequestrated. The conventional allies’ motivation to join depends on cost savings from interior borders and the net value (deducting the costs of political concessions) derived from deterrence. The sole determinant behind the formation of the ANZUS alliance, with its three non-contiguous allies, must be based on the tradeoff of strategic deterrence from the United States in return for political concessions (e.g. Australia’s participation in the Vietnam War).

An application to the NATO alliance Even in its most elementary characterization, the analysis of mutual defense games can provide insights into NATO burden sharing, formation, and expansion. Exterior front-line allies in NATO (e.g. Germany) are at a bargaining disadvantage compared with more interior European allies (e.g. Belgium, Luxembourg, and the Netherlands). Such exterior allies have assumed relatively greater defense burdens during the Cold War. When average beneﬁt shares received are compared with actual defense burden shares paid, West Germany was an overcontributor from 1975 through 1990, except for the height of the Reagan defense buildup in 1985 (Khanna and Sandler 1996, 1997). With the end of the Cold War and the lessening of the eastern threat, Germany has displayed relatively large defense cutbacks (Hartley and Sandler 1999). Non-contiguous NATO members, especially those allies either separated by a large distance or else bordering an enemy, are at a decided bargaining disadvantage, leaving them no choice but to assume larger defense burdens relative to other allies. Prime examples of such allies would include the United States, the United Kingdom, Norway, and Turkey. From the start of the NATO alliance in 1949, the United States has typically carried the largest defense burden in terms of defense spending as a percent of GDP. The United Kingdom also shouldered one of the largest defense burdens. Norway and Turkey have assumed a greater defense burden relative to their cohort of NATO allies.6 Even though Canada has free rode on the United States, the long common border between the two allies makes their alliance mutually beneﬁcial despite lopsided burden sharing. Without Canada as an ally, the United States might have to divert many military resources to guarding its northern border, which has no natural barriers or defenses. Next, consider NATO’s formation in 1949 as a counter to Soviet aggression in Eastern Europe and its take over of satellite states. On December 10, 1948, negotiations began in Washington, DC, on a North Atlantic Treaty, which would tie participants together in a mutual defense alliance.7 Initial negotiations included representatives from Belgium, France, Luxembourg, the Netherlands, the United Kingdom, Canada, and the United States. The ﬁve European countries were best positioned to gain from a mutual defense pact from the creation of interior borders. These seven countries then invited Denmark, Iceland, Italy, Norway, and Portugal to join the negotiations on March 15, 1949. With the exception of Italy, these late participants were not contiguous with the ﬁrst seven nations pushing for NATO and, therefore, could offer less to the alliance. Their inclusion in the original alliance must then be

Alliance formation, expansion and the core

151

based on considerations other than mutual defense – for example, deterring Soviet aggression through a threat of strategic weapon retaliation in return for political concessions. Once the twelve-nation alliance was established on April 4, 1949, it was just a matter of time to ﬁll in missing pieces – Spain and West Germany – since these countries would create large expanses of interior borders. The addition of Greece and Turkey in 1952 added protection to NATO’s southern ﬂank, but these countries were not contiguous to the original twelve members. In Turkey’s case, other beneﬁts (e.g. a listening post to the Soviet Union) came from its inclusion. The analysis of mutual defense games can also provide some insights into the NATO expansion process and which nations might ultimately become members following the recent admittance of the three Visegrad nations.8 The inclusion of East Germany in NATO as part of uniﬁed Germany in 1990 is strongly supported by a mutual defense framework, since this entrant just moves NATO’s perimeter eastward and produces cost savings from added interior borders. The Baltic nations (Estonia, Latvia, and Lithuania), which seek membership, include only a small inner border with Poland, making their inclusion offer little, if any, cost savings from mutual defense. Given the strong political opposition to their inclusion by Russia and this lack of cost savings, there is little or no gain for the current members from including them. Thus, the Baltic states are very unlikely entrants. The geographical location of Slovakia, with its interior borders with the three entrants and Austria, makes Slovakia an attractive candidate for inclusion. If Slovakia is included, there will only be a small exposed border with Ukraine needing defending. Sandwiched between Italy and Hungary, Slovenia is also a likely entrant with just its external border with Croatia needing protection. Romania is, however, a less attractive prospect for NATO membership with its long exposed borders on all sides except in the west, where it borders Hungary. The geographical position of Romania offers virtually no advantage in terms of cost savings from mutual defense, so that its inclusion must be based on other grounds. Without these other grounds, Romania is an unlikely entrant even though it was among the ﬁve considered at the July 1997 Madrid summit. NATO’s new eastern members may be in favor of including members to their east, since this expansion favors these new members most directly with the acquisition of an interior border. The western allies have, however, much less to gain and, since entry decisions must be unanimously approved by NATO allies, further eastward expansion is unlikely. Moreover, such eastward expansion is viewed as a threat by Russia, whose strong opposition raises a serious political concern. A more likely scenario involves ﬁlling in the missing pieces – that is, Slovakia, Slovenia, and Austria. Other neutral countries – Sweden, Switzerland, and Finland – are unlikely candidates owing to their geographical locations and minor cost savings (see later).

Other considerations The mutual defense game representation can be made yet more realistic by extending the analysis to consider natural defenses, ally-speciﬁc risks, guerrilla warfare,

152

Todd Sandler

and transaction costs. For these extensions, we use the baseline case with three unit-square countries arranged in a row (pact a in Figure 9.1). Natural defenses and risk factors Suppose that the right-hand ally (country 3) in the baseline model has some lengths of borders with natural defenses – for example, mountain ranges, cliffs along a coastline. Further suppose that these natural defenses are along one-third of its three exterior sides, so that only three of its four inches of borders require guarding. In this case, its cost of protection is v({3}) = −3 and we are back to the case of an end country with dimension of 21 in. × 1 in. The centroid of the core is then (4/5, 8/5, 8/5), as shown earlier, with the end country drawing equal to the middle country. Hence, natural defenses along an exterior border act to shorten an end ally’s borders needing protection, and, in so doing, afford the ally a bargaining advantage. Next suppose that the natural defenses involve the interior nation (country 2) along its interior boundary with country 1. In particular, suppose that only half of this interior border needs protecting, so that both country 1 and 2 only have to expend 3.5 in cost to protect their borders if they walk from the alliance. Any advantage to the interior country in natural defenses along its perimeter must be also shared by the contiguous end country. The maximum cost savings for this three-nation defense pact is reduced to 3 from the baseline case, since there are now less exposed inner borders that no longer need protecting. The core is identiﬁed9 by the rhombus with vertices at (3/8, 9/8, 12/8), (9/8, 3/8, 12/8), (9/8, 15/8, 0), and (3/8, 21/8, 0), possessing a center of mass at (3/4, 6/4, 3/4). This centroid has the same relative payoffs between the outer countries and the interior country as the baseline model. Total payments differ because cost savings are only three-quarters of the baseline case. Next, suppose that all interior borders of this three-country mutual defense pact have natural defenses, so that only half of either interior border needs defending. Total cost savings are just 2 and the core consists of the rhombus with vertices at (1/4, 1, 3/4), (3/4, 1/2, 3/4), (3/4, 1, 1/4), and (1/4, 3/2, 1/4). The center of mass is now located at (1/2, 1, 1/2), where the relative shares stays the same. Whenever interior borders possess defensive advantages, cost-saving shares at the center of mass will remain one for two between the end countries and the interior country. The latter country is unable to improve its bargaining advantage because the natural defense along its border must also be possessed by the neighboring end country. By limiting the maximum cost savings from mutual defense, natural defenses along interior boundaries curtail the beneﬁts and the need for allying. If both interior borders are impenetrable, then there are no cost savings from mutual defense and the nations are not anticipated to ally – that is, the core is empty. This statement depends on there being no other allying beneﬁt to replace cost savings. Hence, inaccessible mountainous countries in the middle of allied states – that is, Switzerland – are not needed in NATO. Such natural defenses distinguish the common border of Sweden and Norway from that of the United States

Alliance formation, expansion and the core

153

and Canada. For Sweden and Norway, their mountainous interior border presents a natural barrier, limiting any cost savings incentive from allying. This contrasts with the United States and Canada where no such natural barrier exists along their shared border, so that signiﬁcant cost savings from allying arise. Natural defenses have a much different inﬂuence when they are along an exterior or interior border. A related, but different, case concerns borders that are more vulnerable or risky than others. During the Cold War, an attack was expected to come along West Germany’s eastern border, rather than along Norway’s northern border with the Kola Peninsula of the Soviet Union. The former was more hospitable to Soviet tanks rolling across plains. Consider the three-nation linear array where the eastern border of the end country is twice as risky as all other borders, meaning that country 3’s cost of protection is 2 rather than 1 at this eastern border. This case is analogous to a longer end country and can be shown to have a center of mass for the core at (4/3, 6/3, 2/3), where the interior ally maintains its absolute and relative share of the cost savings, attributable to the interior borders. The remaining gain of 2 is now unevenly divided between the two end countries. West Germany’s complaints that it did more than its fair share during the Cold War is not only backed up by statistical analysis (Khanna and Sandler 1996, 1997), but justiﬁed given its more vulnerable position vis-à-vis the mutual defense pact. Position and size matters. Since the end of the Cold War and the much-reduced eastern threat, Germany has displayed one of the largest relative burden decreases in NATO – from 3.0 percent of GDP during 1985–9 to 1.6 percent in 1997 (Hartley and Sandler 1999). This dramatic change can be explained, in part, by the reduced riskiness of its eastern border. Now that Germany has become an interior nation with the addition of Poland and the Czech Republic, further such reductions can be anticipated. The baseline case can also be extended to consider more risky interior borders which enhance the importance of mutual defense, because the associated cost savings from allying can increase dramatically. Suppose that the two interior borders are twice as difﬁcult to protect, thus requiring double the armaments and troops at twice the cost of guarding an equal expanse of exterior border. Cost savings are now 8, or 4 for each of two interior borders. The center of mass for the core is at (2, 4, 2).10 If a mutual defense pact can sequester vulnerable perimeters behind more defensive ones, then mutual defense pacts have great gains to offer. This is a consideration that should be driving NATO’s expansion when mutual defense considerations are germane. As such, Sweden has little to offer NATO as a member, but Slovakia has much. Poland has a great deal to offer Germany but little to offer the alliance. Guerrilla warfare and insurgencies The partnership for peace (PFP) program, established on January 10–11, 1994, is intended to promote security in Eastern Europe and to prepare some PFP members for NATO membership (PFP 1996). The PFP is intended to promote civilian control of armed forces and to foster security arrangements between NATO and PFP

154

Todd Sandler

members. Prospective PFP members are expected to solve border disputes as well as insurgencies within their borders. Both requirements are essential conditions for NATO membership and are sensible from a mutual defense perspective. The existence of border disputes implies more vulnerable borders, while insurgencies imply that protection against attack must be countrywide. Again consider the baseline model with three 1-in. square countries in a linear array. Further suppose that defense costs are normalized to equal 1 for guarding a country’s one-inch square territory. For guerrilla warfare, interior borders offer no cost-saving advantage since a country must protect everywhere including its side of the border, with or without an alliance. Because there are no savings, the core is empty. In fact, an interesting negative externality can develop when each ally attempts to drive the insurgents from its interior and, in so doing, places a neighboring ally at greater risk while overspending (Sandler and Lapan 1988). If, however, all countries drive the guerrillas from their territories into nonallies surrounding the mutual defense pact, then border protection and cost savings are again the essential considerations as nations must keep the insurgents from entering their territory. As such, the previous analysis applies. Thus, the type of warfare is a crucial factor. Transaction costs As a ﬁnal extension, transaction costs are added to the model, so that forming and maintaining a mutual defense alliance is costly. In particular, transaction costs can take the form of decision-making costs, information-gathering costs, interdependency costs (from less autonomy), and enforcement costs (Sandler and Hartley 1999: chapter 8). Although there are myriad ways that these transaction costs can be introduced, only a single scenario is investigated. Suppose that the three-country baseline model applies, where transaction costs for a two-ally mutual defense pact are 1, while they are 2 for three allies. The following equations must be satisﬁed for a core, after normalizing utility to get a zero disagreement point:11 v({1}) = −4 ⇒ 0 ≤ u1 v({2}) = −4 ⇒ 0 ≤ u2 v({3}) = −4 ⇒ 0 ≤ u3 v({1, 2}) = −7 ⇒ 1 ≤ u1 + u2 v({2, 3}) = v({1, 3}) = v({1, 2, 3}) =

(9.9)

u2

−7 ⇒ 1 ≤ + u3 −9 ⇒ −1 ≤ u1 + u3 −10 ⇒ 2 = u1 + u2 + u3 .

In system (9.9), defense cost includes the unit cost of protecting an exposed border and the relevant transaction costs. For example, an alliance between an interior and an end country has six exposed 1-in. sides to protect and transaction costs of 1 for

Alliance formation, expansion and the core

155

a total cost of 7. Obviously, an alliance between the two non-adjoining nations is undesirable, which would be true for all non-contiguous countries. The relevant triangle (not shown) for locating the core has vertices at (2, 0, 0), (0, 2, 0), and (0, 0, 2), since maximum cost savings are 2. The center of mass for the core is (1/2, 1, 1/2) with the middle ally maintaining its relative advantage. If, however, transaction costs were to increase nonlinearly with alliance size, cost savings are curtailed and alliance size would be limited.

Concluding remarks The analysis of the core of a mutual defense game represents a new paradigm for studying defense alliances that holds promise for understanding alliance formation, stability, and expansion. Unlike discussions about NATO expansion in the literature, the mutual defense representation is a beneﬁt-based approach founded on cost savings from interior borders. Other beneﬁts can be introduced (e.g. deterrence beneﬁts) and combined with those of cost savings in the framework introduced here. Additional considerations increase the realism of the model and produces further insights. A number of conclusions can be drawn. A country’s size, location vis-à-vis potential allies, and border attributes matter for alliance formation as well as for the distribution of gains if an alliance were to form. Natural defenses along exterior borders or reduced perimeters of an exterior ally can transfer relative gains from the interior allies to the exterior allies. Natural defenses along interior borders, however, do not change the relative distributions of gains and serve to limit the need for a mutual defense pact. Transaction costs and insurgencies curtail gains from mutual defense pacts. The same number of allies can have vastly different gains and imputations of these gains from mutual defense alliances depending on their geographical conﬁguration. As the number of allies increases, this variability of gains and their distributions can also increase. Spatial consideration can allow the core to grow as the number of participants increases. Three future directions should be pursued. First, transaction costs can be introduced in a more interesting and integrated fashion. Second, alternative solution concepts, such as Shapley value or the nucleolus, can be applied (Luce and Raiffa 1957: 246). The Shapley value generalizes the minimax value and gives a computable unique equilibrium. In contrast, the nucleolus is more akin to distributing beneﬁts according to a Rawlsian maximin principle. These alternative solution concepts may be more appropriate for analyzing the alliance expansion decision which weighs the relative merits of a prospective entrant.12 Third, the analysis can be extended to investigate multiple alliances.

Acknowledgments The author’s research has been funded in part by a NATO Fellowship for 1998–2000. I have greatly proﬁted from comments provided by Daniel Arce,

156

Todd Sandler

Patrick James, Bruce Russett, and an anonymous referee. Any remaining shortcomings and all views expressed are solely those of the author.

Notes 1 On NATO expansion and these cost estimates, see Asmus et al. (1995, 1996), the Congressional Budget Ofﬁce (1996), Bureau of European and Canadian Affairs (1997), US General Accounting Ofﬁce (1997), and Sandler and Hartley (1999: chapter 3). 2 On bargaining games and bargaining solutions, see Luce and Raiffa (1957), Scarf (1973), Binmore (1992), Eichberger (1993), and Gardner (1995). 3 The required transformation for utility is ui = 8(ui + 4)/11 for i = 1, 2, 3. 4 The relevant transformation for utility is now ui = (ui + 6)/2. 5 This equilateral triangle has vertices at (2, 0, 0), (0, 2, 0), and (0, 0, 2). 6 See Sandler and Hartley (1999: 43) for NATO defense burdens for 1970–7. 7 The dates and facts of this paragraph come from Sandler and Hartley (1999: 24–29). 8 Membership also depends on other factors including Russian opposition, the risk of civil unrest, the absence of border disputes, the state of the country’s military forces, and interoperability of these forces with NATO forces (Sandler and Hartley 1999: chapter 3). 9 The transformation of utility is ui = 3(ui + 4)/4. 10 The vertices of the rhombus for the core are (0, 4, 4), (4, 0, 4), (4, 4, 0), and (0, 8, 0). 11 The normalization is the same as that of the baseline case with ui = ui + 4 for every i. 12 Yet another solution concept is that of coalition building, because old members do not tend to drop out as new ones enter (Arce 1994).

References Arce, D. G. (1994) Stability criteria for social norms with applications to the Prisoner’s Dilemma. Journal of Conﬂict Resolution 38, 749–765. Asmus, R. D., Kugler, R. L., and Larrabee, F. S. (1995) NATO expansion: the next steps. Survival 37, 7–33. Asmus, R. D., Kugler, R. L., and Larrabee, F. S. (1996) What will NATO enlargement cost? Survival 38, 5–26. Binmore, K. (1992) Fun and Games. D. C. Heath, Lexington, MA. Bureau of European and Canadian Affairs. (1997) Report to the Congress on the Enlargement of the North Atlantic Treaty Organization: Rationale, Beneﬁts, Costs and Implications. US Department of State, Washington, DC. Congressional Budget Ofﬁce (CBO) (1996) The costs of expanding the NATO alliance, CBO Papers. CBO, Washington, DC. Eichberger, J. (l993) Game Theory for Economists. Academic Press, San Diego, CA. Gardner, R. (1995) Games for Business and Economics. John Wiley & Sons, New York. Hartley, K. and Sandler, T. (1999) NATO burden sharing: past and future. Journal of Peace Research 36, 665–680. Khanna, J. and Sandler, T. (1996) NATO burden sharing: 1960–1992. Defence and Peace Economics 7, 115–133. Khanna, J. and Sandler, T. (l997) Conscription, peace-keeping, and foreign assistance: NATO burden sharing in the post-Cold War ear. Defence and Peace Economics 8, 101–121. Luce, R. D. and Raiffa, H. (1957) Games and Decisions. John Wiley & Sons, New York. McGuire, M. C. (1990) Mixed public–private beneﬁt and public-good supply with application to the NATO alliance. Defence Economics 1, 17–35.

Alliance formation, expansion and the core

157

Morrow, J. D. (1991) Alliances and asymmetry: an alternative to capability aggregation model of alliances. American Journal of Political Science 35, 904–913. Murdoch, J. C. and Sandler, T. (1982) A theoretical and empirical analysis of NATO. Journal of Conﬂict Resolution 26, 237–263. NATO. (1995) Study on NATO Enlargement. NATO, Brussels. Olson, M. and Zeckhauser, R. (1996) An economic theory of alliances. Review of Economics and Statistics 48, 266–279. Partnership for Peace (PFP) (1996) Partnership for Peace: What is it? NATO, Brussels. Russett, B., Oneal, J., and Davis, D. R. (1998) The third leg of the Kantian tripod for peace: international organizations and militarized disputes, 1950–1985. International Organization 52, 441–467. Sandler, T. (1977) Impurity of defense: an application to the economics of alliances. Kyklos 30, 443–460. Sandler, T. and Cauley, J. (1975) On the economic theory of alliances. Journal of Conﬂict Resolution 19, 330–348. Sandler, T. and John F. Forbes. (1980) Burden sharing, strategy, and the design of NATO. Economic Inquiry 18, 425–444. Sandler, T. and Hartley, K. (1999) The Political Economy of NATO: Past, Present, and into the 21st Century. Cambridge University Press, Cambridge. Sandler, T. and Lapan, H. E. (1988) The calculus of dissent: an analysis of terrorists’ choice of target. Synthese 76, 245–261. Scarf, H. (1973) The Computation of Economic Equilibria. Yale University Press, New Hoven, CT. United States General Accounting Ofﬁce (US GAO) (1997) NATO enlargement: cost implications for the United States remain unclear. Testimony before the Committee on Appropriations, US Senate, GAO/T-NSIAD-98-50, 23 October, US GAO, Washington, DC. van Ypersele de Strihou, J. (1967) Sharing the defense burden among Western allies. Review of Economics and Statistics 49, 527–536.

10 The determinants of US military expenditures in the context of arms race Sylvie Matelly

Introduction The end of the Cold War meant the end of a bipolar world in which all international political events could be explained by the opposition of the Soviet and American blocks. One of the principal features of this bipolar world was the arms race. However, with current political and economic changes, past theories must be reviewed. Recent research on arms races has tended to model regional conﬂicts, such as India–Pakistan and Greece–Turkey. But there is also the question of whether Cold War explanations of armaments allow us to understand the post-Cold War defence expenditures of the United States? With this in mind, this chapter develops theoretical speciﬁcations of action–reaction models applied to the United States with the idea that redeﬁning the grievance coefﬁcient, described by Richardson (1960), can permit existing theories of arms race to adapt to the new international context, in which the United States is the only super-power and give alternative explanations to the dynamics of armament and disarmament. Despite numerous studies, articles and works published on arms race, Richardson’s three coefﬁcients (defence, fatigue and grievance) have not been fully analysed. They represent the three kinds of forces that inﬂuence an arms race. The grievance coefﬁcient, in particular, has usually been treated as a constant in the model, or related to the number of acts of conﬂict or cooperation. But it might be interesting to go deeper. The grievance is deﬁned by Richardson (1960) as ‘From the mathematical point of view, g and h (grievance coefﬁcients) represent any motives, which, while affecting warlike preparations, remain constant, independently of the amount of such preparations at home or abroad. From the psychological point of view, such motives could include, deeply rooted prejudices, standing grievances, old unsatisﬁed ambitions wicked and persistent dreams of world conquest or, on the contrary, a permanent feeling of contentment’. Grievance measures the component of military expenditures not explained by the military spending of other countries or its own lagged military expenditures. In general, it is difﬁcult to measure because it is not easy to distinguish what can be attributed to a response to the opponent’s military spending and what are the grievance factors. In this sense, the American example is interesting for several reasons. First, the United States was one of the two central actors of the Cold War. It is currently

Determinants of US military expenditures

159

the most powerful military country in the world, with the largest military budget. Second, the persistence of the American hegemony over the last 50 years confers to American strategic actions a sort of continuity that allows a structural speciﬁcation of American defence spending. In addition, with hindsight, it is possible to conduct a long-term economic analysis of American military decisions (1952–97). Third, the end of the Cold War allows one to separate the part of the US military expenditures justiﬁed by action–reaction with the USSR from the grievance factors. Thus, our study will proceed in several stages to estimate empirically the American grievance term. Empirical veriﬁcations have always been the weak point of action–reaction models. A presentation of ﬁve models will be given as an illustration of this problem. New tests on these models on longer samples do not give good results and indicate how classic linear econometric methods can be inadequate to estimate an arms race. In the case of arms races, military spending of the two potential rivals are rarely stationary and may be cointegrated. In our case, the choice of this model allows estimation of an action–reaction function for the United States during the Cold War (1952–89). We can then forecast what American military expenditures should have been, given the reduction in SovietRussian military expenditure after 1989. It is then possible to extract the military spending generated by the action–reaction process and to deduce grievance, as the part of the observed military expenditures not directly explained by the arms race. Finally, it is possible to consider determinants of this grievance.

Arms race coefﬁcients and Cold War theoretical models The Richardson model of an arms race explains the military expenditures (or arms stocks) of rival countries as an interdependent and competitive action–reaction process, described by a linear differential equation: δM = α1 + α2 M + α3 M , δt where α1 , α2 and α3 are, respectively, grievance, fatigue and defence coefﬁcients, and M and M the military expenditures of the country and its adversary, respectively. The defence coefﬁcient represents the evolution of one country’s military spending as a function of the other country’s military expenditures. In the framework of the arms race, this coefﬁcient is positive. The military expenditures of other countries threaten the nation and incite it to increase its own military expenditures. This hypothesis of positive interdependence between military expenditures is recurrent in arms race models and action–reaction functions. However, these models do not really explain this process, rather they assume it a priori. The fatigue coefﬁcient is accounted for by the fact that military decisions can never be considered as totally independent of national/international economic conditions. The fatigue coefﬁcient, constant and negative, includes all factors that limit the increase of military expenditures. It expresses the fact that arms accumulation is negatively inﬂuenced by the nation’s arms stock because of economic and technical limits of the arms dynamic. The other factors are assimilated into the grievance coefﬁcient.

160

Sylvie Matelly

This confrontation between two countries induces an arms balance that can be either stable or unstable and leads up to a situation of dissuasion or conﬂict as described by Coulomb (1998). There are various criticisms of the model. First, the process described here is deterministic: when the actions and reactions begin, they tend to be self-sustaining. Richardson seems to exclude a real control of the decision from the state and decision-makers. Moreover, this analysis limits the possibilities of speciﬁcity. Thus, observing the Cold War and arms race between the United States and the USSR, it clearly appears that it was an opposition of two opposite systems, with different cultures, structures and historical evolutions. After the Second World War, theoretical developments of the arms race idea multiplied to explain the abnormal increases of defence expenditures during peacetime (Rapoport 1957). The models and their speciﬁcations for strategic relations between countries are numerous. They can be classiﬁed into two classes of models: descriptive and normative. Normative models try to formalise arms race as a rational process where the decision-makers maximise a welfare function and try to insure security. Descriptive models are more mechanical and describe the action–reaction as Richardson did. However, these models are ﬁnally very closed. Brito and Intriligator (1995) survey the arms race as an interaction between two adversary countries. But arms race has been often acknowledged as a fact. It is an a priori assumption of the relations between Russia and the United States. The essential object of theorists is then to understand why the raised levels of arms did not inevitably lead to a conﬂict and permit dissuasion. Therefore, it concerns the dissuasion as a balance of forces, a ‘balance of terror’, and tests its stability over a period of time. Instability is, thus, assimilated to conﬂict. In fact, if the decision to increase the stock of arms is rational, it will result in a procedure of optimisation of a determined collective welfare under the constraints of resources and security. If this rationality is more debatable, the public decision can only be considered as a game between two rivals, or a consequence of a bureaucratic process. The causes and consequences of arms race are always linked to a set of interactions between military expenditures, arms stocks and strategic factors of the countries at stake. Table 10.1 recapitulates ﬁve theoretical models of these actions–reactions between the USSR and the United States. Moreover and as underlined by Wiberg (1990), Richardson’s model does not specify the concepts and the evaluating method used to verify the theory. For example, is it more correct to consider the ﬂow of military expenditures or the stock of armaments as determinant of the arms race? The stock of arms seems to be more realistic but it is difﬁcult to measure and its estimation may be quite erroneous. Military expenditures are ambiguous and often debatable, but it is possible to estimate them. Another essential inconvenience of the models elaborated during the Cold War is that they do not have a sufﬁciently long series in order to be validated. Further, the explanations of the choice of data are not very precise. The regression method is not always deﬁned and the tests are sometimes omitted in the ﬁnal presentation. Then, it is often difﬁcult to recover the veriﬁcations by completing them with supplemental years. Out of the ﬁve equations presented in Table 10.2, only three of them have been tested. They permit a ﬁrst analysis of

Table 10.1 Five theoretical examples of action and reaction processes of armaments dynamics between two opposite countries Model

Action–reaction functions

Empirical methodology

1 Richardson (1960) 2 MacGuire (1977) 3 Ostrom (1977)

(dM1 /dt)t = α1 M2t − α2 M1t + α3

Descriptive statistics and graphics Two-stage least square, 1960–73 Restrictive least square, 1954–69, R 2 = 0.891 Linear regression, 1966–83, R 2 = 0.993, DW = 3.02, F = 526

∗ BUS(t) = 587∗ + 0.689BUS(t−1) + 0.0126[BUSSR(t−1) + BUSSR(t−2) ] Data in current dollars; MUS(t) = ∗ 0.492MUSSR(t−1) +0.677MUS(t−1) Data in 1981 constant dollars; MUS(t) = 2808702∗∗ + ∗∗ 25.04MUSSR(t) + ∗∗ )2 5.864(MUSSR(t) M1t = α1 C1t + α2 T1t = α3 M1t−1 + ε1t , where C1t and T1t are AR(1)

4 Wolfson and Farrell (1987)

5 MacGinis (1991)

No empirical test but vector auto regressive estimation recommended

Notes Mi , i’s military expenditures; Mit , i’s military expenditures in t (MUS , US military expenditures and MUSSR , USSR military expenditures); Bi , i’s strategic forces; Cit , military expenditures opportunity cost; Tit , external threats; ε and εt , residuals of the speciﬁcations. ∗ Signiﬁcant at 5% level. ∗∗ Signiﬁcant at 1% level.

Table 10.2 Empirical results of the ﬁve theoretical relations (estimation period, 1952–97) Endogenous

MUS /(t)

MUSt

Richardson

MacGuire

Ostrom

Wolfson and Farrell

MacGinis

MUS(t) −0.07∗ (2.23) MUS(t−1) 0.894∗∗ (29) 0.893∗∗ (31) 0.999∗∗ (68) MUSSR(t) 0.08∗ (2.8) 1.12∗∗ (3.5) MUSSR(t−1) 0.109∗∗ (4.14) −0.03 (0.86) (MUSSR(t) )2 MUSSR(t−1) + 0.052∗∗ (3.7) MUSSR(t−2) Constant −0.008 (0.134) 0.061 (0.06) 0.041 (0.052) 0.244 (0.71) 0.045 (0.65) Adjusted R 2 F -statistic Durbin test

0.10 4.12∗ 2.64

0.99 3,024∗ 2.84

0.97 3,195∗ 3.09

0.83 111∗ —

0.99 4,631∗ 4.07

Notes t-ratios in brackets. ∗ 5% Signiﬁcant level. ∗∗ 1% Signiﬁcant level of the Student test; F -statistic >4.08 and Durbin <1.96 at 5% signiﬁcant level.

162

Sylvie Matelly

the arms race between the United States and Russia. To be able to compare the theoretical interest of these ﬁve equations and ﬁnally to construct a basic equation of the arms race between the United States and the Soviet Union, we will begin again with the empirical veriﬁcations of these equations, during a longer period (1952–97) and with military expenditures as the variables of interest.

Theoretical models of arms race and empirical problems Arms race models pose many problems for empirical tests. A simple speciﬁcation often reﬂects numerous assumptions such as the determination of the process and perfect information. Therefore, the tests do not appear very signiﬁcant. More complex models use variables and concepts always difﬁcult to evaluate or estimate. Moreover, the basic choices determine the most appropriate econometric method. If the actions and reactions of the United States and Soviet Union form a system of two simultaneous equations, the method used to estimate the ﬁve equations in Table 10.1 is a linear regression of two-stage least square where the Soviet reaction function represents the instrumental variables. If we consider Russian military expenditures as given when the United States decides its own spending, ordinary least square (OLS) is appropriate. If we accept this last assumption, the ﬁve models presented earlier can be tested with OLS. Military expenditures are expressed in logarithm to be stationary and from 1952 to 1997. The three tests (adjusted R 2 , F -statistic and Durbin test) allow us to verify the relationship likelihood and the independence of errors that guarantee, the residuals are normally distributed and convergent. The Durbin test indicates serial-correlation of the errors. They are independent only if D < 1.96 at 5 percent signiﬁcant level. While a large proportion of the level is explained, only a small proportion of the changes is explained. The military expenditures for the previous years appear to be the strongest determinant. These results suggest that classic empirical methods are not well adapted to the arms race process, with major structural changes like the end of the Cold War. Recent developments in the econometrics of non-stationary and cointegrated variables permit us to reconsider arms race models and to test strategic interaction between two potential rival countries. These new methods have several advantages: they do not impose a preliminary theoretical speciﬁcation, they avoid the consequences of non-stationary variables or serially correlated errors and, ﬁnally, they limit the negative effects on speciﬁcation and estimation of the relationships between two series (Bourbonnais 1998).

Soviet versus American strategic interdependence: the choice of the variables Oström (1977) demonstrated that the function of total military expenditures is largely accepted as a symbol of engagements and political choices of the United States. Elsewhere, the function of total military spending is one of the many strategic indicator in comparison to other macro-economic aggregates, which are

Determinants of US military expenditures

163

directly expressed in dollars (e.g. Aben 1992). The evaluation of speciﬁc strategic variables is essentially more difﬁcult as well as more questionable. Military expenditures are relatively easily available. Moreover, some organisations annually estimate these expenditures (see, e.g. Stockolm International Peace Research Institute, SIPRI Yearbook; United States Arms Control and Disarmament Agency (US ACDA), Main Statistical Tables; International Institute for Strategic Survey (IISS), The Military Balance). During the Cold War, an important difference between these sources was noticed. These differences can be in large part explained by various conceptual choices. It, therefore, seems correct not to mix data from different organisations for the same variable. In the case of the American decisions of military expenditures, the US ACDA estimations appear preferable to those of the SIPRI. Essentially, the ACDA is an American administration whose principal information sources on the Soviet Union and Eastern countries during the Cold War came from American information agencies. It seems logical that the American decision-makers or American military industrial complexes used estimations of the strategic forces of different countries that were relatively close to the data available by ACDA. Preliminary tests conﬁrm the validity of ACDA data, even if their estimations of the military spending of the USSR may have been more disputable during the Cold War than later. These ﬁrst remarks demonstrate the subjectivity of decisions about military expenditures and political choices of states involved in an arms race. Military expenditures are measured in current US dollars. This choice is based on the behavioural signiﬁcance of current military expenditures during the Cold War (see Table 10.3). All modiﬁcation of raw data by correcting prices effects by constant prices or purchasing power parities are extremely questionable for such highly strategic values. Indeed, the price indexes of the gross national product (GNP) are not of much signiﬁcance for military variables. The evolutions of the price of military expenditures are dependent on a speciﬁc inﬂation to the constraints of the arms race and to technological process. The technological progress increases the cost and the price of military materials. Moreover, the price increases of armaments relative to the arms race are realised outside of the market, because of speciﬁc constraints. They cannot be assimilated with inﬂation related to civil production (US ACDA 1975). These new observations conﬁrm once more the subjectivity of the political choices and the imprecision of the data used by United States.

Table 10.3 Correlation coefﬁcients of US and USSR military expenditures and armed forces, 1952–89 and 1952–97

Military expenditures in current prices Military expenditures in constant 1995 prices Military expenditures/GNP Armed forces

1952–89

1952–97

0.977 0.682 0.916 0.070

0.812 0.549 0.907 0.506

164

Sylvie Matelly

Table 10.3 gives correlations between various measures of US and USSR/ Russian military activity for the Cold War, 1952–89, and the whole period. It is noticeable that there is effectively a difference between the whole of the period and the arms race during the Cold War: the military expenditures seem to be less signiﬁcantly correlated to the whole period than during the arms race. The armed forces are not signiﬁcantly correlated. The respective structure of the two armies can explain this difference: during the Cold War, the American army was much more capital intensive than its Soviet counterpart. Fontanel et al. (1985) emphasised the importance of such a difference. Consequently, the dissolution of the USSR and disarmament led to a reduction of the strength of the Russian army.

Soviet versus American strategic interdependence, analysis of the data By looking at military expenditures of the United States and the USSR until 1988, it is clear that both countries underwent a similar evolution of their defence spending (see Figure 10.1). The trend of these series changes and decrease in 1988 for the USSR and in 1989 for the United States (the ﬁrst time) and 1992 (deﬁnitely). The following remarks can be made: • • • •

The two series are quite closely correlated between 1952 and 1989. The US military expenditures seem to react to the USSR expenditures because they always have lags compared to USSR military expenditures. The gap between the two series of military expenditures increases after 1989 and can, therefore, depend on a range of variables that are speciﬁc to each country and can be assimilated into some grievance or ambitions. It appears relatively clear that the two series are no longer correlated after 1989. American military spending remains high while the Soviet ones have 350,000 300,000

US ME USSR ME

250,000 200,000 150,000 100,000 500,00 0 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992

1996

Figure 10.1 US and USSR military expenditures, 1952–97. (In millions of current US dollars. After 1991: Russian military expenditures.) Source: US ACDA, main statistical tables.

Determinants of US military expenditures

165

been reduced by much more. We can then suppose that the US expenditures are, between 1952 and 1989, an expression of the actions and the reactions of this country towards its socialist rival. It is then possible to forecast the US military expenditures if the same action–reaction process had continued after 1989, ceteris paribus. In the same way, there is clearly a major structural change in 1989. It is then possible to assume that Russian interactions on US military expenditures stop this year. The correlation coefﬁcients, estimated in the Table 10.3, seem to conﬁrm this possibility: the military expenditure correlation in current, constant and compared to GNP are most important for the period 1952–89 than for the 1952–97. Thus, grievance measures the part of military expenditures not explained by the military spending of the rival countries or its own lagged military expenditures. It is often difﬁcult to observe. The end of the Cold War shows this more clearly for United States. The military expenditures of the United States were justiﬁed, during the Cold War (1952–89 in our case), as a response to the expenditures of the USSR, but when the Soviet spending collapsed, the United States did not follow. The difference between what we would have predicted for US spending on the old arms race model and what we actually observe can be interpreted as a measure of the US grievance. It is then necessary to estimate a standard arms race equation for the United States for 1952–89 (period of action–reaction between the two potential rivals, see Figure 10.1). This equation permits us to predict the US military expenditures for 1990–7, the difference between the actual and predicted expenditures is the part of the expenditures that can be, in fact, explained by other US objectives. Initial tests for unit roots showed the series to be I (2) in levels (see Appendix A) over the period 1952–97. But that includes the large drop after 1989, which will increase the order of integration. Over the earlier period they seem to be I (1). Moreover, some a priori empirical veriﬁcation indicated that the logarithm transformation was preferred over the level equation. So, American military expenditures seem to depend on their own past military expenditures as well as on past USSR values. Table 10.4 presents estimates of the autocorrelation function for US military spending, which shows the typical characteristics of a non-stationary process. Table 10.4 Autocorrelation and partial correlation for US military expenditures between 1952 and 1989 Self-correlation

Partial self-correlation

Probability

1 0.932 2 0.855 3 0.777 4 0.676 5 0.582

0.932 −0.094 0.13 −0.092 0.037

0.000 0.000 0.000 0.000 0.000

166

Sylvie Matelly

The basic function of US action–reaction: observations and methods Granger causality tests give high values for the F -statistic, demonstrating the interdependencies that exist between the two variables. Nevertheless, American military expenditures seem to be more strongly Granger caused than the USSR expenditures. This remark is relatively interesting for our model. The military spending appear stationary in ﬁrst difference. The Johansen test rejects any cointegration relations between the two variables. In this case and after these various tests, a VAR in ﬁrst difference is appropriate. It is assumed that USSR military expenditures are exogenous and may not have an instantaneous impact on American expenditures. Observations on the American decision, indeed, suggest a delay in the impact of USSR military spending. It is necessary to estimate the lag of reaction of the United States after an USSR action. To evaluate the optimal lag of USSR military expenditures (Table 10.5), three methods are used. Correlation coefﬁcients are calculated between the two series of total military expenditures, by differing systematically Soviet military series. The correlation coefﬁcient is not, however, a test for causality. It does not allow us to conclude on a relationship between variables. It can be just observed that American military expenditures are most strongly correlated with two years before USSR military expenditures. Therefore, the Akaike and Schwarz criteria are estimated from successive lagged regressions from 1952 to 1989, such as MUS(t) = α(1)MUS(t−1) + β(1)MUSSR(t−n) + u(t). The best lag minimises these two criteria. Since there is some uncertainty about the appropriate lag order we experiment with two models: model 1, a relatively parsimonious model and model 2, which allows a richer lag distribution. Model 1: MUS(t) = α(1)MUS(t−1) + β(1)MUSSR(t−2) + u(t), Model 2: MUS(t) = α(2)MUS(t−n) + α(3)MUS(t−n−1) + β(2)MUSSR(t−n) + β(3)MUSSR(t−n−1) + v(t), Table 10.5 Evaluation of the optimal lag for USSR military expenditures Lag, n

Correlation

Akaike criterion

Schwarz criterion

0 1 2 3 4 5

0.812 0.986 0.991 0.991 0.988 0.981

−2.69 −2.73 −2.64 −2.84 −2.93 −3.05

−2.6 −2.65 −2.56 −2.74 −2.84 −2.96

Determinants of US military expenditures

167

where u(t) and v(t) are the residuals of these estimations, corresponding to the difference between observed and estimated military expenditures and assimilated later to grievance elements. The estimation for model 1 is presented in Table 10.6. The best lag for the model 2 must be deﬁned. In this ﬁrst estimation, the tests give acceptable and coherent results. Errors are not apparently serially correlated and are normally distributed; let us think that the global relationship may be plausible and coefﬁcients not biased. However, it is not possible to conclude about a decisive impact of Soviet military expenditures on the American military expenditures (the t-ratio does not indicates a coefﬁcient signiﬁcantly different from zero). Assuming that USSR military expenditures are exogenous is perhaps maladjusted. In the case of simultaneous equations, as arms races are often assimilated (see theoretical models), econometric speciﬁcations rarely use different lags. To evaluate the optimal speciﬁcation of model 2, we use the method adopted by Dunne et al. (1999). Making an initial estimate of a VAR(5) for the ﬁrst differences, the likelihood ratio (LR), the Aikaike information criteria (AIC), Schwarz Bayesian criterion (SBC) and unadjusted LR can permit to determine the best lag for model 2. They suggest a VAR(1), which is relatively consistent with the strong self-correlation discussed earlier (Table 10.7). Giving this lag and again testing the cointegration and Granger causality, it is possible to estimate US military expenditures using the equation shown in Table 10.8. Table 10.6 Model 1 MUS(t) = 0.565MUS(t−1) − 0.104MUSSR(t−2) + 0.03 (4.2) (−1.7) (2.1) Adjusted R 2 : 0.38 Durbin Watson: 1.72 F -statistic: 9.79 Tests for normality of residuals Kurtosis: 4.14 Jarque–Bera: 6.55 Sum squared residuals: 0.09

Table 10.7 Determining the lag of model 2 Order

LL

AIC

SBC

LR

1 2 3 4 5

89.41 93.7 102.78 109.92 111.07

89.66 94.19 103.52 110.90 112.29

89.84 94.55 104.06 111.62 113.19

χ 2 (16) = 5.32 χ 2 (12) = 5.03 χ 2 (8) = 4.82 χ 2 (4) = 4.78

Notes Period: 1952–80. Variables included: MUS and MUSSR.

168

Sylvie Matelly Table 10.8 Model 2 MUS(t) = 0.678MUS(t−1) − 0.277MUS(t−2) + 0.159MUSSR(t−1) (4.6) (−1.8) (1.99) −0.084MUSSR(t−2) + 0.031 (−0.56) (2.02) Adjusted R 2 : 0.447 F -statistic: 10.67 Tests for normality of residuals Kurtosis: 4.25 Jarque–Bera: 6.36 Sum squared residuals: 0.079

With the two equations shown in Tables 10.6 and 10.8, it is possible to forecast the American military expenditures after 1989.

The choice of US basic function and forecast The action–reaction functions estimated in the previous section (models 1 and 2) allows us to predict the US military expenditures for 1990–7 on the basis of the estimated arms race having continued after 1989. Given the fall of USSR military spending, it is expected that the US military expenditures would decrease. However, when the Soviet spending collapsed, the US spending did not decrease (see Figure 10.2). The difference between the estimated and the observed military expenditures can then be interpreted as grievance. The two models ﬁt US military expenditure data well between 1952 and 1989, but neither forecasts well. Model 2 forecasts a larger drop than model 1. The existance of grievance factors can be an explanation of the limited reaction of United States to the Soviet military expenditures collapse. Suppose we estimate an arms race equation of the form: MUS(t) = a + bMUSSR(t) + et , neglecting to include the unobserved grievance term Gt , which appears in the true model: MUS(t) = a(0) + b(0)MUSSR(t) + c(0)Gt + ut . Over the Cold War period the grievance inﬂuences on US military expenditures happened to be correlated with Soviet military expenditures, because the Soviets would respond to US hegemonic motives. Say this took the form: MUS(t) = [a(0) + c(0)a(1)] + [b(0) + c(0)b(1)]MUSSR(t) + [ut + c(0)vt ]. Thus, we overestimate the response of US military spending to the Soviet military spending. However, with the end of the Cold War, grievance continues but military spending drops sharply; that is, b(1) goes to zero and the relationship shifts, causing

Determinants of US military expenditures

290,000

169

Observed military spending Model 1 ME Model 2 ME

240,000 190,000 140,000 90,000 40,000 1952 1955 1958 1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997

Figure 10.2 Observed and estimated military expenditures. (In millions of current US dollars. VAR and AR estimated for 1952–89 and forecast after 1989.) Source: US ACDA, Main Statistical Tables.

the estimated arms race to predict a large fall in US military expenditure that does not materialise.

Other possible factors of American grievance or ambition Grievance and ambition are essential elements in the perception a country has of its general environment and its role in this environment. They are complex and speciﬁc to each country. They can determine not only the arms race but also the long-term dynamics of military expenditures. In fact, it has often been considered that arms race was the consequence of the confrontation of two different systems (capitalism and socialism). Thus, this systemic difference was rarely integrated in arms race models, and the foundations and basic equations are often symmetrical for the two countries. Consequently, it is possible and realistic to suppose that political or economic divergences are factors of grievances or ambitions. Indeed, it groups all structural, historical, economic, institutional or cultural factors that could inﬂuence the military expenditures and international relation of the country. This makes the grievance of a country very difﬁcult to estimate. Twenty-two possible grievance elements are evaluated. They are grouped in ﬁve categories: economic size, military implication, economic opening, national and international political inﬂuence. Characteristics of these series are detailed in Appendix B. Economic size The economic importance of the United States is indisputable. The liberalisation of changes and the end of the Cold War do not modify their economic hegemony. Moreover, it coincides with an unequalled military superiority. The impact of

170

Sylvie Matelly

15 10 5 0 1952

1956

1960

1964

1968

1972

1976

1980

1984

1988

1992

1996

–5 –10 ME/GNP

GNP growth

ME growth

–15

Figure 10.3 US economic size and military implication. (Data in constant price 1995.)

military expenditure on national wealth and growth is often studied in economic analysis. Military expenditures can then limit or on the contrary be an advantage in the creation of wealth. The opposite relationship can equally be envisaged. Figure 10.3 shows the regular reduction of the military expenditures in terms of share in the GNP. The possible impact of the economic size on military expenditures depends on the nature of such expenditures. First, military expenditures are public expenditures. The national production is then the ultimate constraint of this decision. A State cannot spend more than the country produces. It is also the object military expenditures have to protect. At the same time, the level of public expenditures or economic growth can be a more pertinent indicator. Comparative structures As arms race is a competition, it is possible to view it just as one factor of this process. Then, some other relative ratios can be signiﬁcant for the grievance between the two countries. We test, for example, the ratios of US to USSR GNPs and of US to USSR armed forces. The ratio of US to USSR armed force shows the relative importance of the armies. It could be a power indicator for a public decisionmaker. In the arms race context, it translates also structural differences about capital-intensive army. Militarisation The militarisation is probably a consequence of the arms race. In return, it creates some rigidities and inertia that inﬂuence military expenditures and limit their decrease. The ratio of US military expenditures to public expenditures indicates not only the importance of military spending in public decision and the sacriﬁce

Determinants of US military expenditures

171

Table 10.9 American defence budget and public deﬁcit (in billions of US current dollars and % of GNP) Defence budget

Public deﬁcit

1985 294.7 212.1 1990 303.3 218.1 1991 296.2 265.5 1992 287.7 326.6 1993 281.1 226.7 1994 263.3 184.6 1995 267.9 146.2 1996 263.9 110.8 1997 255.1 2.4 Defence programmes for 1997–2002 1998 259.3 1999 264.4 2000 270.9 2001 280 2002 288.3

GNP growth 2.5 1.2 −0.6 2.3 3.1 4 3.1 4.3 3.7

Source: IISS, The military balance, 1996–7.

of a welfare state to pursue arm races, but also the choice of military expenditures as a means for economic policies as illustrated in Table 10.9. The US military expenditures to GNP ratio show the economic importance of military efforts. National political factors The institutional structures can increase or reduce the tendency for a country to be aggressive. Political issues showed the tendency for dictatorial regimes to be more aggressive than democracies (e.g. Bueno de Mesquita and Siverson 1997; Gleditsch and Ward 1998). In the same way, public choice economists underline the importance of bureaucratic inﬂuence and military–industrial complex impact on the military expenditures decision (e.g. Dunne et al. 1984; Hartley 1987). In the case of United States, we test the political ideology of the President as a dummy variable where a democrat president is represented by 1 and a republican president by 2. The support of the Congress to the President can equally inﬂuence military expenditures. This concept is developed and estimated by Wittkopf and McCormick (1998), in a retrospective analysis of congressional executive relations. The implication of the State in the nation can equally determine some public decisions. Thus, the share of public expenditure in the GNP is a possible measure of this implication. The ratio of the accumulated public debt to the GNP can traduce some constraints on the public expenditures. It is essential to be prudent using this debt ratio. The arms race during the Cold War is probably one of the

172

Sylvie Matelly

causes of this debt. But we can imagine that this increasing debt incited public decision-makers to limit military expenditure growth and to look for a détente in international relations. The recent increase in military expenditures coincide with a public surplus budget and can support this idea. Economic opening Trade is traditionally considered as a factor of peace, because it makes conﬂicts more expensive. We can then suppose a negative relationship between grievance or ambition and trade. However, trade increases the economic competition between nations, and can equally increase political tensions as explained by Ward (1982). Political and economic powers are not so different from each other and are signiﬁcantly interdependent. The veriﬁcation of this potential links could be interesting. Six variables are included in the economic opening: total exports, total imports, US opening ratio, US export share of arms in world arms exports, raw materials and chromium imports. Imports of raw materials can be strategic in an arms race period. It is essential for a country engaged in an arms race to be independent of other nations’ resources or arms productions. Therefore, raw materials are indispensable. In the same order of ideas, imports of chromium have been essential during the Cold War. It is the only raw material not produced by the United States. The risk to be dependent on the Soviet Union (one of the main producers) during the Cold War was really strong. It even incited the United States to lift the South African embargo on this product. External political factors From a political viewpoint, the number of conﬂicts or the implication of some countries in conﬂicts can exacerbate the grievances. It is difﬁcult enough to measure these political factors. Polachek (1980) deﬁned, for example, a net frequency of conﬂict that expresses, in fact, the number of conﬂicts during a deﬁnite time period. It allowed him to measure the elasticity of conﬂicts. Many external political factors can inﬂuence the ambitions or grievances of a nation. The United Nations peacekeeping operations and number of nuclear countries can give some indication about international tensions. During the Cold War, the number of Warsaw pact or NATO members could have inﬂuenced the perception of the United States about the balanced forces. A dummy arms race variable is integrated. It corresponds to 1 for the 1952–89 period and to 0 after 1989, when a change in USSR military expenditure is observed. As seen before, many elements can explain the grievance of a State. In the case of our model, it is quite difﬁcult to estimate it for the United States, considering the few years since the end of the Cold War (1990–7). Nevertheless, two tests can give ﬁrst information and suggest future researches: a linear regression and a Granger causality test are successively applied to each possible grievance element for the period 1952–97. When they are not stationary, the variables are tested in a logarithm form. A linear regression can be an appropriate method to select

Determinants of US military expenditures

173

abundant information or numerous variables. It often does not allow to optimise the correlation coefﬁcients but it avoids the singularity of the matrix of coefﬁcients in a estimation of the global function of military expenditures (action–reaction + grievance). The Granger causality test is a different approach: with a classic Fisher test on the nullity of estimated coefﬁcients, it enables to verify the possibility of a relationship between two variables. This concept of causality researches if any past values of exogenous variable can determine the present value of American military expenditures; if one grievance element improves the explanation of the US military spending, it is then possible to assume the explanatory variable causes the military expenditures. This causality is tested by applying a 1-year lag to American military spending since it has been demonstrated previously the optimality of this lag. Table 10.10 gives the results of these tests. The linear regression allows to reject the null hypothesis if the t-ratio exceed 2.3, the Fisher test suggests a Granger causality beyond 3.2. Table 10.10 The possible factors of American grievance between 1952 and 1997

Economic size Gross national product Economic growth Public expenditures Comparative structures Ratio of US to USSR GNP Ratio of US to USSR armed forces Militarisation US military/public expenditures Military expenditures/GNP Economic opening Exports Arms export Imports Opening US raw material imports Chromium imports External political inﬂuence Arms race Peacekeeping operations Number of nuclear countries Members of Warsaw pact Members of NATO National political inﬂuence Support of Congress Public expenditure/GNP Debt/GNP

t-Ratio

Fisher

17.6 1.2 20.29

1.19 0.73 3.13

3.08 1.00

1.89 1.01

8.3 6.9

3.51 2.39

−0.8 23.4 −1.61 9.9 10.6 1.6

0.12 4.44 0.46 2.47 1.19 0.08

5.8 6.32 11 4.2 3.9

7.71 0.24 2.12 0.27 0.38

3.7 7.9 0.9

0.27 3.28 2.93

174

Sylvie Matelly

Conclusion The theories on arms races have often remained obsessed by strategic interactions between two rival countries, limiting the debate on the determination of military expenditures to some actions and reactions between these countries. But other factors, which, following Richardson, we have labelled grievance, are also important. While action–reaction explanations of US military expenditure worked well during the Cold War period, they completely fail to predict post-Cold War military expenditures. This suggests that some of the US military expenditures that were attributed to a reaction to Soviet military expenditures were in fact responses to other determinants. We discussed a range of possible causes. Finally, military expenditures are, even in an arms race process, motivated decisions. It depends more on subjective elements appreciated by decision-makers than on objective ones such as arms stocks or strategic forces. These subjective elements can be deﬁned as variables observable and estimated by these decision-makers but also economic or political factors speciﬁc to each country.

Acknowledgements I am grateful to Professor J. Fontanel for his comments on a preliminary draft of this chapter and to Professor R. Smith for his suggestions, his help and the time he devoted to the various versions of this chapter.

Appendix A Data and sources

US military expenditures USSR military expenditures US gross national product Economic growth Public expenditures Public debt Armed forces Exports

Deﬁnition and unit

Notation

Source

Billions of current dollars

MUS

US ACDA

MUSSR Billions of constant 1995 dollars (GNPt − GNPt−1 ) × 100/GNPt−1 Billions of constant 1995 dollars Thousand of soldiers CAF, billions of constant 1995 dollars

MUSSR

IMF

g MUS De AF X

UNO 1952–74, IMF 1974–97 (continued)

Determinants of US military expenditures

Deﬁnition and unit Imports Opening Raw materials imports Chromium imports Peacekeeping operations Number of nuclear countries Members of Warsaw pact Members of NATO pact

Notation

(X +M)×100/(2×GNP) CAF, billions of constant 1995 dollars

M O M1

Unity

Mc Pk

Number of countries

Nuc

175

Source

US Bureau of mines United Nations publications

Va Na

Appendix B Unit root tests Variables

In level

In ﬁrst difference

In second difference

United States data Military expenditures (ME) Gross national product (GNP) Public expenditures (PE) ME/GNP ME/PE Public debt/GNP PE/GNP Total exports Total imports (M) Opening degree Raw materials imports/M Chromium imports Congress support

−0.17 0.79 1.29 −2.52 −1.86 −2.26 −1.57 −5.68∗∗ −4.07∗∗ −0.43 −0.79 −4.05∗∗ −2.43

−2.44 −4.72∗∗ −4.29∗∗ −5.93∗∗ −4.81∗∗ −2.90 −5.7∗∗

−7.66∗∗

Soviet Union data USSR military expenditures

−1.55

−2.53

United States/Soviet Union ratios Armed forces Gross national products

−1.66 −2.45

−4.06∗∗ −3.31∗∗

−6.53∗∗

−3.98∗∗ −6.81∗∗ −6.69∗∗

Notes Augmented Dickey–Fuller test for Zt = φ1 Zt−1 + bt + c + εt . ∗∗ Signiﬁcantly stationary at 90%.

−4.87∗∗

176

Sylvie Matelly

References Aben, J. (1992) Economie Politique de la Défense. Collection Regards sur notre temps, Édition Cujas. Bourbonnais (1998) Econométrie, Dunod. Brito, D. L. and Intriligator, M. D. (1995) Arms races and proliferation. In: Hartley, K. and Sandler, T. (eds), Handbook of Defense Economics. Pays-Bas, pp. 109–164. Bueno de Mesquita, B. and Siverson, R. M. (1997) Nasty or nice? Political systems, endogenous norms, and the treatment of adversaries. Journal of Conﬂict Resolution 41(1), 175–199. Congressional Budget Ofﬁce (1983) Defense Spending and the Economy. Washington, DC. Coulomb, F. (1998) les Théories economiques de la guerre de la paix et de la défense. Des origines à nos jours. Thèse de Doctorat sous la direction de J. Fontanel, Université Pierre Mendes France, Grenoble. Cromwell, J. B., Hannan, M. J., Labys, W. C., and Terraza, M. (1994) Multivariate Tests for Time Series Models. Sage University Paper Series on Quantitative Applications in the Social Sciences, 7(100). Dunne, J. P., Pashardes, P., and Smith, R. P. (1984) Needs, costs and bureaucracy : the allocation of public consumption in the United Kingdom. Economic Journal 94(3), 1–15. Dunne, P., Nikolaidou, E., and Smith, R. (1999) Arms Race Models and Econometric Applications. Conference on the Arms Trade, Security and Conﬂict. Middlesex University Business School, June. Fontanel, J., Humm, A., and Smith, R. (1985) La substitution capital travail dans les dépenses militaires. Arès numero spécial, pp. 129–144. Gleditsch, K. S. and Ward, M. D. (1998) Democratization and war in the context of time and space. Third Pan-European International Relations Conference and Joint Meeting with the International Studies Association, Vienne. Granger, C. W. (1997) On modelling the long run in applied economics. Economic Journal 107(1), 169–177. Hartley, K. (1987) Reducing defence expenditure: a public choice analysis and a case study of the UK. In: Schmidt, C. and Blackaby, F. (eds), Peace, Defence and Economic Analysis. International Economic Association, SIPRI, MacMillan, pp. 399–423. Intriligator, M. D. and Brito, D. L. (1990) Arms race modeling: a reconsideration. In: Gleditsch, N. P. and Njølstad, G. (eds), Arms Races, Technological and Political Dynamics. International Peace Research Institute, Sage Publication, pp. 58–77. MacGinnis, M. (1991) Richardson rationality, and restrictive models of arms races. Journal of Conﬂict Resolution 35(3), 443–473. MacGuire, M. (1977) A quantitative study of the stratégic arms race in the missile age. Review of Economics and Statistics 59(3), 328–339. Oström, C. W. (1977) Evaluating alternative foreign policy decision-making models. Journal of Conﬂict Resolution 21(2), 235–265. Polachek, S. W. (1980) Conﬂict and trade. Journal of Conﬂict Resolution 24(1), 55–78. Rapoport, A. (1957) Lewis Fry Richardson’s mathematical theory of war. Journal of Conﬂict Resolution 1(2), 249–299. Richardson, L. F. (1960) Arms and Insecurity: A Mathematical Study of Causes and Origins of War. The Boxwood Press, Pittsburgh. United States Arms Control and Disarmament Agency (US ACDA) (1975) Measuring Price Changes of Military Expenditures. US Department of Commerce, Bureau of Economic Analysis.

Determinants of US military expenditures

177

Ward, M. D. (1982) Dynamics of cooperation and conﬂict in foreign policy behavior: reaction and memory. International Studies Quaterly 26(1), 87–126. Wiberg, H. (1990) Arms races, formal models and quantitative tests. In: Gleditsch, N. P. and Njølstad, G. (eds), Arms Races, Technological and Political Dynamics. International Peace Research Institute, Sage Publication. Wittkopf, E. R. and McCormick, J. M. (1998) When Congress supports the President: a multilevel inquiry, 1948–1996. Paper presented at the Third Pan-European International Relations Conference, Vienna. Wolfson, M. and Farrell, J. (1987) Economic warfare between superpowers. In: Schmidt, C. and Blackaby, F. (eds), Peace, Defence and Economic Analysis. International Economic Association, SIPRI, MacMillan, pp. 155–181.

11 Arms race models and econometric applications J. Paul Dunne, Eftychia Nikolaidou and Ron Smith

Introduction During the Cold War, the arms race between the United States and the USSR, which dwarfed all previous ones, was the focus of much concern. Since the end of the Cold War, the focus has shifted to regional antagonisms, but these are still analysed using arms race models developed during the Cold War. The standard framework for empirical study of arms races is the Richardson model described below which explains the time-series pattern of military expenditure between potential enemies in an action–reaction framework. A coupled pair of differential equations explains changes in levels of weapons in each of two nations as a function of the weapons of each side. Once the process has started no country is at fault, the escalation is a consequence of systemic interaction rather than aggression. Despite its popularity the results of the Richardson model has given disappointing results when applied to real data, as Sandler and Hartley (1995: 106–107) note. This is partly because there are real problems in moving from the abstract model to an empirical one. Any calibration requires decisions about the measurement of the variables, functional form, length of lags and expectation formation that are not speciﬁed in the theory. There are also likely to be problems with the quality and reliability of the data which make their use questionable. In addition, the estimation of these models – forward looking, dynamic, simultaneous equation systems – presents its own set of issues and problems. Recent developments in econometrics provide the opportunity to reconsider the empirics of arms race models and why they have been less than successful. This chapter reviews the econometric issues and illustrates them by analysing the Greece–Turkey and India–Pakistan confrontations. The section ‘Econometrics issues in estimating arms race models’ considers the basic Richardson model, its developments and the implications of unit roots and co-integration, covered in more detail in Smith et al. (2000). The section ‘The Greece–Turkey arms race’ then analyses the arms race between Greece and Turkey within a vector autoregression (VAR) framework, illustrating the problems with this type of empirical analysis. The section ‘The India–Pakistan arms race’ repeats the analysis for India and Pakistan. The ﬁnal, section provides some conclusions.

Arms race models and econometric applications

179

Econometrics issues in estimating arms race models The ‘structural form’ of the familiar Richardson model can be written in discrete time with the addition of a stochastic error term as m1t = α1 + β1 m2t + γ1 m1t−1 + ε1t ,

(11.1)

m2t = α2 + β2 m1t + γ2 m2t−1 + ε2t ,

(11.2)

where mit is some measure of military preparedness of country i in year t (i = 1, 2). Richardson interpreted αi (i = 1, 2) as exogenous ‘grievance’ terms, βi > 0 as ‘reaction’ terms and γi < 0 as ‘fatigue’ terms. We assume that: E(εit ) = 0, E(εi2 ) = σi2 , E(εit εj t ) = σij , E(εit εj t−s ) = 0, where s = 0 and i, j = 1, 2. These structural shocks, εit , will be driven partly by idiosyncratic factors (events in former Yugoslavia and the Balkans in general for Greece and the conﬂict with the Kurds for Turkey) and partly by common factors (events in the former Soviet Union or NATO modernisation or both). So, we would not expect the structural shocks to be independent. This form is structural in that current values of the endogenous variables appear on the right-hand side of the equation. Sets of equations with this general form can be derived from a variety of different theories, for example, see the discussion in Brito and Intriligator (1995) or Levine and Smith (1997). The structural parameters, αi , βi and γi , will be functions of the deep parameters of the system. For instance, in Levine and Smith (1997), they are functions of depreciation rates for weapons, strategic parameters which describe the nature of conﬂict, discount rates and the elasticity of substitution between security and consumption. Even if one can estimate the structural parameters consistently, this may not allow one to recover the underlying deep parameters. The reduced form of the system, in terms of pre-determined, lagged variables can be written in the vector error correction model (VECM) form of a ﬁrst-order VAR as m1t = δ11 + δ12 m1t−1 + δ13 m2t−1 + u1t ,

(11.3)

m2t = δ21 + δ22 m1t−1 + δ23 m2t−1 + u2t ,

(11.4)

where E(uit ) = 0, E(u2i ) = ωi2 , E(uit uj t ) = ωij , E(uit uj t−s ) = 0, where s = 0 and i, j = 1, 2. Smith et al. (2000) discuss the relationship between the two forms. There is Granger (1969) causality from m1 to m2 if δ22 = 0 and from m2 to m1 if δ13 = 0. Each equation of the reduced form can be estimated consistently by least squares. In the theory the military expenditure variables are treated as stationary variables, integrated of order zero I (0).1 If the variables are I (1), or equivalently contain a stochastic trend, then there is a danger of spurious regression. In a regression of one I (1) variable on another, the R 2 tends to unity with the sample size and the

180

Dunne et al.

t ratio to a non-zero value, even if the two series are unrelated. The requirement for the regression not to be spurious is that the two variables cointegrate. If this is the case then the process can be represented as a restricted form of the error correction model above. If the long-run relationship is m1t = βm2t then the disequilibrium or error correction term is measured by zt = m1t − βm2t . The VECM then takes the form: m1t = δ11 + α1 zt−1 ,

(11.5)

m2t = δ21 + α2 zt−1 ,

(11.6)

where the feedbacks are stabilising if α1 < 0, α2 > 0. Estimation and testing of the co-integrating vectors can be done in a number of ways, including within the maximum likelihood framework suggested by Johansen (1988). Unit root tests and co-integration have been widely adopted, Kollias and Makrydakis (1997) is an arms race example. However, there are a number of problems with the techniques. Both the tests for unit roots (used to determine the order of integration) and the tests for co-integration tend to have low power, so determining the order of integration and co-integration is not straightforward. The tests are also sensitive to the choice of lag order, the treatment of serial correlation, the treatment of the deterministic elements, the presence of structural breaks and various other factors. There are also questions of interpretation, since the order of integration is not a structural property of the series but a description of the time-series properties of a sample. Series which appear I (1) on short spans of data often appear I (0) on long spans, where span refers to the length in time of the series not the number of observations. Over centuries of data, the UK share of military expenditure is clearly I (0), over shorter spans it appears to be I (1). While co-integration allows us to estimate the long-run equilibrium, it does not help in identifying the short-run structural interaction. A major problem with the Richardson model is the lack of a budget constraint. This can be dealt with by, for example, including GDP to reﬂect income. Care needs to be taken in including variables within a VAR, however, as the number of parameters grows very rapidly with the number of variables and the number of lags and the small sample properties of large VARs are rather poor. In addition, inference, for example, Granger causality tests, tends to be sensitive to speciﬁcation, including or excluding variables can change the results. However, adding income may provide more plausible identifying restrictions. For instance, it is possible that countries adjust their military expenditure in response to their own GDPs, but not to the other countries GDP, or have an arms race in shares of military spending in GDP (military burdens). There is a variety of interesting testable system restrictions on the VAR, for example, exogeneity of income, levels versus shares. In principle, starting from a general system and testing system wide restrictions is an appropriate way to develop a model of the arms race process. In practice, it can be difﬁcult to ﬁnd theoretically coherent and statistically acceptable speciﬁcations on a large system through such a procedure. Ad hoc deletion of individual insigniﬁcant coefﬁcients

Arms race models and econometric applications

181

is also unsatisfactory because what look like acceptable single equation restrictions can produce unacceptable systems properties. A further consideration is the expectations process, which is left, unspeciﬁed in the theoretical model. Within the framework discussed above there are a number of possible interpretations, which are discussed in Smith et al. (2000). Having brieﬂy outlined the issue we now consider two applications, Greece– Turkey and India–Pakistan. In any empirical application we have to determine: 1 2 3 4 5 6 7

The number of variables analysed, military expenditures alone or with income. The transformation of the variables. Linear or logarithmic, for instance. The order of integration of the variables. The order of the VAR, number of lags included. The treatment of deterministic elements, dummy variables, trends and intercepts, in the co-integrating VAR. The number of co-integrating vectors The just identifying restrictions required to interpret the co-integrating vectors and any over-identifying restrictions.

As we shall see, this gives a large possible parameter space to search over.

The Greece–Turkey arms race There is considerable debate over the Greece–Turkey arms race as previous studies have given mixed results. Majeski and Jones (1981) and Majeski (1985) using causality analysis, tested for interdependence in the military expenditures of Greece and Turkey for 1949–75 and their results indicated the presence of instantaneous causality. Kollias (1991) applied the classical Richardson model for the two countries over the periods 1950–86 as well as over 1974–86 (the period after the Turkish invasion of Cyprus), but his results were very poor and did not indicate the existence of an arms race. However, by employing speciﬁc indices of military capabilities, he found that Greek military expenditure depends on Turkish military expenditure and on the relative size of the arms forces. Also, Kollias and Makrydakis (1997), using co-integration and causality tests, found evidence of a systematic armaments competition between Greece and Turkey over the period 1950–95. Refenes et al. (1995) using neural networks and indices of military capabilities (ratio of armed forces and military expenditures per soldier) examined the hypothesis of an arms race between Greece and Turkey over 1962–90 and found that Turkey’s quantitative advantage is the most signiﬁcant external security determinant of Greek military expenditure. On the other hand, there are studies that do not provide strong evidence of an arms race between the two countries. Georgiou (1990) tested the hypothesis of an arms race over the period 1958–87 but could not ﬁnd any evidence of the existence of an arms race and Georgiou et al. (1996) using a VAR speciﬁcation ended up with similar conclusions for the period 1960–90. Using SIPRI data (see Figure 11.1) on military expenditure for the two countries a VAR model of the arms race was speciﬁed.2 Initial tests for unit roots

182

Dunne et al. 7,000

MG

6,000

MT

Million $

5,000 4,000 3,000 2,000 1,000 1996

1993

1990

1987

1984

1981

1978

1975

1972

1969

1966

1963

1960

0

Years

Figure 11.1 Military expenditure for Greece and Turkey (in constant 1990 million US$). Source: SIPRI Yearbooks.

showed the series to be I (1). The Turkish invasion of Cyprus had a marked impact on Greek–Turkish relations and this is modelled by a step dummy CD (Cyprus dummy) which takes the value of 0.5 in 1974, of 1.0 for 1975–9 and zero otherwise. Non-nested tests indicated that equations using the logarithms of the data ﬁt better than ones using untransformed data. Starting from a lag length of 5, using the logarithms of military expenditure and using CD as an exogenous variable, adjusted likelihood ratio (LR) tests and the Schwarz Bayesian criterion (SBC) indicate a ﬁrst order VAR, though the Akaike information criteria (AIC) and unadjusted LR tests suggest longer lags. Given the length of the time series we continue with a VAR(1) and investigate whether or not the variables are co-integrated. Using a ﬁrst-order VAR with unrestricted intercepts and restricted trends over 1961–96 trace and eigenvalue tests suggested one co-integrating vector at the 5 per cent level between the logarithms of military expenditures. Normalising on the log of Greek military spending, LMG, the co-integrating vector was: LMGt = −29.1 LMTt , plus trend and Cyprus dummy. This has the wrong sign on the logarithm of Turkish military spending. All the results were very sensitive to speciﬁcation and sample period, suggesting that the estimates are fragile. One possible source of mis-speciﬁcation is failure to take account of the budget constraint. To consider this a VAR model in the logarithms of military expenditure and GDP (which were calculated from SIPRI ﬁgures for military expenditure and shares), was estimated. All variables were tested for unit roots and were found to be I (1) and non-nested tests again suggested that the logarithmic equation ﬁt better than the untransformed equation. In this case the SBC indicated a second-order VAR. A joint LR test for the exclusion of the GDP variables χ 2 (8) = 42.9,

Arms race models and econometric applications

183

which is well above the 5 per cent critical value, suggesting that income is important. Assuming unrestricted intercepts and restricted trends in a VAR(2) with the Cyprus dummy over 1962–96, trace and eigenvalue tests again suggested one co-integrating vector. Again normalising on Greek military expenditure gave the co-integrating vector: LMGt = 4.01 LMTt + 0.70 LYGt − 1.38 LYTt plus trend and Cyprus dummy. This has the right signs but an implausibly large coefﬁcient on Turkish military expenditure. It is possible that this is a relation in shares and this was tested by imposing the implied over-identifying restrictions. This gave: (LMG − LYG) = 4.98 (LMT − LYT) plus trend and Cyprus dummy. Again the coefﬁcient on the logarithm of the share of Turkish military expenditure is implausibly large. The share restrictions were rejected by the data χ 2 (2) = 12.8. The VECM estimates for the just identiﬁed system (t ratios in brackets) are: Gt = 2.42 + 0.10Gt−1 − 0.06Tt−1 − 0.58Yt−1 (2.74) (0.55) (0.39) (2.17) − 0.26TYt−1 + 0.10Zt−1 + 0.12CDt (1.36) (2.68) (2.10) R 2 = 0.29; SER = 0.09 Tt = 5.10 − 0.27Gt−1 + 0.37Tt−1 − 0.59Yt−1 (6.83) (1.86) (2.64) (2.61) − 0.62TYt−1 + 0.21Zt−1 + 0.24CDt (3.80) (6.76) (4.91) R 2 = 0.69; SER = 0.07. There are also equations for GDP which we do not report. For the Greek equation, four of the seven coefﬁcients are signiﬁcant at the 5 per cent level. The coefﬁcients on the error correction term, lagged Z, measure the speed at which any disequilibrium is removed, though the adjustment in the Greek case is in the wrong direction. The speciﬁcation easily passes the tests for ﬁrst-order serial correlation, functional form, normality and heteroscedasticity. The equation for Turkey is a better speciﬁcation in terms of the coefﬁcient estimates, with six of the seven signiﬁcant, but fails the tests for functional form χ 2 (1) = 7.68. Cusum and Cusum squared tests suggest structural stability, while the persistence proﬁles to system-wide shocks are fairly similar, The persistence proﬁle of system-wide shocks to the CVs show a relatively fast convergence, around 3 years. Again, there seems to be some evidence of a co-integration, but not in the form of a long-run arms race model. Experiments with a variety of different formulations

184

Dunne et al.

did not reveal a robust arms race relationship. The results tended to be difﬁcult to interpret and extremely sensitive to sample and speciﬁcation. Given their antagonistic interaction and their reactions to each others military preparations, there has undoubtedly been an arms race between Greece and Turkey. But that arms race has not taken the form of stable Richardson type reaction functions. Given all the other factors intervening in their interaction, this is not perhaps surprising.

The India–Pakistan arms race Unlike Greece and Turkey where the literature is ambiguous, previous studies of India and Pakistan, for example, Deger and Sen (1990) have found evidence of a Richardson type arms race in the subcontinent. Given the results for Greece– Turkey, the starting point for India and Pakistan was a VAR in military expenditures and GDP. The GDP ﬁgures were again calculated from the SIPRI ﬁgures for military expenditures (see Figure 11.2 for the levels of military expenditure) and shares. All these variables appear to be I (1), though shares of military expenditure in GDP appear to be I (0). Both model selection criteria and LR tests indicate that a second-order VAR is appropriate. Both non-nested tests and likelihood criteria indicated that a linear model ﬁts better than a logarithmic one. In the linear second order VAR in the four variables, the LR test statistic for the hypothesis that the income variables do not appear in the military expenditure equations is 15.4. This is just below the asymptotic 5 per cent critical value for a χ 2 (8), which would be appropriate if the variables were stationary. Given that the small sample critical values for non-stationary variables would be rather larger, this suggests that we can treat the income variables as Granger non-causal with respect to military

9,000 MI

Million $

8,000

MP

7,000 6,000 5,000 4,000 3,000 2,000 1,000 1996

1993

1990

1987

1984

1981

1978

1975

1972

1969

1966

1963

1960

0

Years

Figure 11.2 Military expenditure for India and Pakistan (in constant 1995 million US$). Source: SIPRI Yearbooks.

Arms race models and econometric applications

185

expenditures and work with a two-variable VAR. There is an element of judgement in this, since the theory suggests that income should be included to capture the budget constraint and some income variables are individually signiﬁcant in the Indian military expenditure equation. Using unrestricted intercepts and no trends (which were insigniﬁcant) in a second-order VAR, 1962–96, the trace and eigenvalue tests both clearly suggest one co-integrating vector at the 5 per cent level. The eigenvalue test statistic is 26.3, 5 per cent critical value 14.9, trace 26.4 and 17.9. Normalising on Indian military expenditure, the co-integrating vector is MIt = 2.008Pt , which indicates that the long-run relationship is for India to spend about twice the Pakistani level. The estimated VECM (with t ratios) is: It = 469.0 + 0.43It−1 − 0.08Pt−1 − 0.37Zt−1 (3.36) (2.41) (0.17) (2.70) R 2 = 0.23; SER = 343.0 Pt = −62.0 − 0.11It−1 + 0.438Pt−1 + 0.14Zt−1 (1.45) (2.05) (3.07) (3.36) R 2 = 0.37; SER = 105.0. Apart from the intercept in the Pakistan equation and lagged Pakistani spending in the Indian equation all the coefﬁcients are signiﬁcant. Both equations pass tests for ﬁrst-order serial correlation, non-linearity, normality and heteroscedasticity at the 1 per cent level; though Pakistan fails on normality and India on heteroscedasticity at the 5 per cent level. Cusum and Cusum squared tests indicate that both equations are structurally stable. In terms of changes the degree of explanation is low, though the equations explain over 95 per cent of the levels of military expenditure. The coefﬁcients on lagged Z, which measure the speed at which disequilibria are removed, are both of the correct sign and indicate that India adjusts to disequilibrium faster than Pakistan. There seems to be a degree of over-reaction, represented by the negative coefﬁcients on the other countries lagged changes. Convergence back to equilibrium is cyclical, but quite rapid, with adjustment complete in about six years. There is quite a high positive correlation (r = 0.46) between the errors in the two equations, indicating a degree of instantaneous feedback. Unlike the Greece–Turkey case, the results seem quite robust and not very sensitive to sample or the details of speciﬁcation.

Conclusions This chapter has considered some of the econometric issues involved in estimating arms race models and has illustrated the problems using the examples of Greece and Turkey and India and Pakistan. At a common-sense level one would regard both dyads as being involved in arms races, yet in one case it takes the classical Richardson form and in the other it does not. In the case of Greece and Turkey, estimating a VAR with the logarithms of military spending and income for the two

186

Dunne et al.

countries fails to ﬁnd any reasonable result. In contrast, estimating a VAR for India and Pakistan is much more straightforward. There is one co-integrating relation between the two countries which suggests that the long-run relationship is for India to have twice as much military spending as Pakistan. Whereas Greece–Turkey estimates seem fragile, India–Pakistan estimates seem robust. Why should there be such differences between the two cases is unclear. There are a number of possibilities. First, the difference may reﬂect the security environment. Turkey and Greece might be considered to be in a more complex environment. Both are members of NATO and alliance effects Murdoch may be important (1995). Both have other security concerns apart from each other. Both faced a Soviet threat for much of the sample and Turkey has faced a continuing Kurdish question. India and Pakistan might be considered a more straightforward confrontation between two states. However, India has confronted China and Pakistan has had disputes with Afghanistan and Bangladesh. Second, the difference may reﬂect different government and decision-making structures. Following the British model, both India and Pakistan have similar decision-making processes which may make them react consistently to changes in their protagonists military spending. This would make them more amenable to modelling than Greece and Turkey with their very different state structures and history. Third, rivalry takes many different forms and there may be substitution between alternative instruments of rivalry depending on the strategic and political situation. This may mean that the optimal response to military spending by the other country may be investment in some other instrument than military spending, for example support for terrorism in the other country. An interesting issue is whether the historical reaction functions between Indian and Pakistan remain stable following their nuclear explosions, which may change the form of rivalry. Finally, the data are not very reliable and are possibly massaged by the states for their own purposes. However,it is not clear why this should explain the difference in reaction functions for the two cases.

Acknowledgements This chapter is a revised version of a paper presented to a conference on ‘The Arms Trade, Security and Conﬂict’, Middlesex University Business School, June 1999. We are grateful to Paul Levine for comments and Dunne is grateful to the ESRC for support under grant R000239388.

Notes 1 A variable is said to be I (d), integrated of order d, if it must be differenced d times to become covariance stationary. A variable is said to be covariance stationary if its expected value, variances and autocovariances are all constant, perhaps after the removal of a deterministic trend. An I (0) variable is thus stationary. 2 There were some problems with the data as the Turkish GDP were revised and this led to some quite marked changes in the SIPRI estimated shares. Speciﬁcally, in 1998 SIPRI Yearbook the shares are much smaller than those reported in previous yearbooks. This difference in the shares ﬁgures is not due to a change in the levels of military expenditure

Arms race models and econometric applications

187

but to revisions in GDP series. The most recent data was used. Further detail is available from the authors on request.

References Brito, D. L. and Intriligator, M. D. (1995) Arms races and proliferation. In: Hartley, K. and Sandler, T. (eds), Handbook of Defence Economics, Elsevier Science, Amsterdam, Chapter 6. Deger, S. and Sen, S. (1990) Military security and the economy: defence expenditure in India and Pakistan. In: Hartley, K. and Sandler, T. (eds), The Economics of Defence Spending, Routledge, London, Chapter 9, pp. 189–227. Georgiou, G. (1990) Is there an arms race between Greece and Turkey? Some preliminary econometric results. Cyprus Journal of Economics 3(1), 58–73. Georgiou, G., Kapopoulos, P. and Lazaretou, S. (1996) Modelling Greek–Turkish rivalry: an empirical investigation of defence spending dynamics. Journal of Peace Research 33(2), 229–239. Granger, C. W. J. (1969) Investigating causal relations by econometric models and crossspectral methods. Econometrica July, 424–438. Johansen, S. (1988) Statistical analysis of cointegrating vectors. Journal of Economic Dynamics and Control 12, 231–254. Kollias, C. (1991) Greece and Turkey: the case study of an arms race from the Greek perspective. Spoudai 41(1), 64–81. Kollias, C. and Makrydakis, S. (1997) Is there a Greek–Turkish arms race? Evidence from cointegration and causality tests. Defence and Peace Economics 8(4), 355–379. Levine, P. and Smith, R. P. (1997) The arms trade and the stability of regional arms races. Journal of Economic Dynamics & Control 21, 631–654. Majeski, S. (1985) Expectations and arms races. American Journal of Political Science 29, 217–245. Majeski, S. and Jones, D. (1981) Arms race modelling. Causality analysis and model speciﬁcation. Journal of Conﬂict Resolution 25, 259–288. Murdoch, J. C. (1995) Military alliances: theory and empirics. In: Hartley, K. and Sandler, T. (eds), Handbook of Defence Economics, Elsevier Science, Amsterdam, Chapter 5. Refenes, A. N., Kollias, C. and Zarpanis, A. (1995) External security determinants of Greek military expenditure: an empirical investigation using neural networks. Defence and Peace Economics 6, 27–41. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge. Smith, R. P., Dunne, P. and Nikolaidou, E. (2000) The econometrics of arms races. Defence and Peace Economics 11, 31–45.

12 Demand for space expenditure and the space race between NATO and the USSR An econometric analysis Vasilis Zervos Introduction There has been no attempt to simultaneously estimate the demand for space expenditure (SE) for Europe, the United States or Russia, despite the relatively large sums of public money allocated by all three economic areas to their space programmes. This chapter aims at further understanding the process of budget allocations by the European, US and Russian public sectors to their civil and military space programmes. The hypothesis examined in this chapter is that there is a space race between the United States and Russia and no European–US space alliance. Assuming there is a European–US military alliance, and the above hypothesis is veriﬁed, then either the Europeans are free riding on US space capabilities within NATO, or there is an interalliance specialisation, with the United States providing the space component. Given that the characteristics of space goods for all three geographical areas are shown to be quite similar to defence goods the basic theoretical models from the area of defence economics of alliances and arms races (Sandler and Hartley 1995) will be adapted to model the behaviour of the space expenditure (SE) of the Western European countries, the United States and the USSR. These theories are adapted to test for the presence of a space race between the United States and the Soviet Union, or for whether there are space alliance effects within NATO. A general model is then formulated to test the hypothesis following a general to speciﬁc econometric approach. In the following section of this chapter, the sample selection and the deﬁnitions of the relevant variables are presented, followed by the theoretical background and the estimations and subsequent analysis for the space race between the United States and the Soviet Union.

Characteristics of SE in NATO and the USSR NATO is a military alliance and in exploring for alliance impacts on SE there are two important methodological issues to consider in this section. The ﬁrst is whether there should be any distinction between commercial SE and government civil and military space expenditure (MSE) for the purposes of this analysis. The

Space race between NATO and the USSR

189

second refers to the deﬁnition of NATO and the resulting selection of countries. Throughout this chapter only public expenditures are used for two reasons: • •

Government SEs are much higher than commercial SE, while during the ‘Cold War’ era, commercial space expenditures were negligible. The strategic interactions and behaviour of the SE of the United States, Europe and USSR/Russia have more to do with national considerations and the ﬁnancing of public goods than with commercial considerations and markets-speciﬁc factors.

The joint examination of military and government civil space programmes within this framework is justiﬁed by the fact that the Western European countries and the United States formed a military (NATO) as well as an economic alliance against the Soviet Union, with space beneﬁts in the form of meteorological, scientiﬁc, remote sensing and other data from US space programmes spilling to Western European countries and vice versa, as well as technological spill-ins which were restricted from transferring to the USSR.1 The deﬁnition of the country sample for ‘Western Europe’ is crucial in the formation of the testable hypothesis in this chapter and as such is justiﬁed at this early stage. With respect to military expenditure European (or European Union (EU)) countries in this empirical analysis are deﬁned as France, Germany, United Kingdom, Italy and Spain, while with respect to space expenditure all ﬁfteen EU countries are included under ‘EU’, or ‘Europe’. This is justiﬁed on the fact that these ﬁve countries are the major European military powers and in the other ten countries are included neutral countries and smaller countries whose military expenditure reﬂect regional effects. In addition, France, Germany, Italy, Spain and the United Kingdom are the major countries in Europe with space capabilities, and with the exception of France2 are major European countries, members of NATO.

Demand for SE: the theoretical framework In the case where there was evidence of a space race between the United States and the Soviet Union, the determinants of the behaviour of the space expenditure3 of Western European countries would not include the SE of the United States, or the military expenditure (ME) of Western Europe, or the United States.4 In particular: (a) if the SE of the United States is shown to be determined by the SE of the Soviet Union and the ME of the United States; and (b) if the SE of Western Europe is shown not to be determined by its ME or the SE of the Soviet Union; and assuming that the ME of Europe and the United States are determined within the NATO alliance, then as far as the role of Europe is concerned, either the Europeans are free-riding on the US space capabilities within NATO, or the SE of the United States is complementary to the NATO alliance within an interalliance

190

Vasilis Zervos

specialisation. In the case where there is a space race between the United States and USSR and no alliance in space within NATO then (a) and (b) hold true. A general estimation model This analysis follows a general to speciﬁc estimation methodology. This is because there are certain advantages in estimating a general model of the following form. The main advantage rests with the fact that estimations of space alliances and space race models can be mutually inconsistent. The estimation of a space alliance model if the true model is one of a space race leads to incorrect interpretations. Consider a space race model, where following the Richardson model (Sandler and Hartley 1995: 83) there are two rivals competing for supremacy in space: NATO and the USSR. The model consists of two differential equations, each one linking the rate of spending on space programmes by each rival to the stock of spending on its own space programmes in a negative way and to the stock of its rival’s space programmes in a positive way (see Dunne et al. 2003). A long-term equilibrium where the rate of spending would approach zero can be estimated by the following model of equations: aSEUSSR = bSENATO + c,

(12.1)

gSENATO = hSEUSSR + e,

(12.2)

where a, b, g, h, are all assumed greater than zero, and SENATO is the sum of the SE of the United States and the EU. There are no sign restrictions on c, or e, as either or both rivals might have political or historical constraints (e.g. the United States has the policy of pursuing ‘leadership in space’). The coefﬁcients c and e (grievance factors) capture domestic explanatory variables for each rival, such as political pressures from domestic public opinion, and ‘ﬂag-carrier’ effects.5 Assume that the true model is this space race model, but a space alliance model of the following form is estimated for a NATO (US–EU) space alliance in rivalry with the USSR (see Okamura 1991; Sandler and Hartley 1995: 36): SEEU = fEU (YEU , PEU , SEUS , SEUSSR ),

(12.3)

SEUS = fUS (YUS , PUS , SEEU , SEUSSR ),

(12.4)

where SE is total space expenditure by the public sector (budget), Y is income, and P is price of space goods relatively to non-space goods. This set of equations describes the SE of the United States and Europe within a NATO space alliance as a function of each ally’s own income, domestic price of space goods, the SE of the other ally (spill-in) and the rival’s SE.6 If the true model is the space race model of Eqs (12.1) and (12.2) and the space alliance model of Eqs (12.3) and (12.4) is estimated, a negative coefﬁcient when regressing, for example, SEUS on SEEU , would be compatible with the space race model and not indicate the existence of a reaction function within the NATO

Space race between NATO and the USSR

191

alliance. Thus, the estimation of the space alliance model where the coefﬁcient of the ‘ally’ is negative would be interpreted as a case of ‘free-riding’, while this would not be the case. To see why this could be the case, consider the following illustrative example: from Eq. (12.1) assume that the true model is the following space race model: SEUSSR = (b/a)SEUS + (b/a)SEEU + c,

(12.5)

then estimating Eq. (12.1) would result in the spill-in variable having a negative coefﬁcient, while the income variable would be insigniﬁcant (notice that in Eq. (12.6), the coefﬁcient of income ‘k’ equals zero, because it is assumed that the space race model is the true model): SEEU = (kb/a)YEU − SEUS + (a/b)SEUSSR − (ca/b).

(12.6)

This example illustrates how the income variable is a major determinant in interpreting the coefﬁcient of spill-ins, with respect to whether there is a case of free-riding in the alliance model, or whether the behaviour of the variables is better explained by a space race model. A way to examine the validity of a space alliance model therefore is to follow a general-to-speciﬁc approach and estimate the general model: SEEU = fEU (YEU , PEU , SEUS , SEUSSR ),

(12.7)

SEUS = fUS (YUS , PUS , SEEU , SEUSSR ),

(12.8)

SEUSSR = fUSSR (YUSSR , PUSSR , SEEU , SEUS ),

(12.9)

then substituting Eq. (12.9) in Eq. (12.7) or (12.8) and solving for the reduced form coefﬁcients would indicate whether there is free-riding or not (a negative reduced form coefﬁcient of the ‘ally’ variable would indicate the presence of free-riding within an alliance). The space race and the space alliance model can be mutually inconsistent with respect to their respective ﬁndings from empirical applications. What is required is a method to choose between the two modelling approaches, such as examining the empirical results of the two rival models for the SE and comparing their ability to ﬁt the data. The general model Eqs (12.7)–(12.9) can be estimated with 2SLS and the signiﬁcance of the income coefﬁcients and the spill-ins coefﬁcient can be examined. For example, an insigniﬁcant income coefﬁcient would indicate that the space race model is a better ﬁt than the alliance model. If this was the case, then this would change the interpretation of the results, since the alliance theory actually describes a behavioural relationship, unlike the Richardson model-based space race model, which provides little information and insight in the underlying forces causing and determining the race (for a critique, see Sandler and Hartley 1995: 87). Finally, the coefﬁcient of the income variable indicates whether the good is a normal or inferior good: in the case of a normal good the coefﬁcient would be signiﬁcant

192

Vasilis Zervos

and positive in value, while an inferior good would have a signiﬁcant and negative coefﬁcient. It must be noted, however, that the estimation of such a model will probably suffer from serious multicollinearity problems. This is due to the fact that if an alliance effect exists, then the SE of the allies are expected to be correlated with each other and with the ‘rival’ variable. Thus, the presence of two highly correlated variables as explanatory variables in each of the Eqs (12.7)–(12.9) could lead to serious multicollinearity, a point further discussed when the empirical results are presented.

Estimation of the space race between the United States and the USSR The general model Eqs (12.7)–(12.9) was estimated using the 2SLS method. For this analysis, the use of unit roots and cointegration will be applied later in this chapter (see Linden 1995; Dunne et al. 2000). The application of vector autoregression (VAR) as employed in Dunne et al. (2003) for an arms race between two areas (with military expenditure and income variables) was not employed, despite its theoretical beneﬁts in providing information over the preferred method with respect to short-run dynamics, because of the relatively large number of variables in the general model, incorporating all three areas. The model was estimated in linear and log-linear form, using GNP per capita as a proxy for income, and assuming similar cost reﬂecting prices throughout the three economic areas. As with the previous defence model estimations, this assumption is unavoidable, due to lack of data on cost reﬂecting prices on space goods for the United States, EU and USSR. The sample data consists of annual observations ranging from 1972 to 1988, a decision justiﬁed on two grounds: •

•

Prior to the early 1970s, European collaborative space efforts were at an experimental stage. This ended with the formation of European Space Agency (ESA), which leads to the fact that collaborative European space programmes and expenditure through ESA are available only post-1972. Prior to 1972 ‘European space expenditure’ was comprised of the sum of largely uncoordinated funding to national space programmes by the respective European nations. At the early stages of the development of the space sectors of the United States and USSR/Russia, the element of competition in space was predominant with the Soviets launching ‘Sputnik’ in the late 1950s, followed by the US ‘Apollo’ programme. During these initial space programmes, the basic infrastructure for the space programmes of the United States and the Soviet Union was created. This results in the respective early data on SE (prior to 1972) depicting abrupt changes, and lacking reliability due to the increased secrecy involved (see Lovell 1973). This data therefore cannot be used to estimate long-run equilibrium conditions depicted in the space alliance and space race equations.

Space race between NATO and the USSR

193

Despite these problems associated with the United States and Soviet data, the full sample range available for the respective time series will be used (from 1961) in estimations where the European SE is absent. The results of the best estimations are given in Eqs (12.10)–(12.12) (t-values in parenthesis). The estimation method is 2SLS, which eliminates problems of simultaneity of the endogenous variables (unlike OLS). Since the three equations presented are exactly identiﬁed, the 2SLS estimators obtained are identical with the ones that would be obtained using full information maximum likelihood (FIML) and 3SLS. The log-linear data estimations are given below for data range from 1972 to 1988 (t-ratios in parenthesis): log SEEU = 7.739 − 1(log YEU ) + 0.825(log SEUS ) + 0.17(log SEUSSR ), (0.45) (−0.47) (1.26) (0.44) = 0.11 (12.10) log SEUS = −8.56 + 1.525(log YUS ) + 1.187(log SEEU ) − 0.568(log SEUSSR ), (−0.59) (0.58) (0.919) (−1.14) = 0.18 (12.11) log SEUSSR = 24.51 − 2.90(log YUSSR ) + 1.513(log SEEU ) − 0.0288(log SEUS ) (1.85) (−1.67) (1.36) (−0.03) = 0.10 (12.12) log lik = 122.147. The overall performance of the model is poor with no signiﬁcant coefﬁcient whatsoever in either of the three estimating equations (based on the t-statistic which is in all cases less than the absolute value of 2). Furthermore, expanding the sample range from 1970 to 1989, and from 1970 to 1990 did not improve the performance of the model. Estimating the model using OLS for each equation, as well as conducting estimations of both 2SLS and OLS using ﬁrst differences data did not improve the performance of the model. Having ascertained the inability of a space alliance model to explain the variation of the SE of Europe, the United States and USSR/Russia, the next step is to test whether a space race between the United States and the Soviet Union, or/and a space race model between Europe and the Soviet Union can explain the SE of the respective areas. Following a general-to-speciﬁc approach, the structure of the tested model changes into a space race model. In estimating such a model a version of the long-run equilibrium condition of a space race depicted by Eqs (12.1) and (12.2) can be used, where SENATO is replaced by SEUS : SEUSSR = nSEUS + c,

(12.13)

SEUS = pSEUSSR + e.

(12.14)

194

Vasilis Zervos

The estimation of Eqs (12.13) and (12.14) with the use of a simultaneous equations method such as the 2SLS suffers from problems of identiﬁcation. There are two solutions to this problem: The ﬁrst is to add ‘domestic-speciﬁc’ variables, as depicted by ‘c’ and ‘e’, which has the drawback of deciding on which are the most relevant variables and theoretically justifying their inclusion. This decision can, however, completely change the structure of the estimated equations.7 In the absence of a theoretical framework to justify the choice of the ‘grievance factors’, the second solution is the preferred method, which is to estimate either Eq. (12.1) or (12.2) using OLS. This results in loss of information due to the simultaneous nature of the model, but compensates this with a less complicated estimation and theoretical interpretation. To avoid such problems one of Eq. (12.13) or (12.14) was chosen to be estimated with the use of OLS. A way to choose between Eq. (12.13) and (12.14) as the estimated equation is the use of Granger causality tests (which determine precedence, rather than causality), to indicate which of the two SE variables ‘Granger cause’ the other (see Smith et al. 2000). For the case where the space race is modelled between Europe and the USSR, however, such analysis provided quite poor results in terms of explaining the data, with the results of the best estimation (using a sample range from 1970 to 1990 and OLS method of estimation) presented in Table 12.1 for reasons of completeness. The tests for mis-speciﬁcation for the above estimation for the null hypothesis of autocorrelation, autoregressive conditional heteroskedasticity (ARCH) and normality of the residuals, give fairly satisfactory results, since in all cases the null is rejected at 5 and 10 per cent levels of signiﬁcance. However, the information of normality of the residual distribution and the test for functional form mis-speciﬁcation also presented give mixed results, since although the residual seems to be following a standard normal distribution, the null hypothesis of the RESET test of no functional form mis-speciﬁcation is rejected at the 10 per cent level of signiﬁcance (Doornik and Hendry 1995: 395), indicating serious problems of mispeciﬁcation of the estimating equation. The variation of the SE of the

Table 12.1 Modelling LSEEU by OLS Variable

Coefﬁcient

Std. error

t-Value

t-Prob.

Part R2

Constant LSEUSSR s1983

6.38 0.11 0.40

1.36 0.13 0.05

4.69 0.80 6.97

0.00 0.43 0.00

0.55 0.03 0.73

R 2 = 0.74; F (2, 18) = 25.57; [0.0000]; = 0.13; DW = 0.8; RSS = 0.29 for twenty-one observations Diagnostic tests for modelling LSEEU by OLS: Test for autocorrelation of the residuals: AR 1-2F (2, 16) = 2.07 [0.16] Test for autoregressive conditional heteroskedasticity: ARCH 1 F (1, 16) = 0.42 [0.53] Test for normality of residuals: normality χ 2 (2) = 1.45 [0.48] RESET test for functional form mis-speciﬁcation: RESET F (1, 17) = 34.21 [0.00] ∗∗

Space race between NATO and the USSR

195

Soviet Union is insigniﬁcant (based on the t-value) in explaining the variation of the SE of Europe for the respective time period. The introduction of lags did not signiﬁcantly improve the performance of the model. Following the absence of any evidence for the presence of a space race between Europe and the USSR, this chapter next examines evidence on the space race between the United States and the USSR. Granger causality tests In the bivariate case the Granger causality tests are based on two OLS estimated models. The ﬁrst of the two estimations are conducted with LSEUS as the dependent variable and explanatory variables lagged values of LSEUS and LSEUSSR (Table 12.2). The null hypothesis is that LSEUSSR does not ‘Granger cause’ LSEUS . The test is an ordinary F -test on the signiﬁcance of the lagged values of LSEUSSR , the null is rejected if the respective coefﬁcients are signiﬁcant (for a presentation of a general case see Linden 1995: 41). Symmetrically, the second of the two estimations is conducted with LSEUSSR as the dependent variable and explanatory variables lagged values of LSEUS and LSEUSSR (Table 12.3). The null hypothesis is that LSEUS does not ‘Granger cause’ LSEUSSR . With respect to modelling LSEUS by OLS, the F -tests on the signiﬁcance of each variable at the 10 per cent level of signiﬁcance reject the null of the coefﬁcient of the explanatory variable being zero. In contrast, in modelling LSEUSSR by OLS, the F -tests on the signiﬁcance of each variable at the 10 per cent level of signiﬁcance do not reject the null of the coefﬁcient of the explanatory variable being zero. This rejects the null hypothesis that the LSEUSSR does not ‘Granger cause’ LSEUS . Table 12.2 Modelling LSEUS by OLS (sample range 1970–90) Variable

Coefﬁcient

Constant LSEUS (t − 1) LSEUS (t − 2) LSEUSSR (t − 1) LSEUSSR (t − 2)

4.39 0.20 0.25 0.21 −0.64

Std. error

t-Value

t-Prob.

0.94 4.65 0.00 0.08 2.42 0.03 0.24 1.03 0.32 0.31 0.69 0.50 0.29 −2.18 0.04 R 2 = 0.97; F (4, 16) = 176.27 [0.00]; O˜ = 0.059; DW = 2.06; RSS = 0.056 for twenty-one observations Tests on the signiﬁcance of each variable Variable F(num, denom) Value Probability Constant F(1, 16) 21.63 [0.00] ∗∗ LSEUS F(2, 16) 115.68 [0.00] ∗∗ LSEUSSR F(2, 16) 9.72 [0.00] ∗∗ Notes ∗ Signiﬁcant at 5%. ∗∗ Signiﬁcant at 1%.

PartR2 0.57 0.26 0.06 0.03 0.23

196

Vasilis Zervos

Table 12.3 Modelling LSEUSSR by OLS (sample range 1970–90) Variable

Coefﬁcient

Constant LSEUSSR (t − 1) LSEUSSR (t − 2) LSEUS (t − 1) LSEUS (t − 2)

1.54 1.51 −0.72 −0.18 0.25

Std. error 0.66 0.21 0.20 0.16 0.17

t-Value

t-Prob.

PartR2

2.33 7.04 −3.52 −1.13 1.48

0.03 0.00 0.00 0.27 0.16

0.25 0.75 0.43 0.07 0.12

R 2 = 0.97; F (4, 16) = 126.83 [0.00]; O˜ = 0.041; DW = 1.83; RSS = 0.027 for twenty-one observations Tests on the signiﬁcance of each variable Variable F(num, denom) Value Probability Constant F (1, 16) 5.46 [0.03] ∗ LSEUSSR F (2, 16) 62.67 [0.00] ∗∗ LSEUS F (2, 16) 1.64 [0.00] Notes ∗ Signiﬁcant at 5%. ∗∗ Signiﬁcant at 1%.

The respective test and estimated equations indicate that the SE of the USSR ‘Granger cause’ the SE of the United States. In view of this, the preferred estimation is Eq. (12.14). The signiﬁcance of the presence of Granger causality rests not only in deciding which of Eq. (12.13) and (12.14) is best to estimate, but is also important in supporting positive tests for the presence of co-integration. This is so because ‘although causality can occur without co-integration, the latter does necessarily imply some causality’ (Linden 1995: 39). In the bivariate case of a LSE, the presence of co-integration means that the two series behave in a way leading to a long-term stable equilibrium. Following the Engle–Granger (EG) approach (Harris 1997: 52), there are two conditions for the presence of co-integration of order 1 - CI(1, 1) – between two time-series. The ﬁrst is that each of the two series is integrated of order one – I (1) – and ¯ US (t), is I (0) (Linden the second is that the error term: zt = LSEUS (t) − LSE 1995: 38). Unit root tests The ﬁrst requirement for the presence of CI(1, 1) between the SE of the United States and the USSR is that both time series are I (1), evidence in support of this through the use of augmented Dickey–Fuller (ADF) tests (Doornik and Hendry 1995) for the presence of unit roots in the time-series are presented in Table 12.4. The choice of the number of lags is arbitrary, and given the small sample size in this case, the number of lags is set to two (for a critique of this, see Smith et al. 2000). The null hypothesis of the ADF is not rejected supporting the presence of a unit root in the data. Following this, Eq. (12.11) is estimated with the use of OLS to obtain the residual zt and test it for degree of integration. If zt is shown to be I (1),

Space race between NATO and the USSR

197

Table 12.4 Unit root tests for LSEUS and LSEUSSR (sample: 1970–90; critical values: 5% = −3.011; 1% = −3.785; constant included) Variable

t-ADF

Õ

Lag

LSEUS LSEUS LSEUS LSEUSSR LSEUSSR LSEUSSR

−1.851 −1.086 0.008 −1.889 −2.532 −1.619

0.076 0.083 0.107 0.042 0.042 0.073

2 1 0 2 1 0

t-Lag

t-Prob.

2.077 3.652

0.053 0.001

−1.196 6.142

0.248 0.000

there is evidence in support of the presence of cointegration between the SE of the USSR and the United States. Evidence on the impact of SDI on the space race between the United States and the USSR Testing the SE of the United States and the USSR for cointegration within the EG framework requires the estimation of the static model depicted in Eq. (12.14). This can be done with the use of OLS to achieve a consistent estimate of the longrun steady equilibrium condition and all short run dynamics can be asymptotically ignored (Harris 1997: 52). It must be noted, however, that bias is a problem in ﬁnite samples such as the one used here (twenty-one observations), and the EG procedure leads to potentially biased estimates of long-run relationships and statistical inferences using standard t-tests are inapplicable (see Harris 1997: 53, 57). In estimating Eq. (12.14), it is expected a priori that the US SDI8 will cause lack of parameter constancy in the estimation. To test for such SDI-speciﬁc effects, the use of recursive least squares estimation method (see Doornik and Hendry 1995) is chosen to graphically illustrate whether the estimation of Eq. (12.14) with a 1970–90 sample reveals signs of a structural break in 1983. The methodology employed by the recursive method of estimation of a model for a sample with T observations is to apply successive OLS estimates to the model starting with M observations (M < T ), and then ﬁt the model to M + 1, M + 2, . . . up to T observations (Doornik and Hendry 1995: 140). This way a number of successive residual sums of squares are obtained based on which sequence of tests for structural breaks and parameter constancy can be conducted (Doornik and Hendry 1995: 268). The recursive least squares (RLS) estimation of the US–USSR space race estimation for the sample period 1970–90 is given in Table 12.5. Modelling LSEUS by RLS performs poorly in terms of explanatory power and signiﬁcance of the independent variable. The recursive graphics for M = 5 reveal that a major reason for this failure rests with the lack of parameter constancy from 1983 when the SDI is initiated onwards. Overall, the estimation results and graphic analysis support the presence of a structural break in 1983.

198

Vasilis Zervos

Table 12.5 Modelling LSEUS by RLS (the present sample is: 1970–90) Variable

Coefﬁcient

Std. error

t-Value

t-Prob.

PartR2

Constant LSEUSSR

4.64 0.49

3.72 0.37

1.25 1.32

0.23 0.20

0.1207 0.08

R 2 = 0.08; F (1, 19) = 1.75 [0.20]; DW = 0.06; RSS = 2.34 for two variables and twenty-one observations

Table 12.6 Modelling LSEUS by OLS Variable

Coefﬁcient

Std. error

t-Value

t-Prob.

PartR2

Constant LSEUSSR s1983

6.164 0.313 0.647

1.386 0.138 0.059

4.45 2.27 10.96

0.00 0.03 0.00

0.52 0.22 0.87

R 2 = 0.88; F (2, 18) = 66.49 [0.0000]; = 0.13; DW = 1.14; RSS = 0.30 for twenty-one observations Diagnostic tests for modelling LSEUS by OLS: Test for autocorrelation of the residuals: AR 1-2F (2, 16) = 1.49 [0.25] Test for autoregressive conditional heteroskedasticity: ARCH 1 F (1, 16) = 1.420 [0.25] Test for normality of residuals: normality χ 2 (2) = 1.60 [0.45] RESET test for functional form mis-speciﬁcation: RESET F (1, 17) = 7.36 [0.01]∗

Estimating the US–USSR space race static model The method preferred to model the impact of the SDI on the estimated model was to incorporate a step dummy variable from 1983 to take into account SDI effects on the SE of the United States and examine the signiﬁcance of its coefﬁcient. The resulting estimating equation was evaluated using OLS and sample range from 1970 to 1990 (Table 12.6). The performance in terms of the diagnostic test results is quite satisfactory, while the explanatory power is also quite good: R 2 is high and all the coefﬁcients are signiﬁcant and have the expected sign. The value of the coefﬁcient of the rival (the SE of the Soviet Union), would be expected to be much closer to one, but is unexpectedly low at 0.31. By examining the partial R 2 it becomes obvious that SDI can explain much more of the variation of the dependent variable than the variation of the rival’s SE. Thus, the high t-value and partial R 2 of the step dummy of 1983, which is a proxy for SDI effects indicates that SDI had a profound effect on the structure of the space race between the United States and the USSR. Despite the early data reservations presented earlier in this chapter, Eq. (12.14) was also estimated with an extended sample range, from 1961 to 1990, and a forecasted period from 1991 to 1994 (Table 12.7). The reason why this estimation includes a forecasting period is to see whether the collapse of the Soviet Union

Space race between NATO and the USSR

199

Table 12.7 Modelling LSEUS by OLS (the present sample is: 1961–94 less four forecasts, the forecast period is: 1991–4) Variable

Coefﬁcient

Std. error

t-Value

t-Prob.

PartR2

Constant LSEUSSR s1983 i1964

0.99 0.83 0.59 0.38

1.71 0.17 0.09 0.23

0.58 4.96 6.28 1.65

0.00 0.03 0.00 0.11

0.12 0.49 0.60 0.09

R 2 = 0.69; F (3, 26) = 19.74 [0.00]; = 0.22; DW = 0.69; RSS = 1.28 for thirty observations Diagnostic tests for modelling LSEUS by OLS with an extended sample range, from 1961 to 1990: Test for autocorrelation of the residuals: AR 1-2F (2, 24) = 4.32 [0.02]∗ Test for autoregressive conditional heteroskedasticity: ARCH 1 F (1, 24) = 0.18 [0.67] Test for normality of residuals: normality χ 2 (2) = 5.42 [0.06] RESET test for functional form mis-speciﬁcation: RESET F (1, 25) = 1.97 [0.17] Table 12.8 Analysis of 1-step forecasts Date

Actual

Forecast

Y − Yˆ

1991 1 10.02 6.91 3.11 1992 1 10.03 6.83 3.19 1993 1 9.97 6.83 3.14 1994 1 9.91 6.79 3.12 Tests of parameter constancy over: 1991–4: Forecast χ 2 (4) = 801.11 [0.00] ∗∗ Chow F (4, 26) = 5.97 [0.00] ∗∗

Forecast

SE t-value

0.66 0.67 0.67 0.68

4.70 4.74 4.65 4.57

Notes ∗ Signiﬁcant at 5%. ∗∗ Signiﬁcant at 1%.

had the expected effect of negatively affecting the SE of the United States. The estimation of the space race equation with such an extended sample range will reveal whether the respective data supports the hypothesis of a space race between the United States and the USSR for the whole of the ‘Cold War’ era. The impact dummy of 1964 marginally improved the estimation and is therefore included. It depicts the impact of the Apollo space programme, on the year the SE of the United States reached its maximum value. As the analysis of 1-step forecasts shows (Table 12.8), the actual is in all cases higher than the forecasted, this results in the tests for the parameter constancy rejecting the null hypothesis of parameter constancy for the period 1991–4 and supporting the end of the space race. Most important, the high explanatory power and signiﬁcance of the 1983 step-dummy capturing the impact of SDI on the space

200

Vasilis Zervos Table 12.9 Unit root tests for zt (sample range: 1970–90; critical values: 5% = −3.011; 1% = −3.785; constant included) Variable

t-ADF

Õ

Lag

t-Lag

t-Prob.

z z z

−3.209∗ −3.202∗ −3.275∗

0.115 0.113 0.111

2 1 0

0.627 0.432

0.5390 0.6711

Notes ∗ Signiﬁcant at 5%. ∗∗ Signiﬁcant at 1%.

race further supports the hypothesis of the catalytic role SDI had on the space race between the United States and the USSR by altering its structure. The interpretation of the reason why the actual value is higher than the forecasted is that following the collapse of the USSR, its SE break-down the US space programmes do not experience such collapse. Overall, this analysis presents evidence in support of a structural change in the space race between the United States and the USSR due to SDI. Co-integration analysis In order to examine the long-run stability of the equilibrium condition of the space race the relationship will be tested for co-integration (see Smith et al. 2000). Estimation modelling LSEUS by OLS has no lagged right-hand side (RHS) variables and it depicts a long-run equilibrium condition. Following the ﬁrst step of the EG approach, co-integration is obtained if the residual of modelling LSEUS by OLS is I (0) (see Harris 1997: 57). The test for the degree of integration of the residual is an ADF test and its results are given in Table 12.9. The null hypothesis is rejected at the 5 per cent level of signiﬁcance, not supporting at this level of signiﬁcance the presence of a unit root and indicating that zt is I (0). This is evidence in support of the presence of co-integration between the LSEUS and LSEUSSR time-series. In view of the econometric problems discussed earlier with the results of cointegration tests on time-series with small ﬁnite sample size of just twenty-one observations, the above results of evidence of co-integration are best interpreted as qualitative indications, rather than quantitative, conclusive evidence.

The impact of the US military spending on the US–USSR space race The overall performance of the space race model and the fact that the SE of the United States includes the substantial US military space expenditure encourages the introduction of the ME of the US to test whether the US government follows an integrated space and defence policy. This would imply that the variation of its ME would go some way to explaining the variation of its SE. Speciﬁcally, the SDI

Space race between NATO and the USSR

201

Table 12.10 Modelling LSEUS by OLS (sample range 1970–90) Variable

Coefﬁcient

Std. error

t-Value

t-Prob.

PartR2

Constant LSEUSSR s1983 LMEUS

−9.30 0.20 0.21 1.37

2.65 0.08 0.08 0.22

−3.50 2.42 2.75 6.10

0.00 0.03 0.01 0.00

0.42 0.26 0.31 0.69

R 2 = 0.96; F (3, 17) = 146.04 [0.00]; = 0.07; DW = 0.97; RSS = 0.096 for twenty-one observations Diagnostic tests for modelling LSEUS by OLS, sample range 1970–90: Test for autocorrelation of the residuals: AR 1-2F (2, 15) = 2.52 [0.11] Test for autoregressive conditional heteroskedasticity: ARCH 1 F (1, 15) = 0.08 [0.78] Test for normality of residuals: normality χ 2 (2) = 5.01 [0.08] RESET test for functional form mis-speciﬁcation: RESET F (1, 16) = 5.47 [0.03] ∗

involved both military and military SEs and programmes (Gray 1982; Baucom 1999). Assuming that as a result of this the military and space programmes of the United States were well integrated and also that this is reﬂected in the respective ME and SE, the inclusion of the ME as an explanatory variable in modelling LSEUS by OLS would diminish the high explanatory power of the step dummy of 1983. The resulting extended estimation of the space race model is presented in Table 12.10. As Table 12.10 reveals, the introduction of the ME of the United States as an explanatory variable improved the performance of the model, by increasing the R 2 , while this new explanatory variable has the highest partial R 2 and the partial R 2 of the step-dummy of 1983 has signiﬁcantly diminished. This therefore indicates that the SE and ME9 of the United States are well integrated in an overall space and defence policy under the SDI. Furthermore, the coefﬁcient of the rival variable is signiﬁcant and with the expected sign, but its value has further decreased. The signiﬁcance of the ME of the United States in explaining its SE was not duplicated for the cases of Europe or the Soviet Union. For the case of Europe this is attributed to the dependence on the United States in space within NATO, which results in the European space programmes being less connected to the European ME. For the case of the Soviet Union a possible explanation for this is that the massive (relatively to the Soviet economy) Soviet military sector could have more difﬁculty in following the SDI, as opposed to the more ﬂexible and strategically important space sector. Another explanation is that the Soviet data might not be accurate, being largely based on estimates of Western sources (ESD 1997), whose data creation methodology is not rigorously scrutinized. Further research is necessary in order to critically assess such explanations. Overall, it can be argued that evidence from this analysis indicate that there is a space race between the United States and the Soviet Union. The fact that within NATO, only the United States seems to be engaged in a space race with the

202

Vasilis Zervos

USSR/Russia for the time period examined means that either Europe free-rides on US space capabilities, or assuming sufﬁcient complementarity within the NATO alliance, this is an indication of interspecialisation within the alliance (see Sandler and Hartley 1995). In addition, the fact that the United States does not apply the principle of nonexclusion in the provision of its space-based capabilities, but supplies its European allies with space-based earth observation imagery and other sensitive and strategic information on a discretionary basis (UN 1992) indicates that within NATO, the US space-based capabilities are not a pure public good. This evidence points at the presence of complementarity, rather than free-riding as the principle behind the dependence of Western Europe on the United States with respect to space capabilities that match the potential in space of USSR for the respective time period. If this interalliance specialization is such that the United States provides most of the strategic goods within the alliance, such as space based capabilities and nuclear arsenal, while the European allies provide substantial conventional and non-strategic defences against the Soviet Union for the time period in question, the issue arises of the implications of this arrangement for the United States and the European space industries. In particular, the signiﬁcance of this interalliance specialisation for the European and the US space industries and their competitive positions in commercial markets in the post ‘Cold War’ era is that it allows the US space industry a potential cost advantage in commercial space markets vis-à-vis the European space industry (see Zervos 1998a,b).

Conclusions and further research proposals The hypothesis examined in this chapter is whether there is evidence of a space race between the United States and the Soviet Union. The implications of this for NATO is that the behaviour of the European SE was not expected to be explained by the behaviour of the European ME, while the opposite should hold true for the respective US time-series. Both of these propositions were strongly supported by the econometric analysis in this chapter. The hypothesis was supported by the econometric analysis for two time periods: 1970–90 and 1961–90. The two periods include the launch and implementation of SDI by the United States, which as the analysis reveals plays a key role in explaining the SE of the United States, alongside the SE of the Soviet Union and serves as the key ‘grievance factor’, by capturing US-speciﬁc changes in its SE. The empirical analysis in support of this hypothesis was extended to show that the behaviour of the European SE was not explained by the behaviour of the European ME, while the opposite did hold true for the respective US time-series. This qualiﬁcation of the hypothesis led to an important result, showing that during the ‘Cold War’ period the US SE served as a public good within NATO, as deﬁned by Europe and the United States. This dominant position in space of the United States within the NATO alliance and the resulting greater size of the United States

Space race between NATO and the USSR

203

space industry could provide a post ‘Cold War’ advantage to the competitiveness of the US space industry in commercial markets. Further analysis is required to examine the implications of this analysis for the commercial markets and the competitive position of the European and the US space industries, in order to form a more complete picture of how historical analysis of the ‘Cold War’ era can have a major impact on contemporary economic issues in the modern era of space commercialisation.

Acknowledgements I am indebted to Keith Hartley, Karim Abadir and Paul Dunne for their useful comments. The usual disclaimer applies.

Notes 1 For example, the United States had a number of rules prohibiting the export of US technology to allied countries if suspecting its further transfer to the USSR. 2 Furthermore, although since the 1950s France abstains from the military part of NATO, it is included with the other four European nations in the analysis of European ME and SE. This is because during the ‘Cold War’ France behaved in accordance with NATO vis-à-vis the USSR (see Murdoch 1984; Smith 1989), and is a major military power within the Western European Union (WEU), alongside Germany, United Kingdom, Italy and Spain, as well as being a major contributor to European space efforts (ESA). Thus, this selection reﬂects the fact that the major European NATO countries are cooperating in civil space (through ESA), and military space (through WEU) with France, rather than their major NATO ally, the United States. 3 The construction of demand functions for SE for Western Europe, the United States or USSR without considering the interactions of the other two geographical areas appears to be the most straightforward way to proceed with this analysis, especially in the absence of reliable and long-range data. Such demand functions usually incorporate variables such as prices, income, or possibly public opinion variables (Chakrabarti et al. 1992). The theory of alliances uses the demand functions in a more general form, since interactions of the demand functions for SE across geographical areas are also considered. 4 The fact that the SE of Europe is much lower than the respective US variable does not indicate the absence of a space alliance between the two areas, since the relationship of the variations of the relevant variables is important and can be examined using econometrics. 5 In the defence economics literature, c and e are called the ‘grievance factor’ (Sandler and Hartley 1995: 83), meaning that the NATO alliance or the USSR might increase (or decrease) their space spending independently of what’s happening to the rival’s respective spending. 6 All variables are in constant ECU (1991) values. In addition, ME for Western Europe (MEEU ) is the sum of the expenditure of ﬁve major European countries: France, Germany, United Kingdom, Italy and Spain. This is due to the fact that these countries are the major European space programme investors. In addition, per capita income is used as a proxy for GDP, since this has a minimal negative effect in the calculations and regressions due to the high correlation of the two, while keeping close the range of the data, of particular value for graphic analysis. The ME for Europe, the United States, the USSR (until 1990), and Russia (from 1991) were obtained from ACDA (1976, 1982, 1992, 1995). The same data source was used for the populations and GNP/GDP of the USSR and Russia. The USSR variables SE and GNP per capita are constructed based on USSR observations until 1991 and Russian observations from 1991 until 1994. MSE and SE of the United

204

Vasilis Zervos

States was obtained from NASA (1994). EU (1996, 1997) was used for the GDP of the United States and EU. Total SE for the European Union was obtained from ESD (1990, 1992, 1996). The same source was used for the SE of USSR. 7 When the most relevant variables are income and prices the model would be virtually identical to the space alliance model presented earlier. 8 The SDI was initiated in 1983 by the US President Reagan involving R&D to determine the feasibility of strategic defence systems. In 1984, the SDI organisation was established and renamed in 1993 as ‘ballistic missile defense organisation’ (BMDO) (Baucom 1999). 9 Estimation in Table 12.10 was not estimated with an extended sample range, unlike in Table 12.6. This was due to the impact regional crisis such as the Vietnam war had on the behaviour of the US ME.

References Arms Control and Disarmament Agency (ACDA). (1976) World Military Expenditures and Arms Transfers 1976. The Superintendent of Documents, US Government Printing Ofﬁce, Washington, DC. Arms Control and Disarmament Agency (ACDA). (1982) World Military Expenditures and Arms Transfers 1982. The Superintendent of Documents, US Government Printing Ofﬁce, Washington, DC. Arms Control and Disarmament Agency (ACDA). (1991–2) World Military Expenditures and Arms Transfers 1992. The Superintendent of Documents, US Government Printing Ofﬁce, Washington, DC. Arms Control and Disarmament Agency (ACDA). (1995) World Military Expenditures and Arms Transfers 1995. The Superintendent of Documents, US Government Printing Ofﬁce, Washington, DC. Baucom, D. R. (1999) Origins of the U.S. Missile Defense Program. Ballistic Missile Defense Organization, US. Online: http://www.acq.osd.mil/bmdo/bmdolink/html/ history.html. Chakrabarti, K. A., Glismann, H. H. and Horn, J. E. (1992) Defence and space expenditure in the US: an inter-ﬁrm analysis. Defence Economics 3, 169–189. Doornik, A. J. and Hendry, F. D. (1995) PcGive 8.0 An Interactive Econometric Modelling System. Chapman and Hall, London, UK. Dunne, P., Nikolaidou, E. and Smith, R. (2000) The Econometrics of Arms Races. Defence and Peace Economics 11(1), 31–43. Dunne, P., Nikolaidou, E. and Smith, R. P. (2003) Arms race models and econometric applications, Chapter 11 of current volume. European Space Directory. (ESD) (1990) Sevig Press, Paris, France. European Space Directory (ESD) (1992) Sevig Press, Paris, France. European Space Directory (ESD) (1996) Sevig Press, Paris, France. European Space Directory (ESD) (1997) Sevig Press, Paris, France. European Union (EU). (1996) The European Aerospace Industry Trading Position and Figures. European Commission, Capital Goods Industries (DG III), Brussels, Belgium. European Union (EU). (1997) The European Aerospace Industry Trading Position and Figures. European Commission, Capital Goods Industries (DG III), Brussels, Belgium. Gray, S. C. (1982) American Military Space Policy, Information Systems, Weapon Systems and Arms Control. Abt Books, Cambridge, MA. Harris, R. I. D. (1997) Using Cointegration Analysis in Econometric Modelling. Prentice Hall, Englewood Cliffs, NJ.

Space race between NATO and the USSR

205

Linden, M. (1995) Studies in Integrated and Cointegrated Economic Time Series. The Finnish Statistical Society, Helsinki, Finland. Lovell, B. (1973) Origins and International Economics of Space Exploration. Edinburgh University Press, Edinburgh, UK. Murdoch, J. C. (1984) Complementarity, free riding, and the military expenditure of NATO allies. Journal of Public Economics 25(1–2), 83–101. Murdoch, J. C. (1994) Aeronautics and Space Report of the President, 1994. NASA, Washington, DC. Okamura, M. (1991) Estimating the impact of soviet union’s threat on the United States–Japan alliance: a demand system approach. The Review of Economics and Statistics LXXIII(2), 200–207. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge, UK. Smith, P. R. (1989) Models of Military Expenditure. Journal of Applied Econometrics 4, 345–359. Smith, R. P., Dunne, J. P. and Nikolaidou, E. (2000) The Econometrics of Arms Races. Defence and Peace Economics 11(1), 31–44. United Nations (UN). (1992) Access to Outer Space Technologies: Implications for International Security. United Nations Institute for Disarmament Research (UNIDR), Research paper 15, New York. Zervos, V. (1998a) The economics of the European space industry: the impact of the government space market on structure, conduct and performance. In: Haskell, G. and Rycroft, M. (eds), New Space Markets. Kluwer Academic Publishers, Dordrecht Netherlands. Zervos, V. (1998b) Competitiveness of the European space industry, lessons from Europe’s role in NATO. Journal of Space Policy 14(1), 39–47.

Part III

Overview

13 Arms trade, arms control and security Collective action issues Todd Sandler

Introduction In 1998, 2.6 percent of world gross national product (GNP) or about three-quarters of a trillion dollars was devoted to military expenditures, thus continuing the downward trend started at the conclusion of the Cold War.1 Much of this decrease from 1997 can be attributed to reductions in military spending in the Russian Federation and the United States; however, events in Kosovo and Chechnya will surely reverse these two countries’ downward defense spending trends. According to SIPRI (1999: 9), arms production has begun to rise in 1998. Arms transfers of $21.9 billion in 1998 are at about the same low level as the $20 billion in 1994. What is most worrying is the large number of major armed conﬂicts – twenty-seven in twenty-six locations. The world remains volatile and dangerous with the Cold War confrontation of the two superpowers giving way to smaller wars driven by territorial disputes, government power struggles, and ethnic hatreds. When these conﬂicts involve nuclear powers, such as India and Pakistan, even local wars can have wide-ranging and catastrophic consequences. Against this backdrop, the Middlesex conference in June 1999 on “The Arms Trade, Security, and Conﬂict” addressed concerns of great moment. As rapporteur, I have the daunting task to ﬁnd some common themes that run through this interesting set of theoretical and empirical papers. The easiest, and most uninteresting, way to proceed is to assign the papers to various categories and then to discuss each in seriatim. I shall resist this unimaginative approach and offer instead a single theme – collective action – that ties these varied contributions together. Collective action refers to situations where the efforts of two or more agents (e.g. individuals, nations) are required to accomplish an outcome. The seminal work in the area is The Logic of Collective Action by Olson (1965),2 whose basic message is that the pursuit of individual rationality is not sufﬁcient for collective rationality (Sandler 1992: 3). In other words, efforts of individuals in a group or collective (e.g. nations engaged in an arms race, allies deciding defense spending, or nations imposing arms controls) to further their self-interest may not truly advance the group’s interest. A collective identity is not enough for the group to prosper. After reviewing the basic principles of collective action in the next two sections, I shall draw out the collective action aspects of the conference’s ﬁve topics – arms

210

Todd Sandler

trade, arms control, arms production, arms races, and alliance formation – in the ensuing ﬁve sections. Concluding remarks are contained in the last section. My essential message is that to understand each of these ﬁve topics and the allocative issues that they raise, one must rely on the principles of collective action.

Principles of collective action Collective action applies to the provision of public goods of all types as well as the correction of externalities. The principles of collective action evolve around three essential inﬂuences: group size, group composition, and institutional design. In the case of group size, Olson (1965) espoused two principles: (i) as group size increases, the group is less likely to provide itself with the collective good; and (ii) as group size increases, the more sub-optimal is the anticipated equilibrium.3 With large groups, the beneﬁt share going to an individual or subset of contributors may be too small to warrant covering the cost alone. If everyone feels this way, then the collective good will not be provided at all. As group size increases, this beneﬁt share usually decreases and this decrease, other things constant, makes it even less likely for any individual or subset to want to support the collective good alone, so that the collective good is not provided. If transaction costs of group formation are also considered, then they reinforce the problem of large groups not being able to provide themselves with the collective action. The second inﬂuence of group size indicates that the departure between equilibrium and a Pareto-optimal ideal increases with group size. That is, the independent adjustment or Nash equilibrium associated with a group of, say, four is anticipated to be less suboptimal than the Nash equilibrium associated with a group of ﬁve or ten. As participants in larger groups ignore the marginal beneﬁts that their actions confer on others, sub-optimality worsens because the sum of these ignored marginal beneﬁts of others increases with group size for most collective goods.4 The second set of collective action principles relate to group composition based on whether the group members are homogeneous or heterogeneous in terms of tastes and endowments. Heterogeneous groups are anticipated to face an “exploitation hypothesis,” whereby the large participants carry a disproportionally greater burden of the collective good.5 If allies’ tastes are the same and if, moreover, defense is viewed as a normal good with a positive income elasticity, then the large ally will surely want more defense than the small. Defense spending is anticipated to be a normal good, because the more that a country earns, the more that it has to lose from an invasion, and so the more defense it wants. Thus, the demand for defense rises with income, as borne out by most empirical studies (Sandler and Hartley 1999: chapter 2). Consider two allies, one of which is large and rich, and the other is small and poor. Suppose that the large ally wants $300 billion in defense spending when it equates marginal beneﬁts and marginal costs, whereas the small ally wants just $50 billion when it equates these margins. If defense spending is a perfect substitute between allies, then the small ally’s demand for $50 billion of defense will be fully met by the large ally’s spending, thus leading to the exploitation of the large ally by the small free-riding ally.

Arms trade, arms control and security

211

A second inﬂuence of group composition concerns whether or not the group provides itself with any of the collective good. If some group members are rich, then they may derive a sufﬁcient share of the collective good’s beneﬁts to justify covering the costs of the collective good on their own, as seen in the alliance example. This underlying principle that asymmetry among members of the collective may foster some provision assumes that these shares, which are just 1/n for homogeneous groups of size n, are now dependent on the member’s relative importance or size in the group. For example, the United States surely derives a much greater share of the gains from collective defense as compared with the smaller NATO allies. This share was sufﬁciently large for the United States to provide the overwhelming amount of defense spending (i.e. over three-quarters) during the early 1950s.6 Finally, collective action principles also hinge on the use of selective incentives and institutional design (Olson 1965). Selective incentives are private inducements used to motivate individual contributions to the collective good. In defense economics, these selective incentives are referred to as ally-speciﬁc private goods, which are jointly produced or derived from a collective action that also yields an alliancewide public good for the group.7 That is, an ally’s purchase of defense not only fosters alliancewide deterrence, but it also bolsters contributor-speciﬁc activities (e.g. disaster relief, patrolling coastal waters). Institutional design can change the “rules of the game” and, in so doing, can alter payoffs so that individuals’ goals are more akin to those of the collective. For example, the use of matching formulae for the support of NATO infrastructure has effectively reduced the price to each ally, since its contribution is matched by the rest of the alliance. Each dollar contributed effectively yields more units of output.

Recent insights and collective action Although these propositions of collective action are not universally true, they are valid for a large number of cases and this explains why they have been so inﬂuential (Hardin 1982; Sandler 1992; Ostrom 1999). When qualiﬁcations to the principles of collective action are made, they derive from a number of considerations.8 First, the underlying strategic assumption can be an essential factor. Olson’s propositions implicitly assumed a Nash equilibrium. If, for instance, leader–follower behavior applies, then exploitation of the large participant need not result when the latter is the leader (Sandler 1992: 56–58). This follows because the leader will account for the anticipated free-riding behavior of the follower, and will reduce its own contributions to the collective activity accordingly, thus shifting more burdens onto the follower. Second, the intertemporal nature of the interaction makes a difference for the outcome of collective action. If, for example, agents interact repeatedly and resort to some punishment-based strategy, such as tit-for-tat, then an optimum may be achieved. Third, taste differences among contributors can matter greatly and alter the predictions for collective action. For example, a small ally, whose tastes for defense are much stronger than its larger counterparts, may be the one exploited. Israel, surrounded by a sea of hostile countries, spent more than twice

212

Todd Sandler

the percentage of GDP on defense as compared to its ally, the United States. Fourth, the underlying “aggregation technology of collective supply,” which indicates how individual contributions to the collective good adds to the overall level of the good,9 is a crucial determinant of the outcome of collective action. To indicate the inﬂuence of the underlying aggregate technology of collective supply (henceforth, called the aggregation technology), I shall introduce four alternative technologies and brieﬂy indicate their implications on the underlying games in strategic form. The most prevalent aggregation technology, and the one that Olson (1965) had in mind, is that of summation, where each individual’s contribution is merely summed to get the overall level of the collective good, consumed by group members. Suppose that each of two players can contribute one or no units to a collective action (e.g. limiting arms trade, where the collective good derives from reducing the negative externality associated with such trading). Further suppose that each unit contributed confers a beneﬁt of 4 to each player, regardless of whether they contribute or not, at a cost of 6 to the contributor. With these beneﬁts and costs, the payoffs of the underlying game matrix corresponds to those in Figure 13.1 where the ﬁrst payoff in each of the four cells is the net beneﬁt of player A and the second payoff is that of player B. If, for example, only A contributes, then he or she gets a net beneﬁt of −2 (=4 − 6), while the free rider, player B, receives 4. When both contribute, they each get 2 (=4 × 2 − 6) as the cost of 6 from contributing a unit is deducted from the sum of beneﬁts of 8 that they receive in total from the two units contributed. If neither contributes, then the net gain is 0 for each. The resulting game is the Prisoner’s Dilemma with a dominant strategy of not contributing, since 4>2 and 0>−2, so that payoffs are greater for the not-contributing strategy regardless of the other player’s strategy. As each individual plays his or her dominant strategy, the Nash equilibrium results from which neither player would unilaterally want to move. In Figure 13.1, this Nash equilibrium has neither player contributing. Arms races are often characterized by an underlying Prisoner’s Dilemma game (Sandler and Hartley 1995: chapter 4), for which arming dominates disarming. The aggregation technology that underlies many Prisoner’s Dilemmas is that of summation, since each player’s effort adds to the total achieved.

B⬘s Strategy A⬘s Strategy

Contributes

Does not contribute

Contributes

2, 2

–2, 4

Does not contribute

4, –2

NASH 0, 0

Figure 13.1 Summation and Prisoner’s Dilemma.

Arms trade, arms control and security

213

When, however, there is a loss if none of the collective action is achieved, then the payoffs in the bottom right-hand cell would change. Suppose that the absence of any collective action results in a loss of 3 to each player. To capture this alteration, the 0s in Figure 13.1 would have to be replaced with −3s in the bottom right-hand cell (not shown). Now there is no dominant strategy (since neither strategy’s payoffs exceed both of the corresponding payoffs of the other strategy) and the pure-strategy Nash equilibrium are the two off-diagonal cells, where one player contributes and the other free rides. The underlying game is Chicken and its prognosis is more favorable for some collective action. In the case of arms control, the lack of any control may well be a Chicken game, where one or two nations provide the control and everyone else free rides – for example, the US actions in 1994 regarding North Korea’s development of nuclear weapons. Next, consider a best-shot technology where just the greatest effort determines the overall level of the collective action, smaller contributions do not add to the total – that is, the size of the collective good is that of the biggest contribution. If only one unit is needed, then a second unit adds no beneﬁt in a best-shot scenario and should not be contributed. The neutralization of a rogue nation is an example of best shot, as is the case of achieving an R&D breakthrough. To illustrate the underlying game, consider the game matrix in Figure 13.2, in which each player can contribute one or no units to the collective action and only the ﬁrst unit provides 6 in beneﬁts to each player at a cost of 4 to just the contributor. When either player A or B contributes alone, the contributor earns 2 (=6−4) and the noncontributor gets a free-rider beneﬁt of 6. If, however, both contribute, then only the ﬁrst unit contributed yields a beneﬁt of 6 to each player and the second unit is redundant. Thus, the payoffs in the top left-hand cell are 2 (=6−4). Inspection of the matrix indicates that there is no dominant strategy, but there are two pure-strategy Nash equilibria, as indicated in the off-diagonal cells. The underlying game is that of coordination where the two players need to alternate their responses and not both contribute – for example, two allies independently developing the same stealth technology or any identical technological breakthrough represents a waste of resources. This point is certainly in keeping with the conference paper by Hartley (2000).

B⬘s Strategy A⬘s Strategy

Contributes

Contributes

2, 2

NASH 2, 6

NASH 6, 2

0, 0

Does not contribute

Does not contribute

Figure 13.2 Best shot and coordination game.

214

Todd Sandler B⬘s Strategy A⬘s Strategy Contributes Does not contribute

Contributes

Does not contribute

NASH 2, 2

–6, 0

0, –6

NASH 0, 0

Figure 13.3 Weakest link and matching.

The third aggregation technology is that of weakest link, where the smallest contribution to the collective good ﬁxes the level of the good enjoyed by everyone. If, say, a group of nations are trying to keep a rogue nation from acquiring the intelligence to build a weapon of mass destruction (WMD), then the nation expending the least effort may determine the outcome of the rogue’s quest. In regards to this weakest-link technology consider the matrix in Figure 13.3. Once again, each of two players can contribute one or no units of the collective good at a costs of 6 to the contributors. Since the smallest contribution determines the overall level of the collective good, its effective level is still zero when just a single player contributes, so that beneﬁts are also zero. In the off-diagonal cells, the contributor realizes a loss of 6 from covering the unit cost in the absence of any gain, while the free rider gets nothing. When both contribute, they each receives 8 at a costs of 6 for a net gain of 2. There is no dominant strategy, but there are two pure-strategy Nash equilibria along the main diagonal where each player matches the other’s behavior. Matching is the anticipated outcome for weakest link unless some mixed strategy is employed where each strategy is played probabilistically. For example, the mixed-strategy Nash equilibrium for the “assurance” game in Figure 13.3 is for each player to contribute three-quarters of the time and not to contribute one-quarter of the time. A player must then have to be assured that the other player will cooperate more than three-quarters of the time to want necessarily to contribute. If a loss of 4 is imposed on each player when neither contributes, then the mixed-strategy equilibria require each to contribute just half of the time. Faced with bad outcomes from no actions, each player now needs less assurance that the other player will contribute before contributing. Myriad aggregation technologies exist, each of which is associated with some game form and, hence, a prognosis for collective action. This is a crucial insight that goes beyond the classic treatments of collective action of Hardin (1982) and Olson (1965). Relevant technologies for arms control, arms trade, and other issues considered at the Middlesex conference include those previously mentioned and others, such as joint products. For the latter, contributions to a collective activity give rise to both group-wide and contributor-speciﬁc beneﬁts. By motivating

Arms trade, arms control and security

215

the contributor to act, contributor-speciﬁc beneﬁts often transform action into a dominant strategy.

Arms trade and its policy Relevant papers in the conference on this topic include those by Hartley (2000), Jackson (2003), Mouzakis et al. (2000), and Zervos (2003). Arms trade involves collective action concerns owing to a joint product situation. The trading country achieves a net beneﬁt from trade-associated employment, proﬁts, scale economies, learning-by-doing economies, while also creating security externalities to any country placed in harm’s way by its weapons trade. A trading country only considers its share of these security risks and ignores the risks imposed on others when deciding its trade in weapons; consequently, too much arms trade results. The dominant strategy to trade, stemming from the trader-speciﬁc joint products, is not a desirable outcome from the community of nations viewpoint. Next, consider the policy of offsets associated with the import of foreign defense equipment. Offsets are a form of work sharing whereby the nation purchasing foreign defense equipment requires the supplier to allocate some work to the industries of the purchasing nations (Martin 1996; Sandler and Hartley 1999). In developing countries that purchase weapons from the major suppliers, offsets are viewed as a path to development by stimulating local employment and acquiring advanced technologies. Offsets are, unfortunately, better at promoting arms sales than at furthering price competition and improving quality. In many instances, offsets represent an added constraint that diverts trade and limits the achievement of allocative efﬁciency. With approximately 130 nations engaged in some program of offsets. The practice of offsets epitomizes a collective action problem with large numbers, making it exceedingly difﬁcult to address. Although it would augment efﬁciency if all nations abandoned the use of offsets, it is not in the interests of any single supplier to announce unilaterally its unwillingness in the future to discuss offsets as long as its competitors continue to engage in offset arrangements as a non-price competitive tactic. Although these offsets may, at times, open up some markets to the traders, the true efﬁciency gains, if any, must be judged in a second-best world where many other markets still remain close.

Arms control regimes Arms control regimes represent the quintessential collective action problem with the need for a large number of heterogeneous participants to act. At the Middlesex conference, papers by Alonso and Hartley (2000), Mouzakis et al. (2000), Sen (unpublished manuscript), and Smith and Udis (2003) addressed various aspects of arms control regimes. To date, there has been relatively little progress in designing such regimes that both limit arms to unstable regions of the world and protect against the proliferation of WMD. It is in every suppliers’ interests to free ride on the efforts of those who limit their arms sales through arms control regimes, while surreptitiously consummating weapon sales that undercut the regimes. The large

216

Todd Sandler

number of relevant suppliers works against such regimes being formed, and, if ever formed, their efﬁcacy is unlikely to be great. The practice of offsets augments the number of arms suppliers over time as weapon technologies are transferred, and this also exacerbates the collective action difﬁculty in forming such arms control regimes. Principles of collective action indicate that the asymmetry among potential participants of arms control regimes, which arise from the presence of a few relatively large weapon suppliers (i.e. the United States, the Russian Federation, the United Kingdom, France), may offer a glimmer of hope. The major suppliers are relatively small in number and are all members of the UN Security Council. Through their actions on the Council, an initial regime can be created and then extended to other nations starting with the largest remaining weapon producers. A collective action problem still remains in terms of the enforcement of the regime rules. An enforcement mechanism is a pure public good and, as such, is anticipated to face free-rider incentives, so that regime formation merely replaces one collective action concern with another of equal or greater difﬁculty to solve. In fact, a time inconsistency problem is also involved, because a nation may view there to be net beneﬁts at the time of joining a regime prior to learning what its commitments really costs. That is, when the regime affects one of its ratiﬁers at a later date, the ratiﬁer may reassess the net beneﬁts of membership and may renege on its obligations. Veriﬁcation also poses a collective action concern, but one that is probably easier to address than enforcement as monitoring capabilities improve with time. The problems of arms control regimes are also related to the aggregation technology (Alonso and Hartley 2000). In particular, a weakest-link technology applies when arms must be cut off to rogue nations and/or highly unstable regions. The nation doing the least to limit its ﬁrms from exporting to such clients and/or regions determines the effectiveness of everyone’s efforts. This is also true of action taken to inhibit the proliferation of WMD. Saddam Hussein’s success in his pursuit of WMD very much depended on those nations that did the least to deny his quest. This, in turn, implies that an arms control regime has to ensure that every signiﬁcant arms supplier’s effort is above an acceptable minimal standard. A related consideration is whether those nations, not part of the regime, can undo the actions of those that are parties to the agreement (Buchholz et al. 1998). When, however, a rogue nation is confronted after it acquires WMD, a best-shot technology then applies, leading inevitably to an exploitation of the large by the small. If past is prologue, then nations look to the United States and the United Kingdom to do something about the proliferation threat as they sit on their hands, or, worse yet, condemn any subsequent US or UK actions as excessive while secretly being relieved that the action was taken. Arms control regimes are certainly replete with collective action difﬁculties and hypocrisy.

Arms races Papers addressing arms races at the conference include those by Levine and Moraiz (2003), Matelly (2003), and Smith et al. (2003). Arms races may assume various

Arms trade, arms control and security

217

game forms including those of the Prisoner’s Dilemma, Chicken, and assurance depending on the pattern of payoffs of the participants. One hopeful sign for these arms races is that they usually involve two or three nations and, thus, some sort of curtailment may be achievable as has been the case between the United States and the ex-Soviet Union (Dixit and Skeath 1999: 107–110). The inability of much of the empirical literature to identify these races does not help to convince policymakers that arms races must be curbed. Ironically, arms races may result in less security – the exact opposite of their intent – if matching weapon for weapon means that neither country is more secure. If, additionally, these races impoverish the participants’ economies (e.g. the superpowers during the Cold War), then they may actually end up being less secure. Once countries come to this realization, the underlying game can change to that of assurance with a “focal” equilibrium where both countries disarm to some extent, as occurred at the end of the Cold War. A related collective action problem concerns R&D races associated with today’s high-technology weapons. Precision-guided munitions, new generations of jet ﬁghters, advanced (stealth) bombers, and modern warships have escalated R&D costs. These costs have brought up unit costs and have induced nations to seek out customers so as to spread R&D costs over a greater production run. As long as allies insist on not pooling their efforts to develop these advanced technologies, wasteful duplication of R&D will result, as underscored by the development of the Euroﬁghter, the F-22, and the Grippen. Some form of collective action is needed.

Arms production In the Middlesex conference, arms production was addressed in papers by Brauer (2000), Jackson (2003), and Hartley (2000). Numerous issues of arms production overlap with those of arms trade; hence, I shall be brief. Less-developed countries (LDCs), which according to Brauer (2000) have the same capability as that of developed countries to produce some types of arms, must decide if this is a good pathway to development. There are other ways – for example, through foreign direct investment – to acquire technological spill-ins. Once an LDC is heavily invested in arms production, there are incentives to supply beyond its own defense needs to exploit scale economies and learning economies, thus lowering unit costs. These sales then produce negative security externalities whenever the weapons end up with rogue nations, rebels, terrorists, or unstable governments. LDCs may become dependent on their arms industry and resist any overtures to participate in arms control regimes. Much as an economy can get “hooked” on nuclear energy and have no alternative energy supply in the short-run, LDCs can become addicted to their arms industry for foreign exchange. This addiction exacerbates the earlier identiﬁed problems associated with arms control regimes. The use of offsets may nurture this addiction as LDCs ﬁrst start to invest in the arms industry. Investment decisions can lead to irreversibilities and signiﬁcant adjustment costs.

218

Todd Sandler

For developed countries with their more diversiﬁed economies, this addiction is less of a worry; nevertheless, these countries must also resist the temptation to expand arms trade to reduce unit costs during the current era of shrinking budgets. Hartley (2000) offers some insightful views on this issue, since arms trade within NATO does not carry these security risk and also serves to reduce unit costs (also see Sandler and Hartley 1999: chapter 5).

Alliances In the essay on alliance formation, Sandler (1999) demonstrated how countries can improve their well-being by allying so as to sequester borders, thus economizing on protection costs. Alliances successfully address some collective action concerns, while raising others. In the latter categories, alliances are anticipated to undercontribute to purely public deterrence (Olson and Zeckhauser 1966; Sandler and Forbes 1980). If, however, public choice motives are considered whereby vested interests lobby (e.g. defense industry workers) to maintain the size of the military sector even when threat is diminished, then overspending inﬂuences may also be present. Overspending may also derive from opposing alliances responding to external threat posed by their counterpart (Bruce 1990). The proper institutional design of an alliance can do much to correct these collective action difﬁculties. For example, ally-speciﬁc beneﬁts, derived from defense spending, are thought to promote efﬁciency, particularly when these ally-speciﬁc beneﬁts are complementary to the jointly produced alliancewide beneﬁts (Hartley and Sandler 1999). Another well-known collective action problem has to do with the heterogeneity of the allies. With purely public deterrence, this asymmetry of allies leads to an exploitation of the large allies by the smaller ones as explained earlier (Olson 1965; Olson and Zeckhauser 1966). This exploitation tendency is, however, curtailed with conventional armaments whose deployment can result in force thinning along the allies’ borders. Allies that do not adequately guard their own perimeter may draw the conventional attack, thus motivating them to provide more of their own defenses.

Concluding remarks Although the essays of the Middlesex conference dealt with at least ﬁve different aspects of defense economics, the collective action theme permeates and unites all of the papers. These collective action concerns point to a host of market failures that continue to plague arms trade, control, and production. The end to the Cold War has not brought an end to allocative efﬁciency worries about defense decisions.

Acknowledgments Todd Sandler’s research was funded, in part, by a NATO Fellowship. His travel was supported by the Arms Trade Research Group, which sponsored the Middlesex conference on the Arms Trade, Security and Conﬂict during June 1999.

Arms trade, arms control and security

219

Notes 1 All ﬁgures are in 1990 constant dollars. The facts in this paragraph are from Stockholm International Peace Research Institute (SIPRI) (1999). 2 For more up-to-date treatments, see Hardin (1982) and Sandler (1992). 3 Sandler (1992) provides an in-depth discussion of all of these principles, which is not possible in this chapter. For further details, the interested reader should consult this book. 4 That is, when an individual equates his or her marginal rate of substitution (MRS) of the collective good (X) for the private good (Y ) to his or her marginal rate of transformation j (MRTXY ) between these goods, a greater sum of other agents’ MRSs (i.e. j =i MRSXY ) is being ignored. As a consequence, the equalities of MRSiXY = MRTXY associated with the Nash equilibrium imply a greater inequality: n

MRSiXY > MRTXY ,

i=1

5 6 7 8 9

j as j =i MRSXY grows with group size. On the exploitation hypothesis, see Olson (1965), Olson and Zeckhauser (1966), Sandler (1992: chapters 2–3), and Sandler and Forbes (1980). For these ﬁgures, see Hartley and Sandler (1999), Russett (1970), Sandler (1992: 103), and Sandler and Hartley (1995, 1999). On joint products, see Murdoch and Sandler (1982, 1984), and Sandler (1993). These and other qualiﬁcations are established in Sandler (1992). This aggregate technology is addressed in Hirshleifer (1983), Kanbur et al. (1999), and Sandler (1998).

References Alonso, M. and Hartley, K. (2000) Export controls, market structure, and international coordination. Defence and Peace Economics 11(5), 481–503. Brauer, J. (2000) Potential and actual arms production: implications for the arms trade debate. Defence and Peace Economics 11(5), 461–480. Bruce, N. (1990) Defense expenditures by countries in allied and adversarial relationships. Defence Economics 1(3), 179–195. Buchholz, W., Haslbeck, C. and Sandler, T. (1998) When does partial co-operation pay? Finanzarchiv 55(1), 1–20. Dixit, A. and Skeath, S. (1999) Games of Strategy. W.W. Norton, New York. Hardin, R. (1982) Collective Action. Johns Hopkins University Press, Baltimore, MD. Hartley, K. (2000) The beneﬁts and costs of the UK arms trade. Defence and Peace Economics 11(5), 445–459. Hartley, K. and Sandler, T. (1999) NATO burden-sharing: past and future. Journal of Peace Research 36(6), 665–680. Hirshleifer, J. (1983) From weakest-link to best-shot: the voluntary provision of public goods. Public Choice 41(3), 371–386. Jackson, I. (2003) The supply side implications of the arms trade. In: Levine, P. and Smith, R. (eds), Arms Trade, Security and Conﬂict, pp. 78–93. Routledge, London and New York. Kanbur, R., Sandler, T. and Morrison, K. M. (1999) The Future of Development Assistance: Common Pools and International Public Goods, Policy Essay No. 25. Overseas Development Council, Washington, DC.

220

Todd Sandler

Levine, P. and Moraiz, F. (2003) Rational wars with incomplete information. In: Levine, P. and Smith, R. (eds), Arms Trade, Security and Conﬂict, pp. 113–136. Routledge, London and New York. Martin, S. (1996) The Economics of Offsets. Harwood Academic Publishers, Amsterdam. Matelly, S. (2003) The determinants of US military expenditures. In: Levine, P. and Smith, R. (eds), Arms Trade, Security and Conﬂict, pp. 158–177. Routledge, London and New York. Mouzakis, F., Levine, P. and Sen, S. (2000) Arms export controls and emerging domestic producers. Defence and Peace Economics 11(5), 505–531. Murdoch, J. C. and Sandler, T. (1982) A theoretical and empirical analysis of NATO. Journal of Conﬂict Resolution 26(2), 237–263. Murdoch, J. C. and Sandler, T. (1984) Complementarity, free riding, and the military expenditures of NATO allies. Journal of Public Economics 25(1–2), 83–101. Olson, M. (1965) The Logic of Collective Action. Harvard University Press, Cambridge, MA. Olson, M. and Zeckhauser, R. (1966) An economic theory of alliances. Review of Economics and Statistics 48(3), 266–279. Ostrom, E. (1999) Context and collective action: four interactive building blocks for a family of explanatory theories. Unpublished manuscript, Indiana University, Bloomington, IN. Russett, B. M. (1970) What Price Vigilance? Yale University Press, New Haven, CT. Sandler, T. (1992) Collective Action: Theory and Applications. University of Michigan Press, Ann Arbor, MI. Sandler, T. (1993) The economic theory of alliances: a survey. Journal of Conﬂict Resolution 37(3), 446–483. Sandler, T. (1998) Global and regional public goods: a prognosis for collective action. Fiscal Studies 19(3), 221–247. Sandler, T. (1999) Alliance formation, alliance expansion, and the core. Journal of Conﬂict Resolution 43(6), 727–747. Sandler, T. and Forbes, J. F. (1980) Burden sharing, strategy, and the design of NATO. Economic Inquiry 18(3), 425–444. Sandler, T. and Hartley, K. (1995) The Economics of Defense. Cambridge University Press, Cambridge, UK. Sandler, T. and Hartley, K. (1999) The Political Economy of NATO: Past, Present, and into the 21st Century. Cambridge University Press, Cambridge, UK. Sen. S. Arms export controls: can economics succeed where politics fail. unpublished manuscript. Smith, R. and Udis, B. (2003) New challenges to arms export control. In: Levine, P. and Smith, R. (eds), Arms Trade, Security and Conﬂict, pp. 94–110. Routledge, London and New York. Smith, R., Dunne, P. and Nikolaidou, E. (2003) Arms race models and econometric applications. In: Levine, P. and Smith, R. (eds), Arms Trade, Security and Conﬂict, pp. 178–187. Routledge, London and New York. Stockholm International Peace Research Institute (SIPRI) (1999) SIPRI Yearbook 1999: Armaments, Disarmaments and International Security. Oxford University Press, New York. Zervos, V. (2003) Demand for space expenditure. In: Levine, P. and Smith, R. (eds), Arms Trade, Security and Conﬂict, pp. 188–205. Routledge, London and New York.

Index

2SLS method 192–3 action–reaction functions 159, 168, 174 Afghanistan 23, 186 aggregation technology of collective supply 39, 41–3, 214, 216, 219; best shot technology and coordination game 213; summation and Prisoner’s Dilemma 212; weakest link and matching 214 Aikaike information criteria (AIC) 166–7 Airbus Industries 87, 89–91 aircraft: operational at sea 79; shelters and runways 91 alliances: collaborative production and arms trade, literature on 55; conﬁgurations 142; cooperative game theory 138; formation 137–8, 155, 218; see also core Alonso, M. 215–16 Al-Yamamah project 86 America see United States Anderton, C. 11, 76 anti-personnel land-mine agreement 95 ANZUS alliance 138, 150 Argentina 21–2, 31, 55, 103 arms: dynamic, economic and technical limits of 159; producers, deﬁning 22–5; supply restrictions 34 arms control 209; classic theories of 100, 108; regimes 56, 64, 108, 215–16; relevant technologies for 214 arms export 100; aggregation 41; beneﬁts of 7–8; negative externalities on national security 94; public choice models and analysis 7 arms export controls 98–100; allies in 100; classic 96, 99; and emerging domestic producers 55; new challenges to 94; organizations 99, 101 arms production 217–18; capabilities, domestic 55; potential and actual 21; rank order of, potential 30; relevant industries 28 arms race 216–17; coefﬁcients 159–62; in shares of military spending in GDP 180; survey of

160; theories on 174; and war, effects of 135; see also United States and USSR arms race models 185; descriptive 160; econometric applications 178; econometric issues in estimating 179–81; normative 160; theoretical models of empirical problems 162 arms trade 209; in aerospace 89–90; debate, implications for 34; economic literature on 11; hidden costs 91; models of Levine et al. 55; and policy 215; regimes 56, 72; relevant technologies for 214; supply-side implications 78 “assurance” game 214, 217 augmented Dickey–Fuller (ADF) tests 175, 196 Australia 35, 150 Australia Group (AG) 37, 50, 53, 97, 105 Austria 137, 151 autoregressive conditional heteroskedasticity (ARCH) test 194 Axelrod, R. 98, 101 ballistic missile defense organisation (BMDO) 204 Baltic nations 137, 151 bankruptcy boundaries 62–3 bargaining: game of mutual defense 139, 156; with incomplete information, model of 114–15; procedure 114; solutions 156; strength of allies 138 Baucom, D.R. 201, 204 beggar-thy-neighbour problem 68 best shot and coordination game 213 Binmore, K. 140, 156 Biological and Toxin Weapons Convention (BTWC) 97, 104 Boeing Industries 91; civilian market dominance of 87 border lengths, alternative 144, 147 Brauer, J. 1, 21–35, 217 Brazil 21, 55 British Aerospace plc (BAe) 7, 14, 80–3, 89 Brito, D.L. 53, 113, 160, 179 Bruce, N. 218

222

Index

Brzoska, M. 23–4, 76 Bueno de Mesquita, B. 171 Butler, B. 9–10 buyers: engaged in regional conﬂicts 56; governments 1 Campaign Against the Arms Trade 7 Canada 150, 153 Carter embargo on arms sales 55 catch-all or know rule 50 characteristic or coalition function 139 Chemical Weapons Convention (CWC) 97, 104 Chicken game 213, 217 China 39, 42, 76, 94, 96, 98, 104, 186 civilian markets 81–2 club theory, formulation of alliances 137 Cobb–Douglas: production function 115; utility function of consumption 59 Code of Conduct for Arms Exports 106 Cold War 7, 13, 48, 55, 67, 80, 150, 153, 163, 171–2, 178, 189, 199, 209; end of 78–9, 94, 104, 158–9, 165; theoretical models 159–62 collective action: concerns 209, 215; principles of 210–11; problems 98, 218; recent insights and 211–15 commodity control list 50 complete information game 117 compliance costs 101; control 100 conﬂict: behaviour, foundations of 115; elasticity of 172 constant elasticity of substitution (CES) function 115; for aggregation of arms exports 41 consumption, substitution for security 65 contest-success functions (CSFs) concept of 113 controlled items, agreed lists of 48, 50 conventional arms transfer (CAT) control 108 Conventional Forces in Europe (CFE) Treaty 97, 99 conventional weapons 55, 94 cooperative arms control agreement 63 cooperative equilibria 138 cooperative game theory, theoretical preliminaries and deﬁnitions 138–41 Coordinating Committee for Multi-lateral Export Controls (CoCom) 94–7, 101, 104; problems of 100 core: of mutual defense game, analysis 155; proposals 144; solutions 138, 141 Cornish, P. 34, 97 costs: of arms exports 9–10; savings, incentive from allying 153 countries: domestic arms sector, incentives for 73; exported military capability 42; size and location 153, 155 Cournot–Nash equilibrium 63 Croatia 23, 55 Cusum and Cusum squared tests 183, 185 Cyprus 25, 35, 138 Czech Republic 23, 137

defence: budgets, limited 79; contracts 82; economics 188, 218; ﬁrms 91; industry workers 218; manufacturers 83; markets 79, 81; procurement process 83; spending 150, 210 Defence Exports Services Organisation (DESO) 9 Defence Industrial Base (DIB): domestic 79; European 88–9 developing nations: classiﬁcation 21; deﬁning 25 discrete choice model 43–6 Dixit, A. 113, 217 domestic military production 63; choice of 61 domestic production decision model: calibration of parameters 73–6; formal model of 56–72; summary of 71 Doornik, A.J. 194, 196–7 dual-use: export controls 38–9; goods 37, 43, 50–1, 94; systems 95; technologies 101–2 Dunne, J.P. 2, 76, 167, 171, 178–87, 190, 192 economic adjustment 78; effects on aerospace industry 83 economic size 169–70 encryption, controls on 105 Engle–Granger (EG) approach 196 environmental problems 38 equilibrium and Pareto-optimal ideal, departure between 210 equipment life cycle, longevity of 78 ethnic conﬂict 37 Euroﬁghter 2000 (EF-2000) 80, 83, 89, 217; and Tornado production gap 84–7 European aerospace: companies 87; industry, battle of sexes in 88; production, game theoretic model of 87–9 European Commission 49 European Space Agency (ESA) 192, 203 European Space Directory (ESD) 201 European Union (EU) 76, 138, 204; competition rules 81; GDP of 204; system of arms export controls 97 exit military markets 85 exploitation: hypothesis 210, 219; of large ally 210–11; tendency 218 export: of dual-use products 41; licences 7; market 43 export controls 37, 53; game 52; models of 39; multilateral regime 44, 49; quantitative or qualitative 42; regimes, evaluation of 48 Export Controls of Conventional Arms and Dual-Use Goods and Technologies 104 exporters: of dual-use technologies 8; of weapons 37 Falklands/Malvinas war 103 ‘fatigue’ terms 179 feasible mutual security (FMSR) 62–3 Fisher test 173 Forbes, J.F. 137, 218–19

Index 223 former Soviet Republics 217; denuclearization of 107 France 34, 39, 88–9, 100, 104, 150, 189, 216 free-rider: ally, small 210; beneﬁt 213; incentives 216; problem 68 Fudenberg, D. 46, 114–15 full information maximum likelihood (FIML) 193 Future Attack Submarine (FASM) 13–14 Future Unmanned Aircraft (FUMA) 89 game theory and conﬂict: literature 98; model of 114–34 García-Alonso, M. 1, 37–53 Gardner, R. 138, 140–1, 156 GATT, trading formulae of 81 general estimation model 190–2 Germany 39, 86, 88–9, 104, 138, 153, 189 global security function 67 goods and services: civil, UK exports of 11; militarily useful, categories 95; military, high and low tech 56–7 government: civil and military space expenditure (MSE) 188; optimization problem 59, 66; procurement of arms 57; restricting equilibrium 45; SEs 189 Granger causality test 166, 172–3, 194–5 Gray, C.S. 95, 99, 108 Greece 21–2, 28, 31, 39, 138, 151; military expenditure 182–3 Greece and Turkey: arms race 158, 178, 181–5 grievances 172; coefﬁcient 158–9; deﬁnition 158; elements 169; factor 202; terms 179 group: composition, inﬂuence of 210–11; size, inﬂuence of 210 guerrilla warfare and insurgencies 153–4 Gulf War 67, 91, 100 Hardin, R. 211, 219; treatment of collective action 214 Harkavy, R.E. 100–1, 108 Harrier 87, 92 Harris, R.I.D. 196–7 Hartley, K. 1, 5–19, 37–53, 55, 76, 78, 83, 86, 89–90, 99, 138, 150, 153–4, 156, 171, 178, 188, 190–1, 202–3, 210, 212–13, 215–19 Hawk 7, 87, 92 Hendry, F.D. 194, 196–7 High Performance Computers (HPC) 42, 52 high-technology arms 40, 47; buyer of 56–65; exports 39; imports 65; suppliers of 65–70 Hirshleifer, J. 42, 113, 115, 219; production and CSFs 115–16 human capital formation 31–4; measure of 21 human rights records 96 Hungary 137, 151 Iceland 35, 150 import-substituting industrialization, incentives for 55

India 31, 42, 209; military expenditure for 184; nuclear tests 95, 106 India and Pakistan arms race 158, 178, 184–5 Indonesia 12, 31, 96 industrial diversiﬁcation 21, 28–31 information: asymmetries 113–14; incomplete, model of war with 115–33; role of market in 12 International Atomic Energy Agency (IAEA) safeguards 50 International Institute for Strategic Studies (IISS) 23–4, 163, 171 international market for arms 57; partial equilibrium model 73 international negative externalities 41–3 International Standard Industrial Classiﬁcation (ISIC) code 21 international trade 78; cooperation 48 Intriligator, M.D. 53, 113, 160, 179 Iran 96, 98 Iraq 96, 100; invasion of Kuwait 107 Iridium global satellite-communications system 102 Iron Triangle 83, 87 ISIC 28–9; revision 2 25–6, 29; revision 3 25–6 Israel 22, 31, 40, 42, 86, 211 Italy 104, 150, 189 Jackson, I. 1, 78, 215, 217 Japan 39, 101, 138 Jehiel, P. 52–3, 98 job losses, world-wide 91 Johansen, S. 180; test 166 joint products 214, 219 Joint Strike Fighter (JSF) 80, 83, 85, 89 Keller, W. 94, 101 Khanna, J. 150, 153 killer applications 103, 107 Kollias, C. 180–1 Kosovo 55, 209 Kruskall–Wallis H 30 Kuwait 25, 35, 39, 107 laissez-faire (LF) 56, 68, 69, 72 Lanchester coefﬁcient 59, 63, 76 Landmine Treaty 97 leader–follower behavior 211 Leontief technology 115 less-developed countries (LDCs) investment decisions 217 Levine, P. 1–2, 11, 38, 55–76, 95, 108, 113–35, 179, 216 life cycle costs 84 light weapons 95; producers of 23 likelihood ratio (LR) 167 Linden, M. 192, 195–6 Luce, R.D. 140, 155–6 Luxembourg 25, 150

224

Index

make-or-buy strategy of defence procurement 78, 84 Mann–Whitney U 30 marginal rate of substitution (MRS) 219 markets: clearing condition 76; equilibrium and comparison of regimes 70–2; hypothetical, of four identical blocs model of 76; role in information and knowledge 12; structure 37, 39, 66 Martin, S. 1, 5–19, 40, 86, 103, 215 Mastanduno, M. 94, 96, 108 Masten, S.E. 78, 84 Matelly, S. 2, 158–75, 216 merger: activity 87; between BAe and GEC-Marconi 91 Mexico 21–2, 28, 31 Middlesex University Business School Conference on Arms Trade Security and Conﬂict 1, 209 militarisation 170–1 military capability: deﬁnition of 43; investing in 63; production function 57; threshold value of 61 military expenditures 162; and GDP 73, 181–4; positive interdependence between 159 military goods: high-tech 40, 47, 56; low-tech 37, 49, 56 military spending 162, 209 Missile Technology Control Regime (MTCR) 37, 49–50, 97, 103, 105, 107 models: efﬁcient allocation of resources 115; of horizontal differentiation 44; of regional conﬂicts 158; of trade 38 monopoly protection, theoretical analysis of lobbying 113 Moodie 100–1 Moore’s Law 103, 108 Moraiz, F. 2, 113–35, 216 Mouzakis, F. 1, 55–76, 215 Murdoch, J.C. 137, 186, 203, 219 mutual defense games: analysis of 150–1; core of 143; NATO alliance application 150–1 mutual defense pacts: four and ﬁve allies 148–9; three countries 141–8 NASA 204 Nash equilibrium 210–11; in military capability 62–3; mixed-strategy 214; in pure strategies 44, 52; symmetric 45; unique 47 National Defense Authorization Act on High Performance Computers (HPC) 42 national and international defence budgets 78 NATO 6, 11, 76, 80, 99, 137–9, 202, 218; alliance 149, 189; burden sharing 150; defense burdens 156; deﬁnition of 189; expansion 141, 155; formation 150; infrastructure 211; membership 154; new eastern members 151; (US–EU) space alliance 190; and the USSR, space race between 188 natural defenses 155; and risk factors 152–3

negative externalities 45 Neuman, S.G. 100–1, 108 Nikolaidou, E. 2, 178–87 Nolan, J.E. 94, 101, 107–8 non-cooperative game 68 non-cooperative regulation (NCR) 68 Non-Proliferation Treaty (NPT) 97, 104, 107 North Korea 96; development of nuclear weapons 213 Norway 150, 153 no-undercut policy 50 nuclear-powered submarines 13–14, 86 nuclear suppliers group (NSG) 37, 49–50, 97 offsets, use of 216–17 OLS method of estimation 194 Olson, M. 137, 149, 209–12, 218–19; propositions 211; treatment of collective action 214 optimization problem 60 order size, decreasing 79, 83 Organization of Security and Cooperation in Europe 138 Ostrom, C.W. 161–2 overcapacity 79, 92 Pakistan 42, 56, 209; military expenditure for 184; nuclear tests 95, 106; see also India and Pakistan arms race Pareto-ranking of regimes 73 pariah states 96 Partnership for Peace (PFP) 137–8, 153–4 payoffs: individual rationality 140; Pareto optimality 140; pattern of 217; structure of 106; vectors, imputation 140 Perfect Bayesian Equilibrium (PBE) 144 Permanent Members of the Security Council (P5) 95, 97 Pierre, A.J. 94–5, 99, 108 Pikayev, A.A. 97, 108 policies: on arms exports 40; of offsets 215 ‘pork-barrel’ problems 83, 91 Portugal 21–2, 28, 31, 150 post-Cold War era 37, 67; military expenditures 174 potential arms production, the PDC index 25–8 potential defense capacity (PDC): employment share 35; index 21, 25–6, 28, 33–4 Prisoner’s Dilemma game 47–8, 212, 217; cartel stability 98; of non-cooperative behaviour 87 producer cartel 68–9; consisting of coordinated regulation 56 proﬁt-seeking, economic models of 113 proliferation of arms producers 23, 76 public choice: models of economics and conﬂict 113; motives 218 public good model 137 Raiffa, H. 140, 155–6 Rana, S. 23–4, 35 Rand 13–14, 108

Index 225 Rasmusen, E. 87, 135 Rawlsian maximin principle 155 ‘reaction’ terms 179 Reagan defense buildup 150 redundant defence workers 10; studies of 9 regimes: comparison of 71–2; three kinds 68–70 Reinicke, W.H. 97, 100, 106 Research and Development (R&D): costs 44; levies on arms exports 7; military 79, 102; private 102; races in high-technology weapons 217; subsidy 43 restriction game 43–7; continuous choice model 46–7 restrictions: on arms imports 73; on quality 40 Richardson, L.F. 158, 161, 174 Richardson: model-based space race model 191; three coefﬁcients 158 Richardson model of arms race 159–60, 178, 190; criticisms of 160; lack of a budget constraint 180; ‘structure form’ of 179 rogue nation 214, 216; neutralization 213 Rolls-Royce 14, 80 Romania 137, 151 Rotillon, G. 38, 52–3 Russia 39, 42, 76, 94, 100, 104, 138, 216; see also USSR Saferworld 7 safety standards 81 Sandler, T. 1–2, 38–9, 41, 45–6, 55, 76, 78, 83, 92, 98–9, 137–56, 178, 188, 190–1, 202–3, 209–19 Sargent, K. 39, 45–6 Saudi Arabia 12, 39 Scarf, H. 156 Schelling, T.C. 108 school enrollment data 21, 31–3 Schwarz Bayesian criterion (SBC) 166–7 security 209; beneﬁts of control 98; concern, degree of 48, 52; environment 186; externalities 215; grounds 47; perception of countries, index of 42; risk 218; of suppliers 67; of supply 88 Self-Regulatory Organizations (SROs) 106 Sen, S. 1, 184, 215 Shapley value 155 Slovakia 23, 137, 151 Slovenia 25, 137, 151 “small arms”, production of 23 Smith, R.P. 1–2, 38, 52, 55–76, 92, 94–108, 178–87, 192, 194, 196, 203, 215–16 Society of British Aerospace Companies (SBAC) 81–2, 86, 89 South Africa 22, 31, 40, 96 South Korea 12, 39 Soviet Union see USSR space alliance model 190–1 space expenditure (SE): characteristics of NATO and the USSR 188–9; commercial 189; demand for 188; model of behaviour of Western

European countries, United states and USSR 188–9; theoretical framework of demand for 189–92 space race: model 190–1; NATO and USSR 188; United States and USSR 192–200 Spain 21–2, 28, 31, 151, 189 Spearman rank-correlation 31 Stag Hunt game 45 Standard Industrial Classiﬁcation 10 Stockholm International Peace Research Institute (SIPRI) 35, 39, 97, 163, 186, 209, 219; arms export data 12; data 181; ﬁgures for military expenditures 184; Yearbook 49, 51, 53, 163, 182 Strategic Defense Initiative (SDI) 197–8, 204 strategic weapons and deterrence 149–50; ballistic missile submarines (SSBNs) 13 Supplier Agreements 97 suppliers: behaviour 70; exports/military expenditure 74; governments, arms exports 1; groups 101, 104; of high-tech weapons 56; nations, economic beneﬁts to 108; top ﬁve 39 supply-side: controls on transfer of weapons-related technologies 37; implications of arms trade 78 Sweden 151, 153 Switzerland 25, 137, 151 Taiwan 35, 39; sale of F16s to 98 technology: of appropriation 115–16; of conﬂict 115; impact on arms control 101–4; of production 115 terrorism 37 threats: involving conventional weapons 141; type of 139 time-series pattern of military expenditure 178 Tirole, J. 46, 114–15 ‘tit for tat’ strategy 98, 211 Tornado aircraft 83, 91 trade: in dual-use products 38; and employment 215; externalities 7–8, 11–13; linkages 12 ‘trade–ﬂag’ relationship 11 transaction costs 154–5; considerations 139; of group formation 210; hypothesis 84; paradigm 78 Trigger List 50 trigger strategies 98 Turkey 21–2, 25, 31, 39, 138, 150–1; GDP 186; invasion of Cyprus 181–2; military expenditure 183; see also Greece and Turkey arms race Udis, B. 2, 52, 55, 94–108, 215 Ukraine 98, 151 Union of Soviet Socialist Republics (USSR) 101; break-up of 23; dissolution and disarmament 164; military activity for the Cold War 164; military expenditures 163–4, 166, 168; populations and GNP/GDP 203; ‘Sputnik’ 192; see also United States and USSR unit costs, rising 79, 82–3

226

Index

United Kingdom (UK) 34, 39, 89, 104, 150, 189, 216; active combat aircraft 79–80; arms trade 5–6; beneﬁts to 8; Department of Trade and Industry 7; exports of civil goods and services 11; importance of government 6–7; industrial policies 81; international competitiveness 11; membership of international treaty organizations and military alliances 11; military production industries 74; Ministry of Defence (MoD) 7, 9, 15, 17, 19, 78, 87–8; motor car industry 80; resources, alternative-use value of 9; security threat to 9; share of military expenditure 180; submarine industry 13–14, 86; subsidy for defence exports 9 United Kingdom (UK) aerospace industry: battle of the sexes in 88; companies 87, 91; economic adjustment in 82–4; industry 5, 78–82; overcapacity in 89; suppliers 81, 91 United Kingdom (UK) arms export: balance of payments, impact of reduction in 11; beneﬁt–cost framework 7–10; economic impacts of limiting 5, 10–11; market, employment behaviour 15; pilot study 12 United Kingdom (UK) defence industry: balance of trade surplus 6; employment data, model and empirical results 15–19; equipment 7, 12; industrial base 5, 13–15; procurement 91; and security policy 11; spending, reduction in 7 United Kingdom (UK) domestic Defence Industrial Base (DIB) 8, 78, 87–8 United Nations (UN) 138, 202; arms trade registry 34; arms transfer register 97, 99, 105; Charter 95; Department of Economic and Social Affairs 29; embargoes or sanctions 97; Security Council of the United Nations (UNSC) 37, 216 United Nations Industrial Development Organisation (UNIDO) 27, 35 United States 13, 34, 39, 76, 100, 104, 150, 153, 211, 216–17; action–reaction, basic function of 166–8; aerospace industry 90; Apollo programme 192; Arms Control and Disarmament Agency (USACDA) 22–4, 53, 107, 163–4, 169, 203; arms export control process 108; and Britain, 1817 agreement 99; defence budget and public deﬁcit 171; DOD Cooperative Threat Reduction (Nunn-Lugar) Programme 98; economic size and military implication 170; exploitation 105; export share of arms 172; GDP of 204; General Accounting Ofﬁce 156; grievance 159, 165, 169–73; hegemony 159, 168; leadership 105; nuclear umbrella 149; populations and GNP/GDP 203; SDI 197

United States military expenditures 163, 165, 168, 204; determinants 158; forecast 168; impact on ratio to public expenditures 170; value of 173 United States and USSR: actions–relations, theoretical models of 160; armed force ratio of 170; arms control treaties 38; arms race between 160; blocks, opposition of 158; GNPs ratios of 170; military activity for the Cold War 164; military expenditures 163–4; negotiations on the MTCR 108; space race 200–2; space race static model 198; strategic interdependence 162–5 Uruguay Round of GATT 91 Varian, H.R. 52, 135 vector error correction model (VECM) 179–80, 183, 185 Vietnam War 150, 204 Visegrad nations 151 Ward, M.D. 171–2 warfare, type of 154 Warsaw Pact 104 Washington Naval treaty 1992 101 Wassenaar Agreement (WA) 37, 50, 94, 97, 99, 101, 104–6; autonomy, lack of 99; degree of credibility in the system 107; discount rate 107; General Working Group of 105; monitoring capability of system 107; munitions list 105 weakest link see aggregation technology 214 weapons of mass destruction (WMD) 95, 105, 214; proliferation of 215–16 Webb, T. 5, 7, 91 Western European Union (WEU) 203 ‘Western Europe’, deﬁnition of country sample 189 West Germany 150–1 World Bank 35; World Development Report 25, 32 World Development Movement 7 world gross national product (GNP) 209 world market: for defence exports 8; equilibrium for high-tech arms 70 world-market arms suppliers, big 34, 216 World Trade Organization 48 Wulf, H. 21–3, 35 Yugoslavia, former 23, 55, 67, 179 Zangger Committee (ZC) 37, 50 Zeckhauser, R. 137, 149, 218–19 Zervos, V. 2, 188–204, 215

The Arms Trade: Security and Conflict (Routledge Studies in Defenseeconomics, 5)

Security and Post-Conflict Reconstruction (Routledge Global Security Studies)

Security and Post-Conflict Reconstruction (Routledge Global Security Studies)

Conflict Management and African Politics: Ripeness, Bargaining, and Mediation (Routledge Studies in Security and Conflict Management)

The Arms Trade and International Law (Studies in International Law)

Competition and Trade Policies: Coherence or Conflict (Routledge Studies in the Modern World Economy , No 13)

Conflict, Security and the Reshaping of Society: The Civilization of War (Routledge studies in liberty and security)

Corpus Stylistics (Routledge Studies in Corpuslinguistics, 5)

Globalisation, Domestic Politics and Regionalism: The ASEAN Free Trade Area (Routledge Warwick Studies in Globalisation, 5)

The International Arms Trade (WCMW - War and Conflict in the Modern World)

Nuclear Proliferation and International Security (Routledge Global Security Studies)

Nuclear Proliferation and International Security (Routledge Global Security Studies)

Gender and International Security: Feminist Perspectives (Routledge Critical Security Studies)

The Routledge Handbook of Security Studies (Routledge Handbooks)

Power Shifts, Strategy and War: Declining States and International Conflict (Routledge Global Security Studies)

Iran and Nuclear Weapons: Protracted Conflict and Proliferation (Routledge Global Security Studies)

Military Nanotechnology: New Technology and arms Control (Contemporary Security Studies)

Trade and Investment in China: The European Experience (Routledge Studies in the Growth Economies of Asia)

Arming the South: The Economics of Military Expenditure, Arms Production and Arms Trade in Developing Countries

Arms Trade and Economic Development: Theory and Policy in Offsets (Studies in Defence Economics)

Arms and Conflict in the Middle East (Contributions to Conflict Management Peace Economics and Development)

Territorial Changes and International Conflict (Studies in International Conflict Vol 5)

War, Image and Legitimacy: Viewing Contemporary Conflict (Contemporary Security Studies)

Trilateral Security: Defense and Arms Control Policies in the 1980s

Complexity and the Experience of Values, Conflict and Compromise (Routledge Studies in Complexity and Management)

The Legal and Moral Aspects of International Trade: Freedom and Trade: Volume Three (Routledge Studies in the Modern World Economy)

The Economics and Politics of International Trade: Freedom and Trade (Routledge Studies in the Modern World Economy, 10)

Uncertain Partners: Stalin, Mao, and the Korean War (Studies in International Security and Arms Control)

Intra-Asian Trade and the World Market (Routledge Studies in the Modern History of Asia)

5 The Arms of the Kraken

Energy Security in Asia (Routledge Security in Asia Pacific Series)

The Arms Trade: Security and Conflict (Routledge Studies in Defenseeconomics, 5)

Security and Post-Conflict Reconstruction (Routledge Global Security Studies)

Security and Post-Conflict Reconstruction (Routledge Global Security Studies)

Conflict Management and African Politics: Ripeness, Bargaining, and Mediation (Routledge Studies in Security and Conflict Management)

The Arms Trade and International Law (Studies in International Law)

Competition and Trade Policies: Coherence or Conflict (Routledge Studies in the Modern World Economy , No 13)

Conflict, Security and the Reshaping of Society: The Civilization of War (Routledge studies in liberty and security)

Corpus Stylistics (Routledge Studies in Corpuslinguistics, 5)

Globalisation, Domestic Politics and Regionalism: The ASEAN Free Trade Area (Routledge Warwick Studies in Globalisation, 5)

The International Arms Trade (WCMW - War and Conflict in the Modern World)

Nuclear Proliferation and International Security (Routledge Global Security Studies)

Nuclear Proliferation and International Security (Routledge Global Security Studies)

Gender and International Security: Feminist Perspectives (Routledge Critical Security Studies)

The Routledge Handbook of Security Studies (Routledge Handbooks)

Power Shifts, Strategy and War: Declining States and International Conflict (Routledge Global Security Studies)

Iran and Nuclear Weapons: Protracted Conflict and Proliferation (Routledge Global Security Studies)

Military Nanotechnology: New Technology and arms Control (Contemporary Security Studies)

Trade and Investment in China: The European Experience (Routledge Studies in the Growth Economies of Asia)

Arming the South: The Economics of Military Expenditure, Arms Production and Arms Trade in Developing Countries

Arms Trade and Economic Development: Theory and Policy in Offsets (Studies in Defence Economics)

Arms and Conflict in the Middle East (Contributions to Conflict Management Peace Economics and Development)

Territorial Changes and International Conflict (Studies in International Conflict Vol 5)

War, Image and Legitimacy: Viewing Contemporary Conflict (Contemporary Security Studies)

Trilateral Security: Defense and Arms Control Policies in the 1980s

Complexity and the Experience of Values, Conflict and Compromise (Routledge Studies in Complexity and Management)

The Legal and Moral Aspects of International Trade: Freedom and Trade: Volume Three (Routledge Studies in the Modern World Economy)

The Economics and Politics of International Trade: Freedom and Trade (Routledge Studies in the Modern World Economy, 10)

Uncertain Partners: Stalin, Mao, and the Korean War (Studies in International Security and Arms Control)

Intra-Asian Trade and the World Market (Routledge Studies in the Modern History of Asia)

5 The Arms of the Kraken

Energy Security in Asia (Routledge Security in Asia Pacific Series)

Recommend Documents