Crisis and War 9780773561212

Table of contents :
Contents
Tables and Figures
Preface
1 International War and its Causes
2 War and Expected Utility Reconsidered
3 Expected Utility, Crisis, and War
4 Conflict and Cohesion: The Blind Men Meet the Elephant
5 Testing Externalization
6 Balance of Power Theory: The Search for Meaning
7 Weighing the Balance
8 Some Residual Analysis and Conclusions
Appendix: Sources of Data
Notes
Bibliography
Index
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
Y
Z

Citation preview

Crisis and War

Using an analysis of patterns of international crisis and war from 1948 to 1975, Patrick James suggests why some international crises result in war while others do not. Over one hundred cases are used to assess three of the most prominent explanations for crisis escalation to war: (i) war is the result of rational choice by leaders who expect to gain from it; (2) war is the product of the outward projection of political unrest within states; and (3) war is the result of classic balance of power politics. James concludes that the best explanations for war include elements from all three categories. His analysis suggests that certain explanations have not been sufficiently considered in previous works. For example, prior domestic strife can be linked to the escalation of foreign conflict in an impressive number of cases, a discovery which runs counter to the consensus that has emerged over the last two decades among political scientists. Prior research on causes of war has often lacked rigour. James has tried to remedy this through a long-term, comparative approach to the subject matter. While Crisis and War follows the tradition of aggregate research, statistical analysis is always connected to particular events through discussions of situations and leaders in both diplomatic history and contemporary world politics. James' comprehensive and original approach to past theories both clarifies and critiques them. Patrick James is a member of the Department of Political Science, McGill University.

This page intentionally left blank

Crisis and War PATRICK JAMES

McGill-Queen's University Press Kingston and Montreal

© McGill-Queen's University Press 1988 ISBN 0-7735-0574-1

Legal deposit ist quarter 1988 Bibliotheque nationale du Quebec

Printed in Canada on acid-free paper This book has been published with the help of a grant from the Social Science Federation of Canada, using funds provided by the Social Sciences and Humanities Research Council of Canada. Canadian Cataloguing in Publication Data James, Patrick, 1957— Crisis and war Includes index. Bibliography: p. ISBN 0-7735-0574-1 1. War. I. Title. U21.2.J34 1988 355'.027 C87-090281-4

Tables 5.2 and 5.3 are from Edward R. Tufte, Political Control of the Economy. Copyright (c) 1978 by Princeton University Press. Reprinted by permission of Princeton University Press. Tables 2.1, 2.2, 2.3 and 2.4 are from Bruce Bueno de Mesquite, The War Trap. Copyright (c) 1981, Yale University Press. Reprinted by permission of Yale University Press. Figure 1.2 is taken from The Domestic Context of American Foreign Policy by Barry B. Hughes. Copyright (c) 1978, W.H. Freeman and Company. Used by permission. Figure 2.1 is taken from Olson & Bailey, "Positive Time Preference," Journal of Political Economy 89 (1981): 1-25. Used by permission of the University of Chicago Press.

Contents

Tables and Figures vii Preface ix 1 International War and its Causes 3 2 War and Expected Utility Reconsidered

28

3 Expected Utility, Crisis, and War 49 4 Conflict and Cohesion: The Blind Men Meet the Elephant 78 5 Testing Externalization 92 6 Balance of Power Theory: The Search for Meaning 116 7 Weighing the Balance 129 8 Some Residual Analysis and Conclusions 145 Appendix: Sources of Data 156 Notes 158 Bibliography Index 188

176

This page intentionally left blank

Tables and Figures

1.1 2.1 2.2 2.3 2.4 2.5 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 4.1 5.1 5.2

TABLES

International crises: 1948-1975 18 Components of equation 2.1 30 A typology of states 32 Derivation of equation 2.2 33 War and rational choice 37 Participation by intermediaries 46 Outside actors for the regions 51 Decision rules 64 Expected utility and war 71 Expected utility and war: bilateral cases 71 Expected utility and war: identical policies for initiator and defender 71 Value threats for crises with Uy = i 72 Expected utility difference and war 74 Form of government and expected utility 74 Trigger violence and expected utility 75 Duration of crisis and expected utility 75 United Nations activity and expected utility 75 Regional organization activity and expected utility 76 Frequency of crisis participation and expected utility 76 Location and expected utility 76 Rummel's indicators of conflict 80 Tufte's economic indicators 95 Elections and economic acceleration, 27 democracies, 1961-1972 96

viii Tables

5.3 Inflation, unemployment, and presidential elections, 1946—1976 97 5.4 Indicators of manifest conflict 103 5.5 Correlation matrix for indicators of manifest conflict 105 5.6 Latent conflict and war 108 5.7 Manifest conflict and war no 5.8 Bilateral expected utility and war 111 5.9 Form of government and manifest conflict 111 5.10 Trigger violence and manifest conflict 112 5.11 Duration of crisis and manifest conflict 112 5.12 United Nations activity and manifest conflict 112 5.13 Regional organization activity and manifest conflict 113 5.14 Frequency of crisis participation and manifest conflict 113 5.15 Location and manifest conflict 113 7.1 Indices of national power 132 7.2 Polarity and war 138 7.3 Coalitional power asymmetry and war 139 7.4 Crisis environment and war 140 7.5 Form of government and polarity 141 7.6 Trigger violence and polarity 141 7.7 Duration of crisis and polarity 141 7.8 United Nations activity and polarity 142 7.9 Regional organization activity and polarity 142 7.10 Frequency of crisis participation and polarity 142 7.11 Location and polarity 143 8.1 Uniquely correct predictions for expected utility, latent conflict, manifest conflict, and subsystemic polarity 146 8.2 Anomalies 149 8.3 Loglinear analysis 151 8.4 International war: win and loss records of initiators 152 FIGURES 1.1 International conflict 9 1.2 The Hughes foreign policy model 11 2.1 Discounting future utility: the c0 x C1 consumption space 42

Preface

This book is intended as a contribution to knowledge about international crises and their connection to the outbreak of war. The specific objective is to obtain further understanding of the causes of war through a systematic investigation of crises and their processes of resolution. Rather than attempting to formulate new theories of war, this research effort focuses on the clarification and testing of existing theories. A number of theories about the causes of war are reviewed and evaluated to determine which are most useful in distinguishing crises that escalate to war from those that are resolved in other ways. Three frameworks of analysis, which correspond to Waltz's images of war - the individual, state and system - are appraised and restructured to facilitate testing in the context of crisis and war. The theories are, respectively, expected utility, conflict and cohesion, and balance of power. Although other theories might just as easily have been selected, limitations of time and energy dictated that only one representative from each image of war could be considered. Collectively, however, the theories that are included offer a wide field of vision on the subjects of international conflict, crisis and war. Since the primary method of testing is data-based research, mathematical notation and statistical analysis are sometimes presented. For the most part, the statistics are rudimentary measures of association commonly used in the assessment of contingency tables. To assist readers who are not familiar with quantitative methods, the mathematical sections are accompanied by non-technical summaries. Chapter 1 deals with fundamental concepts and describes the scope of the investigation. Chapters 2, 4 and 6 review the research programs in expected utility, conflict and cohesion, and balance of power, respectively. Chapters 3, 5 and 7 test hypotheses derived from each

x

Preface

of the three theories. Chapter 8 contains some auxiliary testing, including the evaluation of a model of crisis outcomes comprised of independent variables derived from all three theories. This final chapter also synthesizes the findings and suggests directions for future research. This book has benefited from a number of sources of intellectual support. Colleagues at McGill and elsewhere have been kind enough to read and criticize this manuscript. I am grateful to Davis B. Bobrow, Stephen Bornstein, Stephen Hill, Baldev Raj Nayar, Mancur Olson, Joe A. Oppenheimer, Warren R. Phillips, Filippo Sabetti, Eric Uslaner and Oran R. Young for their helpful comments. Bruce Bueno de Mesquita of the University of Rochester provided me with some essential components of his data set, for which I am very thankful. There are two scholars to whom I owe special debts. Jonathan Wilkenfeld, my dissertation advisor, read this manuscript at several stages and it bears his imprint. Michael Brecher, my colleague at McGill, also has read and commented extensively on the manuscript. To each of these fine scholars I am especially grateful. This book has been published with the help of a grant from the Social Science Federation of Canada, using funds provided by the Social Sciences and Humanities Research Council of Canada. Several other agencies have provided funding at various stages of this research project. Grants from the University of Manitoba and McGill University assisted in the compilation and analysis of data. With those funds I obtained the valuable services of Daryll Clow, Frances Hisey, Tod Hoffman, Lawrence Katcher, Esther Mayer and Henry Mizrahi as research assistants. McGill's Social Science Computing Centre, and its Associate Director, Susan Czarnocki, have been very helpful also. I am very grateful to the staff at McGill-Queen's University Press for their valuable assistance in the publication of this manuscript. Philip Cercone and Joan McGilvray, the Director and Coordinating Editor, respectively, have been extremely helpful. Ruth Crow also provided useful editorial input. I have been fortunate to receive encouragement and assistance over the years from many sources. It should be emphasized that I alone am responsible for the errors which may be revealed in this book.

Crisis and War

This page intentionally left blank

CHAPTER ONE

International War and its Causes

Concern for man himself and his fate must always form the chief interest of all technical endeavors, concern for the great unsolved problems of the organization of labor and the distribution of goods - in order that the creations of our mind shall be a blessing and not a curse to mankind. Never forget this in the midst of your diagrams and equations. ALBERT EINSTEIN, ADDRESS AT THE CALIFORNIA INSTITUTE OF TECHNOLOGY,

193!

What distinguishes war from international crises that do not result in war? No greater question confronts the scholar of world politics. The devastating impact of war has earned for it more attention than any other political phenomenon. Historians, sociologists, economists and other social scientists have all endeavoured to explain the causes of major international violence, often finding it necessary to transcend the orthodox boundaries of their particular disciplines. A quick glance at the shelves of any college library will make that point obvious. Some scholars have compiled case histories in their efforts to account for war; others have attempted to explain it with aggregate data and quantitative methods.1 Some students of world politics have emphasized a single cause of war; others have examined a range of interrelated causes.2 Some have pursued a cross-sectional research strategy; others have traced the patterns of violent conflict over time.3 The scholarship has been as varied and complex as warfare itself. While impressive in scope and method, the literature on war maintains a wistful character. Even the most optimistic observer would have to concede that a general understanding of the causes of war lies far beyond reach. Research accumulates at an almost maddening pace, yet comprehension seems to lag behind. This state of affairs

4 Crisis and War

often has been attributed to the novelty of systematic research and the ephemeral nature of international relations.4 Since these conditions will not change in the foreseeable future, the concerned scholar is confronted with the problem of how to proceed. Taking a philosophically conservative point of view, Eberwein advocates a sceptical approach to the study of international violence. Scholars, he argues, should strive for greater understanding of "what theories in fact already exist, and their respective potencies" rather than seeking to create a "generation of allegedly 'new' theories of war."5 Theories are indeed available in great numbers, but many are incomplete or uncertain. In some instances the architects of a theory have left crucial assumptions implicit, while in others the range of application of the theory is unclear. It almost goes without saying that there are cases in which both of these difficulties are present. A prominent example would be the Leninist theory of imperialism, for which the basic assumptions and period of relevance have been constant subjects of debate. As a logical implication of this argument about cumulative understanding, there should be a greater emphasis on the clarification of existing theories, along with comparative testing. Such an approach would have the potential to reveal which theories possess the greatest explanatory power and under what conditions they work best. With that objective in mind, this study subjects a subset of the existing theories of war to a critical reappraisal and to individual and comparative testing. The purpose of this exercise is to assess these theories in absolute and relative terms as explanations of war. While some highly interesting theories undoubtedly remain undetected, a better understanding of the past can only enhance prospects for their discovery. Thus the call for a reassessment of what is known does not imply a retreat. According to one recent evaluation of the academic literature on international conflict, it may mean exactly the opposite: theoretical articulation [is] a task that is too rarely attended to in the field of quantitative international relations. Theoretical articulation is the clarification and extension of the theoretical structure of an existing paradigm. Clarification of the theoretical structure enhances the ease of application and instruction of a paradigm and creates a stronger structure to expand; extension of the theoretical structure demonstrates the ability (or inability) of a paradigm to predict, and thus explain, previously inexplicable phenomena. In addition, theoretical development of a paradigm allows the scientific community to test the logical validity and generality of the theories upon which they base their explanations.6

5 International War and its Causes

This exposition implies that a more comprehensive understanding of the causes of war must await the re-evaluation of theoretical axioms and, where possible, extension of their explanatory power. Two classes of questions arise at this point about a research project like the present one. One stratum is conceptual or definitional. For instance, what constitutes an "international war" or the more general class of "international conflict"? Systematic efforts to discover the causes of war date from the models of arms races developed by Richardson and the panoramic historical study by Wright, and each of these scholars developed an operational definition.7 Some facets of their definitions of war continue to enjoy popularity today: for example, a battle-related casualty criterion is guaranteed to appear in any data-based study. The more general point to be made is that comparative research of the kind proposed here requires operational rendering of the subject matter. International war, conflict, and other relevant phenomena will be defined in such terms, thus enabling selection from a wide range of evidence. Another set of questions deals with the methods of inquiry. Which theories should be selected for testing, and by what criteria? Once chosen, how are they to be evaluated? This chapter begins with a discussion of concept formation. Problems related to the methods of inquiry are then dealt with at some length. These discussions are followed by a presentation of the evidence that will be used to assess the theories of war.

CONCEPTS: INTERNATIONAL CONFLICT AND WAR Although some specific types of international conflict (such as crisis, war, and protracted social conflict) have been isolated and described, formal definitions of international conflict in general have rarely been presented.8 The definition used in this study was selected for its plausibility, range of application, and rigorous character. An international conflict is defined here as any interaction delimited in time and space, involving two or more international entities (whether states or transnational actors) which possess non-identical preference orderings over one or more sets of alternative choices.9 This definition has several advantages. First, it is unbiased. It does not entail either a state-centred view of international politics or the perspective of interdependence.10 The definition does not focus attention strictly on states and their interactions, but recognizes also the presence of a wide variety of non-state actors at the international level. These include both individuals, such as the Secretary General

6 Crisis and War

of the United Nations and groups such as the International Labor Organization. Second, the definition is plausible because, to use the language of game theory, it treats conflict as a mixed-motive interaction. It allows for considerable agreement among parties to a conflict, requiring only that their preferences be different over at least one set of alternatives.11 This is in keeping with the observation that in the real world, as opposed to that of the game theorist, games of pure conflict - like those of pure cooperation - are exceedingly rare. Rigorous examination of commonly held notions about conflict confers a third advantage. Consider Davies' formulation of conflict as "the total class of antagonistic personal acts (both non-violent and violent), ranging from a Violent' argument between brothers to a war between two nations."12 The literature on international conflict is consistent with that description, but the use of a term like "antagonistic" carries with it a degree of uncertainty which complicates the process of operationalization. Among other things, the threshold of disharmony that must be reached for a conflict to exist may be difficult to specify. The definition adopted here incorporates intuitive notions, such as those articulated by Davies, without sacrificing precision. There are also some axioms attendant to the definition which to some extent may be viewed as disadvantages. This definition of international conflict necessarily assumes that actors at the international level have identifiable preference orderings.13 It also is assumed that interactions are the result of purposeful behaviour, as opposed to random or unpredictable processes. Although neither of these assumptions is valid in a universal sense, that will not create practical difficulties. As will become apparent, international conflict creates a set of conditions under which both assumptions are closely approximated. Defining war poses a separate conceptual problem. It is resolved here by employing the Correlates of War (cow) Project definition to distinguish wars from other events. The Project's defining conditions for international war, described by Small and Singer, are elaborate. To present them in detail would be rather time-consuming and of little practical value, for they have been evaluated in a number of readily available sources. Only the two general criteria used by the cow Project to distinguish wars will be mentioned here: (i) the participants must be nation-states; and (2) the conflict must have led to'onfe thousand or more combat-related casualties.14 The Small and Singer conditions are consistent with implicit (and explicit) definitions of war used in case studies and other literature. It also should

7 International War and its Causes

be noted that the cow conception of war is compatible with the definition for international conflict used here. It demarcates a subset of international conflicts in which nation-states pass over a threshold of destruction in their efforts to exert influence. Use of the cow criteria should not be construed as uncritical acceptance of its defining conditions. For example, requiring a given level of political status from war participants excludes many violent conflicts. However, these excluded interactions are (i) included in the set of international conflicts; (2) deemed to be qualitatively different from international wars; and (3) available for comparison (if desired) with conflicts that meet the cow conditions of international war. It seems fair to say that the major conceptual problems identified so far — defining international conflict and war — have been dealt with adequately. METHODS OF I N Q U I R Y

Comparing Classes of Events

How to proceed in terms of method is a difficult but ultimately tractable problem. Beginning with the simplest approach and adding complexity only as necessary, it will be assumed that at least some wars have common antecedents. Since preferences and intentions have been attributed to the participants in world politics, resultant patterns of behaviour should be visible. The comparative testing of theories thus will be based upon a wide variety of evidence from the realm of international conflict. This strategy should help to clarify the range of application for each of the respective theories. Although this is an ambitious approach to testing, the objective is not to develop a theory which by itself can explain the occurrence of any war. "It would be nonsense," as Gantzel observes, "to search for one theory of the causes of all wars, i.e. for a theory which does not differentiate types of war and/or specific conditions of historical periods. Such an approach would imply a dubious concept of theory."15 Thus, this analysis is not expected to result in a final answer to the question of war and its causes. Instead, the use of an aggregated approach is based on the desire to expose each theory under consideration to a diversity of evidence, in the process reducing the probability of biased testing. Having adopted an aggregated or "data-based" approach, it is reasonable to identify some set of interactions with which to compare the set of international wars (referred to from this point onward as W), in order to distinguish the unique properties of the latter. (The

8 Crisis and War

set W is a subset of/, the set of international conflicts.) Logically, it would be desirable to discover the differences between members of W and its complement in /. Theories would be evaluated on the basis of their ability to explain points of variation across these two subsets. Although that approach would be thorough, it is not feasible because there are too many international conflicts which are not wars, and too many ways to compare theoretical explanations. Therefore the proposed analysis must be more focused. First, it is necessary to select a more restricted set of events with which to compare W than its complement in /. Consider the possibility of C, the set of international crises. Crisis, like war, is a term for which there exist quite a few implicit and explicit meanings. The definition proposed by Brecher and James, under the auspices of International Crisis Behavior (ICB) Project, resulted from an authoritative review of concepts in the field: An international crisis may be defined as a situational change in an international system characterized by two individually necessary and collectively sufficient conditions: (i) distortion in the type and an increase in the intensity of disruptive interactions, with a high probability of military hostilities; and (2) a challenge to the existing structure of the system.16

The crisis conditions described by Brecher and James precede virtually any war, thus facilitating comparison. Figure 1.1 is a Venn diagram that should clarify the relationships among /, C, and W. From Figure i. i, it is apparent that C and W intersect within /. These subsets define a four-fold typology of conflict: (i) C but not W; (2) W but not C; (3) members of both C and W; and (4) members of neither C nor W. Placement in Subsets i and 2 is self-explanatory, and members of Subset 3 are crises which develop into wars. Subset 4 is somewhat more complex. It includes those international interactions which generate neither wars nor crises, yet clearly are conflicts. One example of a Subset 4 case might be "structural" aggression against a developing country, to use the language made familiar by Galtung.17 This would entail exploitation of a Third World populace by a developed state, the modus operandi being collusion between the elites of the parties involved at the expense of the "periphery" of the weaker country. Couched in more neutral terms, rapid economic and political change in less-developed countries have at times resulted in Subset 4 conflicts that did not lead to a war or crisis. Only Subsets i and 3 have been used in the analysis, not because Subsets 2 and 4 are unimportant but because the issues raised here

9 International War and its Causes Figure 1.1 International Conflict

do not concern such conflicts directly. The comparison of crises that have resulted in war to those that have not is of central interest in seeking a better understanding of the causes of war. The research does not address the determinants of conflicts within Subset 4 because these events fall outside the realm of W as well as C. Similarly, Subset 2 is excluded from the analysis. Again, since the focus is on comparison of Subsets i and 3, only in an indirect sense will that concern the causes of wars which have not resulted from crises.18 In the abstract, this makes the research somewhat narrower than promised at first. Rather than looking at war per se, the objective is to further understanding of the causes of a subset of W — that generated by C. However, as will become apparent, that portion is large and diverse. This approach resolves the problem of selecting a workable subset of interactions with which to juxtapose W. There still exists the problem of how to compare Subsets i and 3, and that is a task with a number of components. Selecting Theories

Some principle of organization is needed to provide a basis for the selection of theories. One set of categories which is both economical and widely used is that proposed by Waltz in Man, the State, and War.

io Crisis and War

Scholarship on international war has clustered around the three focal points educed by Waltz: (i) the nature of man himself; (2) the internal characteristics of the state; and (3) the relationships among nations in the international system. To describe the influence of this taxonomy as paradigmatic would not be much of an exaggeration. For example, in the foreign policy model proposed by Hughes, which appears as Figure 1.2, the categories of independent variables correspond closely to those articulated by Waltz. *9 Many other models display much the same structure, either explicitly or implicitly.20 Even if Waltz's framework is viewed only as a culmination of preexisting tendencies in the literature, it still can serve as a mutually exclusive and exhaustive system of categorization for theories of war. As the two-directional arrows from Figure 1.2 imply, the occurrence of war (or foreign policy behaviour generally) cannot be explained entirely by factors operating at any one level. Theories are placed in a given category from Waltz to reflect their emphasis. All theories must deal in some way with individuals, states (including their internal structures), and interstate relations. Singer makes that point elegantly in his discussion of levels of analysis. He argues that the problem is really not one of deciding which level is most valuable to the discipline as a whole and then demanding that it be adhered to from now unto eternity. Rather, it is one of realizing that there is this preliminary conceptual issue and that it must be temporarily resolved prior to any given research undertaking.21

Singer regards the processes within states and the external relations among them to be the primary options, with one or the other providing a point of departure for virtually all studies of international politics. At least one critic has challenged Singer's line of reasoning and, by implication, that of Waltz, too. Charging that conventional approaches have oversimplified the international system, Isaak advocates the "phenomenological" method as a solution to the levels of analysis problem. In specific terms, he claims that Singer's "deterministic constructs" inhibit contextual understanding of world politics.22 But at this point logic must intervene on behalf of Singer. Even if nation-states have been characterized by Singer and others as sentient beings (i.e., Waltz's third level), must this imply that their actions somehow are predetermined? A focus on the "state" does not imply the use of historical necessity to explain behaviour, and the former may be useful when trying to generalize from results. The level of aggregation selected is (and should be) a function of

11 International War and its Causes Figure 1.2 The Hughes Foreign Policy Model

Nation-State Source: Barry B. Hughes, The Domestic Context of American Foreign Policy (San Francisco: W.H. Freeman and Company 1978), 3.

the objectives of the researcher, with an inevitable trade-off regarding descriptive detail and the generality of propositions and findings.23 Theories from each of the three levels isolated by Waltz were selected for testing in this study, to assure a wide range of perspectives on the causes of war. The three levels will be described in more detail, relying upon several theories and hypotheses to illustrate their respective properties. This should promote understanding of the origins of, and logic behind, the theories that were selected. Adopting the perspective of the first image of conflict, some have attributed the occurrence of war to the characteristics of individuals.

12

Crisis and War

As a rule, they have obtained theories from two somewhat overlapping sources: sociology and psychology. Researchers have assessed the impact of roles, expertise, and past experience on policy choices in a small-group setting, and quite a few of these studies have focused directly on the sociology of the crisis milieu.24 With a somewhat different perspective on the individual, political psychologists have investigated the effects of drives, beliefs, and personality factors on the propensity toward conflict behaviour.25 Among the personality traits which have been analyzed in this vast literature are self-confidence, desire for achievement, anxiety, dogmatism, risk-propensity, and militarism.26 The quest for greater understanding of individual decision making has ranged across continents and centuries. While this study does not include a systematic review of the literature, it can be said that the first-image branch of inquiry has contributed to the understanding of conflict. To cite one illuminating example, Stoessinger's analysis of leadership serves as a counterpoint to the often exaggerated role of the Zeitgeist in explaining war.27 He has gathered convincing evidence that the personalities of leading statesmen may precipitate conflict, as in the case of Nasser in 1967, or the pairing of Hussein and Khomeini in the 1980 war between Iraq and Iran. As the research of Stoessinger and others demonstrates, the role played by the individual makes war anything but a matter of destiny. More than anything else, the first image would benefit from an enhanced ability to explain war in general terms.28 Insights derived from case studies can be fascinating, but they also may be difficult to apply to new instances of international conflict, because the actors and circumstances keep changing. Renewed testing of first-image theory should place an emphasis on propositions with an extensive range of application. In other words, hypotheses should be linked more closely to the general (as opposed to contextual) occurrence of war, even if that means sacrificing some descriptive detail. Concentrating on the internal workings of the state as an actor, second-image scholars have tried to discover the causes of war through observation of national attributes. Many of the studies in this category have focused on intra-state conflict as a potential cause of international strife. Riots, politically motivated strikes, and other forms of internal unrest have been proposed as an explanation of conflict among states, as an elite facing domestic turmoil may "project" such conflicts outward in an attempt to restore unity by creating an external menace. Among such responses to domestic pressure, war would be the most extreme. Other state-oriented approaches have emphasized the more stable

13 International War and its Causes

characteristics of nation-states and their connection to international conflict. A prime example is the effort by Rosenau to develop a "pretheory" of foreign policy (although not of conflict or war) based o'n national attributes. In his scheme of classification for independent variables, "societal" and "governmental" categories incorporate variables such as degree of industrialization and value orientations on the one hand, and regime type and centralization on the other.29 Subsequent research on the effects of these and other relatively static attributes on foreign policy has produced varying degrees of success.30 For the second image, an explanation of war based on some variables which change quickly and others that evolve slowly would be a step forward. This would counter charges of state-centred "determinism," because foreign policy behaviour would be accounted for without absolute reliance on relatively static attributes such as degree of militarization or class structure. Stable characteristics of the state could be used to complement internal behavioural variables, which have not demonstrated an independent capacity to account for international conflict. Accordingly, the integration of behavioural and national attribute variables will be a priority in the further testing of second-image theory. Theories derived from Waltz's third image have invariably dealt with the distribution of military capabilities and its presumed impact on the international system's propensity for war. The literature on balance of power31, systemic polarity32, arms races33, and alliances34 is vast in scope. Propositions about war have been tested repeatedly, with analytical techniques ranging from the traditional case study to the limits of quantitative sophistication. In fact, the volume of systemic research on "power politics" greatly exceeds that of the two other images combined. However, third-image theory is troubled by its very pervasiveness. There is so much literature that it is difficult to assess cumulative progress, or even to keep up with terminology. These difficulties may be connected, because different measurements of the same variable can cause contradictory results from testing. The greatest opportunity for improvement rests in the clarification of basic concepts and their role in theorizing about the choice between peace and war. This completes the overview of Waltz's three levels of analysis. Practicality dictates that only one theory from each level be tested. There are'two primary reasons why more extensive testing would not be feasible. To permit thorough evaluation (or "theoretical articulation," using Morrow's terminology), each theory must be expressed as a logically interconnected set of propositions about the

14 Crisis and War

resolution of international crises.35 Extensive reviews of the literature will be required to accomplish that objective, because it will be difficult to find a definitive presentation of a theory that can be readily adapted to the analysis of crises. A less grandiose consideration is the time and effort required to compile data. More theories bring new variables with them, and data collection becomes especially arduous if there are many cases to consider. Having settled on one theory per level of analysis, the next problem is to decide which theories are worth another look. In the name of cumulative research, the menu will be restricted to theories which already have been perceived as interesting enough to warrant testing. As noted previously, to evaluate the foundation upon which further research is to build, it is essential to re-examine existing theory, and even then sampling is required. Following is an exegesis of the theories to be tested — expected utility, conflict and cohesion, and balance of power — including a brief assessment of how each is viewed today and a rationale for its inclusion in the present study. One especially innovative first-level theory is that developed in recent years by Bueno de Mesquita.36 Using a set of axioms borrowed from microeconomics, he derives a theory of war based on expected utility. It assumes that the decision to initiate a war is based on hypothetical calculations by national decision makers of the expected value or utility of such an action. When the utility calculus projects a net benefit from engaging in warfare, war is regarded as a rational policy. The components of the utility calculus are elaborate. They go far beyond an impressionistic comparison of an adversary's military capability with one's own, and include considerations ranging from intermediaries to geography. Given the complexity of the utility calculus, and of the axiomatic basis from which it is derived, only a cursory outline of Bueno de Mesquita's model is provided here. Suppose there exists a state whose central decision makers are trying to decide whether to attack another state. This theory suggests that they may attack when prospective utility from the conflict is positive, and would not do so otherwise. The overall utility level for the initiator is based on a summation of utilities associated with a set of events, each value being discounted by the probability of the event's occurrence. Arguments in the summation (such as those corresponding to third-party involvement) are assigned utility scores on the basis of their expected impact on the outcome of a war with the target state if one should occur. Bueno de Mesquita thus obtains an aggregate measurement of expected utility for the initiator and the target state of each war in his data set, which spans the period 1816-1974.

15 International War and its Causes

Bueno de Mesquita tests a set of hypotheses derived from his expected utility theory of decision making, in the process obtaining some impressive results. Among the propositions he evaluates, the most general asserts that the initiator of a war will have a positive expected utility score from the conflict, whereas that of the target state will be negative. This hypothesis received overwhelming confirmation from the cow data set, as did several others. Expected utility also enjoyed success relative to other theories. For example, expected utility outperformed a competing "balance of power" theory in predicting the characteristics of war initiators and targets. While 78 percent of the initiators enjoyed an advantage in power, a full 86 percent had non-negative expected utility relative to their targets.37 Although relatively new, the expected utility theory follows in the established tradition of using rationality postulates to explain political behaviour.38 Unlike most other first-image theories, expected utility stresses the common traits among decision makers, and that should facilitate testing under a wide range of conditions. Bueno de Mesquita's theory also has provoked intellectual warfare among scholars of international politics, with reactions ranging from enthusiastic endorsement to vigorous criticism.39 Given its innovative and controversial nature, the expected utility theory merits further attention, particularly in view of its apparent success in explaining the antecedents of war. Accordingly, critical commentary and testing are presented in Chapters 2 and 3, respectively. Conflict and cohesion (or internal-external conflict linkage, externalization, etc.) is a time-honoured theory from the second image of war which proposes that external aggression by states is linked to internal strife. At an intuitive level, this linkage seems plausible. For example, foreign policy crises tend to increase support for national leaders, at least in the short term. Therefore, leaders may be motivated to direct the public eye away from domestic problems and toward foreign affairs. Such a strategy could be carried out by exploiting external conflicts or even manufacturing them. Despite the plausibility of conflict linkage, encouraging results have not been forthcoming.40 Researchers have discovered some highly complex relationships involving certain types of internal and external conflict, but these patterns exist only under very specific conditions. There is, however, reason to believe that a new approach to operationalization and testing could produce a challenge to the conventional wisdom in this area. By whatever name — projection, externalization, or conflict and cohesion — the theory in question entered a phase of popularity which coincided with the crest of the behavioural revolution in the 19605. Since that time, testing has been

16 Crisis and War

coloured almost uniformly by the behavioural brush. Much of the research in question has been insulated from theorizing by social scientists, and this may have led to a misleading consensus on the implications of the results. In other words, conflict and cohesion appears to be a prime candidate for theoretical articulation. Chapters 4 and 5 attempt to integrate theory with the practice of testing externalization. Third among the theories to be evaluated is that commonly referred to as "balance of power." More a melange of theories than anything else, its pertinence still is undeniable. Balance-of-power theory has been around for centuries, and it continues to find more support than any other interpretation of international relations. From the viewpoint of power politics, war is the result of distributional imbalances among members of the international community. The most common generalization, as Zinnes observes, is that peace will be preserved when states are "balanced" against each other, in the sense that no single actor (or coalition) is capable of imposing its will on the others in the system.41 But competing interpretations of the balance of power have focused upon systemic polarization and distributional dynamics, among other issues. In more general terms, it seems fair to say that the status of balance-of-power theory as an explanation of war has been evolving toward uncertainty. It contains basic conceptual problems which demand attention. For example, what does power mean in contemporary political relations? There are also difficulties on the operational front. Can power be quantified on the basis of physical capabilities, such as a state's economic production or military expenditures? Even if power can be measured in a valid and reliable manner, what does it mean to say that a "balance" of power exists? These and other issues are explored in detail in the review and testing of balance-of-power theory presented in Chapters 6 and 7. On an individual basis, expected utility, conflict and cohesion, and balance of power appear to be worthwhile candidates for theoretical articulation and subsequent evaluation. But what will this testing contribute to the overall understanding of international conflict? It is clear that all of the theories follow in the philosophical tradition of realpolitik, otherwise known as power politics or the "classical paradigm."42 Expected utility theory views war as a product of elite calculations of prospective loss and gain from armed conflict, the conflict and cohesion theory posits that international war may result from "power politics" within a state, while the balance-of-power theory is concerned directly with configurations of interstate power. Thus, the degree of success achieved by these theories in testing will say

17 International War and its Causes something about the capacity of realpolitik to explain the transition from crisis to war, regardless of the level of analysis concerned. THEORIES AND EVIDENCE

Data on international crises and wars have been gathered from numerous primary and secondary locations, listed in the appendix. The principal source is the International Crisis Behavior (ICB) Project data set. The Handbook of International Crises, Volume I by Brecher, Wilkenfeld and Moser thoroughly describes those events from 1929 to 1979 which meet the Project's definition of an international crisis. For the present study, a shorter period of time was selected. Table 1.1 displays the 132 international crises from 1948 to 1975 contained in the ICB data set and analyzed in this study, along with the actors involved in each crisis and its duration.43 The 17 wars in the data set appear in italics.44 For each international crisis, Table 1.1 indicates the initiator (i), the defender (/) and other actors (K). These designations require some explanation. An international crises encompasses actor-level (or foreign policy) crises, as defined by Brecher: A crisis [actor-level] is a breakpoint along the peace-war continuum of a state's relations with any other international actor(s). A crisis is a situation with four necessary and sufficient conditions, as these are perceived by the highest level decision-makers of the actor concerned: i. a change in its external or internal environment, which generates 2. a threat to basic values, with a simultaneous or subsequent 3. high probability of involvement in military hostilities, and the awareness of 4. finite time for their response to the external value threat.4^ This definition determines which states are included in the list of actors. The defender (j) in an international crisis is the first state to perceive an actor-level crisis, while the initiator (i) is the state responsible for/s perception of a crisis. The states listed as potential war participants (the ks) are either (i) states which have perceived an actor-level crisis or (2) involved actors, which have perceived one or two of the conditions of a foreign policy crisis. These criteria for inclusion and the associated notation will be explained further in Chapters 2 and 3. Actor-level crises determine the points in initiation and termination for international crises. For example, Kuwait Independence began with Kuwait's perception of a crisis on 25 June 1961. That crisis resulted from the claim by Iraq's prime minister that Kuwait

Table 1.1 International Crises: 1948-1975 Actors Case 1 Communism in Czechoslovakia 2 Soviet Note to Finland I 3 Israel Independence 4 Berlin Blockade 5 Costa Rica/Nicaragua I 6 Indonesian Independence III 7 Pushtunistan I 8 Luperon 9 Soviet Bloc/Yugoslavia 10 Korean War I 1 1 Tel Mutillah 12 Punjab War Scare 13 Suez Canal 14 Catalina Affair 15 Burma Infiltration 16 Invasion of Laos I 17 East Berlin Uprising 18 Trieste II 19 Qibya

i

j

k\i

k\2

CZE USR ISR USA NIC NTH PAK GUA USR RON ISR PAK UNK USR CHI VTN GME ITA ISR

USR FIN EGY USR UNK FRN COS INS AFG DOM YUG KOS CHN SYR IND EGY SWD BUR USA FRN USR YUG JOR

&i3

Duration ^21

IRQ

k%2

LEB

&23

JOR

USA

CHN LAO EGY

SYR

LEB

&24

SYR

Termination

Initiation 02/13/48 02/22/48 05/15/48 06/07/48 12/11/48 12/19/48 03/—/49 06/19/49 08/19/49 06/25/50 03/15/51 07/07/51 07/30/51 06/16/52 02/08/53 03/24/53 06/17/53 10/08/53 10/14/53

-

02/25/48 04/06/48 07/20/49 05/12/49 02/21/49 12/27/49 10/05/50 06/21/49

_ _ -

09/30/50 05/14/51 08/—/51 01/30/52 Q7/—752 10/15/54 07/11/53 12/05/53

20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

Guatemala Taiwan Straits I Costa Rica/Nicaragua II Baghdad Pact Gaza Raid Pushtunistan II Goa I Qalkilya Poland Liberalization Hungarian Uprising Suez-Sinai Campaign Nicaragua/Honduras Jordan Regime Tunisia/France I Syria/Turkey Border Ifni West Irian I Berlin Deadline Formation of UAR Tunisia/France II Sudan/Egypt Border Indonesia — Aborted Coup Lebanon/Iraq Upheaval Taiwan Straits II Cambodia/Thailand Mexico/Guatemala Fishing Rts

USA CHI

GUA HON CHN USA

YEN COS UNK EGY EGY ISR

PAR IND

ISR POL USR ISR HON EGY FRN SYR MOR

NIC IRQ

AFG POR

JOR USR HUN EGY UNK FRN NIC JOR SYR TUN TUR USR SPN

NTH GMW USR USA INS

EGY FRN EGY USA EGY CHN

SYR

IRQ TUN SUD INS LEB CHI

CAM TAI

MEX GUA

SYR

SYR

UNK

IRQ

IRQ USA

USR USA USA

FRN

JOR

USA USA

UNK JOR

12/12/53 08/—754 01/08/55 02/24/55 02/28/55 03/27/55 08/10/55 09/13/56 10/—/56 10/23/56 10/29/56 02/26/57 04/04/57 05/31/57 08/18/57 1 1/23/57 12/01/57 12/15/57 02/01/58 02/08/58 02/09/58 02/21/58 05/08/58 07/17/58 07/24/58

_ _ _ _ -

06/29/54 04/23/55 01/20/55 10/W55 06/23/55 Ill— 755 09/06/55 10/—/56 10/22/56 11/14/56 03/12/57 05/09/57 05/03/57 06/27/57 10/29/57 _

_

09/15/59 02/14/58 06/17/58 02/25/58 05/20/58 10/W58 10/23/58 02/06/59

12/29/58 - 02/01/59

Table 1.1 '(continued) Duration

Actors Case

46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Cuba/Central America I India/China Border I Shatt-al-Arab I Rottem Ghana/Togo Border Assassination Attempt/ Venezuela President Congo I: Katanga Mali Federation Cuba/Central America II Ethiopia/Somalia Pathet Lao Offensive I Bay of Pigs Pushtunistan III Kuwait Independence Bizerta Berlin Wall Vietcong Attack West Irian II Breakup of UAR Soviet Note to Finland II

i

j

CUB IND IRN ISR TOG

PAN CHN IRQ EGY GHA

DOM COP SEN CUB SOM VTN USA AFG IRQ TUN GMW VTN NTH SYR USR

VEN BEL MLI NIC ETH USA CUB PAK KUW FRN GME USA VTS INS EGY FIN

kn

kiz

&13

^21

^22

NIC

DOM HAI

SYR

USR

GUA

USA

TAI USR

LAO

UNK UNK

FRN

USR USA

^23

^24

Termination

Initiation 04/25/59 08/25/59 11/28/59 02/15/60 03/—760

-

127— 759 04/19/60 01/04/60 03/08/60 04/01/60

06/24/60 07/05/60 08/20/60 11/09/60 12/26/60 03/09/61 04/15/61 05/19/61 06/25/61 07/17/61 07/29/61 09/18/61 09/26/61 09/28/61 10/30/61

_ _

09/—760 02/15/62 09/22/60 12/07/60 _h

-

05/16/61 04/24/61 01/29/62 07/13/61 09/29/61 10/17/61 11/15/61 08/15/62 10/05/61 11/24/61

66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91

Goa II Mauritania/Mali Taiwan Straits III Pathet Lao Offensive II India/China Border II Yemen War I Cuban Missiles Malaysia Federation Jordan's Internal Challenge Dominican Republic/Haiti Algeria/Morocco Border Venezuela/Cuba Kenya/Somalia Cyprus I Jordan Waters Dahomey/Niger Rwanda/Burundi Panama Canal East Africa Rebellions Ogaden I Gulf of Tonkin Congo II Rann of Kutch Dominican Republic Guinea Regime Rhodesia UDI

IND MLI CHI VTN CHN YEM USR INS EGY HAI MOR CUB SOM GYP ISR DAH BUI USA TAZ SOM VTN CON IND DOM IVO RHO

POR MAU CHN TAI IND JOR USA MAL ISR DOM ALG YEN KEN TUR EGY NIE RWA PAN UNK ETH VTS COP PAK USA GUI ZAM

USA

USA SAU

EGY CUB

JOR USA UNK

GRC

LEB

JOR

USA USA

BEL

UGA KEN MLI

USR

BUI

UNK

SYR

12/11/61 03/29/62 04/22/62 05/06/62 09/08/62 09/26/62 10/16/62 02/1 1/63 04/21/63 04/26/63 10/01/63 11/01/63 11/13/63 1 1/30/63 12/11/63 12/21/63 12/21/63 01/09/64 01/19/64 02/07/64 08/02/64 08/04/64 04/08/65 04/24/65 10/09/65 1 1/05/65

_ _ -

12/19/61 02/18/63 06/27/62 06/12/62 01/23/63 04/15/63 11/20/62 08/09/65 05/04/63 06/03/63 11/04/63 12/01/63 03/04/64 08/10/64 05/05/64 01/04/64 Q4/—/64 01/12/64 01/30/64 03/30/64 08/-V64 12/30/64 01/10/66 08/31/65 12/—/65 04/27/66

Table 1.1 (continued) Duration

Actors

i

Case 92 Yemen War IV

93 El Samu 94 Che Guevara 95 Six Day War

96 97 98 99 100 101 102 103

Cyprus II Pueblo Karameh Prague Spring Essequibo Territory Pre-War of Attrition Beirut Airport Ussuri River

104 War of Attrition I

105 EC-121 Spy Plane 106 Shatt-al-Arab II 107 Football War

108 109 110 111

Cairo Agreement Black September Cienfuegos Base Portuguese Invasion of Guinea

j

YEM JOR CUB EGY GYP RON JOR CZE YEN EGY ISR CHN EGY RON IRQ

^11

SAU EGY ISR BOL ISR SYR TUR GRC USA ISR GME GUY ISR LEB USR ISR USA IRN HON ELS SYR LEB JOR SYR ISR USR USA CUB POR

GUI

^12

USR

^13

JOR

^21

USA

USR

USA

^22

POL

^23

^24

Termination

Initiation

10/14/66 11/12/66 03/23/67 05/17/67 11/15/67 01/22/68 03/18/68 HUN BUL 04/09/68 07/09/68 09/07/68 12/28/68 03/02/69 03/08/69 04/15/69 04/15/69 06/15/69 10/22/69 09/15/70 09/16/70

_ _ -

09/26/67 11/15/66 10/10/67 06/11/67 12/04/67 12/23/68 03/22/68 10/18/68 08/—/68 11/07/68 oi/—/69 10/20/69 07/28/69 04/26/69 10/30/69 07/30/69 11/03/69 09/29/70 10/23/70

1 1/22/70 - 12/11/70

112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132

Bangladesh Chad/Libya I Caprivi Strip Uganda/Tanzania I Uganda/Tanzania II North/South Yemen I Zambia Libyan Plane Iraq Invasion - Kuwait Israel Mobilization Cod War I October - Yom Kippur War South Yemen/Oman Cyprus III Mayaguez Angola Moroccan March — Sahara Belize I Sahara Cod War II East Timor

PAR CHA SAF UGA TAZ YES RHO IBY IRQ EGY ICE EGY YES TUR CAM ANG

BAN

IND

IBY ZAM TAZ UGA YEN ZAM ISR

KUW ISR IRQ UNK

ISR SYR OMA GYP USA

ZAM MOR SPN

USR

MOR ALG

MAU

GUA UNK ICE POR

UNK INS

IBY USR

SAU USA GRC

CUB

ZAI ALG

SAF

USA

03/25/71 05/24/71 10/05/71 10/20/71 09/17/72 09/26/72 01/19/73 02/21/73 03/20/73 04/10/73 05/14/73 10/05/73 11/18/73 07/15/74 05/12/75 07/12/75 10/16/75 11/01/75 11/14/75 11/23/75 1 1/28/75

_ _ -

12/17/71 04/17/72 10/12/71 11/25/71 10/05/72 11/28/72 02/03/73 02/21/73 06/08/73 06/—773 11/13/73 05/31/74 03/11/76 02/24/75 05/15/75 03/27/76 11/14/75 11/30/75 Q4/—776 06/01/76 07/17/76

Table 1.1 (concluded) Notes: A few points of clarification are in order regarding the contents of this table. (1) The 132 cases in the data set include all international crises initiated in the years 1948—1975. (2) For crises in which a precise point of termination cannot be specified, a dash appears in place of a date. (3) Crises that resulted in war are distinguished by italics. (4) The acronyms from the table are listed below, along with the crisis participants they represent. AFG ALG ANG BAN BEL BOL BUR BUI CAM CHA CHI CHN CON COP COS CUB GYP CZE DAH DOM EGY ELS ETH FIN

Afghanistan Algeria Angola Bangladesh Belgium Bolivia Burma Burundi Cambodia (Kampuchea) Chad China-Taiwan China-People's Rep. Congo-Brazzaville Congo-Kinshasa (Zaire) Costa Rica Cuba Cyprus Czechoslovakia Dahomey Dominican Republic Egypt El Salvador Ethiopia Finland

GUI GUY HAI RON HUN ICE IND INS IRN IRQ ISR ITA IVO JOR KEN KON KOS KUW LAO LEB IBY MAL MLI MAU

Guinea Guyana Haiti Honduras Hungary Iceland India Indonesia Iran Iraq Israel Italy Ivory Coast Jordan Kenya Korea-North Korea-South Kuwait Laos Lebanon Libya Malaysia Mali Mauritania

PAN POL POR RHO RWA SAU SEN SOM SAP USR SPN SUD SWD SYR TAZ TAI TOG TUN TUR UGA USA YEN VTN VTS

Panama Poland Portugal Rhodesia (Zimbabwe) Rwanda Saudi Arabia Senegal Somalia South Africa Soviet Union Spain Sudan Sweden Syria Tanzania Thailand Togo Tunisia Turkey Uganda United States Venezuela Vietnam-North Vietnam-South

FRN GME GMW GHA UNK GRC GUA

France East Germany West Germany Ghana Great Britain (UK) Greece Guatemala

MEX MOR NTH

NIC NIE OMA PAR

Mexico Morocco Netherlands Nicaragua Niger Oman Pakistan

YEM YEN YES

YUG ZAI

ZAM

Yemen Yemen-North Yemen-South Yugoslavia Zaire Zambia

26 Crisis and War

was an "integral part" of Iraq. On 30 June, the United Kingdom also perceived a crisis, when Kuwait requested assistance under the provisions of their defense treaty. This second - and final — actorlevel crisis ended on July 13, as did Kuwait's crisis, when an Iraqi representative to the United Kingdom denied Iraqi intentions to attack Kuwait or otherwise threaten its security.46 System-level disruption thus began with the initiation of the first actor-level case, and ended with the termination of the last actor-level case.47 Brecher's definition of an actor-level crisis merits some commentary. It is consistent with others in the literature, therefore enhancing prospects for validity and reliability in the designation of actors in an international crisis.48 Although many definitions of an actor-level crisis resemble that of Brecher, there is one which merits special attention because of its points of difference. Hermann's treatment of crisis has been influential among students of world politics, and in fact served as a point of departure for Brecher's concept formation.49 It may prove instructive to compare their defining conditions for crisis. According to Hermann, when decision makers experience an external threat to high priority goals, under conditions of time pressure and surprise, an international crisis is said to be at hand. Among others, Holsti and Paige use these defining conditions explicitly in their classic case studies.50 For several reasons, however, the ICB Project definition may be viewed as an improvement over that of Hermann. By incorporating internal threats, Brecher allows for the fact that an international crisis may be triggered by developments which are not in the realm of military security or diplomatic relations. For example, the instability of a regime might elicit outside intervention in its affairs. "Short" time has been replaced with "finite" time, which takes account of pressure from the scope of the objectives desired as well as clock time. Furthermore, Brecher drops the element of "surprise" from the list of defining attributes, because it has not been especially useful in the explanation of crisis decision making.51 By comparison, the defining attributes delineated by Brecher appear to offer a more valid and reliable assessment of crisis conditions.52 Definitions from the ICB research program thus underlie the structure of the data set which appears as Table i. i. As noted, the time frame for the analysis is 1948—1975, with 1948 marking the advent of some essential sources of data from the United Nations and elsewhere. This period also overlaps with the years most commonly used in previous testing of the three theories, which facilitates the comparison of results.53

27 International War and its Causes

Some of the major conceptual and operational problems have been identified for more detailed study. Expected utility, conflict and cohesion, and balance of power are evaluated both separately and comparatively against the evidence. (The techniques which enable these judgments to be made require a separate discussion, presented in Chapter 3.) Chapter 2 surveys expected utility as a theory of war, while Chapter 3 tests a series of hypotheses derived from it. Chapter 4 reviews conflict and cohesion, while Chapter 5 contains the corresponding analysis. Chapter 6 examines balance-of-power theory, which is tested in Chapter 7. The comparison of theories and testing of a combined model are described in Chapter 8, along with an analysis of the most poorly predicted cases. This final chapter also integrates the results and suggests some directions for further inquiry.

CHAPTER TWO

War and Expected Utility Reconsidered

War and expected utility make strange bedfellows. "War" is a term replete with the imagery of death and destruction. By contrast, "expected utility" is a phrase drawn from microeconomic theory which suggests optimizing agents (whether firms or otherwise) interacting in a milieu devoid of violent or irrational behaviour. On the surface, war as a result of expected utility maximization would appear to be a contradiction in terms. Given its counter-intuitive nature, the predictive success enjoyed by Bueno de Mesquita's expected utility theory becomes even more interesting to those concerned with explaining war.1 At the most general level, his model may be summarized as follows: The selection of war or peace is a choice that is initiated, conducted, and concluded by individual leaders who must accept responsibility for their decisions.... The choice of war or peace depends on the choices of individuals and not on compulsion by circumstance.2

Bueno de Mesquita has adopted the time-honoured position that social reality is a product of contingent behaviour, rather than being predetermined in some Hegelian manner. He does not see military conflict as the end result of a process which lies beyond the influence of decision makers. As opposed to secular changes in socioeconomic structure (or other long-term developments), the immediate calculus of decision is considered to be of utmost importance. The emphasis placed by Bueno de Mesquita on the judgments of individual leaders follows in the tradition of the first image from Waltz. Although circumstances constrain action, statesmen decide the fate of nations, and the nature of decision making by the individual leader remains the ultimate "cause" of armed violence.3

29 War and Expected Utility Reconsidered

In order to understand expected utility theory in the context of war, the means and ends of foreign policy must be distinguished from each other. Utility theory in this context is concerned with the relative costs and benefits associated with resorting to arms. Thus, war can be a rational choice if it is perceived by a national leader to be the optimal means toward some end. The moral value of the objectives desired - conquest of territory, control over foreign populations, and so forth - is not at issue. An expected utility model seeks to explain, not to judge, decision making at the brink of war, and it uses the rationality postulates of economic theory in so doing. Since the focus is on the means as opposed to the ends of foreign policy, war can be reconciled with the notion of rational choice. One further qualification should be made prior to presentation of the expected utility theory. The theory about to be described is one that has been substantially revised since its appearance a few years ago. However, it is appropriate to outline the initial version, because its relative simplicity will facilitate understanding of the various critiques and refinements. THE EXPECTED UTILITY CALCULUS

What, then, are the circumstances under which a leader may rationally choose to go to war? According to Bueno de Mesquita, the resort to arms will depend upon the expected value or utility of doing so as perceived by the central decision-maker of a given state. Suppose that the leader of a state i is debating whether to attack another state j, and that both states are nonaligned. The calculus of decision for i in this context is conveyed by Equation 2.i 4 :

Leaving aside questions about measurement, some further explanation of Equation 2.1 is in order. Table 2.1 replicates Bueno de Mesquita's mathematical presentation of the equation's components. Although that description of the calculus is entirely sound, a less formal treatment also would be useful. A non-technical summary of Equations 2.1 and 2.2 should facilitate understanding of the critique that follows. Since the war between i and j is assumed to involve only these two

30 Crisis and War Table 2.1 Components of Equation 2.1 E(Ui)b

ua

Uij

(Uu - Uij)to

(Uij - UiOto

A(U B - Uy)^

A(Ug - UM)t(rtn

Pi

1 - Pi

: i's expected utility from a bilateral war withy. : i's utility for i's preferred view of the world. U;i = 1 by definition. : i's utility for fs policies. Uy can vary between 1 and — 1, representing complete agreement and disagreement respectively. : i's perception of what might be gained by succeeding in a bilateral conflict withy in which i can then impose new policies onj. This term reflects i's current evaluation of the difference between the policies that i currently desires j to hold and i's perception of f s current policies (hence it is evaluated at time to). Thus, the greater the perceived similarity between the policies i desires for^ and/s current policies (i.e. Uy —» UH), the less the utility i expects to derive from altering fs policies. : i's perception of what might be lost by failing in a bilateral contest withy in which j can then impose new policies on i. This term reflects i's current evaluation of how much j could shift i's policies away from its world view to make them more in line with/s interests as perceived by i. Like the previous term, this term is evaluated based on current policies (at t0). : i's perception of anticipated change in the difference between fs world view and/s policies over the time period t0 (the present) to some future time, tn. : fs perception of anticipated change in how muchj would want to alter fs policy outlook in the future compared to fs current perceived policy differences with i. Thus this term represents fs perception of anticipated future potential policy losses (over the period to to the future period tn) to j, while the previous term represents fs perception of future potential policy gains to be derived from imposing i's will onj. Both this and the previous term represent fs estimates under the assumption of no war. : fs current perception of his probability of succeeding against j in a bilateral conflict. And finally, : i's current perception of his probability of losing against j in a bilateral conflict.

Source: Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press 1981), 47-8.

31 War and Expected Utility Reconsidered

states, there are no terms in the calculus which take into account the role of intermediaries. Thus, in the first line of the calculus, winning (Pi) and losing (i — PI) are mutually exclusive and exhaustive probabilities. These terms serve as weights for the policy differentials, referring respectively to (U,i — Uy-) and (U y — Uy) for victory and defeat. The utility i has for its own policies is assumed to be at the theoretical maximum, which is set arbitrarily to one unit. Thus, U^ = i in the equation. Perfect agreement withj would mean Uy = i, with complete disagreement represented by Uy = — i. Therefore the maximal gain or loss anticipated from warfare is two units. The weighted differentials — which are identical in magnitude and range from o to 2 in absolute value - are aggregated to produce an overall score for expected utility. However, there is more to the calculus than present (i.e., t0) considerations alone. The decision to initiate a war depends also upon how i sees the current situation in the context of the future. These concerns are represented by the latter two lines of Equation 2.1. From an attack at t0 (with the war ending at t x ), i can expect with a given probability (P,) to prevent unfavourable developments in/s policies over the next time interval (t1 to tn).5 These unfavourable changes appear as A(U y — Uii). Of course, i must also consider the scenario of an unsuccessful attack (the probability of this being (i — Pj) in the equation), consequently enabling enforcement by j of changes in i's policies.6 According to Bueno de Mesquita, multilateral war creates a more complex setting comprised of seven types of actors, five of which are intermediaries (see Table 2.2). It also poses the question of risk propensity with respect to the decision-making environment. Of course, these new considerations render Equation 2.1 incomplete. Potential intermediaries result in the following utility calculus for i in an environment of risk, a term to be explained momentarily:

The components of Equation 2.2 are explained in formal terms by Table 2.3. Actor i is assumed to be able to evaluate its expected utility relative to^ and each of the five types of intermediary defined in Table 2.2. These calculations result in an overall level of expected value, represented by E(Ui) in the equation. Under conditions of risk, the decision maker performs the relevant calculations within a

32 Crisis and War Table 2.2 A Typology of States 1 The potential initiator, and i hereafter, 2 The potential defender, and^ hereafter, 3 Those nations that are friendly toward the initiator but not toward the defender (called k\ hereafter), 4 Those nations that are friendly toward the defender but not toward the initiator (called k2 hereafter), 5 Those nations friendly toward both the attacker and the defender (called k$, k's or k'z hereafter, where £3 prefers i toj, k'$ prefers j to i, and k's is indifferent to i andj. All other notation using a single or double prime follows the same preference pattern), 6 Those allied nations not friendly toward either the attacker or the defender (called A4, k'4, and kl hereafter), and 7 All nonaligned nations (called k$, k'^, and #5 hereafter), whose policies reflect no expressed commitment to any nation in the international system. Source: Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press 1981), 49.

stable coalition structure, and two kinds of decision rules emerge once the value of E(Ui) is known. One is that war becomes a viable option if and only if there is a non-negative value for E(Ui) from Equation 2.2. Bueno de Mesquita would decribe this as a "riskacceptant" utility calculus. The "risk-averse" decision maker, on the other hand, would demand further assurance before considering war as a potential choice. Risk aversity would require also that E(Ui)b ^ o. In other words, the "cautious" decision maker would want to expect victory in a bilateral contest as well, rather than relying on intermediaries to provide the margin of superiority.7 Within the context of an unstable coalition structure (or uncertainty, the obverse of risk), the risk-averse leader assumes that all of the nations appearing to favour his nation i actually are indifferent.8 On that basis, these actors are excluded from the computation of expected value. A "risk-acceptant" decision maker, by contrast, will incorporate intermediaries in the calculus without that bias. Under uncertainty, the risk-acceptant leader excludes states which are friendly or unfriendly to both i and j, along with the states which are nonaligned (i.e., categories 5, 6 and 7 from the list in Table 2.2), because there is no reliable basis for differentiating their preferences in a potential wartime situation.9 One further point concerning the calculus requires elaboration

33 War and Expected Utility Reconsidered Table 2.3 Derivation of Equation 2.2 The first component of Equation 2.2 represents the expected utility for i from a bilateral war with^'. Derivation of the second component follows. If each third party k is treated as supporting i, then i's expected utility from this outcome is:

and if each third party is treated as supporting j then i's expected utility from the outcome is:

where: U;ki = i's perception of the utility to be derived from each third party actor; Uikj = i's perception of the utility to be derived by j from each third party actor; P;k = i's perception of its probability of success against j, given that third party k aids i; 1 — P;k = i's perception of its probability of failing against j, given that third party k aids i; Pjk = i's perception of its probability of failing against j, given that third party k aidsj; 1 — Pjk = i's perception of its probability of succeeding against 7, given that third party k aids j; AUikito_^tn = i's perception of anticipated future changes in the utility i can expect to derive from k (with the change in utility being evaluated as U;ki at tn minus Uiki at to); AUikjto_^tn = i's perception of anticipated future changes in the utility j can expect to derive from k (with the change in utility being evaluated as U;kj at tn minus Uikj at to); to = the time at which i is calculating expected utility. Thus any terms with a t0 subscript are computed based on current values; to—*tn = the time span over which i estimates expected changes in the utility values. This span is assumed to be fairly short, reflecting the fact that policy makers rarely have a long time horizon. tn then is the near future. All estimates of change in utility are evaluated as the relevant utility term at tn minus the relevant utility term at to;

34 Crisis and War Table 2.3 (continued) = z's overall expected utility from a multilateral conflict with j, given that i treats all third parties k( (where / includes third parties of types 1 through 5, as denned above) as potentially aiding i; and = z's overall expected utility from a multilateral conflict withj, given that i treats all third parties k( as potentially aiding j. Summing across all k's, i's estimate of the net expected value from the support decisions of third parties is:

After substituting, multiplying through, and rearranging terms:

This completes the derivation of the second component of Equation 2.2. Note that the subscript "i" is used by Bueno de Mesquita. In subsequent equations, "r" will be used because it is easier to read. Source: Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press 1981), passim.

before moving on to the propositions about war and subsequent critical commentary. Discussion so far has focused on immediate payoffs. The other two components of Equation 2.1, and their analogues in the intermediary terms of Equation 2.2, account for potential changes in the value each state places on the other's policies. (The time frame is from t0 to tn selected arbitrarily.) Bueno de Mesquita has argued that the latter two components of Equation 2. i ought to be considered equal to zero, as a general rule. The justification he has provided is based on the assumption that states cannot predict changes in utility levels, and therefore terms like A(Uii — Uy) in Equation 2.2 will approximate a zero value. Exceptions occur only when arguments such as U^, Uiiy, and Uy take on extreme

35 War and Expected Utility Reconsidered

values. To illustrate these unusual circumstances, suppose that UH = U,j or UH = — Uy, approximately. Then, according to Bueno de Mesquita, any random "drift" in the first case will move the interests of i andj apart, while in the second it will move them closer together. Random fluctuations at the extreme values (i.e., i and — i) are expected to be toward zero, because there is so little room for change in the other direction.10 PROPOSITIONS

From properties of the expected utility calculus, Bueno de Mesquita has derived and tested a series of hypotheses about the conditions necessary for war. At this point, some of these propositions will be reviewed, with emphasis on those that can be adapted to the present research on crisis and war.11 The results obtained by Bueno de Mesquita from his tests will be appraised at a later point. Proposition i describes a condition identified by Bueno de Mesquita as essential for the outbreak of war: Proposition i: War initiators will have positive expected utility with respect to the nation they attack.

To initiate a war on a rational basis, i must expect to benefit from doing so. This condition is satisfied when E(Uj) ^ o. A confirming example from Bueno de Mesquita's list of interstate wars would be the Soviet Union's attack on Finland in 1939, the expected value for the initiator being 0.842.12 Proposition 2 deals with a subset of cases: Proposition 2: A non-aligned state cannot rationally attack a stronger nonaligned state.

This applies specifically to cases in which neither the initiator nor the defender can expect to receive assistance from intermediaries. Under such conditions the direct comparison of national capabilities can be used to calculate the probability of victory for the initiator. Earlier it was suggested that A(Uji - Uy) might have a non-zero value under particular circumstances. Proposition 3 deals with the possible effects this term may have when i and j possess virtually identical interests at t0: Proposition 3: If i and j follow essentially identical policies on all issues of

36 Crisis and War relevance to each other or to others (i.e., Uy = Uy), then E(U;) must be greater than or equal to zero if winning seems more likely than losing (i.e., if P > i — Pj). A war thus may be initiated on a rational basis.

The line of reasoning is that potential intermediaries would not interfere with the outcome of a war between allies with seemingly identical interests. If the initiator is stronger than the defender and has reason to expect deterioration in relations withj, then the value of a war withj must be positive. Therefore, an attack may rationally occur. Proposition 4 takes account of the fact that war is an expensive undertaking: Proposition 4: If z's positive expected utility from a war with j is less than /s expected loss from a war with i (so that |E(Ui)| — |E(Uj)| < o), both i and j are better off negotiating rather than fighting. Consequently, war cannot occur on a rational basis.

If one party expects to gain less from a fight than the other will lose, some sort of deal is preferable for both, assuming they are rational actors. This completes the description of a sample of the propositions about necessary conditions for war derived by Bueno de Mesquita from the initial version of his theory.13 THEORETICAL ARTICULATION

There are certain conceptual problems with the preceding formulation of expected utility theory which require some consideration. First, there is the issue of whether it is better to search for necessary, as opposed to sufficient, conditions for war. A second area of difficulty is the unitary actor assumption and its role in the description and explanation of international conflict. The third problem is linked to the problematic nature of nationality in the context of the warpeace dilemma. Fourth and finally, Bueno de Mesquita's particular conception of utility theory merits some further attention. The discussion to follow is intended primarily to explain the significance of each problem. Potential solutions will be appraised and implemented at the stage of operationalization. Necessary and Sufficient Conditions for War

With respect to the issue of causation, the goal in comparing classes of events should be to distinguish their consequences in an econom-

37 War and Expected Utility Reconsidered Table 2.4 War and Rational Choice A. Expected utility difference and war Outcome

Expected utility

+ _

War

Other

65

108

11

67

2

x = 14.03 , p < .001 B. Expected utility difference and rational choice Choice behaviour

Expected utility

+ _

Rational

Irrational

65

108

11 67 x = 37.43 , p < .001 2

Source: Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press 1981), 168.

ical fashion. Bueno de Mesquita takes the position that students of international politics should search for the conditions necessary for war to occur, because it may be impossible to specify sufficient conditions.14 But this point is not self-evident. Leaving aside for the moment whether they could be enumerated, at least one finite set of conditions sufficient for war must exist, because wars have occurred. By contrast, it is easy to see that the set of conditions necessary for war may be more difficult to specify. Consider the many trivial members it will have. There must be at least two participants, each with some population, boundaries, and the like. Deciding which attributes are essential with regard to explanation and prediction of war may be a rather formidable task. Imagine, for example, that war and positive expected utility for the initiator show a virtually constant conjunction. Does this help to explain why many other disputes which also satisfy the inequality E(Ui) ^ o have not degenerated into warfare? Consider Table 2.4. Part A is the table used by Bueno de Mesquita to test Proposition 4. The "dependent" variable in the new table (Part B) is choice behaviour, with rational choice defined as z's pursuit of war as a strategy when |E(Ui)| - |E(Uj)| ^ o and de-escalation when |E(Ui)|

38 Crisis and War

— |E(Uj)| < o. This is a defensible conception of choice because if decision-makers are utility-maximizing agents, they ought to behave that way regardless of context.15 It is the calculus - not some exogeneous factor - that is supposed to determine choice. Part B of the table does not support that assertion. It would appear instead that information (i.e., results from the utility calculus) which argues in favour of peace more frequently induces "rational" choice, while that which suggests the value of war leads to "irrational" choice. How can it be acceptable to the theory when so many decision makers disregard the calculus? The answer is, of course, that something outside the balance equation explains these choices. Majeski and Sylvan have suggested that many potential wars do not occur because the benefits from them would not be worth the trouble.16 Wars involve opportunity costs that are not factored into Bueno de Mesquita's equations. At any given time there may exist a number of states against which an initiator would have positive expected value, yet war remains relatively uncommon. The isolation of necessary conditions will not tell much about when a conflict will escalate and which parties will be involved. Consider as examples the wars between France and Mexico in 1862, Russia and Turkey in 1877, and Spain and Morocco in 1909. In each case the initiator had a very high level of expected value.17 Since these calculated values do not change much over the course of a few years, the theory cannot explain why the wars occurred at the times they did. In other words, the necessary conditions for war — in terms of expected utility — can persist for years or even decades without producing a war. Thus, in practice, those looking for necessary conditions may do little more than add to the set of defining attributes of the events which they hope to explain. A quest for sufficient conditions is easier to defend on practical grounds. The State as an Actor

Second among the areas of conceptual difficulty is the unitary actor assumption. This should not be confused with the notion of "statecentrism." The "unitary actor" assumption implies that the state acts as an integrated whole, whereas "state-centrism" suggests that international relations may be analyzed in terms of interactions among states. To be sure, Bueno de Mesquita is correct in pointing out that decisions concerning war and peace tend to induce hierarchical decision-making processes, even in democracies.18 But even a cursory glance at Allison's study of the Cuban Missile Crisis is enough to suggest that the unitary actor assumption should be used with cau-

39 War and Expected Utility Reconsidered

tion.19 The decision-making process indeed may become centralized, but the potential for miscalculation and misunderstanding is high in the crisis environment. Accidents thus may affect the outcome. The existence of this random aspect of decision making, which lies outside the boundaries of the expected utility calculus, raises two related problems. These twin complications will be discussed in the context of the standard criteria for evaluating theories: a theory can be judged as (i) a descriptive mechanism; (2) a means of explanation; or (3) a predictive device.20 One obvious problem with the unitary actor assumption is that its description of the decision making process is not entirely accurate. According to the expected utility theory, behaviour is the result of decisions made by unitary actors who decide as if they calculated expected value. While no leader actually will perform the elaborate calculations which comprise the essence of the theory, decisions at the brink of war may be consistent with their results. In other words, the highly complicated expected utility calculus is regarded as an approximation to impressionistic cost/benefit assessments made by national leaders. Explanation also presents a problem for the expected utility model, given the presence of the rational actor assumption. In that context, consider a theory with one or more invalid axioms. It is a fundamental point of logic that it is impossible to derive any statement (including those which contradict each other) from such a set of assumptions. The explanatory value of the expected utility theory thus becomes doubtful on logical grounds. Is war a product of rational expectations concerning its outcome or something altogether different? This question is beyond the scope of the present research and would require in-depth examination of individual cases. But it should be kept in mind that explanation is not currently a strong point of the expected utility theory. Prediction of behaviour is the area in which expected utility has enjoyed its greatest success. Using cow data spanning over 150 years, Bueno de Mesquita tested propositions i to 4 and produced uniform confirmation for them. He found, for instance, that 65 of 76 war initiators had positive expected utility, with only 14 percent of the defenders holding such an advantage. Supporting Proposition i, this asymmetry yielded an impressive score of 0.94 for Yule's Q, a standard measure of association between nominal scale variables. There is a similar degree of confirmation available in the case of Proposition 4, to cite one other example. Of 78 two-country conflicts in which the defender's expected losses outweighed the initiator's expected gains, 67 did not escalate to war. By contrast, as noted by

40 Crisis and War

Bueno de Mesquita, "of the remaining 173 conflicts in which the opponent did not expect to lose more than the initiator expected to gain, 65, or 38 percent, became wars." Such a frequency differential would occur less than once in a thousand times by chance alone.21 From the preceding discussion it is apparent that the expected utility theory could be improved through modification of the role played by its unitary actor axiom. This would be especially welcome if the predictive power of the current theory could be preserved. War and Rational Choice So far, expected gains and losses from conflict have been defined in terms of postwar changes in policy. No mention has been made of the fact that, even in a war which is not particularly severe, both sides will experience some material loss. For any conflict, therefore, some alternative should exist that could prevent escalation to war. Rather than choosing the path of physical destruction, the parties involved could bargain over policy changes. The latter option would leave resources intact, while still allowing for a policy-based compromise.22 While correct in theory, the preceding argument becomes problematic in concrete terms. A bargain which is Pareto-preferred to war (i.e., involves no losses for all concerned and gains for at least one) may well exist, but expecting decision makers to find it imposes some strong assumptions about the availability of information. It may be easier for i to estimate its potential power relative to^, along with differences over/s policy, than to anticipate the extent of wartime losses.23 For example, those making the decisions in Europe during July of 1914 knew what grievances existed and had some sense of their adversaries' capabilities. What they did not anticipate, however, was the magnitude of the suffering about to take place. In practice, considerations of material loss may occupy a secondary role in planning for war, because leaders often tend to exaggerate their prospects for rapid and relatively painless victory. Some prominent historical cases of that would include the German invasion of the Soviet Union in 1941, the 1961 us-backed intervention in Cuba, and the more recent Soviet efforts to quell Afghani rebels.24 The conditions which would promote the discovery of a mutually beneficial bargain are not currently specified by the expected utility theory. This is a problem that should be confronted more directly at the stage of testing.

41 War and Expected Utility Reconsidered Bueno de Mesquita's Conception of Utility Theory

Fourth among the conceptual problems is the particular usage of utility maximization as a model of decision making. The quarrel here is not with the attempt to apply the theory to choice at the international level.25 Instead, difficulties arise because certain aspects of utility theory have not as yet been incorporated. As a theoretical illustration, consider the case of the consumer trying to maximize the following utility function:

where U

= consumer's utility function to be maximized at time o, subject to a budget constraint; Ct = level of consumption at moment t; V(Ct) = utility at moment t; T = time horizon for planning; and n = a constant, the rate per period of positive time preference. All but the last component of Equation 2.3 should be self-explanatory. "Positive time preference" means that the consumer discounts future consumption relative to that of the present at a constant pace.26 To discover this rate of discount, Olson and Bailey contend that it is essential to compare at two different times (t0 and tt) the consumption levels between which the consumer is indifferent. Figure 2.1 displays an indifference curve U0 in the commodity space C0 X C1( where C0 represents consumption at t0 and Ct likewise at tl. The consumer is indifferent to each of the commodity bundles at U0, and this reveals the value of n, with the slope of the tangent ABD to U0 being — (i + n). For utility to be a meaningful concept, it is necessary to have some idea of what this slope represents at a practical level, and that information can be obtained only through comparison of consumption levels over time. However, Bueno de Mesquita derives the utility functions of national leaders in a static manner. He looks at the degree of similarity between the alliance commitments of a given i and j at t0, and uses a statistical measure of association to assign a value to Uy. Leaving

42

Crisis and War

Figure 2.1 Discounting Future Utility: The c0 X d Consumption Space

Source: Mancur Olson and Martin J. Bailey, "Positive Time Preference," Journal of Political Economy 89 (1981): 6.

aside operational considerations (such as the value of intersecting alliances as a measurement of shared interests), this does not constitute intertemporal comparison of "consumption," and the failure to incorporate a discount rate creates a gap in the utility calculus as proposed by Bueno de Mesquita. Suppose, for example, that i is faced with the choice between war and peace with j at time t0. Since war is costly, it has negative implications for future domestic consumption within some time frame, perhaps several years in duration. Consequently, in order to predict

43 War and Expected Utility Reconsidered

whether i rationally chooses war, knowledge of the trade-off function between future and present concerns is needed. The leader of state i may expect to benefit in some way from a war against^' (in Bueno de Mesquita's terms, E(Ui) > o), and yet decide not to attack because of domestic considerations. Perhaps the elite of i might be willing to pay the price of war at one time and not another, although they expect victory at both times. If true, that could help to explain an earlier puzzle: Why have many conflicts been resolved short of war even though the potential initiators had reason to be confident of victory? Perhaps some of the "irrational" de-escalations from Part B of Table 2.4 are not so difficult to understand after all. There is another conceptual problem regarding Bueno de Mesquita's appropriation of utility theory. A central tenet of microeconomic theory is that the value attached by a consumer to further increments of a commodity declines with the amount already possessed.27 By way of illustration, it is reasonable to assume that a consumer blessed with a large store of margarine is less likely to be concerned with obtaining more of this commodity than someone with no margarine. (Each increment of a commodity is somewhat less valuable to a consumer than the one before, despite possessing the same intrinsic worth.) This can be expressed mathematically by the following pair of inequalities:

where U = utility of the consumer; and A = any given commodity. By contrast, the utility scores in The War Trap are generated by a measure of association between categorical variables (tau b). As a result, utility becomes a linear function of policy change. This does not reflect the usual assumptions made by economists about marginal utility. The most important effects of that practice pertain to Proposition 3. When the policies of i andj are virtually identical, it is only natural to expect that any change in /s policies will be for the worse as far

44 Crisis and War

as i is concerned. How can things get better? But this does not imply that an attack would be "rational," even if i expects to win. Imagine that at time t0 the potential adversaries agree on virtually everything of importance to i, and also that i has some reason to believe that /s policies will change somewhat for the worse by 11. Since marginal utility declines with the amount of agreement already possessed, decrements at the margin could be discounted enough to be without impact upon z's thinking. In addition, since future payoffs are discounted at t0, that also should contribute to the initiator's potential lack of concern about "drift" in the policies of an ally.28 But is the concept of marginal utility relevant to foreign policy? Will the marginal evaluation of agreement with a rival state actually decline with the amount of agreement already possessed? According to Bueno de Mesquita's formulation it does not. But observed patterns of behaviour do not support his position. If the marginal utility of agreement does not decrease, then why do leaders allocate their resources across a range of objectives, rather than pursuing one goal relentlessly? When states concur in principle on important issues, they direct their energies toward influencing other system members, as would be predicted by decreasing marginal utility of agreement. Developments leading to a war between allies may be distinguished on the basis of the hypothetical reasoning of i in the following instance: 1 /s policies could be on an unfavourable course; 2 a successful war withy would prevent this from continuing; 3 the sizeable opportunity costs of war are deemed to be less than the potential cost of unfavourable policies pursued by j; and 4 the decision is made to go to war. In other words, it is rational to attack an ally only if its anticipated deviations are worth preventing. Since the value of the policy changes to be obtained will be discounted heavily when/s policies are virtually identical to those of i, the inequality from step 3 often will be reversed. This reformulation of Proposition 3 is more in line with the best traditions of applied microeconomic theory, because opportunity costs and declining marginal utility are accounted for in the process of decision making. Mathematical considerations raise further problems with Bueno de Mesquita's treatment of utility. In an extended review of the expected utility theory, Wagner notes several complications in that area. Consider the following equation, used by Bueno de Mesquita to represent the initiator's expected value from a war with the defender:

where

45 War and Expected Utility Reconsidered

= the probability that k will ally with i; = i's perception of the utility to be derived from k; i — P;k = the probability that k will ally with j; and Uikj = i's perception of the utility to be derived by j from k's support.

Pik Uiki

"This," as Wagner observes, "is supposed to represent the expected value to i of a war with j if k fights on the side of i. Although Pjk is properly cast as the conditional probability of i's winning, the rest of the equation is more difficult to put in perspective. It should include, as Wagner notes, "the utility to i from defeating j and the utility to i from losing toj.n*9 Although Ujk;, for example, is supposed to convey the value to i of k's potential war effort, the impact of k has already been covered by its "effect on the probability that i will win the war, and is thus included in the conditional probability term [ Pik]."30 After revealing some other discrepancies in the calculus, Wagner demonstrates that the central equation could be restated as follows:

As Wagner notes, this version is very similar to a balance of power formulation. The first term in Equation 2.7 accounts for i's power relative to that of j. After the summation, the first fraction will be positive or negative depending on whether k's policies are closer to those of i orj respectively, while the second fraction corresponds to k's capabilities as a proportion of all these states (i,j, and k) combined. Equation 2.7 thus differs only in a marginal way from direct comparison of one coalition's capabilities with those of the other. Therefore, the findings based on the above equation do not discredit a "power politics" hypothesis, as Bueno de Mesquita claimed, but instead discredit the assumption that "only the capabilities of the aggressor and defender are important in determining the balance of power."31 While damaging to the theory found in The War Trap, the preceding criticisms do not preclude its successful reformulation. The treatment of probability is the essence of the problem identified by Wagner. Consider a case in which there is one potential interme-

46 Crisis and War Table 2.5

Participation by Intermediaries

Participating Probability of intermediaries occurrence

Probability of winning for initiator

Notes: cap; (read as "capabilities of z") is discounted over the distance toj. The capabilities of intermediaries are discounted over the distance from each respectively to i or j, depending on which is closer.

diary. Under such circumstances, Wagner argues that z's probability of winning should be calculated as follows:

where Pki, Pkj» Pkn = the probability that k will ally with i, that k will ally withj, and that k will remain neutral. Piki Pikj Pikn = the probability that i will defeat j if k fights as an ally of i, if k fights as an ally of _;', and if k remains neutral. P and (i — P) can then be used as weights for (i — Uy-) and (Uy — i) respectively to yield an overall score for expected utility.32 With additional intermediaries, the conditional probabilities from Equation 2.8 will become more complex, though not necessarily beyond calculation. To illustrate, Table 2.5 depicts the case of two intermediaries, with k favouring i while m prefers j. In the pre-crisis environment, the relevant actors express their preferences regarding the initiator and defender.33 This eliminates the need for one set of probabilities. For example, since k favours i, Pjy is considered to be zero. Two potential intermediaries therefore will require the calculation of four probability terms regarding participation, as depicted by the second column of Table 2.5.

47 War and Expected Utility Reconsidered Each of the terms from the second column of Table 2.5 must be weighted by the probability that, with that particular combination of participants, the initiator will defeat the defender. This in turn will permit the summation required to generate values for P and i — P. The third column in Table 2.5 displays the four probability terms which will be used as multipliers for those in the second column. Each is based on the capabilities possessed by »', together with its active allies, as a proportion of the capabilities held by all active participants. Having clarified matters somewhat through the two-intermediary example, the expression for expected utility in the general case (i.e., n potential third parties) may be presented. With some algebraic manipulation it can be reduced to the following:

where E(U;) = expected utility for i from a war with j; n, 2" = number of war intermediaries and possible combinations thereof, respectively; pr = probability that the rth combination of intermediaries will participate in a war with _;'; and qr = probability that i will win a war against j with the rth combination of intermediaries. As in the expression used by Bueno de Mesquita, when there are no potential third parties to consider, positive versus negative expected utility will depend directly upon the power of the initiator relative to that of the defender. CONCLUSION

For the most part, the expected utility theory of war is consistent with the claims of its creator. This first-image theory is indeed economical in its assumptions and extensive in its range of application. The theory has been used by Bueno de Mesquita to generate interesting (and even counter-intuitive) hypotheses about conflict. When tested using data on interstate threats, interventions, and wars, these propositions received overwhelming confirmation. Some conceptual problems have been identified, however, and in

48 Crisis and War

a few instances tentative solutions have been offered. For example, conditional probabilities of participation by intermediaries should appear in the utility calculus. Other areas of difficulty that have been noted, such as the role of decreasing marginal utility in the decision calculus, cannot be resolved without reference to measurement strategies and the revisions of the theory made by Bueno de Mesquita and others. Rigorous testing of expected utility in the context of crisis and war is the ultimate objective of theoretical articulation. Thus, in the following chapter, hypotheses about the transition from crisis to war are developed and tested. Conceptual refinement and improved measurement strategies occupy a prominent place in the process of evaluation.

CHAPTER THREE

Expected Utility, Crisis, and War

There never can be a definitive or critical evaluation of a theory. Some tests, such as comparative case studies, provide depth, while other, data-based tests, have the advantage of breadth. In either case, variables may be selected in more than one way, and the fidelity of any choice with respect to an underlying referent can always be disputed. The selection of evidence - along with measures of association — is subjective also. Caution is especially appropriate when evaluating a new theory, because varied approaches to concept formation and testing may be needed before the theory can achieve a measure of success comparable to its rivals. Compared to more established research programs such as balance of power, expected utility has not accumulated a wide range of confirmation from testing. Setbacks at the empirical level, if in fact they do occur, should therefore not be viewed as decisive.* A set of propositions is delineated here, followed by applications and testing. PROPOSITIONS

Bueno de Mesquita's formulation of expected utility occupies a role similar to that of Thucydides' exegesis of balance of power. Both scholars could lay claim to an original and authoritative exposition of a theory. But revisions and extensions invariably follow. In this instance, a revised expected utility theory is used to predict the outcomes of international crises. The propositions discussed earlier have been reformulated in terms that are appropriate for crisisoriented testing. There are four propositions to be tested: Proposition 3.1: War is more likely to occur in cases where the initiator of

50 Crisis and War a crisis has positive expected utility relative to the defender than in cases where it is zero or negative. Proposition 3.2: In the absence of intermediaries, war is more likely to occur when the initiator is the stronger of the two states involved in the crisis. Proposition 3.3: If the two states follow essentially identical policies on relevant issues, that is Uy = Uy = i, then the initiator's expected utility, E(Uj), will be greater than zero when i is stronger than j. However, opportunity costs and declining marginal utility will inhibit the outbreak of war in this subset of cases. Proposition 3.4: If the crisis initiator's positive expected utility is greater than the defender's expected loss (|E(Ui)| > |E(Uj)|), war is more likely to result than when the inequality is reversed.

The theoretical problems raised earlier are addressed at least in part by these reformulated propositions. According to Proposition 3.1, expected utility must be positive, not merely non-negative, for war to occur on a rational basis. This is a logical choice to make, since sufficient conditions for war should be more restrictive than necessary conditions. Furthermore, war as a "rational" choice is less problematic when something better than a stalemate is expected, although opportunity costs remain a potential source of complication. Proposition 3.2 is intended to be a test of expected utility under a more specific set of crisis conditions. Proposition 3.1 might be confirmed in the general instance, but what if that masks an inability to predict the outcomes of crises which are direct confrontations? In other words, if expected utility is a valid predictor of resolution for complex, multilateral crises, it should be able to deal at least as well with less complicated, two-actor crises. Proposition 3.3 also requires further explanation. Since the defender and initiator cannot be distinguished by mutual allies on the basis of their policies, these potential intermediaries are expected to be indifferent to the outcome of a war. Under such conditions Bueno de Mesquita predicts that a stronger initiator would attack its ally. But the restructured proposition is based on the potential effects of opportunity costs and declining marginal utility. These factors suggest that the price of a war between allies often will be greater than the ostensible gains, thus resulting in a prediction of de-escalation in such cases. Operationalization will unfold in stages. Role designations for states and geographic boundaries for crises must be described and explained. Measurements have to be derived for the components of

51 Expected Utility, Crisis, and War Table 3.1 Outside Actors for the Regions Europe

United States

Asia

France Netherlands Portugal United Kingdom United States

Middle East France

Soviet Union United Kingdom United States

Western Hemisphere

Soviet Union United Kingdom Africa

Belgium Cuba France Portugal Soviet Union United Kingdom United States

the utility calculus. Finally, environmental conditions and risk propensity must be expressed in concrete terms. REGIONS AND ROLES

Geographic Region

Expected utility calculations are made with respect to other actors within a given international subsystem. For purposes of this analysis there will be five world regions: Europe, Asia, the Middle East, the Western Hemisphere, and Africa. These correspond to the regional designations used in The War Trap. According to Bueno de Mesquita, regional membership is held by any state physically located in the region and by any state that has taken a sustained and active interest in the affairs of one or more portions of the physical region.2

Table 3.1 displays the membership for each subsystem. The outside actors for a crisis in a given region will be a subset from the appropriate list in this table. Role Designations

With regard to initiators and defenders, the following designations will be used:

52 Crisis and War

i, initiator: the state that triggers an international crisis. j, defender: the first actor to perceive itself to be in an international crisis.3 These descriptions are significantly different from Bueno de Mesquita's, even allowing for the shift from war to crisis. He considers the initiator to be the state which had the last "reasonable" chance to avoid war, as evidenced by its first use of force on foreign territory. Although he adds the caution that the term "initiator" should not be equated with "aggressor," his use of "initiator" elsewhere implies that the two terms are synonymous.4 Initiators plan the "war trap"; defenders merely react to it. Furthermore, initiators (and not defenders) are hypothesized to have positive expected utility. Some examples will be used to demonstrate the value of an alternative scheme of classification. Consider the Cuban Missile Crisis of 1962. For Bueno de Mesquita, the United States played the role of crisis initiator, with the Soviet Union the apparent defender, applying the same logic for role designation as that used for the outbreak of war. But these labels do not seem compelling. To be sure, President Kennedy's speech demanding the withdrawal of the missiles could be considered the first overt action of the conflict. And the United States almost certainly would have been considered the initiator of the war, according to Bueno de Mesquita's decision rule, if an invasion of Cuba had occurred in the face of Soviet refusal to withdraw the missiles. Under these hypothetical circumstances, the United States would have played the role of initiator, because it had the last "reasonable" chance to prevent war. But casting the United States in the role of initiator (of either the actual crisis or a hypothetical war) seems mistaken. Khrushchev and his associates in the Kremlin realized that the United States would discover what had happened, preferably much later than in actual practice. The missile ploy challenged the existing nuclear balance and was aggressive in nature, at least to the United States.5 It therefore is more convincing to designate the United States as a defender and the Soviet Union as an initiator. Cuba fills out the set of actors as an intermediary siding with the Soviet Union. These designations show greater consistency with Bueno de Mesquita's theory than do the roles derived hypothetically using his own rule. The point is that the state with the last "reasonable" chance to prevent a war often turns out to be a poor choice for the role of initiator. Looking past this specific example, there are two systematic prob-

53 Expected Utility, Crisis, and War

lems associated with Bueno de Mesquita's designation of the initiator as the party with the last reasonable opportunity to avert the outbreak of war. Both are related to the use of the term "reasonable." One can be illustrated by the Football War which resulted from a crisis involving Honduras and El Salvador in 1969. Salvadoreans had been migrating for years from their overcrowded country to work on the banana plantations of sparsely populated Honduras. For this and other reasons, a certain amount of hostility had developed between the two countries. Over the course of two soccer matches, one played in each capital city, the underlying tension erupted into rioting. After the riots, large numbers of Salvadoreans who had come to Honduras as refugees fled back to El Salvador and reported that they had experienced mistreatment and confiscation of property. These events triggered a crisis which began on 15 June 1969. Both sides broke off diplomatic relations soon after, and border incidents proliferated. On 14 July, the Honduran air force attacked various cities and ports in El Salvador, while Salvadorean troops crossed into Honduras.6 Bueno de Mesquita designates El Salvador as the initiator and Honduras the defender in this war. According to his coding rules, El Salvador was the last party to have a reasonable chance to avert war. But this is when the controversy begins in earnest. The crisis data set has coded El Salvador as the first state to perceive a threat to basic values, high probability of military hostilities, and a finite time for response. Regarding the evidence, with both sides engaging in hostile acts around the border, it is difficult to say how unreasonable an invasion of Honduras was in the context of Salvadorean national security. Furthermore, it could be argued that the treatment of Salvadorean refugees had been "unreasonable" and that Honduras therefore had foregone the last opportunity to prevent war. If the war is instead treated as the escalation of a crisis initiated by Honduras, there is no need to speculate on which side behaved reasonably at the outbreak of war. With this coding rule, the question becomes one of determining which actor first perceived itself to be in a crisis and for what reasons. Although that may be difficult in some instances, at the very least it has a more objective basis. Roles are to be assigned through a reconstruction of the past, as opposed to an estimation of the reasonable options available at a given time. Even if the last reasonable opportunity to prevent war could be estimated for a wide range of crises, there is a further problem with Bueno de Mesquita's coding. While he has not made the error of attributing aggressor status to initiators as so defined, his expected utility model implies that they have planned their actions. As noted previously, it would be illogical to expect initiators to win wars (as

54 Crisis and War

does Bueno de Mesquita) unless at some level they are thinking aggressively. This creates a problem in light of the following observation by Schelling: In bargaining, the commitment is a device to leave the last clear chance to decide the outcome with the other party, in a manner that he fully appreciates; it is to relinquish further initiative, having rigged the incentives so that the other party must choose in one's favor.?

Another problem with Bueno de Mesquita's designation of roles is that the last reasonable opportunity to prevent war may rest with one party simply because the other has made an irrevocable commitment. The Six Day War is a case in point. On 17 May 1967, two actions by Egypt triggered a crisis for Israel. Eygptian aircraft flew over Israel's nuclear research centre at Dimona and two additional Egyptian divisions were dispatched into Sinai. (Others had crossed the Canal in the preceding two days.) The next day, United Nations Secretary General U Thant acceded to President Nasser's persistent demands to withdraw the UN Emergency Force from the Sinai. This action intensified the crisis for Israel, which correctly anticipated the events to come. On 23 May Egypt announced the closure of the Straits of Tiran, blockading Eilat, Israel's only port on the Indian Ocean. On 30 May, a defence pact was signed between Egypt and Jordan. Israel responded to the perceived danger of this alliance by destroying the Egyptian air force on 5 June, triggering a crisis for Jordan and Egypt. The scope of the conflict expanded soon after, with Jordan and Syria entering the fray and the superpowers exchanging threats.8 Israel is regarded by Bueno de Mesquita as the state which initiated the 1967 war.9 When it attacked Egypt, Israel passed over the last reasonable opportunity to prevent a war. But the strategic and tactical situation faced by Israel on June 5 could have become untenable without decisive action. The blockade and defence pact would not have been the end of renewed pressure from the Arab states. Egypt, not Israel, had plans to revise the existing map of the Middle East through the use of force. In the context of an expected utility theory, Egypt is the logical choice for the role of crisis initiator. It would appear that the crisis-oriented formulation of the initiator and defender roles is a more convincing approach. While it is easy to imagine a crisis initiator planning a threat and "calculating" its value, the same cannot be said of a state that is confronted with a choice between war and the last chance to avoid it, reasonable or not. Thus the crisis-based role designations appear to be more appropriate for Bueno de Mesquita's theory than the coding he actually used.

55 Expected Utility, Crisis, and War THE EXPECTED UTILITY CALCULUS

Probability

Having dealt with geographic boundaries and role designations, there now is the utility calculus to consider. It will expedite matters to continue with the four-actor example from the previous chapter, with i and j being joined by potential intermediaries k and m. Recall that there are eight probability terms from Table 2.5 to be operationalized. Four of these pertain to the initiator's chances to win a war with various combinations of intermediaries. Their operational rendering will be discussed in the context of the approach used in The War Trap. Bueno de Mesquita relied upon the Correlates of War (cow) Project's National Capabilities Index to operationalize his probability terms for the initiator's winning and losing a war. The Index as adapted by Bueno de Mesquita has two principal advantages as a means of measuring power. One is that it is "carefully coded and precisely documented."10 The other is that capabilities are discounted by Bueno de Mesquita over distance through a "loss of strength" gradient. He adds that feature in order to allow for the fact that military operations in distant locations are more difficult to conduct. * 1 A state's power is therefore at a maximum when fighting on its home territory, because the costs and risks of transporting war materiel are at a minimum. This Index will not, however, be used in the analysis to follow. The cow Index measures the distribution of power in the system at five-year intervals.12 The expected utility scores based on that Index may be criticized as insufficiently sensitive to change. Estimates by leaders at the brink of war are likely to be made on the basis of more current information about their adversaries.13 This suggests that annual figures for capabilities would be more in keeping with the theory being tested. Yet a shift to yearly statistics would entail a mammoth collection of additional data, because the cow Index has six components: demographic (population and its concentration in urban centres), military (expenditures and men under arms), and industrial (energy consumption and iron and steel production). Fortunately, there is a more streamlined indicator which has demonstrated a high level of correspondence with the cow Index: gross national product (GNP). Using the same set of countries as Singer and his associates, Organski and Kugler discovered that "the two measures, while not identical, were highly correlated, with a coefficient of determination

56 Crisis and War

of .86."14 Replacement of the cow Index by GNP also has some other advantages. The use of GNP reduces the complexity of the variable and attenuates the problems associated with missing data, which would be severe in the case of a six-point index compiled on a yearly basis.15 The further assessment of GNP as a measurement of power will be postponed until the discussion of balance-of-power theory. While GNP has its benefits as a substitute for the cow Index, the use of a "loss of strength" gradient for distance requires that capability scores be converted to fractions. The composite probability scores used by Bueno de Mesquita are proportions in the range of o to i, based on the share each nation-state has of all the power in the system. When raised to a logarithmic exponent which increases with distance, these decimal fractions decrease as intended. But untransformed GNP scores (expressed in us dollars) would increase when raised to such an exponent because they are not fractions ranging from o to i. To permit discounting by distance, the GNP of each actor in a crisis will be expressed as a fraction of the summed scores for all of the actors in that crisis. To continue, the function that is borrowed from Bueno de Mesquita and used to discount capabilities appears below:

where axy = c = dxy = e =

adjusted capabilities for state x at location y; composite capabilities for state x; distance from state x to location y; and epsilon, a constant.

To derive this function as Bueno de Mesquita did, with extensive references to diplomatic and military history, would be beyond the scope of this work. However, the principal reasons behind the structure of Equation 3.1 must be explained. The exponent allows for a certain distance from location y before state x's capabilities begin to decay. If a military operation is within three day's travel (about 1500 miles in the modern era), the capacity to act is assumed to be undiminished. Once the three-day threshold has been assumed, the larger the base being exponentiated, the more slowly the adjusted value declines. This is in keeping with the belief that large countries enjoy economies of scale in distant military ventures, which is consistent with the record of military history.16 Operationalization of the four probability terms from the third

57

Expected Utility, Crisis, and War

column of Table 2.5 now becomes straightforward. For each scenario, the capability scores of the actors involved are discounted over distance, except for the defender, which is presumed to fight on its own territory. The initiator's probability of winning with each combination of actors corresponds to the fraction of capabilities possessed by its coalition. To continue with the case of two potential intermediaries, there are four other probability terms which must be determined. They correspond respectively to the chance that each combination of third parties — ranging from none to both - actually will appear in a war between the initiator and defender. To explain these probability terms adequately, it is necessary to deal with two more general issues beforehand. One concerns which states are to be included in the set of eligible third parties. The other is related to the estimation of conditional probabilities of participation. Bueno de Mesquita included every member of a region in the calculation of expected utility. On grounds of descriptive validity this appears to be excessive. For a given war (or crisis), there are some states which almost certainly will remain neutral, despite being located in the region concerned. Thus it is proposed to restrict the set of potential war participants to states which, based on historical sources, had some involvement in the crisis. For other states, the probability of involvement is considered to be zero. Although some cases may be more complex, the pair of examples offered at this point should be sufficient to demonstrate that a streamlined set of potential war participants has greater validity. In the crisis over Goa in 1955, India and Portugal would be the lone participants. It would seem unrealistic to include China, Indonesia, or the Soviet Union as potential intermediaries simply because they are in Asia, as these states demonstrated little or no interest in the colonial dispute in question. Or consider the Unilateral Declaration of Independence by Rhodesia in 1965, which created a danger to Zambian security when the Rhodesians displayed troops on the border and indirectly threatened the hydroelectric power supply of their landlocked neighbour. The United Kingdom became involved on the side of Zambia when it agreed to provide Royal Air Force planes and personnel for air defense.17 On the basis of the historical record it is clear that, aside from the antagonists, the United Kingdom was the only country with a non-negligible role from the point of view of military security. The inclusion of every African state in the calculus would bias the result away from the contributions made by the three actors that were really involved in the crisis. For each crisis in the data set, the intermediaries in the calculus

58 Crisis and War

therefore will be restricted to states which perceived at least one of the defining conditions for a crisis. This will facilitate a more realistic assessment of expected value for the initiator. While there probably are some cases for which an inclusive designation would be appropriate, the average level of distortion should be at a minimum when only the interested third parties are counted. Table 1.1 displays the potential intermediaries for each international crisis in the data set. Actors designated k ir (r= i,...,3) favour the initiator and those listed under k2S (s= i,.--4) favour the defender. The maximum number of intermediaries supporting an initiator is three (e.g., Berlin Deadline), while the most supporting a defender is four (e.g., Israel Independence). Once the set of eligible third parties has been decided, conditional probabilities for participation may be estimated. Utility scores are used to estimate intensity of preference and, consequently, the likelihood of entry into war by each intermediary. Combining the absolute values of the utility scores for an intermediary with respect to i andj provides an assessment of how concerned that actor is with the outcome of the crisis. The total utility scores for an intermediary with regard to i and j range from zero to two in absolute value. At one end of the scale, a combined score of zero would mean that the potential third party is indifferent to the defender and initiator alike. At the other end, a score of two would indicate complete agreement with one antagonist and absolute hostility to the other (i.e., U^ = i and Uiiy = — i or U^i = — i and Uikj = i, using Bueno de Mesquita's notation). Division by two transforms these total scores into probability levels for participation. When there is a value of one, intervention is regarded as certain. With a value of zero, entry into war is not seen as a possibility. Of course, nothing is "certain" or "impossible" in reality, so the extreme levels (o and i) should be regarded as approximations to the true probabilities. Each intermediary is considered to have two potential alignments in a hypothetical wartime situation resulting from a crisis: (i) siding with one of the antagonists and (2) neutrality. The third logical possibility — lining up with the other antagonist - is treated as functionally non-existent, as noted previously in strictly theoretical terms. An example will be used to demonstrate at the operational level why this simplifying assumption is tenable. Burma Infiltration, 1953-54, has the same array of participants as the hypothetical case displayed by Table 2.5. Taiwan was the initiator, with Burma as the defender and the United States and the People's Republic of China as intermediaries k and m, respectively. Details from this case should help to explain why the third probability

59 Expected Utility, Crisis, and War

score for a given intermediary (e.g., the probability of the PRC assisting Taiwan) is considered to be zero for practical purposes. Consider the setting of Burma Infiltration. Chinese Nationalist forces crossed into Burma in 1949 as the Chinese Communists solidified their position in the south. The Nationalists supplied arms to insurgents hostile to the Burmese regime. This provoked the PRC, which by 1952 had over 10,000 troops near the border with Burma. On 8 February 1953, Burma experienced a crisis over an armed attack by insurgents and Chinese Nationalists on the city of Loikaw. The Burmese Government, aware of the involvement of the United States in Nationalist Chinese operations in Burma, announced its intention to sever assistance agreements with Washington. Serious clashes between Burma and Nationalist troops continued throughout the summer of 1953, although neither the PRC nor the United States was directly involved in the fighting. The crisis ended in October 1954 when Nationalist troop withdrawals from Burma reached a level which restored the latter's sense of long-term security.18 With respect to determining the roles played by the United States and the PRC, evidence from the case history (i.e., the pre-crisis environment) is compelling. Antagonism between the Nationalist and Communist Chinese, in conjunction with the assistance provided by the United States to Taiwan in its Burmese military operations, limits the number of credible alignments. Thus the probability that the PRC would intervene in a war to help the Nationalists is considered to approximate zero, and the prospect of a Burma-pRC-United States coalition opposed to Taiwan also is regarded as negligible. Of course, not every case is as straightforward as Burma Infiltration. However, the intrinsic properties of an international crisis will invariably restrict coalitional arrangements. An international crisis begins with a crisis for the first actor when a state (i.e., the defender) experiences a threat to basic values, along with finite time for response and the prospect of military hostilities. This makes it very difficult to imagine an intermediary which would have the potential both to approve and disapprove of the same value threat. In other words, foreign policy preferences are held firmly enough in the short term to virtually rule out such developments. Returning to the example of two potential third parties, it also is necessary to calculate a probability level for each combination of actors. For any given combination there is a probability of occurrence based on the product of probability scores for the relevant set of independent events. Take, for example, the contingency in which k assists i while m does not help j. The probability of these events occurring simultaneously is given by the product of their respective probabil-

6o Crisis and War

ities, Pki and (i — Pmj). The former is estimated by ((Uyyl + |Uiiy|)/2, while the latter corresponds to (i — (|Uimi| + |Uimj|)/2). This completes the measurement of the eight probability terms from Table 2.5, and the calculations for the w-intermediary case are analogous. Measuring Utility Some questions arise concerning Bueno de Mesquita's measurement of utility in The War Trap. He assesses the utility of one state for the policies of another in terms of their similarity with respect to alliance commitments. While he is on safe ground when rejecting membership in intergovernmental organizations as an alternative, since this has little to do with foreign policy orientation, his dismissal of the use of economic linkages is more questionable. He offers four arguments against using such indicators to measure congruity of interests.19 His first point is that alliances are a direct product of government action, while economic relations are not. Although it is true that governments (and not corporations or other private actors) enter into alliance agreements, it also is correct to observe that economic interests influence which pacts will be signed. With regard to utility considerations, it may even be unimportant whether governments have brought about the economic linkages which exist among them. Suppose that the corporations or other economic agents of one state are closely connected to those of another. It is reasonable to assume that government decision makers of the first state will have an interest in the viability of the economy of the second, even if the government itself did not consciously promote the accumulation of economic linkages. Commercial association thus may increase the probability that these states will pursue mutual security arrangements. Bueno de Mesquita's second argument against the use of economic indicators begins with the observation that trade relations reflect a mixture of public and private policies, with the proportions varying over time and location. He claims that economic indicators as a result will be ephemeral and not of much use in the measurement of utility. But if governmental decision making fluctuates significantly from one state to the next, it is reasonable to expect that the degree of control over coalition formation also will lack a uniform quality. The existing system of rule in a given nation-state affects both economic and political behaviour abroad. There is no reason to assume a priori that governments everywhere and always will dictate political and military policies (such as alliance aggregation) while lacking authority over the economic sphere (for example, trade agreements). These

6i

Expected Utility, Crisis, and War

compartments for "high" and "low" politics seem out of date in the current world system. Bueno de Mesquita contends that economic measurement of utility is inappropriate for a third reason: it is not responsive to subtle shifts in foreign policy. But this confuses the issue at hand. An indicator of common interests should reveal those interests. In other words, it is reasonable to demand that the indicator predict the nature of foreign policy behaviour, and not the reverse, as Bueno de Mesquita has suggested. There is a final justification offered by Bueno de Mesquita for excluding economic indicators: the scarcity of data prior to World War I. Although the present analysis concerns a more recent period, the problem of inadequate information persists. Gathering data on trade and investment, to cite but two examples, would present overwhelming difficulties. For many Third World countries, and members of the Soviet Bloc as well, estimates are either unavailable or extremely suspect. Since complete and reliable figures exist only for OECD member nations and a few other countries, the use of trade or investment flows would bias the analysis; cases involving other states would have to be excluded. At some point it certainly would be interesting to use economic statistics for the available cases and to compare the results generated by the two sets of data. But for now it will be necessary to rely upon alliance data alone. ENVIRONMENTAL CONDITIONS AND RISK

PROPENSITY

Environmental Conditions This is a concept in need of clarification, primarily because Bueno de Mesquita's operationalization is not derived from the axiomatic basis of his theory. He claims that the presence of risk versus uncertainty should be inferred from "systemic tightness." In qualitative terms, "systemic tightness" measures the degree to which states in the system are organized into identifiable coalitions. When they are, the system is considered stable and the environment is one of risk. Under such conditions a decision maker can estimate probabilities of participation for intermediaries in a potential war. By contrast, discernible changes in alliance aggregation are expected to induce uncertainty in a given regional subsystem.20 With loyalties in transition, the prospective cast of characters for a war becomes problematic. While some departure from the deductive approach is understandable when trying to estimate the beliefs of political leaders, the

62

Crisis and War

path traveled by Bueno de Mesquita may not be the best one available. Coalitional stability at the systemic level could be of limited relevance to the environment of an individual conflict-, because strategic interaction may have a much greater effect on the beliefs held by decision makers about potential actions by others. Contextual information available in the crisis data set is therefore used here to categorize environments. For each crisis, researchers affiliated with the ICB Project have generated a descriptive summary.21 These case histories include an assessment of the expectations of each state's decision makers about potential third parties to a war. This specific information will be used to decide for each individual case whether conditions of risk or uncertainty existed in the minds of those making decisions. Environmental conditions are considered to be those of risk unless the case summary reveals a statement by one (or more) involved crisis actors questioning or denying support to the initiator in the event of an attack on the defender. Such an assertion would create conditions of uncertainty because the initiator would no longer be able to define the field of conditional probabilities. One instance of each type of environment is provided. The vast majority of cases resemble the first example to be discussed. Invasion of Laos I (1953), in that conditions of risk prevailed there. In March of 1953, Vietminh troops gathered steadily at the Laotian border and initiated an offensive on April 5. These events triggered a crisis for France, which had been defending its interests in Indo-China for some time, and for Laos also, once the attack began in earnest. Although NATO denounced the invasion, the Vietminh had to contend with two adversaries at most, facing only a remote chance of participation by others.22 For a contrast with the Vietminh case, there is the People's Republic of China crisis over Taiwan (Taiwan Straits in), with the United States playing an uncertain role. In April 1962, Chiang Kai-Shek threatened to invade the mainland and conscripted additional manpower for the army. At the outbreak of the crisis, Taiwan had no guarantee from its ally, the United States, that the latter would intervene on its behalf if military initiatives took place against the PRC. However, Taiwan's foreign minister had suggested that us support would be forthcoming, and the People's Republic feared that this might be the case.23 But on June 23 the American Ambassador to the PRC, John Cabot, denied that the United States would actively support an invasion of the mainland by Taiwan. President Kennedy reinforced that assertion on June 27 at a news conference. Taken together, these circumstances suggest that the set of actors

63 Expected Utility, Crisis, and War

involved in a potential war had an uncertain character from the perspective of the initiator, Taiwan. The next section examines the practical implications of risk versus uncertainty in the crisis environment through a discussion of risk propensity. Risk Propensity

With regard to risk-taking, Bueno de Mesquita considers a riskaverse state to be one which has constructed a foreign policy with the following property: there are fewer states in the region with positive expected utility relative to it than the reverse. In other words, more states rationally could be attacked by this state than could rationally initiate an attack. For a risk-acceptant decision-maker, exactly the opposite will be the case. Unfortunately, this treatment of risk propensity is biased when applied to major powers. As even Bueno de Mesquita admits, powerful countries are certain to appear risk-averse.24 But there is a more basic problem with the inference of risk propensity from asymmetric levels of vulnerability: does relative safety necessarily imply caution? Two more credible explanations come to mind. One, as noted, is that, all other things being equal, the more powerful states are likely to be relatively safe from attack. The other is that adroit diplomacy may lead to security, while incompetent statesmen can find themselves facing any number of serious threats, regardless of their orientation toward risk. An alternative approach thus will be used to discern risk propensity. Imagine an individual of ordinary means presented with a choice between a cash award of $1000 and a fifty-fifty chance of winning $2000 or nothing. Undoubtedly all but a few respondents would prefer the sure payoff, although the expected value from the gamble is exactly the same. In a sense, decision makers at the international level are similarly placed: foreign policy "gambles" in which they may engage will have important consequences, win or lose.25 Furthermore, even a successful venture in foreign policy can produce undesirable domestic consequences, such as economic dislocation, civil unrest and, in extreme cases, revolution. It is reasonable to expect that the risks attendant on initiating a crisis will not be attractive to the overwhelming majority of decision makers. In sum, the notion of risk acceptance seems virtually irrelevant to choice at the international level. This speculation raises the question of whether decision makers then will be neutral or averse to risk with respect to crisis decisions. The answer is indeterminate, because only very detailed case studies (of a type that would transcend the International Crisis Behavior

64 Crisis and War Table 3.2 Decision Rules Type of environment

Type of decision Riskneutral Riskaverse

Risk Calculus

E(UO

Rule

E(UO > 0

Calculus Rule

Uncertainty

Only states with Ujy < appear in the calculus. E(UO > 0 E(Ui)b > 0

uikj

summaries) could deal with this subject properly. In the absence of more complete information, it will be assumed that risk propensity interacts with type of environment to determine the nature of the decision rule regarding the utility calculus (Table 3.2). Under conditions of risk, a decision maker is assumed to perform the standard calculations which have been discussed up to this point. Uncertainty induces risk aversion, and the decision rule becomes more conservative: war can be selected rationally only if expected utility is positive and the margin of superiority does not depend on help from intermediaries. In other words, both overall and bilateral expected utility must be favourable. This completes the operationalization of environmental conditions and risk propensity. To breathe some life into the rather abstract discussion above, expected utility will be operationalized for the Formation of the United Arab Republic (1958). This case has the same configuration of actors as the hypothetical two-intermediary example in Table 2.5. Egypt, the initiator, is supported by Syria (kn). Iraq, the defender, has Jordan (k 21 ) as an ally in the crisis. There are four logical possibilities with respect to war participation: (i) Egypt and Iraq; (2) Egypt, Iraq, and Syria; (3) Egypt, Iraq, and Jordan; and (4) Egypt, Iraq, Syria, and Jordan. The probability of participation for the initiator and defender is set to i.o in each case, since one state would be attacking and the other is assumed to defend itself. For the intermediaries, the likelihood of entry into a war between Egypt and Iraq is based upon their utilities for these two states. The probability of war participation for Syria and Jordan is the same, .71. The terms pki, Pkm Pmj> and pmn from Table 2.5 therefore are equal to .71, .29, .71, and .29, respectively. This makes it possible

65 Expected Utility, Crisis, and War

to estimate the likelihood of each of the four combinations of actors listed above with p x = .084, p2 = .206, p3 = .206, and p4 = .504. For example, pj represents the probability that both intermediaries will remain neutral. This quantity is equal to the product of pkn and Pmn, or .29 X .29 = .084. For the initiator, the probability of victory in each instance is based on the fraction of overall capabilities held by its coalition. The resultant scores are q x = .87, q2 = .91, q3 = .91, and q4 = .90. For example, q2 represents the probability that Egypt would win a war against Iraq, with Syria as the lone intermediary. The following equation derives q2, with all but the capabilities of Iraq being discounted by distance:

Under conditions of risk, Egypt's expected utility from a war with Iraq may be obtained by substituting the p and q terms into Equation 2.9:

The positive value would lead to a prediction of escalation for this case. But the crisis evolved differently. King Hussein of Jordan and King Feisal of Iraq reacted to the threat to their regimes from the oversized United Arab Republic by forming the Arab Federation. Nasser condemned the Federation as a coalition of reactionaries, but no violence took place and the crisis de-escalated.26 STATISTICAL METHODS AND THREATS TO VALID INFERENCE

The statistical procedures to be used in testing are explained at this point. The analysis relies upon contingency tables, with war versus other forms of crisis resolution as categories of the dependent var-

66 Crisis and War

iable. Inferences about these tables are based upon the tau b statistic, a standard measure of association between nominal variables. The tau b statistic measures the percentage reduction in error in predicting categories of the dependent variable. For example, tau b = 0.25 would mean that use of the independent variable in question leads to a 25 percent reduction in error when predicting the outcome variable. Put differently, the 0.25 score indicates that the independent variable does 25 percent better than chance in anticipating values of the dependent variable. Statisticians regard a value such as 0.25 for tau b as moderately strong.2? While there are other statistics available for the analysis of nominal data, the skewed distribution of data on war, the dependent variable, reduces the number of viable measurement options.28 It should be noted that with skewed data it is extremely difficult to generate high values for tau b or other measures of association. In the present data set, chance alone would succeed 115 out of 132 times — 87.1 percent of the time — simply by guessing de-escalation in all cases. Thus percentage reductions in error are unlikely to be overwhelming. In sum, the task of distinguishing wars from crises resolved otherwise should be recognized for its high level of difficulty. Threats to valid inference must be dealt with during the course of the statistical analysis. When classes of events are compared, spurious differences may emerge due to the presence of extraneous factors. Fortunately, techniques have been developed to cope with that problem. Quasi-experimentation is one such method. It offers a systematic approach to dealing with threats to valid inference. According to Cook and Campbell in their authoritative work, quasiexperiments have treatments, outcome measures, and experimental units, but do not use random assignment to create the comparisons from which treatment-caused change is inferred. Instead the comparisons depend on non-equivalent groups that differ from each other in many ways other than the presence of a treatment whose effects are being tested.29

In the present context, the treatments would correspond to conditions associated with escalation of crisis to war. War and other forms of resolution would be the outcome measures, with the experimental units referring, for example, to initiators having positive versus negative expected utility. These analogies have been made for heuristic purposes. The presentation of threats to inference follows the format and appropriates the language of quasi-experimentation, but the statistical analysis itself is not based on a quasi-experiment in any rigorous sense. The

67

Expected Utility, Crisis, and War

testing of hypotheses involves the straightforward use of control groups, as opposed to the more complex research designs that are encompassed by quasi-experimentation. Nevertheless, quasi-experimentation offers a useful structure within which to delineate intervening variables, and it also stands in the way of tendencies toward "data-mining." Many of the threats to valid inference discussed by Cook and Campbell are irrelevant, because they concern individuals as subjects (i.e., experimenter effects). But others must be confronted, and they fall into several categories: statistical conclusion, internal, and external validity.30 Statistical conclusions are threatened by random heterogeneity among respondents. The latter refers to the potential for differences among subjects (in this instance crisis initiators) which are correlated with the outcome variable. The salient possibility is the initiator's system of rule. Two categories of government are used to control for random heterogeneity: democratic and authoritarian. Democratic regimes are distinguished by "competitive elections; pluralist representation in the legislature; several autonomous centres of authority in the political system; competitive parties; and a free press."31 Other regimes are characterized as authoritarian. Initiators in the control (i.e., zero or negative expected utility) and experimental (i.e., positive expected utility) groups are compared with respect to form of government. Internal validity refers to the establishment of a connection between independent and dependent variables. Threats to internal validity include history, maturation, selection, and interactions with selection. History is a problem because an observed effect might be the result of an event (other than the treatment: i.e., the conditions associated with escalation of crises to war) that has occurred between the preand post-test. Perhaps crises which begin with violent acts are more likely to result in war. To confront that possibility, an intervening variable — the level of violence of the environmental change triggering the crisis - is introduced. Violence is considered (i) low if none of the actors experienced violence at the trigger; (2) medium if some but not all of the actors had violent triggers; and (3) high if all the actors experienced violence at the trigger. Maturation refers to changes in the subjects - in this case the crisis initiators - over the course of the experiment, excluding treatmentinduced effects. Longer crises might be more prone to such contamination, so the average lengths of control and treatment cases will be compared. Selection refers to non-random differences between experimental

68 Crisis and War

and control cases other than whether or not they have received a treatment. One possibility is intervention by supranational organizations. The United Nations and regional organizations have sometimes assumed prominent roles in crises, and their activities may have been more common in one type of case than the other. The frequency of involvement by supranational bodies thus will be compared for the control and experimental groups. Interactions with selection refer to spurious effects resulting from selection and history. It is conceivable that the frequency of crisis participation could interact with extraneous events to produce spurious effects. For example, if the experimental group includes crisis actors with a high incidence of participation in crises, that could lead to spurious treatment effects based on experience. Control and experimental groups are compared for frequency of participation among their members, measured in the following way:

where F = frequency score for the actors in a crisis; z = the number of actors in a crisis (v = i, ... , z); and iv = the number of crises in the data set which involve actor v. External validity refers to the ability to make generalizations from results. Interaction of setting and treatment is the principal threat in this domain. A treatment (i.e., a condition such as positive expected utility) could have a tendency to appear in one region more than another, but that would be impossible to discern from aggregate figures. The control and experimental cases therefore are compared on a regional basis. Analysis of the threats to inference is straightforward. Statistical significance is set at 5 percent, or p < 0.05, a standard level in aggregate research. The chi-square statistic will be used as a test of significance for the respective tables. In other words, a threat to inference is regarded as noteworthy if the chi-square statistic surpasses the threshold necessary for p < 0.05. This completes the presentation of threats to valid inference.

69 Expected Utility, Crisis, and War Table 3.3 Expected Utility and War Crisis outcome Expected utility

War (%)

Other (%)

Total (%)

Positive Zero or negative Total

10(18.5) 4 (5.9) 14(11.5)

44 (81.5) 64(94.1) 108 (88.5)

54 (44.3) 68 (55.7) 122 (100.0)

Notes: There are two points of explanation concerning the format of this table and others which have the same structure. (1) Regarding the entries in the first two columns of numbers (i.e., those listed under "War" and "Other"), the figures in parentheses correspond to the percentage of cases in the row represented by the cell frequency. For example, 4 out of 68 cases with zero or negative expected utility resulted in war. Similarly, 108 out of 122 (or 88.5%) cases in the table did not have war as a crisis outcome. (2) In the third and final column ("Total"), the figures in parentheses correspond to the percentage of cases in the column represented by the cell frequency. For example, the initiator had zero or negative expected utility in 68 out of 122 (or 55.7%) of the cases in the table. Many of the subsequent tables have the same format. The parenthetical figures in the final column will be the percentage of cases in that column represented by the cell frequency, while the parenthetical figures in the other columns will be the percentage of cases in the row represented by the cell frequency. EXPECTED UTILITY AND CRISIS OUTCOMES

Table 3.3 offers support to Proposition 3.1. Of the cases in which the initiator had positive expected utility, 10 out of 54 (18.5 percent) resulted in war, while only 4 of the 68 (5.9 percent) other cases resulted in war (tau b = 0.20, p < 0.02). It also is interesting that the initiator's expected value was negative in 68 of the 122 crises, or 55.7 percent. On average, the initiator of a crisis will be intent on bringing about a change in the existing order. Since the status quo reflects the distribution of capabilities, the states which initiate crises often will not have leverage with the defenders they have challenged. In other words, expected value calculations usually favour continuation of the present order, and de-escalation prevails in most such cases. The Ghana/Togo Border (1960) and Pushtunistan in (1961) crises are useful illustrations of that point. Early in March 1960, a draft constitution for the Republic of Togo triggered a crisis for Ghana, because the document claimed the Volta region, inside the Ghanian border. Soon after, a plot by irregular forces (comprised of inhabitants of the Volta region) trained in Togo to infiltrate and attack Ghana was discovered. Prime Minister Nkrumah of Ghana responded by strengthening security forces and car-

yo Crisis and War

rying out army exercises in the Volta region, and in April an additional army, battalion was sent to the border. After the date of the anticipated invasion passed without incident, Togo ceased its efforts to obtain the Volta region from Ghana and tensions subsided.32 Togo, with an expected utility of -0.97 from a war with Ghana, had an incentive to pursue peaceful means to gain concessions from its adversary. Evidence that Togo did so is the bargaining over an extensive range of issues which emerged after the crisis. The negotiations culminated in a union for economic and political cooperation, signed on 12 June ig6o.33 Afghanistan and Pakistan experienced a similar but longer-lasting border-related crisis beginning in May 1961, when Pakistan became aware of an infiltration of i ,000 Afghani troops into the disputed territory of Pushtunistan and the Pakistani air force bombed the presumably occupied areas along the border. An Embassy spokesman for Afghanistan in Karachi denied the charges of territorial violation and tensions faded somewhat. But in August, Pakistan accused Afghani consulates and trade agencies of engaging in subversion, and demanded that they be closed. This triggered a crisis for Afghanistan, and on September 6 these states severed diplomatic relations. Further violence did not occur, however, and in January 1962 the frontiers reopened.34 This outcome is not difficult to understand, since Afghanistan had an expected value of —1.01 from a war over the disputed territory. Pakistan's resolve deterred Afghanistan, the weaker of the two states, from escalating the crisis. Once again, the existing order survived the challenge of a revisionist state. Togo and Afghanistan wanted to change some aspect of their relations with Ghana and Pakistan, respectively. They pressed their adversaries but found them willing to resist. The revisionist states responded rationally to the negative expected value of warfare, and turned to non-military means to advance their interests. On a less positive note, there are four wars which defied predictions of de-escalation: Yemen War i (1962), Yemen War iv (1966), Football War (1969) and Angola (1975). With the exception of Angola in the final case, the initiator in each instance had an expected value of zero from going to war. These cases may be regarded as anomalies in a weak sense, since the initiator would not have expected losses from the war in question. Proposition 3.2 is supported by Table 3.4, which includes the bilateral crises in the data set. Of the cases in which the initiator had positive expected utility, 5 out of 31 (16.1 percent) resulted in war, while only i out of the remaining 44 crises (2.3 percent) had that

71 Expected Utility, Crisis, and War Table 3.4 Expected Utility and War: Bilateral Cases Crisis outcome Expected utility

War (%)

Other (%)

Total (%)

Positive Zero or negative Total

5(16.1) 1 (2.3) 6 (8.0)

26(83.9) 43 (97.7) 69(92.0)

31 (41.3) 44 (58.7) 75 (100.0)

Table 3.5 Expected Utility and War: Identical Policies for Initiator and Defender Crisis outcome Policy differential

War (%)

Other (%)

Total (%)

Zero Other Total

3(11.1) 14 (13.3) 17(12.9)

24(88.9) 91 (86.7) 115(87.1)

27 (20.5) 105 (79.5) 132 (100.0)

outcome (tau b = 0.25, p < 0.02). Although not overwhelming, this discrepancy suggests that expected utility can be used to help distinguish the outcomes of direct confrontations. It complements the success enjoyed by Proposition 3.1 in the general instance of crisis. Proposition 3.3 is not supported by Table 3.5. Crises in which Uy= i are no more or less likely than others to result in war. In fact, this result is unfriendly to both Bueno de Mesquita's hypothesis about war between allies and the suggested revision of it. However, it is interesting to note the sheer number of crises that feature coalition partners as initiator and defender. Out of 132 cases in the analysis, 27 had that property. There is reason to believe that the frequency of these crises can be explained in terms of the threats involved. Table 3.6 displays the 27 cases concerned, along with the highest value threatened in each instance. There is not one crisis in which a threat to existence or grave damage occurred. Allies may get involved in crises rather frequently but even in cases of war, the intensity of conflict is limited. It is not surprising that coalition partners vie for influence and quarrel over territory and politics, because recurrent interactions will produce conflict in some areas. However, even in crises, some basic common interests continue to be recognized, such as survival and safety from grave damage.

Table 3.6 Value Threats for Crises with Uy = 1 Highest Value Threatened Case

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Communism in Czechoslovakia Costa Rica/Nicaragua I Luperon Costa Rica/Nicaragua II Nicaragua/Honduras Lebanon/Iraq Upheaval Mexico/Guatemala Fishing Rts. Cuba/Central America I Cuba/Central America II Assassination Attempt/Venezuela President Breakup of UAR Yemen War I Dominican Republic/Haiti Algeria/Morocco Border Ogaden I Yemen War IV Prague Spring Football War Cairo Agreement

Existence

Grave damage

Influence

Territory

Political system

Low

X X X X X X X X

X X

X X

X X

X X

X

X

X

Table 3.6 (concluded) Highest Value Threatened Case

20 21 22 23 24 25 26 27

Black September Uganda Tanzania I Uganda Tanzania II North/South Yemen I Cod War I South Yemen/Oman Sahara Cod War II

Existence

Grave damage

Influence

Territory

Political system

Low

X X X

X

X

X X X

Notes: The six categories of value threat are listed below, in descending order of gravity. 1. Threat to existence — Survival of population, genocide, existence of entity, total annexation, colonial rule, occupation. 2. Threat of grave damage — Large casualties in war, mass bombings. 3. Threat to influence in the international system - Declining influence in global and/or regional systems, diplomatic isolation, cessation of patron aid. 4. Threat to territorial integrity - Integration, annexation of part of a state's territory, separatism. 5. Threat to political or social system — Overthrow of regime,change of institutions, replacement of elite, intervention in domestic politics, subversion. 6. Threat to economic interests or low threat - Integration of economies, control by another actor's economy, requisition of resources, loss of markets, blocked access to resources or markets. Source: Michael Brecher and Jonathan Wilkenfeld, Crisbank: Data Bank of the International Crisis Behavior Project (unpublished manuscript 1981)

74 Crisis and War Table 3.7 Expected Utility Difference and War Crisis outcome

Expected utility difference

War (%)

Other (%)

Total (%)

Positive Zero or negative Total

6 (19.4) 4 (10.0) 10(14.1)

25 (80.6) 36 (90.0) 61 (85.9)

31 (43.7) 40 (56.3) 71 (100.0)

Table 3.8 Form of Government and Expected Utility Form of Government Expected utility

Democratic (%)

Authoritarian (%)

Total (%)

Positive Zero or negative Total

25 (47.2) 20 (30.3) 45 (37.8)

28 (52.8) 46 (69.7) 74 (62 2)

53 (44.5) 66 (55.5) 119(100.0)

Note: x2 = 2.88, P < .10.

Table 3.7 lends only marginal support to Proposition 3.4 (tau b = 0.13, p < 0.14). When the initiator's expected gain outweighs the defender's anticipated loss, war is somewhat more likely than in crises for which the asymmetry is reversed — 6 out of 31 cases (19.4 percent) versus 4 out of 40 (10.0 percent). The logic of utility trade-offs appears to transcend the boundaries of the calculus as it exists now. Threats to inference will be assessed in the context of Proposition 3.1. It would be beyond the scope of this investigation to present these data for each of the three other hypotheses, because that would generate 21 more tables. Such constraints also dictate that one hypothesis be selected from the other theories for the same purpose. Those propositions will be exposed to the threats to inference in Chapters 5 and 7. Tables 3.8—14 correspond to the various threats to valid inference. Several of these tables require discussion. Table 3.8 shows a connection between form of government and expected utility (chi-square = 2.88, p < o.io). Negative expected value is linked with the authoritarian regime type (69.7 percent), whereas crisis initiators with positive expected utility more frequently have democratic regimes (47.2 percent). The difference reflects the larger economies which,

75 Expected Utility, Crisis, and War Table 3. 9 Trigger Violence and Expected Utility Trigger violence Expected utility

High (%)

Medium (%)

Low (%)

Total (%)

Positive Zero or negative Total

26 (48.1) 28 (41.2) 54 (44.3)

13 (24.1) 17 (25.0) 30 (24.6)

15 (27,.8) 23 (33, .8) 38 (31 •1)

54 (44.3) 68 (55.7) 122 (100.0)

Note: x2 : = 0.69, P •c .71.

Table 3.10 Duration of Crisis and Expected Utility Expected utility

Mean duration of crisis (in days)

Positive

Zero or negative

147.8

118.4

Note: 95% confidence interval, based on the assumption of 118.4 as the population mean: 77.9 to 158.5. The 147.8 average for the Positive cases thus falls within an acceptable range.

Table 3.11 United Nations Activity and Expected Utility Activity Expected utility

Absent (%)

Present (%)

Total (%)

Positive Zero or negative Total

15 (27.8) 34 (50.0) 49 (40.2)

39 (72.2) 34 (50.0) 73 (59.8)

54 (44.3) 68 (55.7) 122 (100.0)

Note: x2 = 5.29, P < .03.

76 Crisis and War Table 3.12 Regional Organization Activity and Expected Utility Activity Expected utility

Absent (%)

Present (%)

Total (%)

Positive Zero or negative Total

31 (58.5) 32 (47.1) 63 (52.1)

22 (41.5) 36 (52.9) 58 (47.9)

53 (43.8) 68 (56.2) 121 (100.0)

Note: x2 = 1.14, P < .2 9.

Table 3.13 Frequency of Crisis Participation and Expected Utility Expected utility Positive

Zero or negative

11.0

8.7

Mean frequency of participation

Note: 95% confidence interval, based on the assumption of 8.7 as the population mean: 7.3 to 10.2. The 11.0 average for the Positive cases thus does not fall within an acceptable range.

Table 3.14 Location and Expected Utility Location Expected utility Positive Zero or negative Total

Americas (%)

Africa (%)

Asia (%)

Europe (%)

Middle East (%)

5 (9.3) 10 (18.5) 11 (20.4) 10 (18.5) 18 (33.3)

Total (%) 54 (44.3)

15(22.1) 17 (25.0) 13 (19.1) 8(11.8) 15 (22.1) 68 (55.7) 20 (16.4) 27 (22.1) 24(19.1) 18(14.8) 33 (27.0) 122 (100.0)

Note: x2 = 5.95, P < .21.

77 Expected Utility, Crisis, and War

on average, characterize the democracies. This source of contamination is not serious, however, because most of the wars in the data set have been initiated by authoritarian regimes. Thus democracy, crisis initiation, and war cannot be linked together. Table 3.11 connects the presence of United Nations activity to positive expected utility (chi-square = 5.29, p < 0.03). However, the specific direction of the linkage effectively removes any threat that might be posed. UN intervention in crises is intended to reduce tension. If anything, the organization's more frequent presence in cases where expected utility is positive (72.2 versus 50.0 percent) should decrease the likelihood of escalation. Table 3.13 links positive expected utility with higher frequency of crisis participation. In a sense, this is a plausible finding, in that initiators with positive expected value may be issuing serial threats to the same targets. More intensive, case-oriented research would be required to assess the effects of frequency of participation on crisis outcomes. CONCLUSION

Expected utility has obtained confirmation for its more basic propositions about crises. War is more likely when the initiator of a crisis has positive expected utility than when the expected value is zero or negative. This pattern also holds true for the subset of bilateral cases. It is interesting to note that crisis initiators usually have negative expected value. International crises reflect conflicts of interest, and very often it is the weaker party that initiates a crisis with the intention of changing the status quo. By contrast, wars tend to result from crises in which a revisionist power has pressed an existing advantage. Testing of auxiliary hypotheses (3.3 and 3.4) has not met with uniform success. War is no more or less likely when the initiator and defender are coalition partners than when they are not. These cases can be distinguished in another way, however. The threats involved are not at the highest levels, referring to grave damage and existence. Crises between allies are relatively frequent but less intense than others, even in cases of war. The second auxiliary proposition, which links differences in expected value for initiator and defender to the outbreak of war, found marginal support from the data. The further implications of these findings and the relevant threats to inference are discussed in Chapter 8.

CHAPTER FOUR

Conflict and Cohesion: The Blind Men Meet the Elephant

Freud hypothesized that an individual might cope with internal conflict by projecting it outward. Perhaps a nation-state in turmoil could be expected to react in much the same way.1 "Externalization" or "projection," the notion that leaders will engage in foreign conflict in order to promote domestic cohesion, was popularized initially by sociologists.2 According to sociological theory, an elite faced with social disintegration might attempt to restore order by diverting the public eye toward an external menace. Domestic strife would then be expected to subside, because those within the group will put aside their differences, at least temporarily, in order to pursue the higher goal of national preservation. Since war is seen as a direct consequence of changes in the internal processes of the state, it would be difficult to find a more explicit example of a theory from the second image of Waltz. Students of international politics became intrigued with this state-oriented line of reasoning, seeing it as a potential explanation for the pervasiveness of conflict. Several pioneers of international relations theory commented directly on externalization as a process. Wright asserted that integration has often been effected through the organization of opposition. By creating and perpetuating in the community both a fear of invasion and a hope of expansion, obedience to a leader may be assured.3

Similarly, Haas and Whiting argued that elites would try to preserve their power in the face of domestic threats by uniting their followers against external danger. They predicted that such behaviour would occur in "periods of rapid industrialization and largescale social change."4 At an equally general level, from his analysis of the nineteenth and twentieth centuries Rosecrance concluded that

79 Conflict and Cohesion

"there tends to be a correlation between international instability and the domestic insecurity of elites." He thus conceived of domestic tranquility as a "vehicle of international stability and external peace."5 From the outset, the theory of conflict linkage found a receptive audience among those concerned with explaining war. It provided a plausible explanation for aggressive actions which could not be traced to outstanding grievances at the international level. The theory also had the potential to be tested in a straightforward manner. Two decades ago, the interest in conflict and cohesion crystallized in the form of a sustained effort to obtain confirmation through aggregate research. THE ERA OF SYSTEMATIC TESTING

What is most remarkable about the theory of externalization is its practically uniform record of failure when tested on an aggregate basis. Seldom has so much common sense in theory found so little support in practice. In that regard, Rummel's study of domestic and foreign conflict linkage is an obvious point of departure.6 He conducted the first data-based research on projection and discovered some surprising patterns. Rummel collected data for 77 states over the period 1955-57, usmg 22 separate indicators for domestic and foreign conflict. (Table 4. i displays the indicators, which are typical of those used in subsequent research.) Relying upon factor analysis and some related statistical techniques, he inferred from his data that (i) discrete dimensions of foreign and domestic conflict behaviour could be identified, and (2) domestic and foreign strife have no apparent connection to each other. Soon afterward, Tanter used data from other years to replicate these findings, although the introduction of a time lag in his study did produce a minor level of association between internal and external conflict.7 Prominent historical cases like the French Revolution and the RussoJapanese war of 1905 had created a predisposition to believe in conflict linkage. Rummers study and the confirmation by Tanter contradicted the conventional wisdom. The new findings made the entire process of externalization seem questionable, stimulating a wave of scholarship. Subsequent research efforts have followed two not entirely separate paths, the common being quantitative analysis. One branch of inquiry uses different statistical tests and new data as a means of checking the reliability of previous findings, although of course that also involves somewhat different formulations of the principal hypothesis. The other branch focuses on the nexus of internal and external conflict in specific geographic regions. These studies are assessed here in two stages, in a survey which

80 Crisis and War Table 4.1 Rumrnel's Indicators of Conflict Domestic indicators Assassinations General strikes Presence or absence of guerrilla warfare Major government crises Purges Riots Revolutions Anti-governmental demonstrations Deaths from all forms of domestic violence Foreign indicators Anti-foreign demonstrations Negative sanctions Protests Countries with which diplomatic relations were severed Ambassadors expelled or recalled Diplomatic officials of less than ambassador's rank expelled or recalled Threats Presence or absence of military action Wars Troop movements Mobilizations Accusations Deaths from all forms of foreign conflict behavior Note: Unless otherwise noted, the indicators refer to absolute frequencies.

is representative but not exhaustive. The first stage consists of reliability and regional studies conducted prior to 1973. In that year, Wilkenfeld's Conflict Behavior and Linkage Politics was published. Since this study had a profound impact upon the direction of research, scholarship after that time is reviewed separately. Like their predecessors - the factor-analytic research designs of Rummel and Tanter — the reliability studies failed to reveal a general linkage between foreign and domestic conflict, although some specific connections were discovered. Phillips found that under conditions of domestic strife, modernized states which had experienced political instability tended to "overreact" to conflict encountered. In other words, they responded to foreign hostility in a more extreme fashion than otherwise would be expected.8 However, that result did

81 Conflict and Cohesion

not constitute evidence that warfare would be initiated by a state in response to internal disruption. Hazlewood derived composite yearly measurements of domestic and foreign strife for the years 1964 and 1965, and used these data to test three models which predicted foreign conflict. He based each model on a different kind of internal conflict, with mass protest, elite instability, and structural war (meaning armed violence on a massive scale) comprising the categories of domestic strife. Hazlewood incorporated curvilinear relationships into his models, in contrast to the usual practice of associating the two levels in an incremental fashion. By way of example, according to the model for elite instability, foreign conflict sent by a state is expected to reach a maximum under moderately unstable conditions. Hazlewood reasoned that to a certain point, elite dissension could be diverted outward. However, as threats to the incumbent elite become more severe, the utility of conflict management through diversion slowly decreases. Ultimately, the elites and their factions become so divided that the use of diversion mechanisms is no longer feasible. Once this threshold has been crossed, an inverse relationship between extreme elite instability and foreign conflict is likely to occur.9

Hazlewood obtained mixed but generally unfavourable results from testing the three models. The hypothesized breakpoints (such as that for elite instability) found little support, although Hazlewood enjoyed some degree of success in predicting less severe types of foreign conflict.10 Once again, however, international warfare failed to show a connection with domestic strife over a wide field of observation. Through their research, Wilkenfeld and Zinnes obtained two insights about conflict linkage. First, they confirmed Rummel's findings, and did so with Markov analysis, an innovative statistical technique. Since their study did not rely upon factor analysis alone as the means of pattern recognition, this improved the credibility of the existing body of research through a different method of testing. Second, Wilkenfeld and Zinnes demonstrated that linkages could be discerned under particular sets of conditions, and this had the effect of promoting interest in the role of mediating variables. (In their research, Rummel and Tanter had failed to allow for the possibility that outside factors might intervene.) Specifically, they found that certain types of domestic strife predicted foreign conflict successfully, although not for all states. Wilkenfeld (and later Zinnes and Wilkenfeld) classified members of the international polity according to type of government, the categories being Polyarchic (Western, industrialized democracies), Centrist (centralized regimes), and Per-

82

Crisis and War

sonalist (dictatorships based on an individual personality). In the presence of this intervening variable, some specific connections emerged. For example, in the case of Polyarchic states, external belligerence could be traced to the presence of internal warfare.11 But even these results could not be cited as evidence that war as a distinct event can be associated with the prior occurrence of domestic conflict. Following a different approach, some members of the research community hypothesized that linkages would have a regional basis. Two studies focused on the Middle East. Burrowes and Spector monitored the foreign and domestic political behaviour of Syria from 1961 to 1967, in the aftermath of the breakup of the United Arab Republic. However, they discovered no connection between serious international conflict and internal strife for the newly-formed Syrian state.12 Wilkenfeld's study of the principal Middle Eastern states involved in the Arab-Israeli conflict had a similar outcome, revealing that foreign conflict in a given period predicted subsequent foreign conflict with considerable accuracy, while domestic strife (with some minor exceptions) demonstrated relatively little value in forecasting. Only in the case of Jordan did internal conflict have a notable impact on conflict sent to the outer environment.13 Research on other developing areas produced similar results. In his study of Africa, Collins discovered that violent foreign behaviour could not be explained in terms of prior domestic strife. He did find, however, that unofficial anti-foreign behaviour and complaints about foreign interference in internal affairs could be predicted from the level of domestic disorder.14 Onate examined the conflict behaviour of the People's Republic of China from 1950 to 1970, and his results suggest a minor relationship involving specific kinds of internal and external conflict.15 Like the other research efforts reviewed so far, these regional studies were able at best to reveal linkages only under highly specific conditions. These connections lacked the straightforward character that the theorists had been expecting. Taken together, the two branches of scholarship do not support the theory of externalization on a grand scale, meaning the association of serious international conflict with domestic strife. But this lack of a general linkage does not mean that the early years of the research program should be viewed as an era of failure. To discover that something may not exist is no less difficult an enterprise than finding covariation among variables, and often no less important. The "first wave" of quantitative scholarship culminated in the appearance of Wilkenfeld's Conflict Behavior and Linkage Politics (of

83 Conflict and Cohesion

which some portions already have been reviewed). No comparable summary of research has appeared since then, although a number of extensive reassessments have been published before and after.16 Some general observations should be made regarding the Wilkenfeld collection and the studies which appear prior to its publication, in order to set the stage for the second phase of the review. First, and foremost, the studies did not support the concept of a uniform linkage involving internal and external conflict, although some scholars found specific connections. The doubts raised by Rummel and Tanter received overwhelming reinforcement, and the collective consciousness among scholars changed with respect to externalization. A linkage had previously been assumed to exist between foreign and domestic conflict, and Wilkenfeld's authoritative compilation of research dramatically altered the conventional wisdom on that subject. Second, there is no reason to believe that the results generated by the first wave were a product of bias toward either a cross-sectional or longitudinal approach. The research program included several examples of each type of study. Third, the findings suggest that efforts to link absolute levels of foreign and domestic conflict may have been misplaced. Perhaps elites are not uniform in their sensitivity to internal strife, meaning that the nexus of external and internal conflict will change from one state to the next. Fourth and finally, the "first wave" of quantitative analysis offers implicit support to the position that factors which operate beyond the borders of a nation-state will affect its propensity to externalize conflict. The initial wave of research, catalyzed by Rummel, left the theory of conflict linkage in a problematic state. Since that time activity has declined precipitously. A "second wave" has yet to develop, although there have been some innovative studies in the wake of Wilkenfeld's compilation. Interest has shifted to societal factors such as militarization and population density, along with environmental constraints on the projection of conflict. The more recent studies suggest some interesting directions to follow in crisis-based theorizing about externalization. THE AFTERMATH

Kegley and his associates conducted an analysis of the foreign and domestic conflict behaviour of 73 states in the system during the 19605. They used Gurr's well-known index of civil strife to measure internal conflict, and relied upon variables from the World Event Interaction Survey (WEIS) on the external side. The Kegley group

84 Crisis and War

found no correlation when they lagged foreign conflict behind that which occurred at the domestic level (r = .02), once again failing to support the general hypothesis. However, they introduced an intervening variable (the degree of militarization of a society, measured through its level of military expenditures expressed as a percentage of GNP) and obtained an intriguing result. For highly militarized states, foreign conflict showed a negative association with prior domestic conflict (r = —.49). By contrast, less militarized countries exhibited no linkage whatsoever. The obvious explanation for these different patterns is that the government in a militarized society may be in a better position to restore order through coercion of its own population, as opposed to relying upon a potentially dangerous foreign war to do so. It is reasonable to assume that such regimes also are less likely to be inhibited about the use of force against their own citizens.17 Vincent attempted to replicate the study by Kegley and his associates, and in the process claimed to have overturned their results. For the years 1963—67, Vincent factor-analyzed attribute data for 128 states and obtained two dimensions of domestic conflict, labeled Turmoil i and 2. The first factor incorporated such diverse events as political assassinations and constitutional changes, while the second ranged from riots to executive adjustments, in all cases expressed as mean frequencies. Vincent discovered that Turmoil 2 correlated significantly (p < .01) with external conflict (also based upon events data). He also found that the — .49 correlation (from the Kegley study) between internal and external conflict for highly militarized states had disappeared.18 However, as pointed out by Richardson and Kegley, these results should be viewed with scepticism. In a larger factor-analytic study by Vincent, Turmoil i and 2 had ranked third and ninth in explaining the variance in domestic conflict, with 4.2 and 2.5 percent, respectively. Thus it is reasonable to doubt, as did Richardson and Kegley, whether Turmoil 2 has any practical relevance, and to observe also that any statistical analysis incorporating Turmoil 2 as a variable "would, in turn, seem to be of dubious credibility."19 On grounds of method, the results obtained by Kegley and his associates are more convincing. With others, Eberwein conducted a cross-sectional analysis of internal and external conflict for 125 nations over the period 1966— 67. This study relied upon the usual sources of data (such as WEIS) and featured a standard set of variables at both levels. The Eberwein team factor-analyzed its data and found a significant correlation between dimensions of foreign and domestic conflict. However, the linkage virtually disappeared when size of population was introduced

85 Conflict and Cohesion

as an intervening variable. The initial correlation had been spurious, with larger states simply having more events at both the internal and external levels.20 Wilkenfeld and his associates assessed the impact of a set of societal factors as part of a more comprehensive model of foreign policy behaviour. Societal attributes included governmental instability, social unrest, the level of economic performance, and the population growth rate. The evidence suggested that these domestic factors had a "modest" effect on foreign policy, while behaviour received from abroad served as a much better predictor of behaviour sent.21 Indirectly, the research of Wilkenfeld and his associates reinforced the argument that externalization should be examined in a more general context that includes the prior foreign experiences of the nationstate. With others, Bremer tested an intriguing variation of the domesticforeign conflict linkage. To be more specific, this research group evaluated the "crowding-combat" hypothesis, which holds that conflictual behaviour at the international level is a product of Malthusian population pressure. Bremer and his colleagues used cow Project data on foreign conflict and demographic change over the period 1816—1965 to test this proposition, and found no support for the notion that international violence is a function of objective (or even subjective) crowding within the nation-state.22 Their results suggest that long-term changes at the domestic level (such as population growth) may not be able to explain foreign conflict any better than short-term, behavioural indicators, at least in isolation. More recently, Ward and Widmaier have provided a further clue about why the pursuit of conflict abroad as a means of restoring domestic tranquility may be less than automatic. They observed that some historical context for interstate conflict probably must exist "if a decision-maker were to think about mobilizing it to help alleviate political instability at home." Furthermore, according to Ward and Widmaier, the decision to initiate a war would require some expectation of success. On the strength of these precautionary observations, they inferred that domestic and foreign conflict would not necessarily have a uniform connection, and their findings are consistent with that line of reasoning. Ward and Widmaier charted protests and civil war over the period 1948-76 and found no general connection between these events and the frequency of serious international disputes. However, for larger powers in the system some covariation could be observed between internal and external conflict. Taken together, the results suggest that more capable states might be able to pursue external conflict in order to restore internal cone-

86 Crisis and War

sion, while such policies would be self-destructive for militarily weak states.23 THEORETICAL

ARTICULATION

At first glance, the implications of research to date are rather bleak for the externalization theory. The findings appear to suggest that creative energies ought to be directed elsewhere. But a review of the conceptual basis of the research program will produce a different recommendation altogether. To be sure, the first wave of scholarship demonstrated that something probably did not exist. But that something may not have been the process of externalization. One fundamental point of contention with the regime established by the first wave of testing concerns the static conception of conflict. To begin, it is not clear that the impressionistic writings of Simmel and other theorists implied that there ought to be a mechanistic connection involving high absolute levels of foreign and domestic strife. Some countries experience internal unrest quite frequently, and their elites may not react automatically by trying to "export" this conflict. For example, although the current strife in El Salvador is unusual even by Latin American standards, the leadership of that country may have little inclination - or even capability - to seek a solution beyond its own borders. By contrast, the elite of an unusually stable system member may feel threatened even by a very low level of conflict, if it represents a significant departure from the societal norm. Having said this, an initial restructuring of the general hypothesis would be as follows: A significant increase in the internal conflict experienced by a given state may result in a significant increase in the foreign conflict behaviour of that state. Change in foreign policy, according to this dissenting viewpoint, is a product of evolving domestic reality, although naturally it will depend on international factors also. As Sullivan comments with respect to research on population dynamics, "theoretically the idea that changes in attributes will affect behavior is more pleasing than the simple, gross attribute-behavior linkage because it views national systems as changing, living entities."24 Absolute levels of conflict or cooperation are likely to be devoid of implications for subsequent behaviour, unless notable by comparison to some meaningful standard.25 Fragmentation of conflict is a second conceptual problem that has plagued systematic research from the beginning. Is it reasonable to expect that isolated manifestations of conflict at the international level will be linked to highly specific forms of domestic strife? Why

87 Conflict and Cohesion

should events such as politically-motivated strikes be useful in predicting the withdrawal of ambassadors or other singular actions? Among those who have engaged in aggregated testing, the norm has been either to concentrate on such fragmentary linkages directly, or to rely upon statistical techniques (invariably meaning factor analysis) to demarcate classes of conflict behaviour. Of course, aggregation of action-categories through an inductive method can be a useful exercise if the goal is to identify dimensions of behaviour at the two levels in question. Rummel saw descriptive analysis along those lines as the highest priority, and argued that theoretical relevance could not be used to select measurements of domestic and foreign conflict, since "comprehensive theories of conflict capable of test do not yet exist."26 But is it likely that such theories will be discovered through an approach that almost completely eschews deductive reasoning? An effort to correlate dimensions of conflict across the foreign and domestic levels is misplaced without some convincing reasons why they should be connected to each other in the first place. Consider the breadth and depth of external conflict that might be necessary for a regime to obtain renewed support from its citizens. Among others, Barbera has referred to the "sustained massive impact" of war on the nation-state, with the changes involved having dramatic implications for the domestic political process. In time of war, governments as a rule are able to achieve greater centralization of power. A regime at war also can expect higher levels of cooperation and conformity from its population, at least initially.27 It is less reasonable to expect that loyalty will be restored to a government that is engaged in a lesser conflict beyond its borders. If the objective of an elite is to restore (or preserve) order at home, the process of externalization often should result in war or at least a serious confrontation with some other state. Many of the fragmentary linkages that have found a place in previous testing become implausible when viewed in such terms. Two central conceptual problems have been identified in the research program so far, each focusing on the meaning of conflict itself. One is that the sociologists who popularized the externalization theory probably never meant that the absolute degree of domestic conflict would be useful in forecasting an absolute level of foreign conflict. The other is that a general mapping of conflict at one level onto the other ought to be explored, as opposed to the fragmentary approach seen to date. There are, in addition, some other conceptual difficulties which merit attention. These are problems which have been confronted to some degree at various points during the second

88 Crisis and War

phase of research, though in all cases more questions were raised than answered. One criticism raised by Stohl in his review of the literature is that researchers have neglected the role of the state in its internal processes. Is government coercion of private citizens actually a component of domestic conflict which may contribute to later foreign strife? Or should repression by a regime be viewed instead as a substitute for externalization? Either way, to properly assess social disruption, it is essential to incorporate more fully the role of the government as a participant within the state.28 Kegley and his research team provided support for that position when they found that militarization seems to have an impact on the projection of conflict abroad. Their discovery makes it all the more urgent to assess the effects of government behaviour in explicit terms. Along with Bueno de Mesquita, Zinnes has isolated a further conceptual problem, although each describes it differently. Zinnes made the following generalization about research derived from the second image of Waltz: Neither slow-changing nor fast-changing attributes of nations by themselves appear to be directly related to a nation's war behavior. When considered in combination, however, these attributes can predict some violent behavior of nations.... the environment is a critical ingredient for understanding the violent activities of states.29

This observation is true especially with regard to externalization. "The domestic/foreign conflict proposition," as Bueno de Mesquita has noted, "fails to isolate those circumstances in which generating conflict is likely to distract attention from domestic unrest from those in which foreign conflict might exacerbate such unrest."30 Together, Zinnes and Bueno de Mesquita have pointed out that environmental constraints on externalization are not thoroughly understood at present.31 While very little can be said with confidence about the impact of the international environment, existing research provides some clues. First of all, there is no reason for a beleagured elite to look "instinctively" beyond its own borders for salvation. Its members are not lemmings and cannot be expected to choose an external solution unless it holds some promise of success. This is implicit in the results obtained by Ward and Widmaier, because stronger states have a greater tendency to engage in what appears to be externalization. Incorporating the environment in more explicit terms, Bueno de Mesquita has proposed a cost-benefit analysis to explain decision

89 Conflict and Cohesion

making by an elite under domestic pressure. He begins by assuming that either a domestic or a foreign strategy could be selected by the leadership, and that each is associated with benefits, costs of implementation, and a probability of success. The expected values of the strategies are to be calculated as follows:

where E(F) = E(D) = B = B' = P = Q = C = K =

expected value of a foreign strategy, F; expected value of a domestic strategy, D; benefits from success of F or D; benefits from failure of F or D; probability of success for F; probability of success for D; costs of F; and costs of D.

The choice of externalization would require that E(F) ^ E(D). This is a step forward in explaining the "poor empirical fit" of previous studies, because E(D) may have been greater than E(F) in many cases.32 The equations developed by Bueno de Mesquita can be improved through a minor modification. The benefits from the strategies should not be treated as equal. The benefits derived from a successful implementation of a foreign strategy would include concessions from the enemy which could not be obtained from a domestic response.33 In the latter case, restoration of domestic order would be the principal benefit. When the strategies fail, the benefits will be unequal, too. The "benefits" from failure under strategy F might include destruction of the government itself, as opposed to continued existence of the elite in some form under failure of strategy D. In sum, it would appear that Equations 4.1 and 4.2 ought to be reformulated as follows:

where Bf = the benefits from successful pursuit of F; B'f = the benefits from a failed pursuit of F;

go Crisis and War

C = the Ba = the B'a — the K = the

costs associated with F; benefits from successful pursuit of D; benefits from a failed pursuit of D; and costs associated with D.

These equations effectively summarize the concerns about environmental factors. The choice of projection will be influenced by the "fringe benefits" and risks associated with foreign conflict, in addition to domestic considerations. At the stage of operationalization, the logic underlying Equations 4.3 and 4.4 will play an important role. Cross-national comparison of conflict is a conceptual difficulty which critics are fond of noting but, sadly, not of solving.34 They have contended for years that the absolute frequency of events is not a commensurate indicator of conflict from one state to the next. For example, is a given number of political strikes in one country necessarily equivalent to the same number in another as a measurement of internal strife? The "weight" of an event will differ from one country to the next, if for no other reason than diversity in population, as discovered by the team headed by Eberwein. A focus on change (as opposed to absolute value) in event frequencies provides a starting point in dealing with this problem. Examining conflict change in proportional terms is one approach that would attenuate the bias inherent in size differences. However, this conception is too extreme and creates a bias in the opposite direction. Consider as an illustration the case of two hypothetical states, one small (X) and the other large (Y). In year "t" state X receives a frequency score of 3 for domestic conflict, and this score rises to 6 the next year, while state Y has scores of 500 and 550 respectively over the two years in question. For X, the transformed domestic conflict score would be 100 percent, while for Y it would be 10 percent. Even though Y experienced far more manifestations of conflict at home in each year (and even a greater absolute increase), the change experienced by X over the same period is more dramatic on a percentage basis. These extreme results suggest that some role should be preserved for the absolute level of conflict. Rather than using percentages, the absolute changes could be exposed to a logarithmic transformation:

where C

= conflict change;

gi

Conflict and Cohesion

Ct = conflict in year t; and C t _! = conflict in year t— i. If Ct — C t _! = o then C = o also, which explains the increment of one unit in the logarithmic expression. The transformation itself reduces high absolute frequencies, thus taking cross-national differences into account in a manner less extreme than a percentage score. For states X and Y in the previous example, the scores for C would be 0.60 and 1.71 respectively. If the increases had been from 3 to 9 and from 500 to 600, the corresponding scores would be 0.85 and 2.00. The disparity in the C-scores for X and Y has therefore increased marginally from one case to the next. By contrast, a percentage-based conception would yield scores of 300 percent for X and 20 percent for Y, an outcome with little or no face validity. One further aspect of Equation 4.5 requires an explanation. The sign of the expression for C will depend on whether Ct — C t _! is greater or less than zero. In sum, the use of an expression such as Equation 4.5 will facilitate an operationalization which incorporates both absolute and relative change. CONCLUSION

Several conceptual problems have been identified in the preceding review. First, rigorous evaluation of the theory of externalization demands a dynamic as opposed to static conception of conflict. Second, attempts to connect fragmentary manifestations of internal and external conflict have been misplaced. Third, environmental constraints on the appearance of the expected linkage are understood poorly at present. Fourth, the role of the state in domestic conflict has been neglected in efforts to test the theory. Fifth and finally, in past studies the cross-national meaning of conflict has been problematic, primarily because indicators have been based on absolute frequencies. For researchers, the experience with externalization has been much like that of the fabled blind men with the elephant. The primary cause of failure rests in the scope of the investigation. Preconceived notions have led to an excessive concern with individual manifestations of conflict, while research on the environment suitable for projection has suffered from neglect. Some highly specific patterns linking domestic and foreign strife have been discovered, but understanding of fundamental processes of change has remained elusive. The next logical step is to ope rationalize and test a theory of the conflict nexus that is consistent with the concerns expressed in the preceding review.

CHAPTER FIVE

Testing Externalization

Now that some basic conceptual difficulties have been resolved, a theory of the conflict nexus can be described and explained. The state as an actor will project its internal conflict outward under specific conditions. The regime must be threatened by a deterioration of the domestic setting. Furthermore, the changes referred to must include an extensive range of developments, as opposed to an isolated increase in some type of internal conflict. As will become apparent, disintegration can either take the form of observed behaviour or be latent within the society. Pursuit of a foreign strategy to restore cohesion will then be constrained by the external environment. In other words, when both the motive and the opportunity exist, projection can be expected to occur. Based on the preceding line of reasoning, two propositions will be tested. The first (and more unconventional of the two) links internal conflict and an environmental constraint to the escalation of crises: Proposition 5.1: When the initiator of a crisis has experienced an increase in latent conflict and also has positive expected utility relative to the defender in a bilateral conflict, war is more likely to occur than in cases which lack one or both of these characteristics.

Initially, this notion of latent conflict may seem a little mysterious. But some potentially important sources of domestic strife may not have visible referents. For example, a "national malaise," to borrow a phrase popularized in recent years, might be hard to discern using behavioural indicators, but still could appear dangerous in the eyes of a ruling elite. Naturally, this path of reasoning might provoke a response of the following sort: latent conflict is irrelevant to the study of externali-

93 Testing Externalization

zation. Dissent, if it really is serious, will appear in some open form of expression. Its manifestation ultimately must be political. But this is precisely where the behavioural bias inherent in research on externalization comes into play. There is no reason to assume that groups dissatisfied with the performance of a government — or anything else - will combine readily in order to further their objectives. There are costs associated with organizing such factions. Many societies may contain "latent" groups which have failed to crystallize because their potential constituents lack the incentive as individuals to create an organizational structure. In contrast to the costs referred to a moment ago, the benefits which accrue from coordinated action are distributed generally. Individuals under conditions of strategic interaction may be slow to combine their efforts, because they hope that others will assume the burdens of organization.1 An elite therefore could perceive danger in the absence of behavioural manifestations of domestic unrest if it believed organized opposition to be developing. In other words, national leaders might have a pre-emptive perspective on domestic disorder. Crisis initiators which have experienced unfavourable trends in latent conflict therefore might have an incentive to pursue escalation in order to promote domestic cohesion, subject to an environmental constraint. As already noted, escalation of a crisis to war requires both a motive and an opportunity to act. Since decision makers are not assumed to be suicidal, even a disintegrating domestic situation would not be sufficient to make war a viable option. Non-negative bilateral expected value distinguishes opportunity in the most straightforward manner that is feasible. War cannot be a rational strategy to maintain domestic cohesion unless the leadership expects to survive its struggle with the defending state. The bilateral expected value criterion also increases the degree of confidence that, in Equations 4.3 and 4.4, E(F) will be greater than E(D). Under such conditions, the foreign strategy of war will promise gains over and above preservation of internal stability, the lone benefit from a domestic response. The other proposition to be tested focuses on manifest or explicitly behavioural conflict: Proposition 5.2: When the initiator of a crisis has experienced an increase in manifest conflict and also has positive expected utility relative to the defender in a bilateral conflict, war is more likely to occur than in cases which lack one or both of these characteristics.

Internal disruption, as manifested through an increase in forms of behaviour threatening to a regime, provides a more obvious mo-

94 Crisis and War

tive for crisis escalation. Whether an opportunity is judged to exist will depend upon the expected value of a bilateral war with the defender. MEASURING

DOMESTIC CONFLICT

The most complex problem of operationalization is the measurement of conflict itself. A common practice among those concerned with externalization has been to construct indices, and the behavioural indicators used by Rummel are representative in that sense (see Table 4.1). Examples on the domestic scene range from mass political events such as general strikes to highly specific occurrences like assassinations. The manifestations of foreign strife are varied also; they extend across categories as diverse as anti-foreign demonstrations and deaths from foreign conflicts. Moreover, some of the action categories contain violence, while others do not. It is fair to say that Rummel and those who followed in his footsteps have used a wide range of indicators in their investigations. Despite their variety, these lineaments of conflict share at least one other common trait: each could be described as behavioural. Even the researchers who tried to identify categories of strife - through factor analysis or other means - have referred to dimensions which represent behavioural manifestations of conflict, such as "turmoil" and "internal war." The search for indicators has been biased toward what can readily be observed. It thus will be necessary to look elsewhere for indicators of latent conflict. Latent Conflict

Even if it is granted that latent conflict at the domestic level may affect externalization, as in Proposition 5.1, how can something like that be discerned? This situation is made no easier by the fact that dynamic considerations are viewed as crucial to the measurement of conflict. To pose this problem somewhat differently, how are changes in something latent to be measured? Despite such obstacles, the search for operational referents may not be quixotic after all. One logical measurement strategy is to focus on changing economic conditions, because deterioration in that area should reduce the effective political capital of any regime. Facing economic problems, an otherwise stable government might engage in political manipulation of the economy, or even foreign adventure, in an effort to counteract trends which it perceived to be dangerous to social cohesion. In other words, the elite would be trying to prevent

95 Testing Externalization Table 5.1 Tufte's Economic Indicators 1 Real disposable income: The per capita disposable income, adjusted for the consumer price index. 2 Employment rate: The percentage of the labour force wholly or partially employed, excluding those laid off. 3 Inflation: Based on expansion of the money supply, including money and reserve money. Source: United Nations, Statistical Office, Statistical Yearbook, 1979/1980 (New York: United Nations 1981).

the further organization of groups which had been expecting more than the government delivered to them. Tufte has produced evidence suggesting that political maneuvering of the economy may be a fact of life in more than a few democratic regimes. 2 His findings associate fluctuations in certain macroeconomic indicators, listed in Table 5.1, with the timing of elections. Table 5.2 compares election to non-election years from 1961 through 1972 for a number of countries, the criterion of interest being the change in real disposable income. Tufte notes that Combining all the experience of the 27 countries over the period 19611972 reveals that real disposable income growth accelerated in 64 percent of all election years (N = 90) compared to 49 percent of all the years without elections (N = 2O5).... for those 19 countries whose economies ran faster than usual in election years, the effect was substantial: real disposable income growth accelerated in 77 percent of election years compared with 46 percent of years without elections.3

Of course, these aggregate findings do not establish the presence of a political business cycle in the 19 countries concerned. More evidence - like pertinent information about the political manipulation of economic policy in each country — would be required to fulfill that objective. But the fundamental point to be discerned from the aggregate data, as Tufte correctly observed, is that "70 percent of the countries showed some signs of a political business cycle."4 Like real disposable income, unemployment and inflation also appear responsive to the timing of elections. Nordhaus monitored rates of employment in Germany, New Zealand, the United States, France, and Sweden, and found that "unemployment has tended to reach a low point around election time," although the strength of the association varied from one country to the next.5 Focusing on the tradeoff between inflation and unemployment in the United States, Tufte

96 Crisis and War Table 5.2 Elections and Economic Acceleration. 27 Democracies, 1961-1972 Percentage of years in which rate of growth of real disposable income increased

Australia Austria Belgium Canada Chile Costa Rica Denmark Finland France Germany Iceland India Ireland Israel Italy Jamaica Japan Luxembourg Netherlands New Zealand Norway Philippines Sweden Switzerland U.K. U.S.A. Uruguay

Election years (%)

75 25 67 100 50 100 25 67 60 33 33 50 67 67 33 100 100 100 50 75 100 60 67 67 67 83 33

N

Years without elections (%)

N

4 4 3 5 2 2 4 3 5 3 3 2 3 3 3 2 4 2 4 4 2 5 3 3 3 6 3

29 86 63 57 44 50 43 50 33 38 75 43 63 50 50 44 29 56 57 43 33 67 50 50 38 40 50

7 7 8 7 9 8 7 8 6 8 8 7 8 8 8 9 7 9 7 7 9 6 8 8 8 5 8

Did acceleration in real income growth occur more often in election years compared to years without an election?

Yes No Yes Yes Yes No No Yes Yes No No Yes Yes Yes No Yes Yes Yes No Yes Yes No

yes

Yes Yes Yes No

Source: Edward Tufte, Political Control of the Economy (Princeton, New Jersey: Princeton University Press 1978), 12.

97 Testing Externalization Table 5.3 Inflation, Unemployment, and Presidential Elections, 1946—1976 Yearly change in unemployment rate and inflation (real GNP deflator): Less unemployment and less inflation Less unemployment but more inflation Less inflation and more unemployment More inflation and more unemployment

Presidential election years (%)

All other years (%)

50 13 38 0

9 30 43 17

101

99 (23)

(8)

Note: Percentages do not equal 100 because of rounding. Source: Edward Tufte, Political Control of the Economy (Princeton, New Jersey: Princeton University Press 1978), 22.

discovered a remarkable difference between presidential election years and all others. Table 5.3 shows that unemployment and inflation declined in 50 percent of presidential election years, while this happy state of affairs occurred in only 9 percent of the other years. Perhaps even more striking was the finding that in none of the presidential years did inflation and unemployment both increase. If anything, the fact that many of these results are from the United States may bias the effects of political manipulation downward. Elections there are held at fixed intervals. Interventions, therefore, will be more predictable and vulnerable to accusations of political meddling. In democracies that allow the party in power to choose when it will go to the polls, blatant acts of manipulation may be quite feasible. (Israeli election behaviour from a few years ago is a good example of this.) At a more general level, the us results provide further evidence that democratic governments are not above politically-motivated intervention in their economies. But should a government concerned with its political standing bother to manipulate the economy? The motive to intervene appears to be real. Citizens care about economic performance, although the manner in which concern is expressed varies somewhat from one country to the next. Most of the available evidence pertains to OECD countries. In the case of New Zealand, Ursprung discovered a linkage between party standing and indicators like inflation, unemployment and balance of payments.6 Paldam surveyed OECD member nations and found that inflation, national income, and unemployment collectively accounted for 50 percent of the variance in governmental popularity

98 Crisis and War

ratings. But more specific studies focusing on the United Kingdom and France uncovered different patterns. Alt and Chrystal discovered a covariation between inflation, unemployment and balance of payments on one hand, and the popularity of British ruling parties on the other. By contrast, Frey and Schneider found that unemployment and real disposable income levels could be used to predict the popularity of British governments, while inflation had negligible effects. The French findings are different yet again. Hibbs and Vasilatos compared economic conditions to mass support levels for two recent presidents of France, and discovered that the political fortunes of Pompidou and Giscard d'Estaing could be linked to the effects of prices on per capita real personal income.7 While there is considerable variation by country, a connection between economic conditions and the political standing of regimes is consistently evident. However, the review is biased because comparable information is difficult to obtain for developing countries, many of which have radically different political systems by comparison to the countries surveyed. Is it tenable, then, to argue that all regimes will have the same incentives to engage in political manipulation of the economy? Even those elites which are not responsible to election polls must accumulate some political capital in order to govern. Widespread malaise, sabotage, vulnerability to foreign threats, and even revolution may await the government which ignores the wishes of its population. A dictatorship, to consider the most extreme example, may have an incentive to bribe its citizens once in a while in order to preserve power or achieve some other goal. Evidence suggests that in the Soviet Union, for example, leadership succession has led to politically motivated changes in economic policies. To expand their influence, new Soviet leaders have instituted agricultural and other programs designed to promote their personal popularity.8 In sum, although the internal workings of authoritarian regimes are difficult to monitor, it will be assumed that virtually all elites engage in manipulation of economic processes in order to enhance their political position. So far, political intervention has been discussed in the context of macroeconomic "fine-tuning." Economic problems perceived as being more serious may elicit more extreme responses from concerned elites. One possibility is that the potential for escalation of a crisis to war would be regarded as an opportunity rather than a danger. A government at war can extend its control over society to an extent that is difficult to achieve in time of peace. Suspension of civil liberties may be justified as a response to the danger of subversion, perceived

99 Testing Externalization

or real. Action then can be taken against potential assailants from within. Authoritarian regimes already will have such powers to a considerable degree, but even they can benefit from a heightened national consciousness. Furthermore, even citizens who are unhappy with a government will be likely to favour its continued existence over foreign occupation. War also offers a short-term boost to employment, along with economic activity in more general terms. A successful war may even result in larger markets for domestic produce and increased access to raw materials and resources.9 War is not, however, the only response available to an elite in crisis that also is troubled by deteriorating economic conditions. Even at the brink of armed conflict, leaders still may opt for a domestic solution to internal problems. Rather than escalating an international dispute, those in power could try to counteract disruption by removing the potential agents of domestic strife. Arrests and other forms of persecution might be employed toward that end. Although such activities are associated more commonly with authoritarian governments, they are not without precedent in democratic states. Under conditions of economic deterioration at home and crisis abroad, domestic repression and foreign conflict could be viewed as potential substitutes for each other. Latent conflict available for export will be the amount that has not been "absorbed" by a pre-emptive domestic response by a regime. The breadth of latent conflict is conveyed by Equation 5.1:

where change in breadth of latent conflict for i; consumer price index in the year of the crisis, t; consumer price index in the preceding year, t — i; number of governmental sanctions in the crisis year; and g t _ j = number of governmental sanctions in the previous year. lc ct c t _! gt

= = = =

Changes in the level of employment measure the depth of latent conflict. When lc is positive and unemployment has been stable or increasing, latent conflict is said to be on the rise. Along with its structure, the specific variables which appear in Equation 5.1 - the consumer price index and number of governmental sanctions for

ioo Crisis and War

the crisis initiator — will require some explanation. So, too, will the choice of unemployment to measure the depth of latent conflict. There are three reasons why the consumer price index has been selected to measure the effects of economic deterioration on the initiator. The first and most mundane reason is the availability of data. Price figures for some African and Asian countries are scarce, but these missing data problems are minor when compared to those encountered with composite indicators such as real disposable income, which Tufte and others employed in their analysis. The latter requires that population figures, national income, and inflation rates be combined into a single index. In effect, the use of real disposable income or an array of indicators would result in the virtual exclusion of Third World cases from the analysis, thereby limiting the relevance of any findings about crisis escalation. Another point in favour of the price index as an indicator of latent conflict is that it has an impact upon virtually the entire population of a country, because only the very wealthy are immune to its effects. Unlike the unemployment rate or the level of expenditures on transfer payments, consumer prices provide information about the wellbeing of the general public. Consumer price changes, therefore, can be used to estimate the breadth of latent conflict in a society. Prices convey at least one further advantage as a measurement option. Scholarship on economic factors and their connection to the political process is not inconsistent with the use of consumer prices to assess regime standing. Real disposable income and inflation levels — each of which is linked directly to the consumer price index — have been associated with the popularity of governments and their propensity to intervene in economic matters for political reasons. While the evidence is not entirely clear on this subject, consumer prices are a viable choice for measurement on the basis of current knowledge. One criticism that might be made of the price index is that its computation varies from one country to the next. This objection can be dismissed, however, since the variability of calculations does not necessarily pose cross-national measurement problems. The components of price indices vary according to the internal conditions of each country. The index for the United Kingdom incorporates housing and transportation along with other goods and services while the index for India includes commodities such as cloth and rice.10 In both cases the various components reflect the needs of the respective populations. These components variations, therefore, do not reduce the capacity of the price index to provide cross-national comparisons of changing economic conditions at the domestic level. There is another potential problem associated with the use of

ioi Testing Externalization consumer price indices to measure economic changes. Changes in relative prices induce substitution effects. For example, "if tea has a price rise relative to coffee, rather than buying still the original bundle, the same utility would be obtained at less cost by sacrificing some tea and buying more coffee."11 A given index could be rendered obsolete by such changes. However, in the sources that have been consulted, price indices are revised periodically to reflect prevailing consumption patterns. The distortion caused by consumer replacement of commodities thus is expected to be minimal. Unemployment is a natural choice to measure the depth of latent conflict. Of all the interest groups in a society, the unemployed have the most obvious reason to disapprove of the government.12 A significant increase in employment, therefore, should enhance the political standing of a regime. If instead the rate of employment remains stable or declines, an increase in the breadth of latent conflict would have greater potential to worry those in power. Unemployment levels in various countries have also been linked with the popularity of governments and their tendencies to engage in macroeconomic intervention. While those out of work comprise only one potentially dangerous group, changes in "membership" can be measured with some reliability in this case. The threats posed by other dissatisfied elements, such as ethnic minorities, are much more difficult to appraise on a cross-national basis. This is true especially of the manner in which they are perceived by leaders. In sum, when theory is in its formative stages and data are not plentiful, face validity should be a primary consideration in the choice of indicators. Both consumer prices and unemployment are satisfactory on that basis, because each can be linked to economic conditions, which in turn affect the political capital of governments.13 Governmental sanctions will be used to measure domestic repression, as seen in Equation 5.1. As defined by the World Handbook of Political and Social Indicators, these sanctions are actions taken by the authorities to "neutralize, suppress, or eliminate a perceived threat to the security of the government, the regime, or the state itself."14 Political arrests and censorship would be typical of the activities suggested by this description. Governmental sanctions focus directly on acts of repression taken by the regime of the initiator, and this measurement carries the further advantage that data are available for virtually all system members. Turning to the structure of Equation 5.1, several observations are in order. For the initiator, the change in the level of governmental sanctions is subtracted from the change in prices, the difference

io2 Crisis and War

being the breadth of latent conflict which is "available" for projection outward. Since, on average, some states will have very high (or very low) consumer price indices or levels of governmental sanctions, the differentials have been converted to logarithms in order to reduce the impact of high absolute frequencies.15 Each differential also has been increased in absolute value by an increment of one, because the logarithm of a zero entry otherwise would be undefined. Thus an absolute value of one within a component would result in a score of zero, as it should. The sign of each logarithmic expression will depend upon the direction of the change within it. For example, if consumer prices have declined over the year preceding a crisis, the first expression would have a negative value. Change in the level of unemployment determines depth, the other component of latent conflict. When breadth is increasing and depth is stable or increasing, latent conflict is said to be on the rise. In other instances latent conflict is considered to be declining or stable. Consider a recent case which may correspond to the scenario of projected internal conflict: the Falkland Islands War of 1982. It is quite possible that the British and Argentine governments had similar motives for entering the crisis, each hoping to rebuild its domestic political capital at the expense of a foreign adversary. Lacking better sources of information, it is inadvisable to speculate further on the domestic scene in Argentina. However, in the case of Britain economic conditions had appeared to be deteriorating prior to the crisis, and the war in the South Atlantic could be viewed as a ploy by the Conservative Government to divert public attention away from home. Crisis conditions may have interacted with an underlying fear of open disruption, resulting in a foreign strategy designed to preserve domestic order through attainment of foreign gains. Rather than relying upon domestic sanctions to maintain stability, the regime had the option of a war with Argentina.16 Manifest Conflict There is an extensive record from which to draw when considering measurement strategies for overt strife. Data are available on an extensive range of political interactions, and theoretical relevance is the primary consideration under such circumstances. In separate investigations, both Gurr and Hibbs have concluded that there are two essential kinds of domestic conflict expressed in manifest terms: protest behaviour and civil war.17 Their research also has produced empirical support for the existence of these discrete dimensions, which have been used frequently in subsequent studies. One rigorous application is that of Ward and Widmaier, who used

103 Testing Externalization Table 5.4 Indicators of Manifest Conflict 1 Protest demonstrations: A protest demonstration is a nonviolent gathering of people organized for the announced purpose of protesting against a regime or government or one or more of its leaders; or against its ideology, policy, intended policy, or lack of policy; or against its previous action or intended action. 2 Political strikes: A political strike is a work stoppage by a body of industrial or service workers or a stoppage of normal academic life by students to protest a regime and its leaders' policies or actions. 3 Armed attacks: An armed attack is an act of violent political conflict carried out by (or on behalf of) an organized group with the objective of weakening or destroying the power exercised by another organized group. 4 Deaths from domestic violence: This records the number of persons reportedly killed in events of domestic political conflict. The data refer to numbers of casualties and not events in which deaths occur. Source: Charles Lewis Taylor and David A. Jodice, World Handbook of Political and Social Indicators, Vols. 1-11, 3rd ed. (New Haven: Yale University Press 1983), II: 19, 21, 29, 37, 43.

variables from the World Handbook to construct indices for each of the two dimensions identified by Hibbs and Gurr. To operationalize protest behaviour, they relied upon protest demonstrations and political strikes, while for civil war they used armed attacks and deaths resulting from domestic political conflict.18 These dimensions demonstrated a connection with external conflict for the more powerful states in the system. This positive finding increases confidence in the ability of protest behaviour and civil war to encompass factors that are relevant to externalization in the crisis environment. Operationalization of manifest conflict will be based on an aggregation of the carefully derived indicators from the Ward and Widmaier study. (Formal definitions of these indicators appear in Table 5.4.) Each component of the index will be structured in the manner described for protest demonstrations:

where ni!

= change in the number of protest demonstrations for i;

104 Crisis and War

mlt

= the number of protest demonstrations in the year of the crisis (t); and m i(t-i) — the number of protest demonstrations in the previous year (t — i). The other components are: m2 m3 m4

= change in the number of political strikes; = change in the number of armed attacks; and = change in the number of deaths from domestic violence.

Manifest conflict change mc thus will be indexed as shown below:

This index has the mathematical properties discussed in the context of Equation 5.1. It measures change across a wide range of behavioural indicators and takes into account cross-national variations in absolute levels of conflict. Of course, it is reasonable to ask why this index should be considered fundamentally different from the various factor-produced dimensions, which were criticized rather severely. The present index has the advantage of being sharply focused while, by comparison, the constituents of past indices usually have been rather amorphous.19 For example, it is easy to see that protest demonstrations and political strikes represent protest behaviour. The same cannot always be said for the components of the various factordeveloped dimensions, which are difficult to put in a larger context, at least for those unfamiliar with quantitative methods. Among the constituents of an index, positive and moderate levels of association are desirable. As Table 5.5 demonstrates, the components of the summation in Equation 5.3 have that property. The correlation coefficients range from r = 0.060, p < 0.503 to r = 0.699, P ^ o.ooi. Higher correlations would imply redundancy among the indicators. Negative values also would be inappropriate, because the components are supposed to be measuring different aspects of the same phenomenon. This completes the operationalization of domestic conflict. Yugoslavia's crisis with Italy over Trieste (1953) provides a concrete illustration of latent and manifest conflict change. Both countries held strong claim to the city of Trieste, which was demilitarized and divided into two zones at the end of World War 11. On 8 October

105 Testing Externalization Table 5.5 Correlation Matrix for Indicators of Manifest Conflict

mi mg

ms

W&4

.3564 p E(F) might hold true for the British Tories, whereas the inequality could be reversed for the Argentine junta at the time of crisis. 17 Rigorous derivations of these categories may be found in Gurr, Why Men Rebel, and Hibbs, Mass Political Violence. 18 See Ward and Widmaier, "The Domestic-International Conflict Nexus." Formal definitions for these indicators appear in Table 5.4. 19 The dimensions obtained by Vincent in "Internal and External Conflict" would be prominent examples of that problem. For a critique of Vincent's dimensions, see Richardson and Kegley, "Cumulation and the DomesticForeign Conflict Hypothesis." 20 ICB Project Case Summary no. 106. 21 Campbell, Successful Negotiation, 152. See also Novak, Trieste, 1941—1954. 22 Consult Scolnick, "An Appraisal of the Studies of the Linkage Between Domestic and International Conflict"; Mack, "Numbers Are Not Enough"; and Dougherty and Pfaltzgraff, Contending Theories. 23 Taylor and Jodice, World Handbook, and Brecher, Wilkenfeld, and Moser, Handbook of International Crises. Each ICB Project Summary includes a bibliography of primary and secondary sources. 24 See Taylor and Jodice, World Handbook. 25 Brines, The Indo-Pakistani Conflict, 215. 26 The analysis of conditions within Egypt is based on consultations with area experts. 27 Due to missing data, there are 118 cases in Table 5.6, which makes it difficult to compare with Table 5.8. A comparison between Tables 5.7 and 5.8 is appropriate because the N is virtually the same in each instance. CHAPTER SIX

1 Dahl, Modern Political Analysis, 41. Dahl's definition may be viewed as an adaptation of one presented in Lasswell and Kaplan, Power and Society,

75-

2 Cline, World Power Assessment 1977, 6—7. 3 Keohane and Nye, Power and Interdependence, 11. For other definitions in the same vein, see Blau, Exchange and Power in Social Life, 117; Dean

172

Notes to pages 117-121

and Vasquez, "From Power Politics to Issue Politics," 13; Hirschman, National Power and the Structure of Foreign Trade, 13; Organski and Kugler, The War Ledger, 5; and Stoessinger, The Might of Nations, 27. 4 For a few relevant examples, consult Cable, Gunboat Diplomacy, 1919— 7979; George and Smoke, Deterrence in American Foreign Policy; Morgan, Deterrence: A Conceptual Analysis; and Schelling, The Strategy of Conflict. 5 Sullivan, Theories and Evidence, 160. Among others who have identified the same dichotomy are Baldwin, "Power Analysis and World Politics," 170; and Knorr, Power and Wealth, 13—14. 6 Singer and Small, "The Composition and Status Ordering of the International System," 237. 7 Morgenthau, Politics Among Nations, 186. For a discussion of ideology as a component of power in the context of the US-Soviet balance, consult Cherne, "Ideology and the Balance of Power." 8 Among the many analyses of balance of power which have treated power as an attribute are those of Kaplan, System and Process in International Politics; Liska, Quest for Equilibrium; Newman, The Balance of Power in the Interwar Years; Penrose, The Revolution in International Relations; and Schrodt, "Richardson's N-Nation Model and the Balance of Power." 9 Beloff, The Balance of Power, 4. 10 Hermann, "Some Issues in the Study of International Crisis," in Hermann, International Crises, 13. 11 Baldwin, "Power Analysis," 165. See also Knorr, Power and Wealth, 13— 15; and Hart, "Three Approaches to the Measurement of Power in International Relations," 290. 12 Keohane and Nye, Power and Interdependence, 24—5. 13 Johnsson, "Bargaining Power," 250. 14 Keohane and Nye, Power and Interdependence, passim. 15 Conversion will be addressed during the discussion of indicators of power in Chapter 7. 16 Nagel, The Descriptive Analysis of Power, 16. 17 Luttwak, The Grand Strategy of the Roman Empire, 197. 18 Nagel, The Descriptive Analysis of Power, 24. It could be observed that sometimes the actor concerned will have assessed the preferences of the more powerful state incorrectly. Thus the observed change in behaviour might be something that the stronger state might not have preferred, leading to an absurd attribution of power. This contingency is ruled out by the requirement that the behavioural change must occur in a direction preferred by the state wielding power. 19 Newman, Balance of Power, 8—9. 20 Claude, Power and International Relations, 13-39. 21 In the domain of international crisis, six components of an operating system have been described in detail by Brecher and James. These con-

173 Notes to pages 122—7

22 23 24 25 26 27 28 29

30 31 32 33 34 35 36 37

38 39 40

stituents collectively encompass the structure and process of a dynamic system: (i) the number of actors in the crisis; (2) extent of superpower involvement; (3) geostrategic salience of the location; (4) heterogeneity among participants; (5) range of issues in dispute; and (6) extent of violence among adversaries. See Crisis and Change in World Politics, 31—2. Waltz, Theory of International Politics, 124. Claude, Power and International Relations, 13. See also Sullivan, Theories and Evidence, 166. Waltz, "The Stability of a Bipolar World," 882-7. Rosecrance, "Bipolarity, Multipolarity, and the Future," 315, 316. Ibid., 317. This line of reasoning is derived from Deutsch and Singer, "Multipolar Power Systems and International Stability," 396. Rosecrance, "Bipolarity, Multipolarity, and the Future," 318. Ibid., 322. One major source of difficulty is the uncertain meaning of the term "pole" in this context. The problem of measuring polarity will be dealt with in Chapter 7. For an interesting discussion of the concept of polarity, consult Nogee, "Polarity: An Ambiguous Concept." Bartlett, Familiar Quotations, 615. Hardin, "The Tragedy of the Commons." Singer, Bremer, and Stuckey, "Capability Distribution, Uncertainty, and Major Power War, 1820—1965." Cannizzo, "Capability Distribution and Major-Power War Experience." This study also relied upon cow Project data. Wallace, "Alliance Polarization, Cross-Cutting, and International War, 1815-1964," 576, 589, 597. Haas, "International Subsystems," 99, 100. Ibid., 121. Alexandroff s analysis of diplomacy among six European states (Great Britain, France, Germany, Russia, Austria-Hungary, and Italy) from 1870 to 1890 is an exception to this rule. He found that dyads with many interactions exhibited less symmetric conflict behaviour than those with few interactions. This result, according to Alexandroff, suggested the value of a "mixed international system" with many, flexible points of contact. See Alexandroff, The Logic of Diplomacy, 138, 158. Organski and Kugler, The War Ledger, 8. Ibid., 51. See Zinnes, "An Analytical Study of the Balance of Power Theories." CHAPTER SEVEN

i These indices have been obtained from German, "A Tentative Evaluation of World Power"; Alcock and Newcombe, "The Perception of National

174 Notes to pages 130—8 Power"; Ferris, The Power Capabilities of Nation States; Cox and Jacobson, The Anatomy of Influence', Rosecrance, et al., Power, Balance of Power, and Status in Nineteenth Century International Relations; Russett, Power and Community in World Politics; Rummel, Understanding Conflict and War; Singer et al., "Capability Distribution, Uncertainty, and Major Power War, 1820— 1965"; and Organski and Kugler, The War Ledger. 2 Olson and Zeckhauser discovered that the percentage of national income devoted to defense spending by NATO alliance members correlated positively with the absolute size of their economies. While this study focused on the sociology of resource allocation as opposed to economies of scale, at the very least it is consistent with the presence of the latter. See Olson and Zeckhauser, "An Economic Theory of Alliances." 3 Cox and Jacobson, The Anatomy of Influence, 437, 438, 441. 4 These and other correlations among indices of power can be found in James, "The Measurement of National Power," mimeo, 1985. 5 Rosecrance et al., Power, Balance of Power and Status in Nineteenth Century International Relations, 11—14. 6 Small, "Doing Diplomatic History by the Numbers," 28. 7 Kravis, Heston and Summers, "Real GDP Per Capita." 8 See Kravis et al., "Real GDP Per Capita"; and Isenman, "Inter-Country Comparison of 'Real' (PPP) Incomes." 9 The terms "international system" and "subsystem" follow the definition developed by Brecher and James, Crisis and Change in World Politics, 10: "an international system is comprised of actors that are situated in a configuration of power (structure), are involved in regular patterns of interaction (process), are separated from other units by boundaries set by a given issue, and are constrained in their behavior from within (context) and from outside the system (environment)." 10 Organski and Kugler, The War Ledger, 19—22. 11 The breakpoints are as follows: embryonic to tight bipolarity, the Czechoslovakia Crisis of 1948 leading to its entry into the Soviet bloc; from tight bipolarity to a transitional phase commonly termed "loose bipolarity", the Taiwan Straits Crisis of 1958, marking the PRC'S open break with the Soviet Union and, in the Western bloc, the beginning of military withdrawal from NATO by the French under deGaulle; and the emergence of polycentrism following the Cuban Missile Crisis and the SinoIndian Border War at the end of 1962. See Brecher and James, Crisis and Change in World Politics, 140. 12 ICB Project Case Summary no. 134. CHAPTER EIGHT i Lebow, Between Peace and War, 41.

175 Notes to pages 147-155 2 3 4 5

ICB Project Case Summaries nos. 107, 117, and 164. Lebow, Between Peace and War, 54. Ibid., 57. Latent conflict will be excluded from the analysis of anomalies and loglinear testing. The comparatively high frequency of missing cases for latent conflict (relative to the other three propositions) is not conducive to such analysis. 6 ICB Project Case Summaries nos. 112, 133, and 185. 7 Superpower involvement does not imply that either the United States or the Soviet Union automatically had the status of an "involved actor". For example, in the case of Rhodesia's Unilateral Declaration of Independence, a Zambian ministerial mission visited Moscow with the hope that the USSR would take a position on the Declaration favourable to Zambia (ICB Project Case Summary no. 185). This low-level, political involvement on the part of the Soviet Union does not qualify it as an involved actor — i.e., a state which perceived either a threat to basic values, finite time for response or a high probability of military hostilities. 8 SPSS, SPSS-X User's Guide, 569. 9 Bueno de Mesquita, The War Trap, 153. 10 Stoessinger, Why Nations Go to War, 4th ed., 203-4. 11 Keohane, After Hegemony, 25—6. 12 Ibid., 26. 13 Out of 132 crises in the data set, 17 resulted in war. All other things being equal, there is a corresponding probability of 0.129 tnat a giyen crisis will lead to war. Relative to crisis, then, war may be regarded as an uncommon event. 14 For a point of view which emphasizes the importance of social institutions, see Young, Resource Regimes. 15 The classic discussion of middle range theories appears in Merton, Social Theory and Social Structure. 16 Bartlett, Familiar Quotations, 891.

Bibliography

Afriat, S.N. The Price Index. London: Cambridge University Press 1977. Albertini, Luigi. The Origins of the War of 1914. Volumes I-III. trans, and ed. by Isabella M. Massey. London: Oxford University Press 1952— !957Alcock, Norman Z., and Alan G. Newcombe, "The Perception of National Power," Journal of Conflict Resolution 17 (1970): 335—43. Aldrich, John H., and Forrest D. Nelson, Linear Probability, Logit, and Probit Models. Beverly Hills and London: Sage Publications 1984. Alexandroff, Alan S. The Logic of Diplomacy. Beverly Hills and London: Sage Publications 1981. Allison, Graham T. The Essence of Decision. Boston: Little-Brown and Company 1971. Alt, James, and K. Alec Chrystal. "The Criteria for Choosing a PoliticalEconomic Model, Forecast Results for British Expenditures 1976—1979: A Reply to Frey and Schneider." European Journal of Political Research 11 (1983): 115-23. Andriole, Stephen J. "The Levels of Analysis Problems and the Study of Foreign International, and Global Affairs: A Review Critique, and Another Final Solution." International Interactions 5 (1978): 113—33. Arrow, Kenneth J. Social Choice and Individual Values. 2nd ed. New Haven and London: Yale University Press 1963. Azar, Edward E. "Peace Amidst Development: A Conceptual Agenda for Conflict and Peace Research." International Interactions 6 (1979): 123—43. Bailey, Martin J., Mancur Olson, and Paul Wonnacott. "The Marginal Utility of Income Does Not Increase: Borrowing, Lending, and FriedmanSavage Gambles." American Economic Review 70 (1980): 372—9. Baldwin, David A. "Power Analysis and World Politics: New Trends versus Old Tendencies." World Politics 31 (1979): 161-94. Barbera, Henry. Rich Nations and Poor in Peace and War. Lexington, Massachusetts: D. C. Heath and Company 1973.

177

Bibliography

Bartlett, John, ed. Familiar Quotations. i§th ed. Boston and Toronto: Little, Brown and Company 1980. Beloff, Max. The Balance of Power. Montreal: McGill University Press 1967. Bergquist, C. "Some Relationships Between Foreign and Domestic Conflict: An Examination Using a Cross Lagged Quasi-experimental Design." Unpublished manuscript. Blalock, Hubert M., Jr. Social Statistics. Rev. 2nd. ed. New York: McGrawHill 1979. Blau, Peter M. Exchange and Power in Social Life. New York: John Wiley and Sons, Inc. 1964. Brecher, Michael. "Toward a Theory of International Crisis Behavior: A Preliminary Report." International Studies Quarterly 21 (1977): 39—74. , with Benjamin Geist. Decisions in Crisis. Berkeley and Los Angeles: University of California Press 1980. , and Jonathan Wilkenfeld. Crisbank: Data Bank of the International Crisis Behavior Project. Unpublished manuscript, 1981. , and Patrick James. Crisis and Change in World Politics. Boulder, Colorado: Westview Press 1986. , Jonathan Wilkenfeld, and Sheila Moser. Crises in the Twentieth Century, vol. i. Handbook of International Crises, Oxford and New York: Pergamon Press, forthcoming 1987. Bremer, Stuart, J. David Singer, and Urs Luterbacher. "The Population Density and War Proneness of European Nations, 1816-1965." In J. David Singer, et al., eds. Explaining War: Selected Papers from the Correlates of War Project. Beverly Hills: Sage Publications 1979. Brim, O. G., et al., eds. Personality and Decision Processes. Stanford, California: Stanford University Press 1962. Brines, Russell. The Indo-Pakistan Conflict. London: Pall Mall Press 1968. Brubaker, Earl R. "Economic Models of Arms Races: Some Reformulations and Extensions." Journal of Conflict Resolution 17 (1973): 187—205. Brunk, Gregory G., and Thomas G. Minehart. "How Important Is Elite Turnover to Policy Change?" American Journal of Political Science 28 (1984): 559-69Bueno de Mesquita, Bruce. "Systemic Polarization and the Occurrence and Duration of'War." Journal of Conflict Resolution 22 (1978): 241-67. . The War Trap. New Haven: Yale University Press 1981. . "An Expected Utility Theory of International Conflict." American Political Science Review 74 (1980): 917—31. . "A Critique of 'A Critique of The War Trap'." Journal of Conflict Resolution 28 (1984): 341—60. . "The War Trap Revisited: A Revised Expected Utility Model." American Political Science Review 79 (1985): 156-77. , and David Lalman. "Reason and War." American Political Science Review 80 (1986): 1113—29.

178 Bibliography Bunce, Valerie. Do New Leaders Make a Difference? Princeton: Princeton University Press 1981. Cable, James. Gunboat Diplomacy, igig-igjg: Political Application of Limited Naval Force. 2nd ed. New York: St. Martins Press 1981. Cambpell, John C., ed. Successful Negotiation: Trieste ig$4 — An Appraisal by the Five Participants. Princeton: Princeton University Press 1976. Cannizzo, Cynthia. "Capability Distribution and Major-Power War Experience, 1816—1965." Orbis 21 (1978): 947—57. Chatterjee, Partha. "The Classical Balance of Power Theory."Journal of Peace Research i (1972): 51—61. Cherne, Leo. "Ideology and the Balance of Power." Annals of the American Academy of Political and Social Science 442 (1979): 46—56. Choucri, Nazli, and Robert C. North. Nations in Conflict. San Francisco: W.H. Freeman 1975. Claude, Inis L., Jr. Power and International Relations. New York: Random House 1962. Cline, Ray S. World Power Assessment igjj: A Calculus of Strategic Drift. Boulder, Colorado: Westview Press 1977. Cohen, Raymond. Threat Perception in Internationl Crisis. Madison, Wisconsin: University of Wisconsin Press 1979. Cole, Charles L. Microeconomics: A Contemporary Approach. New York: Harcourt-Brace-Jovanovich, Inc. 1973. Cole, Harry L. The War of 1812. Chicago and London: University of Chicago Press 1965. Cook, Thomas D., and Donald T. Campbell. Quasi-Experimentation: Design and Analysis Issues for Field Settings. Chicago: Rand McNally College Publishing Company 1979. Coser, Lewis A. The Functions of Social Conflict. Glencoe, Illinois: The Free Press 1956. Cox, Robert W., and Harold K. Jacobson. The Anatomy of Influence. New Haven and London: Yale University Press 1974. Crow, Wayman J., and Robert C. Noel. "An Experiment in Simulated Historical Decision Making." In Margaret G. Hermann with Thomas W. Milburn, eds. A Psychological Examination of Political Leaders. New York: The Free Press 1977. Dahl, Robert A. Modern Political Analysis. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. 1963. Davies, James Chowning. "Aggression, Violence, Revolution, and War." In Jeanne N. Knutson, ed. Handbook of Political Psychology. San Francisco: Jossey-Bass Publishers, Inc. 1973. Dean, P. Dale, Jr. and John A. Vasquez. "From Power Politics to Issue Politics: Bipolarity and Multipolarity in Light of a New Paradigm." Western Political Quarterly 29 (1976): 7—28.

179 Bibliography Deane, Phyllis. "War and Industrialisation." In J. M. Winter, ed. War and Economic Development. Cambridge: Cambridge University Press 1975. Deutsch, Karl W., and J. David Singer. "Multipolar Power Systems and International Stability." World Politics 16 (1964): 390—406. Dougherty, James E., and Robert L. Pfaltzgraff, Jr. Contending Theories of International Relations. Philadelphia: J. B. Lippincott Company 1971. Downs, Anthony. An Economic Theory of Democracy. New York: Harper and Row 1957. Dror, Yehezkel. Crazy States: A Counterconventional Strategic Problem. Lexington, Massachusetts: Heath Lexington Books 1971. Eberwein, Wolf-Dieter. "The Quantitative Study of International Conflict: Quantity and Quality?: An Assessment of Empirical Research. "Journal of Peace Research 18 (1981): 19—38. , et al. "External and Internal Conflict Behavior Among Nations, 1966—1967." Journal of Conflict Resolution 23 (1979): 715—42. Farrar, L. L., Jr. "Cycles of War: Historical Speculations on Future International Violence." International Interactions 3 (1977): 161—79. Feierabend, Ivo K., Rosiland L. Feierabend and Ted Robert Gurr, eds. Anger, Violence, and Politics: Theories and Research. Englewood Cliffs, N.J.: Prentice-Hall 1972. Feinstein, C. H. National Income, Expenditure and Output of the United Kingdom, 7855—1965. Cambridge: Cambridge University Press 1972. Ferris, Wayne H. The Power Capabilities of Nation States: International Conflict and War. Toronto and London: Lexington Books 1973. Frank, Jerome D. Sanity and Survival: Psychological Aspects of War and Peace. New York: Random House 1967. Freud, Sigmund. Civilization and Its Discontents. Joan Riviere, trans. 4th ed. London: Hogarth Press 1949. Frey, Bruno S., and Friedrich Schneider. "Politico-Economic Models in Competition with Alternative Models: Which Predict Better?" European Journal of Political Research 10 (1982): 241—54. Frohlich, Norman, and Joe A. Oppenheimer. "Entrepreneurial Politics and Foreign Policy." World Politics 24 (1972, supplement): 151—78. . "I Get By With a Little Help From My Friends." World Politics 23 (1970): 104-20. and Oran R. Young. Political Leadership and Collective Goods. Princeton, New Jersey: Princeton University Press, 1971. Galtung, Johan. "A Structural Theory of Aggression. "Journal of Peace Research i (1964): 95—119. Gantzel, Klaus Jiirgen. "Another Approach to a Theory on the Causes of Internatonal War." Journal of Peace Research 18 (1981): 39—55George, Alexander L., and Richard Smoke. Deterrence in American Foreign

180 Bibliography Policy: Theory and Practice. New York and London: Columbia University Press 1974. German, F. Clifford, "A Tentative Evaluation of World Power, "Journal of Conflict Resolution 4 (1960): 138—44. Gilpin, Robert. War and Change in World Politics. Cambridge: Cambridge University Press 1981. Gough, Paul A., and Gregory G. Brunk. "A Re-examination of the Relationship Between Economic Conditions and Elections." Political Science 33 (!98l): 155-62. Gourevitch, Peter, Andrew Martin, George Ross, Christopher Allen, Stephen Bornstein and Andrei Markovits. Unions and Economic Crisis: Britain, West Germany and Sweden. London: George Allen and Unwin 1984. Gray, Colin S. "The Arms Race Phenomenon." World Politics 24 (1971): 39-79Gurr, Ted Robert. Why Men Rebel. Princeton: Princeton University Press 1970. , ed. Handbook of Political Conflict: Theory and Research. New York: The Free Press 1980. Haas, Ernst B. "The Balance of Power: Prescription, Concept, or Propaganda?" World Politics 5 (1953): 442-77. , and Allen S. Whiting. Dynamics of International Relations. New York: McGraw Hill 1956. Haas, Michael. "International Subsystems: Stability and Polarity." American Political Science Review 64 (1970): 98—123. . "Societal Approaches to the Study of War." Journal of Peace Research 2 (1965): 307-23Hardin, Garrett. "The Tragedy of the Commons." In Garrett Hardin and John Baden, eds. Managing the Commons. San Francisco: W. H. Freeman and Company 1977. Hart, Jeffrey. "Three Approaches to the Measurement of Power in International Relations." International Organization 30 (1976): 289—305. Hazlewood, Leo. "Diversion Mechanisms and Encapsulation Processes: The Domestic Conflict-Foreign Conflict Hypothesis Reconsidered." In Patrick J. McGowan, ed. Sage International Yearbook of Foreign Policy Studies, Volume III. Beverly Hills: Sage Publications 1975. Hermann, Charles F. Crises in Foreign Policy: A Simulation Analysis. Indianapolis: Bobbs-Merrill 1969. , ed. International Crises: Insights from Behavioral Research. New York: The Free Press 1972. Hibbs, Jr., Douglas A. Mass Political Violence: A Cross-national Causal Analysis. New York: Wiley 1973. , with the assistance of Nicholas Vasilatos. "Economics and Politics in France: Economic Performance and Mass Political Support for

181 Bibliography Presidents Pompidou and Giscard d'Estaing." European Journal of Political Research 9 (1981): 133-45. Hirschman, Albert O. National Power and the Structure of Foreign Trade. Berkeley: University of California Press 1969. Hirshleifer, J. Investment, Interest and Capital. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. 1970. Holsti, K. J. The Dividing Discipline: Hegemony and Diversity in International Theory. Boston: Allen and Unwin 1985. Holsti, Ole R. Crisis Escalation War. Montreal: McGill-Queen's University Press 1972. Hopple, Gerald W. Political Psychology and Biopolitics: Assessing and Predicting Elite Behavior in Foreign Policy Crises. Boulder, Colorado: Westview Press 1980. Hughes, Barry B. The Domestic Context of American Foreign Policy. San Francisco: W. H. Freeman and Company 1978. Huntington, Samuel P. "Patterns of Violence in World Politics." In Huntington, ed. Changing Patterns of Military Politics. New York: International Yearbook of Political Behavior Research 1962. Intriligator, M. D., and D. L. Brito. "Can Arms Races Lead to the Outbreak of War?" Journal of Conflict Resolution 28 (1984): 63—84. Isaak, Robert A. "The Individual in International Politics: Solving the Level-of-Analysis Problem." Polity 7 (1974): 264-76. Isard, Walter, and Charles H. Anderton. "Arms Race Models: A Survey and Synthesis." Conflict Management and Peace Science 8 (1985): 27—98. Isenman, Paul. "Inter-Country Comparison of 'Real' (PPP) Incomes: Revised Estimates and Unresolved Questions." World Development 8 (1980): 61-72. James, Patrick. "The Measurement of National Power." (unpublished, 1985). Janis, Irving L. Victims of Groupthink. Boston: Houghton-Mifflin Company 1972. Johnsson, Christer. "Bargaining Power: Notes on an Elusive Concept." Cooperation and Conflict 16 (1981): 249—57. Kamerschen, David, and Lloyd M. Valentine. Intermediate Microeconomic Theory. Cincinnati, Ohio: Southwestern Publishing Co. 1977. Kaplan, Morton A. System and Process in International Politics. New York: Wiley 1957. Kegley, Jr., Charles W., Neil R. Richardson, and Giinter Richter. "Conflict at Home and Abroad: An Empirical Extension. "Journal of Politics 40 (1978): 742-52. Keohane, Robert O. After Hegemony: Cooperation and Discord in the World Political Economy. Princeton, New Jersey: Princeton University Press 1984. , and Joseph S. Nye. Power and Interdependence. Boston: LittleBrown 1977.

182 Bibliography Kluckhohn, Clyde. Mirror for Man. New York: McGraw-Hill 1949. Knorr, Klaus. Power and Wealth: The Political Economy of International Power. New York: Basic Books, Inc. 1973. Kravis, Irving B., Alan W. Heston, and Robert Summers. "Real GDP Per Capita for More Than One Hundred Countries." Economic Journal 88 (1978): 215-42. Kugler, Jacek. "Terror Without Deterrence: Reassessing the Role of Nuclear Weapons." Journal of Conflict Resolution 28 (1984): 470—506. Kuper, Adam J., and Jessica Kuper, eds. The Social Science Encyclopedia. London and Boston: Routledge and Kegan Paul 1985. Lakatos, Imre. "Falsification and the Methodology of Scientific Research Programmes." In Imre Lakatos and Alan Musgrave, edc. Criticism and the Growth of Knowledge. Cambridge: Cambridge University Press 1970. Lasswell, Harold D. World Politics and Personal Insecurity. New York: The Free Press 1965. , and Abraham Kaplan. Power and Society: A Framework for Political Inquiry. New Haven: Yale University Press 1950. Lebow, Richard N. Between Peace and War: The Nature of International Crisis. Baltimore and London: Johns Hopkins University Press 1981. Lee, K. W. "A Conceptual Framework for International Political Analysis." Political Science Review 21 (1982): 129-48. Levi, Ariel, and Phillip E. Tetlock. "A Cognitive Analysis of Japan's 1941 Decision for War." Journal of Conflict Resolution 24 (1980) 195—211. Liska, George. Quest for Equilibrium: America and the Balance of Power on Land and Sea. Baltimore and London: Johns Hopkins University Press 1977Lorenz, Konrad. On Aggression. New York: Bantam Books 1967. Luce, R. Duncan, and Howard Raiffa. Games and Decisions. New York: John Wiley and Sons, Inc. 1957. Luterbacher, Urs. "Last Words About War?" Journal of Conflict Resolution 28 (1984): 165-82. Luttwak, Edward N. The Grand Strategy of the Roman Empire. Baltimore and London: Johns Hopkins University Press 1976. Mack, Andrew. "Numbers are Not Enough: A Critique of Internal/External Conflict Behavior Research." Comparative Politics 7 (1975): 597—618. Majeski, Stephen J., and David J. Sylvan. "Simple Choices and Complex Calculations: A Critique of The War Trap." Journal of Conflict Resolution 28 (1984): 316-40. McCloskey, Donald M. The Applied Theory of Price. New York: Macmillan Publishing Co. Inc. 1982. McCormick, James M. "International Crises: A Note on Definition." Western Political Quarterly 31 (1978): 352—8. McDougall, William. "The Instinct of Pugnacity." In Leon Bramson, and

183 Bibliography George W. Goethals, eds. War: Studies from Psychology, Sociology and Anthropology. New York: Basic Books 1968. McGowan, Patrick J., and Robert M. Rood. "Alliance Behavior in Balance of Power Systems: Applying a Poisson Model to Nineteenth-Century Europe." American Political Science Review 64 (1975): 859—70. Merton, Robert K. Social Theory and Social Structure, rev. and enl. London: Free Press of Glencoe 1957. Morgan, Patrick M. Deterrence: A Conceptual Analysis. Beverly Hills: Sage Publications 1977. Morgenthau, Hans J. Politics Among Nations: The Struggle for Power and Peace. 2nd ed. New York: Alfred A. Knopf 1954. Morrow, James D. "A Continuous-Outcome Expected Utility Theory of War." Journal of Conflict Resolution 29 (1985): 473—502. Mouritzen, Hans. "Selecting Explanatory Level in International Politics: Evaluating a Set of Criteria." Cooperation and Conflict 15 (1980): 169—82. Mukherjee, M. National Income of India: Trends and Structure. Calcutta: Statistical Publishing Society 1969. Nagel, Jack H. The Descriptive Analysis of Power. New Haven and London: Yale University Press 1975. Nelson, Stephen D. "Nature/Nurture Revisited I: A Review of the Biological Bases of Conflict. "Journal of Conflict Resolution 18 (1974): 285-335. Newman, William J. The Balance of Power in the Interwar Years, 1919—1939. New York: Random House 1968. Nogee, Joseph L. "Polarity: An Ambiguous Concept." Orbis 18 (1975): 1193-224. Nordhaus, William D. "The Political Business Cycle." Review of Economic Studies 42 (1975): 169-90. Novak, Bogdan C. Trieste, 1941-1954: The Ethnic, Political, and Ideological Struggle. Chicago and London: University of Chicago Press 1970. Olson, Mancur. The Logic of Collective Action. Cambridge: Harvard University Press 1965. , and Martin J. Bailey. "Positive Time Preference." Journal of Political Economy 89 (1981): 1—25. , and Richard Zeckhauser. "An Economic Theory of Alliances." Review of Economics and Statistics 48 (1966): 266—79. Onate, Andres. "The Conflict Interactions of the People's Republic of China, 1950—1970." Journal of Conflict Resolution 18 (1974): 578—94. Oppenheimer, Joe A. "Small Steps Forward for Political Economy." World Politics 32(1980): 121-51. Organski, A. F. K., and Jacek Kugler. The War Ledger. Chicago and London: University of Chicago Press 1980. Paige, Glenn D. The Korean Decision. New York: The Free Press 1968. Paldam, Martin. "A Preliminary Survey of the Theories and Findings on

184 Bibliography Vote and Popularity Functions." European Journal of Political Research 9 (1981): 181-99. Penrose, E. F. The Revolution in International Relations: A Study in the Changing Nature and Balance of Power. London: Frank Cass and Co., Ltd. 1965. Phillips, Warren R., and Richard Rimkunas. "Crisis Warning." University of Maryland, mimeo. Plott, Charles R., and David M. Grether. "Economic Theory of Choice and the Preference Reversal Phenomenon." American Economic Review 6 9 (!979): 623-38. Rawls, John. A Theory of Justice. Cambridge: Belknap Press of Harvard University 1971. Ray, J.L. "Status Inconsistency and War Involvement in Europe, 1816— 1870." Peace Science Society (International) Papers 23(1974): 69-80. , and J. David Singer. "Measuring the Concentration of Power in the Internatonal System." Sociological Methods and Research i (1973): 403-37Reuck, Anthony de, and Julie Knight, eds. Conflict in Society. Boston: LittleBrown 1966. Reynolds, H. T. Analysis of Nominal Data. 2nd ed. Beverly Hills: Sage 1984. Richardson, Lewis F. Arms and Insecurity. Pittsburgh: Boxwood Press 1960. . Statistics of Deadly Quarrels. Pittsburgh: Boxwood Press 1960. Richardson, Neil R., and Charles W. Kegley, Jr. "Cumulation and the Domestic-Foreign Conflict Hypothesis. "Journal of Politics 43 (1981): 143-51Rosecrance, Richard N. Action and Reaction in World Politics: International Systems in Perspective. Boston: Little-Brown 1963. . "Bipolarity, Multipolarity, and the Future." Journal of Conflict Resolution 10 (1966): 314-27. , Alan Alexandroff, Brian Healy, and Arthur Stein. Power, Balance of Power, and Status in Nineteenth Century International Relations. Beverly Hills: Sage Publications 1974. Rosenau, James N., ed. Linkage Politics: Essays on the Convergence of National and International Systems. New York: The Free Press 1969. . "Pre-theories and Theories of Foreign Policy." in R. Barry Farrel, ed. Approaches to Comparative and International Politics. Evanston, 111.: Northwestern University Press 1966. Rummel, R. J. "Dimensions of Conflict Behavior Within and Between Nations." General Systems Yearbook 8 (1963): 1—50. . "Dimensions of Foreign and Domestic Conflict Behavior: A Review of Empirical Findings." In Dean G. Pruitt and Richard C. Snyder, eds. Theory and Research on the Causes of War. Englewood Cliffs, N. J.: PrenticeHall, Inc. 1969.

185 Bibliography . "Libertarianism and International Violence. "Journal of Conflict Resolution 27 (1983): 27—71. . "The Relationship between National Attributes and Foreign Conflict Behavior." In J. David Singer, ed. Quantitative International Politics. . Understanding Conflict and War, 5 Vols. Beverly Hills: Sage 1975-81. Russett, Bruce M. "The Calculus of Deterrence." Journal of Conflict Resolution 7 (i963): 97-!°9, ed. Peace, War, and Numbers. Beverly Hills: Sage Publications 1972. . Power and Community in World Politics. San Francisco: W. H. Freeman and Company 1974. Schelling, Thomas C. The Strategy of Conflict. Cambridge: Harvard University Press 1963. Schrodt, Phillip A. "Richardson's N-Nation Model and the Balance of Power." American Journal of Political Science 22 (1978): 364—90. Scolnick, Jr., Joseph M. "An Appraisal of Studies of the Linkage Between Domestic and International Conflict." Comparative Political Studies 6 (1974): 485-509. Sherif, Mustafer, and C. W. Sherif. Groups in Harmony and Tension. New York: Harper 1953. Simmel, Georg. Conflict and the Web of Group-Affiliations. Kurt H. Wolff and Reinhard Bendix, trans. Glencoe, Illinois: The Free Press 1955. Singer, J. David. "The Level-of-Analysis Problem in International Relations." World Politics 14 (1961): 7 7-9 2. . "Accounting for International War: The State of the Discipline." Journal of Peace Research 18 (1981): 1-18. , ed. Quantitative International Politics: Insights and Evidence. New York: Free Press 1968. , and Melvin Small. The Wages of War, 1816—1965: A Statistical Handbook. New York: John Wiley 1972. . "The Composition and Status Ordering of the International System: 1815—1940." World Politics 18 (1966): 236—82. et al. "Capability Distribution, Uncertainty, and Major Power War, 1820—1965." Beverly Hills: Sage 1979. In Singer, et al., Explaining War. Beverly Hills: Sage 1979. Siverson, Randolph M., and Joel King. "Alliances and the Expansion of War." In J. David Singer and Michael D. Wallace, eds. To Augur Well: Early Warning Indicators in World Politics. Beverly Hills and London: Sage Publications 1979. Small, Melvin. "Doing Diplomatic History by the Numbers: A Rejoinder." Journal of Conflict Resolution 21 (1977): 23—34. . "The War Proneness of Democratic Regimes, 1816-1965. "Jerusalem Journal of International Relations 1(1976): 50-69.

i86 Bibliography , and J. David Singer. Resort To Arms: International and Civil Wars, 1816—ig8o. Beverly Hills: Sage Publications 1982. Smoke, Richard. War: Controlling Escalation. Cambridge and London: Harvard University Press 1977. SPSS, Inc. SPSS-X User's Guide, and ed. New York: McGraw-Hill 1986. Starr, Harvey. "The Quantitative International Relations Scholar as Surfer: Riding the Fourth Wave." Journal of Conflict Resolution 18 (1974): 335-68. Stein, Arthur A. "Conflict and Cohesion: A Review of the Literature." Journal of Conflict Resolution 20 (1976): 143—72. Stoessinger, John G. The Might of Nations: World Politics in Our Time, grd ed. New York: Random House 1969. . Why Nations Go To War. grd, 4th eds. New York: St. Martin's Press, 1982 and 1985. Sullivan, Michael P. International Relations: Theories and Evidence. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. 1976. Tanter, Raymond. "Dimensions of Conflict Behavior Within and Between Nations, 1958-60." Journal of Conflict Resolution 10 (1966): 41-64. Taylor, Charles Lewis, and David A. Jodice. World Handbook of Political and Social Indicators, V. I, II. 3rd ed. New Haven: Yale University Press

1983Thompson, William R., Raymond D. Duvall, and Ahmed Dia. "Wars, Alliances, and Military Expenditures: Two Pendulum Hypotheses." Journal of Conflict Resolution 23 (1979): 629—54. Trask, David F. The War with Spain in i8g8. New York and London: Macmillan, Inc. 1981. Tufte, Edward R. Political Control of the Economy. Princeton, New Jersey: Princeton University Press 1978. Ursprung, Heinrich W. "Macroeconomic Performance and Government Popularity in New Zealand." Comparative Political Studies 16 (1984): 457-77Vincent, Jack E. "Internal and External Conflict: Some Previous Operational Problems and Some New Findings. "Journal of Politics 43 (1981): 128—42. Volkan, Vamik D. Cyprus: War and Adaptation — A Psychoanalytic History of Two Ethnic Groups in Conflict. Charlottesville, Virginia: University Press of Virginia 1979. Wagner, R. Harrison. "War and Expected Utility Theory." World Politics 36 (1984): 407-23. Wallace, Michael D. "Alliance Polarization, Cross-Cutting, and International War, 1815—1964: A Measurement Procedure and Some Preliminary Evidence." Journal of Conflict Resolution 17 (1973): 575—604.

187 Bibliography Waltz, Kenneth N. Man, the State and War. New York: Columbia University Press 1959. . "The Stability of a Bipolar World." Daedalus 93 (1964): 881-909. . Theory of International Politics. Reading, Massachusetts: AddisonWesley 1979. Ward, Michael D., and Ulrich Widmaier. "The Domestic-Foreign Conflict Nexus: New Evidence and Old Hypotheses." International Interactions 9 (!982): 75-101Weber, Max. The Theory of Social and Economic Organization. A.M. Henderson and Talcott Parsons, trans. New York: Free Press of Glencoe 1964. Whiting, Allen S. China Crosses the Yalu. New York: Macmillan 1960. Wilkenfeld, Jonathan. "Domestic and Foreign Conflict Behavior of Nations." Journal of Peace Research 5 (1968): 56—69. . "Some Further Findings Regarding the Domestic and Foreign Conflict of Nations." Journal of Peace Research 6 (1969): 147—56. , ed. Conflict Behavior and Linkage Politics. New York: David McKay Company, Inc. 1973. . "A Time-Series Perspective on Conflict Behavior in the Middle East." in McGowan, Sage International Yearbook. , Gerald W. Hopple, and Paul J. Rossa. "Sociopolitical Indicators of Conflict and Cooperation." in Singer and Wallace, To Augur Well: Early Warning Indicators in World Politics. , and Stephen J. Andriole. Foreign Policy Behavior: The Interstate Behavior Analysis Model. Beverly Hills: Sage Publications 1980. Wohlstetter, Albert. "Theory and Opposed-Systems Design."Journal of Conflict Resolution 12 (1968): 302—31. Wright, Quincy. A Study of War. and ed. Chicago: University of Chicago Press 1965. Young, Oran R. "Anarchy and Social Choice: Reflections on the International Polity." World Politics 30 (1978): 241-63. . Resource Regimes: Natural Resources and Social Institutions. Berkeley: University of California Press 1982. Zagare, Frank. "Review of The War Trap." American Political Science Review 76 (1982): 738-39Zinnes, Dina A. "An Analytical Study of the Balance of Power Theories." Journal of Peace Research 4 (1967): 270—88. . "Research Frontiers in the Study of International Politics." In Fred I. Greenstein and Nelson W. Polsby, eds. Handbook of Political Science, Volume 8: International Politics. Reading, Massachusetts: Addison-Wesley Publishing Company 1975. , and Jonathan Wilkenfeld. "An Analysis of Foreign Conflict Behavior of Nations." In Wolfram F. Hanrieder, ed. Comparative Foreign Policy: Theoretical Essays. New York: David McKay Company, Inc. 1971.

This page intentionally left blank

Index

Acton, Lord, 124 Afghanistan, 40, 70 aggregate data. See quantitative methods Allison, Graham T., 38 Alt, James, 98 Angola, 140; Crisis, 70, 139, i6an Arab-Israeli conflict, 82, no, 143, 147 Argentina, 102, 17 in Arrow Paradox, i65n Austria-Hungary, 152, i?3n Baghdad Pact Crisis, 148-50, 154 Bailey, Martin J., 41 balance of power, 16, 45, 49, 116, 118, 121-4, 126-30, i6in, i72n; theory, 13-16, 27, 49, 56, 120, 123, 127-30, !35, !37, »43> M5 Baldwin, David A., 119 Bangkok, 138 Bangladesh, 140; Crisis, 108, 109—10, 139 Barbera, Henry, 87 Belgrade, 105 Beloff, Max, 118 Berlin Deadline Crisis, 58 bipolarity, 123—6, 135—9, 141, 143-4, i74n Brazil, 107 Brecher, Michael, 8, 17, 26, 120, 137

Bremer, Stuart J., 85, 124,134 Bueno de Mesquita, Bruce, 14—15, 28-9, 31-2, 34-45, 47-58, 60-3, 71, 88-9, 148, 152-3, i67n Burma, Infiltration Crisis, 58-9 Burrowes, Robert, 82 Cabot, John, 62 Cambodia, 138 Campbell, Donald T., 66-7 Cannizzo, Cynthia, 125 Chiang Kai-Shek, 62 China, People's Republic of, 57, 58-9, 62, 82, 109, 138, 149, i74n; Civil War, i62n Chrystal, K. Alec, 98 Claude, Inis L., Jr., 122-3 Cline, Ray S., 117 Collins, John N., 82 conflict and cohesion. See externalization theory conflict, internal, 78-88, 90—4, 102-4, 1 o7-ii, 114; latent, 92-4, 99-102, 104-10, 114, 145, 147—8, 17511; manifest, 93, 102—7, 109—11, 145—8, 151 conflict, international, 15-16, 28, 36, 43, 47, 78-88, 90-1, 94, 99,

103, 107, 114, 117, 124, 126, i6gn; definition of, 5—7; literature on, 3—4, 6, 11-13, 87, 90; types of, 5, 6, 7-9 conflict linkage. See externalization theory conflict, protracted social, 5, 140, 143, is8n Cook, Thomas D., 66-7 Correlates of War (cow) Project, 6-7, 15, 39, 55-6, 85, 107, 124-5, 135, i5gn, i62n, i66n, i67n,i73n Cox, Robert W., 131, 134 crisis, actor level, 17, 26, 53-4, 58-9, 62, 67-70, 105, 107, 129, 135-8, 140, 147, 149-50, isgn, i62n, i63n, i64n, i66n crisis, international; definition of, 8; escalation of, 3, 9, 48, 74, 92—4, 98, 105, i6in Cuba, 40, 52, i62n, i66n Cuban Missile Crisis, 38, 52, 137, i74n Czechoslovakia Crisis, i74n Dahl, Robert A., 117 Davies, James Chowning, 6 defender: in a crisis, 17, 5°-4, 57-9' 62. 6 4> 6971, 74, 77, 92-4, 106,

190 108, 114, 129, 135-9,

144, 150, 152, i66n; in a war, 14-15, 35-6, 39, 44-7. 52-3. 152 de Gaulle, Charles, 17411 Dien Bien Phu Crisis, 16211 Dimona, 54 Dougherty, James E., 106 Downs, Anthony, 16in Eberwein, Wolf-Dieter, 4, 84, 90 Egypt' 54- 64-5, no, 139. i47» H9, i6an, lyan

Eilat, 54

Einstein, Albert, 3 El Alamein, 16in El Salvador, 53, 86 expected utility, 14-15, 28-9, 31-2, 35-9, 44-52, 55, 57-8, 63-71, 74, 77, 89, 128, 137, 145, 147-8, 151, 153-5, i65n, i&7n; bilateral, 29, 31, 64, 92—4, 106, 108, 111,114—15 expected utility calculus, 14, 28-9, 31-2, 35, 38, 39, 42, 45, 48, 51, 55, 57, 64, 74, l65n expected utility theory, 14-16, 27-9, 36, 39-40, 44, 47, 49, 53- 77, M5J revised 50, 54 externalization theory, 14-16, 27, 78-94, 102-3, 105—6, 109—11,114 Falkland Islands War, 102, i 7 i n Feisal, King, 65 Ferris, Wayne H., 131,

134- 135 Finland, 35 Football War Crisis, 53, 70, 139 formation of the VAR Crisis, 64; Arab Federation in, 65 France, 38, 62, 95, 98,

Index

I73n-i74n; Revolution, 79

Freud, Sigmund, 78 Frey, Bruno S., 98 Galtung, Johan, 8 Gantzel, Klaus Jurgen, 7 Garandal, 147 Germany, 40, 95, 152, i6in, i73n Ghana, 69—70; Togo Border Crisis, 69 Giscard d'Estaing, Valery, 98 Goa, i66n; Crisis, 57 gross national product (GNP), 55-6, 84, 131, 1 34-5. !38, 14° Guardian, 17 in Gulf of Tonkin Crisis, 139-40; and Intra-war crisis: Pleiku, i62n; Tet Offensive, i62n; Vietnam Spring Offensive, i62n; Invasion of Cambodia, i62n; Vietnam-Ports Mining, i62n; Christmas Bombing, i6sn; Final North Vietnam Offensive, i62n Gurr, Ted Robert, 83, 102-3, 1O7 Haas, Ernst B., 78 Haas, Michael, 125 Hanoi, 140 Hardin, Garrett, 124 Hazelwood, Leo, 81 Hegel, Georg W.F., 28 Hermann, Charles F., 26, H9 Hibbs, Jr., Douglas A., 98, 102-3, 1O7 Hitler, Adolf, 152 Hobbes, Thomas, 120 Holsti, Ole R., 26 Honduras, 53 Hughes, Barry B., 10 Hungarian Uprising Crisis, 139 Husan, 147

Hussein, Saddam, 12 Hussein, King, 65, 147 Hyderabad Crisis, i62n India, 57, 100, 108—9, 139, 140, 152 India/China Border Crisis n, 139, i74n Indonesia, 57, i62n, i66n;Independence Crisis in, 139, i62n initiator: of a crisis, 17, 49-54, 55, 57-8, 62-7, 6 9-7 !> 74, 77, 92-3, 100-1, 105, 108, 110-11,

114, 129-30, 135-7,

139-40, 144, 147-9; of a war, 14-15, 35-40, 43-4,46-7, 52-5, 81,

145, !52-3 interdependence, 5, 116, 119-20, i 5 gn International Crisis Behavior (ICB) Project, 8, 17, 26, 62, 63, 107, 125, i5gn, i6in, i64n, i7in International Labor organization, 6 intra-war crisis, 16in, i62n Iran, 12, 149 Iraq, 12, 17, 26, 64-5, 149; Invasion-Kuwait Crisis, 147 Isaak, Robert A., 10 Isenman, Paul, 134 Israel, 54, 97, no, 140, 147,152;Independence Crisis, 58, 139; Mobilization Crisis, 147 Italy, 104-6, i6in, i73n Jacobson, Harold K., 131, 134 James, Patrick, 8, 137 Jordan, 54, 64-5, 82, 147; Internal Challenge Crisis, 147

Karachi, 70 Kashmir n Crisis, i62n

igi Index Kegley, Jr., Charles W., 83-4, 88 Kennedy, John F., 52, 62, !55 Keohane, Robert O., 116-17, 119-20, 153 Khan, Yahya, 110 Khomeini, Ayatollah, 12 Khrushchev, Nikita, 52 Koptic Church, no Korea, North, 152 Korean War: Crisis u, i62n; Crisis in, i62n Kravis, Irving B., 134 Kugler, Jacek, 55, 126, 127, 131, 137 Kuwait, 17, 26; Independence Crisis, 17, 147, i62n

Laos: Invasion Crisis i, 62; Invasion Crisis u, i62n Lebanon, 147 Lebow, 146—7 Loikaw, 59 Luttwak, Edward N., 121 Mack, Andrew, 106 Majeski, Stephen J., 38 Malthusian population pressure, 85 Markov analysis, 81, iGgn Mexico, 38 Morgenthau, HansJ., 118 Morocco, 38 Morrow, James D. 13 Moser, Sheila, 17 multipolarity, 123—6, 129, 135-6' M3. M8, 153 Nasser, Gamal Abdel, 12, 54, 65, 149 Netherlands, i62n, i66n New International Economic Order (NIEO), HQ Newman, William J., 121-2 New Zealand, 95, 97, i7on Nkrumah, Kwame, 69

Nordhaus, William D., 95 North Atlantic Treaty Organization (NATO), 62, i74n Nye, Joseph S., 116—7, i19—20 Olson, Mancur, 41 Onate, Andres, 82 Organization for Economic Cooperation and Development (OECD), 61, 97 Organski, A.F.K. 55, 126, 127, 131, 137 Paige, Glenn D., 26 Pakistan, 70, 108—10, 140, 149,152 Paldam, Martin, 97 Pareto-preferred, 40, 136, 149, i6sn Pfaltzgraff, Jr., Robert L., o ' J 106 phenomenology, 10, i6on Phillips, Warren R., 80 Pnom Penh, 138 polarity, 13, 16, 125-7, 7

7

129-3°. 135-6. Mi-3. 145, 148, 151, 154, i73n polycentrism, 137, i74n Pompidou, Georges, 98 Portugal, 57, i66n power, 77, 78, 85, 116-24, 126-31, 13440, 144, 154-5, l 6 l n > i&4n, i72n, i74n; as an attribute, 117-18, i72n; as a relation, 117, 121; national, 15, 16, 40, 45, 47, 55-6, 116-8, 121-3, 128-31,134-7 power politics, 13, 16-17, 45, 119-20, 145, 154 Pushtunistan, 70; Crisis in, 69 projection. See externalization theory Qalkilya Crisis, 147 Qibya Crisis, 147 quantitative methods, 3-5,

7. 13. 49. 79. 82-3. 104> 107, 114, i58n quasi-experimentation, 66-7

Rann of Kutch Crisis, 108-09, *39 realpolitik. See. power politics Rhodesia, 57, 149; Unilateral Declaration of Independence (UDI) Crisis, 57, 148-50, 154, i75 n Richardson, Lewis F., 5 Richardson, Neil R., 84 risk, 31-2, 61-5, 90, i&4n; acceptant, 32, 63; averse, 32, 63-4, i&7n; neutral, 63, i&4n Rome, 105 Rosecrance, Richard N., 78, 123-4, 131, 134, 136, 143 Rosenau, James N., 13 Rummel, R.J., 79-81, 83, 87, 94, 131 Russett, Bruce M., 131 Russo-Japanese War, 79 Sadat, Anwar, 110 Scelba, Mario, 105 Schelling, Thomas C., 54 Schneider, Friedrich, 98 Scolnick, Jr., Joseph M., 106 Shastri, Lai Bahadur, 109 Simmel, Georg, 86 Sinai, 54; Incursion Crisis, i62n Singer, J. David, 6, 10, 55, 117-18, 124-5, *34 Six Day War Crisis, 54, 110, 139, 147 Small, Melvin, 6, 117—18 Smith, Ian, 149 Spain, 38 Spector, Bertram, 82 South Africa, i62n Soviet Union, 35, 38, 40, 52, 57, 61, 98, 119, 127,

192

Index

!37> 149-50. l66n > l 73n~lT5n state-centrism, 5, 13, 38, 120, 15911, i6in Stoessinger, John G., 12, 152-3 Stohl, Michael, 88 Straits of Tiran, 54 Strategic Arms Limitation Treaty (SALT i), 119 Stuckey, John, 124, 134 Suez Nationalization, i62n Suez-Sinai Campaign Crisis, 110, 147 Sullivan, Michael P., 86, 117 Sweden, 95 Sylvan, David J., 38 Syria, 54, 64—5, 82

United Arab Republic (UAR), 65, 82 United Kingdom, 26, 57, 98, 100, 102, 105, 120, 149, i6in, i 7 i n , i73n United Nations, 6, 26, 54, 68, 77, 138, 148 United States, 40, 52, 56, 58-9, 62, 95-7, 105, 109, 118—20, 127, 134—5, 137, 140, 149-50, 16311, i66n, i75n Ursprung, Heinrich W., 97 U Thant, 54 utility, 14, 31, 34, 38, 41, 43-5, 48, 50, 58, 60-1, 64, 74, 101, i&5n utility theory, 29, 36, 41, 43, i6sn

Taiwan, 58—9, 62—3, 149; Straits Crisis n, 137, 148-50, 154, i74n; Straits Crisis in, 62 Tanter, Raymond, 79-81, 83 Thailand, 138 theoretical articulation, 4, 13, 16, 48, 86, 114, 121 Thucydides, 49 Tito, Josip B., 105 Togo, Republic of, 69-70 Tragedy of the Commons, 124 Trieste Crisis, 104—6; Allied Military Government, 105 Tufte, Edward R., 95-6, 100 Turkey, 38, 149

Vasilatos, Nicholas, 98 Vietminh, 62 Vietnam: North, 118, 140; South, 118, 140 Vietnam War, See Gulf of Tonkin Crisis Vincent, Jack E., 84 Volta region, 69-70

uncertainty, 32, 61-4, 124, 126

Wagner, R. Harrison, 44-6 Wallace, Michael D., 125 Waltz, Kenneth N., 9-11, 13, 28, 78, 88, 114, 122-3, 136, 143 war: causes of, 3-7, 9-17, 27-9. 34-40, 44, 47, 49-50, 66-7, 78-9, 924, 116, 121-7, 129—30, !36-7> !39> l 6 i n ; definition of, 5, 6—8, i58n; findings on, 69-71, 74, 77, 106, 108—11,

114-15, 138-44, 14655; impact of, 3, 28, 40, 42-4, 50, 84, 87, 98-9, 119, 127; necessary and sufficient conditions for, 35-7, 148; potential participants in, 17, 47, 57-9, 61-5, 137; scholarship on, 3, 6, 10, 118 Ward, Michael D., 85, 88, 102-3 War of Attrition: Crisis i, no, 139, 147; Crisis n, i62n Whiting, Allen S., 78 Widmaier, Ulrich, 85, 88, 102-3 Wilkenfeld, Jonathan, 17, 80-3, 85 World Event Interaction Survey (WEIS), 83, 84, 107, i6gn World War i, 61, 152 World War n, 104, 127, 137, i6in Wright, Quincy, 5, 78 Yemen War: Crisis i, 70, i62n; Crisis n, i62n; Crisis in, i62n; Crisis iv, 70, i62n Yom Kippur War Crisis, October, 110, 139, 147 Yugoslavia, 104-6 Zambia, 57, 140, 149—50, 175" Zeitgeist, 12 Zinnes, Dina A., 16, 81, 88, 127