The power-conflict story: a dynamic model of interstate rivalry 9780472111916, 9780472027415

Table of contents :
Frontmatter
List of Figures (page ix)
List of Tables (page xi)
Acknowledgments (page xiii)
Chapter 1. An Introduction to Storytelling (page 1)
Chapter 2. Gathering Pieces of the Story (page 10)
Chapter 3. The Power-Conflict Story (page 55)
Chapter 4. The Moral (page 85)
Chapter 5. Verifying the Story (page 111)
Chapter 6. Epilogue (page 149)
Appendix A. Simulation Results (page 165)
Appendix B. Power Transitions Among the Major Powers (page 166)
Notes (page 169)
References (page 177)
Index (page 187)

Citation preview

The Power-Conflict Story

Blank Page

The Power-Conflict Story A Dynamic Model of Interstate Rivalry

Kelly M. Kadera

Ann Arbor

THe UNIVERSITY OF MICHIGAN PRESS

Copyright © by the University of Michigan 2001 All rights reserved Published in the United States of America by The University of Michigan Press Manufactured in the United States of America ©) Printed on acid-free paper

2004 2003 2002 2001 43 2 1] No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, or otherwise, without the written permission of the publisher. A CIP catalog record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Kadera, Kelly M., 1965— The power-conflict story : a dynamic model of interstate rivalry / Kelly

M. Kadera.

p. cm. Includes bibliographical references and index. ISBN 0-472-11191-4 (cloth : alk. paper) 1. International relations — Political aspects. 2. Balance of power.

3. War. I. Title.

327.1'01 —dc21 00-051173

JZ1310 .K33 2001

ISBN13 978-0-472-11191-6 (cloth) ISBN13 978-0-472-02741-5 (electronic)

This book is dedicated to the Dynamic Duo for sharing their love of modeling with me and with future generations of modelers.

Blank Page

Contents

List of Figures 1X List of Tables Xl

Acknowledgments X11

Chapter 1. An Introduction to Storytelling 1 Chapter 2. Gathering Pieces of the Story 10

Chapter 3. The Power-Conflict Story 55

Chapter 4. The Moral 85 Chapter 5. Verifying the Story 111 Chapter 6. Epilogue 149

Appendix A. Simulation Results 165

Notes 169 References 177 Index 187 Major Powers 166

Appendix B. Power Transitions Among the

Blank Page

Figures

1. The population growth model 2 2. Three stages of development 3 3. An intersection of two nations’ power curves 4 4. Doran and Parsons’s (1980) cycle of relative power 17

5. The BOP and PT explanations as step functions 28 6. The BOP and PT explanations as linear functions 31

concentration 32 concentration 32

7. A U-shaped relationship between war and power

8. An inverted U relationship between war and power

9. Bueno de Mesquita and Lalman’s probability of war 33

distributions 34 11. Three regions of conflict behavior 62 10. Probability of violence as a function of dyadic power

12. Growth in national power according to three

different growth rates 71 13. A sample trajectory in the p, versus p, plane 83 14. An example of a bull and gnat transition 88

15. An example of a tortoise and hare transition 89 16. An example of a David and Goliath transition 90

17. Varying the B, parameter 96

18. D&G trajectories for B, = .08 98

19. B&G trajectories for B, = .167 100

20. D&G and T&H trajectories for B, = .9 101 21. D&G and T&H trajectories for B, = .5 103

22. Major power CINC scores, 1817-72 125 23. Major power CINC scores, 1873-1928 126 24. Major power CINC scores, 1929-84 127 25. Logistic decline in B&G conflict difference 133

Blank Page

Tables

Distributions and War 25 2. Contenders Only 26 1. Organski and Kugler’s First Look at Power

War 44 Model §2 Rivalries 30

3. War Probabilities for Regional and Global Dyadic

4. Alliance Power Distribution and the Outbreak of

5. A Review of the Parameters in the Power-Conflict

6. Baseline Deductions from the Power-Conflict Model 106

1816-Present 112

Scores 118

7. Membership in the Major Power System,

8. Major Powers Excluded in the Calculation of CINC

9. Three Potentially Useful Conflict Data Sets 120

Gnats 133

10. Empirical Power Transitions between Major Powers 128

11. Comparing Conflict Levels for Bulls and Gnats 131 12. Patterns in Conflict Differences over Time: Bulls and

and Hares 135

13. Comparing Conflict Levels for Tortoises and Hares 134 14. Patterns in Conflict Differences over Time: ‘Tortoises

15. Comparing Conflict Levels for Davids and Goliaths 136 16. Timing of Second Peak in Joint Use of Force Cases 139 17. Timing of Second Peak in Joint Force Cases that

Support T1, B1, or D1 139

Blank Page

Acknowledgments

Generous thanks are due to a multitude of people for their help and patience. Dina Zinnes, Robert Muncaster, Paul Diehl, Frank Zagare, Jacek Kugler, Jack Levy, T. Clifton Morgan, Patrick James, Gretchen Hower, Gary Segura, Paul Hensel, Rebecca Morton, Jerry Sorokin, Charles Shipan, William Reisinger, Elizabeth Martin, and Jerry Loewenberg and have all contributed useful comments and advice. Erik Gartzke provided an independent classification of all of the empirical transitions

against which I could check my own classification decisions. Kendra Holtzman ran the B parameter sensitivity analyses. Kris Beck compiled the adjusted MID data for major powers. Daniel Morey compiled the index and proofread the final copy. Martha and Robbie Diehl were kind enough to provide hours of fun for Miss Maddy, without which I could not have completed many of the simulations in chapter 4. Finally, Philippe LePrestre lured me into this fascinating world of international relations years ago at Wells College.

Blank Page

CHAPTER 1

An Introduction to Storytelling

This book is primarily a story. It is a story about how nations become more and less powerful, how rival nations compete with one another, and how the relationship between two opponents’ power levels is intimately connected to their competitive behavior. In this introductory chapter, I explain the storytelling process. This includes a presentation of early interests that first suggest the story’s plot, an explanation of how the story is written, a discussion of why the writing process adopted here is a useful one, and a glimpse at the story’s ending and some potential sequels. Early Interests

The intellectual roots of this project strongly influenced the nature of not only a book but a more comprehensive research agenda as well. ‘The work grew out of my simultaneous exposure to three interesting sets of

ideas: (1) the use of a first-order differential equation to understand population dynamics (Mesterton-Gibbons 1989; Heiss, Knorr, and Morgenstern 1973; Kuznets 1966; Pearl 1924), (2) the speculation that na-

tional power development evolves over time according to an S-curve similar to that generated by the population growth model (Doran 1971, 1989a, 1989b; Doran and Parsons 1980; Organski 1958; Kadera 1990), and (3) two seemingly opposed beliefs relating the likelihood that two rival nations will go to war with the relationship between their power levels —the balance of power explanation of war (Kissinger 1979, 1994; Claude 1962; Morgenthau 1985)! and the power transition explanation (Organski 1958; Organski and Kugler 1980). Immediately I was struck by the potential for research in the first two areas to inform that done in the third. Moreover, I was surprised by the failure of international relations scholars, especially those interested in the relationship between the

distribution of power and interstate war, to draw on the population modeling literature to better understand the growth in national power levels over time.

2 The Power-Conflict Story Population

Time Fig. 1. The population growth model

The Plot: Population, National Power, and Power Transitions

The story presented here begins with a query. How might population models tell us something about modeling the growth of national power? In order to answer that question, I briefly present the typical biological model, which produces a picture of a species’ population evolving over time in a manner such as that depicted in figure 1. ‘The equation responsible for generating this picture is of the form:

a 8(1- x (a)

ds k) —=rAs{1-— , 1.1

where: s is the total number of individuals in a species’ total population at a given time, ds/dt is the change in the population level over time,

A is a positive constant representing the instantaneous birth rate, and

K is the carrying capacity or the population level resources can, in the long run, sustain. Equation (1.1) can be rewritten as:

An Introduction to Storytelling 3 Power

Power maturity

Transitional growth

Power potential

Time

dt K (1.2) Fig.2. Three stages of development

ds A —=As——s"’. 1.2

The first term, As, is responsible for growth in the species’ population. The second, As’/K, represents a force causing the population to diminish. The two terms compete, in a sense, for dominance over ds/dt. Note that as s approaches K, the carrying capacity, the second term’s influence has an increasing effect. This biological process 1s self-driven; it depends on no external factors. A population naturally expands over time until it reaches the maximum level that resources can sustain. A. EF. K. Organski (1958) proposes that growth in national power is similarly governed. According to him, a nation generally experiences three stages of development: (1) power potential, (2) transitional growth, and (3) power maturity. Each stage is similar to a portion of the graph in figure 1. If Ichange the vertical axis to “power,” this can be demonstrated as in figure 2. In chapter 3, I borrow a portion of the biological model and apply it to national power growth. Specifically, I use the first, or natural growth, term of equation (1.2). Rewriting just that piece of the equation for national power yields:

— = ap, 1.3 4 oP (1.3) dp

4 The Power-Conflict Story Power

--- NationA — Nation B

= Time “ “ “

a,

—_— a s

Fig. 3. An intersection of two nations’ power curves

where: p is anation’s power level at a given time, dp/dt is the change in a nation’s power level over time, and a 1s a positive constant representing the instantaneous growth rate.

Equation (1.3) is the first step in modeling the patterns in national power development over time. Given this assumption that the rise in national power is a function of the level of power itself, I will then ask what forces might pull the level of national power down. Rather than relying on the exhaustion of resources as a nation approaches some maximum power level, I tell a different story. That story begins with the realization that a nation’s power develop-

ment does not take place in a vacuum. In other words, it is only a partially self-driven process. The literature review in chapter 2 suggests

that at least two other factors are involved. The first factor is a rival nation’s power level, which also develops in stages and impinges on that of the initial nation. This second nation’s power level, as it grows, may overtake that of the first, as shown in figure 3. The point at which nation A is surpassed by nation B is called a power transition (Organski 1958;

Organski and Kugler 1980). If these two nations are competing for domination of one another, as in an interstate rivalry, nation B will try to

diminish nation A’s growth and vice versa. Nation B’s success (and nation A’s) will be highly dependent on its own power level. In other words, nation A’s growth, dp ,/dt, is dependent on B’s power level, p,, as well as on its own, p,.

An Introduction to Storytelling 5 What method of reducing nation A’s power growth does B have at its disposal? In chapter 3, I propose that conflict is the primary means of competing in this manner. The change in nation A’s power 1s there-

fore also driven down by the conflict nation B directs at it, c,,. The negative effects of nation B’s power and conflict behavior can be appended to the natural growth component in equation (1.3):

— = ap, — f(Pp.Cea); (1.4) d

where: Pp, 1s nation A’s power level, Pz 1S nation B’s power level, Cp, 18 the conflict B directs at A, and f(Pzg,Cz,) IS an as yet unspecified function.

Similarly, B’s growth, dp,/dt, is dependent on its own power level, Dz, that of its opponent, p,, and the conflict the opponent directs at it, c,,. At the same time, the conflict behavior of each rival is influenced

by, among other things, the distribution of power between A and B. International conflict scholars from two opposing schools of thought have proposed different relationships between conflict behavior and the DP, versus pz, discrepancy. Those interested in S-curves for national power development believe that dyadic warfare, a special form of conflict, is most likely at the power transition point (Organski 1958; Organski and Kugler 1980). Those favoring the balance of power explanation argue that war tends to occur when one nation is preponderant, far to the right in figure 3 (Morgenthau 1948; Kissinger 1979, 1994; Claude 1962). Despite the debate between these two academic camps, one commonality emerges. Both portray conflict behavior as the result of the dyadic power relationship. In chapter 2, I explore the power transition— balance of power debate in order to develop a better understanding of just how the p, versus p, discrepancy might cause c,, and c,, to increase and decrease. Acting on suggestions gleaned from the theoretical and empirical literature on the balance of power and power transition explanations of war in chapter 2, I design a mathematical model in chapter 3. This model

relates four variables: a nation’s power level, a rival nation’s power level, conflict the nation directs at its opponent, and conflict directed at the nation by the opponent. None can be thought of as the dependent variable. The change in any one over time is a function of at least two of

6 The Power-Conflict Story the others. Hence the distinction between premise and conclusion, cause and effect, antecedent and consequence, becomes blurred. Writing the Story

I have become convinced that the only way researchers will understand what is going on in such a complex world is to use more sophisticated tools than they have in the past. Some (e.g., Bull 1969) claim that it is impossible to use rigorous methods to understand international politics because the very nature of the subject is too complicated. I believe, however, that this is precisely why we need more advanced, less simplistic tools. We observe relationships among multiple variables that are nonlinear, nonmonotonic, and change over time or under various cond1tions. In order for us to comprehend such a vast, interwoven, and complicated set of relationships, we need to progress beyond our statistical building blocks. Specifically, as I shall argue in chapter 2, the investigation of the relationship between power distributions and the outbreak of war needs to move away from determining the strength of the statistical association between two nominal or ordinal scale variables and from estimating the beta coefficients in a standard linear regression “model,” a linear probability “model,” or a logit or probit “model” of the outbreak of war. Only then will the elaborate relationship between power distribution and conflict behavior be understood. The tool I choose in order to adequately capture this engaging and complicated set of processes is a system of differential equations. In chapter 2, I expand on the rationale for this choice, noting its appropriateness for the substantive matter at hand.

In Defense of a Storytelling Tradition

In building a system of equations to represent dyadic power-conflict relationships, I have followed the Richardson tradition by telling a sequence of related small stories, each of which is carefully expressed mathematically. The result is a well-constructed and interesting formal model. To give a taste of this process, I include the following short passage depicting how Richardson uses storytelling to build the first component of an arms race model in Arms and Insecurity (1960a).

Permit me to discuss a generalized public speech, fictitious, but typical of the year 1937. The Defense Minister of Jedesland, when introducing his estimates, said:

An Introduction to Storytelling 7 The intentions of our country are entirely pacific. We have given ample evidence of this by the treaties which we have recently concluded with our neighbors. Yet, when we consider the state of unrest in the world at large and the menaces by which we are surrounded, we should be failing in our duty as a government if we did not take adequate steps to increase the defenses of our beloved land.

We now have to translate that into mathematics... . [T]he simplest representation of what that generalized defense minister said is this:

a’

dx

where f is time, x represents his own defenses, y represents the menaces by which he is surrounded, and & is a positive constant, which will be named a “defense coefficient.” (14) Richardson accomplishes a remarkably beautiful storytelling style. Moreover, his completed story, a formal model, is parsimonious and charm-

ing. I depart somewhat from the Richardson tradition because rather than developing my own story de novo I borrow heavily from stories that

already exist in the literature. Specifically, I take two stories that are commonly presented as opposed and assume that each is right some of the time. My contribution to storytelling is found in the choices I make concerning which tale to draw on at which point in time or under which conditions. The effect is to refocus the reader’s attention on a larger, more comprehensive story. At the same time, a less myopic look at the power-conflict relationship makes a contribution to the substantive understanding of international conflict by integrating two opposed schools of thought. The power transition and balance of power explanations of war are typically

portrayed as competing stories. In their review of the distribution of power literature, Siverson and Sullivan state that the two theories “make completely opposite predictions about the effect of the equality of power in the international system” (1983: 474).2 Not only are the theo-

ries treated as competitors, but so are the implications gleaned from empirical data. When authors find support for one story, they believe it simultaneously contradicts the other story. Some examples of this type of reasoning (Houweling and Siccama 1988 and Organski and Kugler 1980) are presented in the next chapter. It is also evident in the theme of

8 The Power-Conflict Story Siverson and Sullivan’s piece. After assessing the quality of the research bolstering each explanation of war, they then determine which side in the dispute claims the “greater weight of evidence” (1983: 475). Findings

placed on the balance of power side of the scale offset those placed on the power transition side and vice versa. This type of reasoning is comfortable because it is akin to the standard social scientific practice of confirming a working hypothesis by rejecting the null hypothesis. Such an approach may allow us to cumulate knowledge in what Zinnes (1976) calls an “additive” fashion, but it does not contribute much to the “integration” of findings into unifying theories. Most and Starr levy the same complaint: . . . [A]nalysts’ descriptive sense of international relations and foreign policy phenomena has substantially expanded in quantity and quality. New data sets exist. Scholars cite one another. The understanding of a variety of analytical techniques has greatly improved.

In terms of... “integrative cumulation,” however, the record is generally much less impressive... . Many argue that theoretical understanding has not been greatly advanced. The results do not seem to add up very readily; there is great difficulty in synthesizing seemingly disparate work. Researchers do not seem to be identifying solutions to the theoretical, methodological and policy problems that challenge them. The field seems incapable of bringing closure to important theoretical and empirical questions. (1989: 1-2)

This book does not unify the entire field of international relations, but it does accomplish the task of integrating one specialized area of study,

namely, the distribution of power between two rival nations and its relationship to conflict. The method of integration used here may, however, suggest a useful technique for similarly plagued areas of research. The Denouement After constructing a formal model by carefully piecing together an elabo-

rate power-conflict story, I ask why that story matters. There are a number of important substantive questions the model is able to address, such as:

1. What types of patterns in power growth and dyadic conflict behavior can be expected prior to, during, and after power transitions?

An Introduction to Storytelling 9 2. When, or under what conditions, does dyadic conflict decrease?

3. How might a nation preserve some minimal level of national power and simultaneously avoid high levels of conflict with its opponent? 4. What types of situations lead to power transitions? 5. Under what circumstances can they be avoided?

This book is almost exclusively concerned with the first question. In addition, the writing and modeling processes generated a number of side-tales, which must be temporarily shelved. Because I am continually

forced to make simplifying assumptions in order to progress with the model building and analyses, interesting questions such as the role of alliance partners are set aside. Hence, the footnotes are littered with even more potential story lines. The primary questions that are not addressed in this book, numbered (2) through (5) above, together with these side-tales, can be thought of as potential sequels to this story, that is, as a future research agenda. Recapitulation

Up to now, I have only indirectly discussed the organization of this work. Here it is spelled out explicitly. In the next chapter, I present the theoretical and empirical literature concerned with the balance of power and power transition explanations of war initiation. The debate over which explanation is best, along with some of the directions suggested by the empirical work, are drawn on to build a formal model of power and

conflict in chapter 3. In order to answer question (1) above, chapter 4 analyzes the behavior of the model near power transitions using simulation techniques. An assessment of how well the model’s simulations match empirical reality is found in chapter 5, which combines systematic data analysis with three historical case studies. Chapter 6 assesses the model’s performance and makes recommendations for future research.

CHAPTER 2

Gathering Pieces of the Story

In the late nineteenth century, Germany experienced unprecedented economic and military growth. By the early twentieth century, Germany had approached and then surpassed Britain as the dominant European

power. This transition in power among the European nations led to World Wars I and II, two of the most extensive and bloody wars in modern history. Germany’s provocative strategy ultimately resulted in its total defeat; the nation was partitioned, not to be rejoined until 1990. During the Cold War, the United States and the Soviet Union shared approximate parity and joint dominance of the international system. Despite (or due to) their persistent rivalry and shared ability to inflict harm on one another, they never fought a large scale, direct, military confrontation. In the early 1990s, the Soviet Union dissolved, and a

weaker Russia took its place as a major player in the international arena. These two seemingly different stories, as well as many others, lead me to ask what 1s the underlying relationship between dyadic power distributions and conflict behavior. This relationship has long been central to the field of international

relations. Power, in the tradition of realpolitik, is thought of as the currency of politics among nations, and international conflict is the nor-

mal outgrowth of power relations. Our understanding of the powerconflict relationship has been hindered by the failure to resolve a long-

standing debate between two schools of thought, balance of power (BOP) theory and power transition (PT) theory. Scholars favoring the BOP theory argue that approximate parity brings peace, as in the case of the Cold War. Those favoring the PT theory believe power parity is a war-prone condition, as in the case of Germany’s challenge to British superiority. So entrenched is this debate that scholars have characterized the two sides as “opposing opinions” (Garnham 1976a: 231), “fundamentally different hypotheses,” “diametrically opposed,” and apparently “in-

compatible” (Bueno de Mesquita 1989: 151). In their review of the power distribution literature, Siverson and Sullivan state that the two theories “make completely opposite predictions about the effect of the equality of power in the international system” (1983: 474).! 10

Gathering Pieces of the Story 11 In order to see how these two theories seem to produce fundamentally different conclusions, it is necessary to examine them more closely. The first section of this chapter accomplishes that task by briefly summarizing the original BOP and PT arguments. My goal, however, is not to settle the BOP-PT debate. In the next four sections, I use the debate asa

means of suggesting key variables in the power-conflict story and the ways in which these variables might be related. In the first, I consider three areas of research that emphasize the dynamic nature of the powerconflict relationship: empirical studies, a cyclical theory of national power, and rational choice models. In the second, I examine empirical results that yield mixed support for the two theories, underscore the puzzles these results present, and propose a solution based on three regions of conflict behavior. ‘Third, I argue that the traditional treatment of war as a dichotomous variable could be improved upon by shifting to the more general concepts of conflict and cooperation. Fourth, the possi-

bility of incorporating two variables commonly thought to impact the power-conflict relationship is considered. The last section summarizes the specific modeling implications reached in the previous sections and lays the foundation for building the formal model presented in chapter 3. A Tale of Two Tales

The Balance of Power Explanation

In general, it seems that the BOP explanation in an explicitly developed form is somewhat elusive. It has evolved in bits and pieces, claiming roots in classical international relations theory. In the modern era, writers such as Morgenthau (1985) and Dougherty and Pfalzgraff (1981) have provided synthesized, complete versions of the BOP theory. Scholars of this persuasion (e.g., Kissinger 1979, 1994; Liska 1962; Waltz 1979; Morgenthau 1985) argue that an approximately equal distri-

bution of power (or capabilities) across a system of nations tends to produce a peaceful equilibrium. Extension of this argument from a system of nations to a dyad has been achieved in the deterrence literature (e.g., George and Smoke 1974; Snyder 1961). Here the view is that two approximately equal rivals will deter each other from initiating an attack, so war is unlikely. Justification for moving from systemic to dyadic analysis is presented by Siverson and Sullivan (1983: 474), who claim that “the underlying rationale for the hypotheses at both levels is similar enough — at least in terms of general theory building — that it would be advantageous to compare them.” Additionally, the authors concur with

12 The Power-Conflict Story Singer’s (1980: 359) proposition that most system properties come from characteristics of its parts. Last, it may simply be the case, as it is here,

that we are interested in the behavior of a (major power) dyad as opposed to the behavior of a system of nations. Such a focus is not at all unusual. Wayman and Singer, in a review of the vast and diverse research associated with the Correlates of War Project, note that “whereas the [initial] outlook led one to expect that the systemic level would be

more important than the dyadic, it now appears that it might be the other way around” (1990: 2). Scholars of BOP reason that when the power distribution is roughly equal each alliance’s or nation’s capabilities serve as a check and balance against ageression by the opponent(s). Since neither nation has a clear

advantage, both are uncertain about their chances of winning a war. Finding war too risky in this sense, both rivals choose not to initiate one. If the balancing mechanism breaks down, however, and one alliance or

nation becomes preponderant, war is more likely than it was during equality. Thus, preventing an opponent from gaining an advantage 1s crucial to self-survival because a preponderant nation will aggress against its competitor. We might liken this to the para bellum adage, “if you want peace, prepare for war.” According to this logic, war may occur in one of two ways. First, the theory postulates that the reason balance is essential for peace is that by nature nations are aggressive and seek to maximize power. Nations with unchecked power are likely to exercise these aggressive tendencies, potentially by initiation of war with weaker nations. The logic of the balancing mechanisms certainly does not exclude, however, the possibility that

war is initiated by the weaker state(s) in an effort to return the power distribution to equilibrium. If other balancing efforts fail to prevent domination or are insufficient, lower-ranked nations may possess no alternative to war. According to this scenario, conflict is a method that can be used to adjust the power relationship. In sum, the classic propositions of the BOP explanation are: (1) under relatively equal distributions of power, war is less likely; (2) under unequal distributions of power, war is more likely; and (3) war is typically initiated by the stronger nation. The Power Transition Explanation

The PT explanation is based on: (1) an S-shaped development of a typical nation’s power over time and (2) a logic governing the interaction of nations at different positions in their development (Organski 1958).

Gathering Pieces of the Story 13

According to Organski (1958), the natural processes of development, modernization, and industrialization internally drive a nation’s power growth. A typical nation thereby experiences three stages of devel-

opment: (1) potential power, (2) transitional growth in power, and (3) power maturity (340). As was pointed out in the first chapter, these stages are descriptive of portions of an S-curve. This curve, as shown in

figure 2, measures a nation’s power across time. Initially a nation’s power grows gradually, while its economy and political system are still

underdeveloped. In the transitional stage, a nation experiences rapid growth as its economy industrializes and its government bureaucracy expands. Finally, the growth levels off, and possibly declines, vis-a-vis other nations experiencing a rapid accumulation of power in their second stage of development.°?

Because all nations do not develop simultaneously, at any given time there will be a mix of nations at various points in their stages of development. The PT explanation is principally concerned with nations at the most developed stages.* Some of these nations will not be “satisfied with the way the international order functions and the leadership of the dominant nation” (Kugler and Organski 1989: 73), and as their power levels approach that of the dominant nation they will be increas-

ingly able to act on this dissatisfaction. It is at this time, when the dominant nation is overtaken by a dissatisfied powerful nation, that war

is most likely. The overtaking nation initiates such a war because it “anticipates greater benefits and privileges if a conflict is successfully waged than if the current status quo is preserved” (175). War is least likely when the dominant nation’s power far exceeds that of the others because the dominant nation has no reason to initiate a war (it gains nothing) and because the others do not possess enough power to initiate war (the probability of success is low). This rationale is similar to that found in Bueno de Mesquita’s (1981) expected utility analysis of war initiation. Transition scholars expect not only that war is most likely near power

transitions but that it is most likely just prior to them. For example, Organski and Kugler suggest that “it is an attempt to hasten this passage that leads the faster-growing nation to attack” (1980: 28). By implication, the rising challenger uses conflict as a method of shifting the dyadic power distribution more quickly than it would shift on its own. Furthermore, the challenger’s motivation to attack exists only before the transition takes

place. The empirical evidence I discuss in this chapter refers only to information concerning the general prediction that wars are most frequent near transitions. A discussion of Organski and Kugler’s subsequent investigation of the historical record on the more precise expectation 1s

14 The Power-Conflict Story reserved for chapter 6, where I directly compare the performance of my model’s predictions with those of the PT explanation. In sum, the classic propositions of the PT explanation are: (1) war is least likely when one nation’s power is clearly dominant, (2) war is most

likely when another nation’s power threatens to overtake that of the previously dominant nation, and (3) the overtaking nation is the likely war initiator.

Why Use a Dynamic Model?

The stories told by Organski in The Stages of Development (1958) and by Morgenthau in Politics among Nations (1985) are rich in dynamic descrip-

tions of national growth and competition. The former details the processes of economic, social, and political development. The latter elaborates the forces pushing nations to strive toward power preponderance and the forces counteracting them in order to maintain an equitable balance. These two stories have in common the idea of change over time. Contemporary authors attempt to capture these fluctuations by incorpo-

rating pseudodynamic features into their statistical investigations (Organski and Kugler 1980; Houweling and Sicamma 1988; Kim 1989; Singer, Bremer, and Stuckey 1972), by developing and testing a model of national power (Doran 1971, 1989a, 1989b; Doran and Parsons 1980; Spiezio 1993), or by introducing a time component into a rational choice model (Morrow 1996; Kim and Morrow 1992; Powell 1996). None of these approaches has thus far proved satisfying. Dynamic Elements in Empirical Research

Investigators using statistical approaches attempt to determine the effect of dynamic components of the two power theories by devising variables that measure some dynamic element at various points in time. Examples

include whether or not the challenger’s growth rate is larger than the dominant nation’s (Organski and Kugler 1980; Houweling and Sicamma

1988), the relative growth rate of the challenger and dominant state (Kim 1989), and the level of change in the power distribution (Singer, Bremer, and Stuckey 1972; Mansfield 1992). These empirical studies reveal strong dynamic effects but do little to explain how those effects work. Organski and Kugler (1980), Houweling and Sicamma (1988), and de Soysa, Oneal, and Park (1997) acknowledge the dynamic features of a power transition by specifying two kinds of equality, one in which neither

Gathering Pieces of the Story 15 competitor is gaining an advantage over the other and a second in which one nation’s rate of power growth is larger than the other’s rate. These

two types of parity are labeled “equal, no overtaking” and “equal and overtaking.” A case 1s coded as an overtaking if the nation that lags at the

beginning of a test period is ahead at the end. Because the resulting contingency tables indicate that dyads that experience overtaking are most likely to go to war, this type of change is clearly important. Similar tables constructed by de Soysa, Oneal, and Park lead them to report that “overtaking is the strongest predictor of war” in a variety of empirical tests (520). None of these tables demonstrates the workings of the change itself. This type of research also limits any notion of change to the transition alone, even though changes in both nations’ power levels have certainly preceded the crossing. Other authors develop more precise measures of power change. For example, a dyad’s “relative growth rate” is assessed by subtracting the lower growth rate from the higher one for the two rivals, where each

nation’s absolute growth rate is approximated with the percentage change in its mean power level over two sequential time periods (Kim 1989; Kim and Morrow 1992). “Velocity of capability change” is simi-

larly calculated with power ratios instead of differences (Schampel 1993). Whether or not the challenger is “committed to change” is indicated by unprecedented increases in military expenditures (Werner and Kugler 1996; Lemke and Werner 1996). The notion of “power shifts” measures capability convergence or divergence (Wayman 1989, 1996; Gochman 1990; Geller 1993). In their classic research on the effects of systemic capability distributions on the amount of war, Singer, Bremer, and Stuckey (1972) develop two variables to measure shifts in the power distribution among the major powers. The first, ACON, detects changes in the systemic concentration of power. The second variable, MOVE, picks up the movement of shares among nations, even when that move-

ment does not result in a change in the systemic concentration. Although the relative growth rate variable does not play a statistically significant role in predicting war (Kim 1989; Kim and Morrow 1992), the velocity of change, commitment to change, power shifts toward conver-

gence, ACON, and MOVE variables do (Schampel 1993; Werner and Kugler 1996; Lemke and Werner 1996; Wayman 1989, 1996; Gochman 1990; Geller 1993; Singer, Bremer, and Stuckey 1972). Power changes in

general, not only those producing transitions, are important predictors of war.

Although many dynamic variables indicate a meaningful impact of power change on war, the mechanics of power shifts are not yet well understood. Part of the problem may well be that the change variables

16 The Power-Conflict Story are coarsely defined and measured, sometimes with information taken from only two snapshots in time. The dynamic power-conflict model seeks to more closely examine both the fundamental idea of power change and its impact on conflict behavior. ‘The conceptual development of the dynamic variables and the empirical evidence of their worth suggest the necessity of dynamic features in any model purporting to examine the power-conflict relationship. Dynamic Elements in a Cyclical Theory of National Power

None of the researchers presented in the preceding review of the empirical literature specifies how one should expect either power or conflict to change over time or what the driving mechanisms behind those changes

might be. I now turn to work that addresses both how national power levels change over time and how conflict is related to those changes. Charles Doran (1971, 1989a, 1989b) developed a historical-socio-

logical theory to explain how great nations experience cycles in their relative power levels over time. By “relative power,” Doran means a nation’s power defined in terms of its own capabilities in relation to the capabilities of all the nations in a particular referent group, namely, the major power system. This is very similar, then, to the Correlates of War’s system share of capabilities indicator, presented in chapter 5. A nation’s power cycle is “marked by ascendancy, maturation, and decline” (Doran 1989b: 85). This cycle parallels Organski’s three stages of development and was popularized in Kennedy’s 1987 book, which

traces the rise and fall of great economic powers from 1500 to the present. The approach 1s a relatively broad one. Instead of focusing on a single, short-lived type of change, such as a power transition, it en-

compasses the general pattern of change in the lifespan of a nation. The generalized theory is depicted in figure 4, an adaptation of Doran and Parson’s (1980) figure 1. Here, growth in relative power is selfdriven; that is, it depends only on itself, much like the population model in chapter 1. Four “critical points” along this curve represent unique moments in

a nation’s development. At the two turning points, t = a and t = ¢, all change in a nation’s relative power has temporarily stopped, that is, the velocity equals zero. In figure 4, this can be seen as a leveling out of the power curve at t = a (and later at t = a’) andt = c. At the two inflection

points, t = b and t = d, the rate of increase peaks and the rate of decrease reaches its lowest level, respectively. In both cases, acceleration, the change in the change of relative power, equals zero. Doran and Parsons (1980) argue that at each of these critical points a

Gathering Pieces of the Story 17 Relative Power Turning point

Inflectior/point Inflectiompoint

Turning/point Turning point

a b c d qa’ time Fig. 4. Doran and Parsons’s (1980) cycle of relative power. (Adapted from C. Doran and W. Parsons, “War and the Cycle of Relative Power,” American Political Science Review, vol. 74, fig. 1, p. 948, by permission of Charles F. Doran and the American Political Science Association.)

nation 1s anticipating a continuation of the pattern immediately preceding that time period and is somewhat surprised by a sudden shift in either the direction or the rate of change. The following serves as an example. Beginning at t = a, where a nation’s relative power level experiences neither growth nor decline, the change in relative power continues

to increase until ¢ = b, when acceleration ceases. But the nation in question expects power changes to continue to become larger and larger and is therefore shocked to see that increases actually become smaller and smaller. A similar scenario, in which expectations based on shortterm linear projections are not met by a curvilinear reality, ensues for each subsequent critical point. This presents problems for national decision makers.

[A]t the critical points... the linear or monotonic projections of the state leaders suddenly become counter to the real trend in direction as well as degree. . . . Forced unexpectedly to search for new foreign policy roles, the state feels threatened by uncertainty and is vulnerable to over-reaction. (Doran and Parsons 1980: 951)

This overreaction often manifests itself as a war initiation. Later, Doran expands on the notion of overreaction to include that of other states in the referent group.

18 The Power-Conflict Story Reactions of the state to the system (and, conversely, of the system to the state) are jointly responsible for the misperception and anxiety that heighten the likelihood of involvement in a major war when the state abruptly finds itself in a position that radically alters its view of where it will stand in the future. (Doran 1989b: 90)¢

One would expect, therefore, that the incidence of both war initiation and war involvement will be high near critical points. Doran and Parsons (1980) investigate the relationship of critical points and initiations of extensive wars. After a cyclical curve is fit using time-series data on relative power for each of the nine major powers, the curve for each nation is analyzed in order to determine the timing of the critical points. A statistical analysis reveals that at critical points the probability that a state will initiate an extensive war is the greatest (963). A later investiga-

tion (Doran 1989a) uses a more sophisticated curve-fitting technique and bolsters support for the power cycle theory. Doran’s theory makes important contribution in two ways: it progresses past the one or two snapshots in time to a full-length picture of power cycles and it 1s supported by empirical evidence that demonstrates the ability of power change concepts to explain interstate war. There are drawbacks, however. Some are discussed below. First, Doran and Parsons (1980) and Doran (1989a) use a statistical

model that is only related to the theoretical model in the similarity of curves that both produce. The structural form of the equation does not represent the theoretical story of cycles in relative power told by Doran in 1971. Because it still demonstrates cyclical behavior, it is used as a simple curve-fitting device (see the discussion of this issue in chapter 1). Incorporating aspects of theoretical frameworks into a formal model is one of the major contributions of this book. Second, war initiation (or involvement) is not directly built into the statistical model. Relative power levels clearly change over time, but the relationship of these changes to conflict dynamics is not formally developed. To explain a nation’s propensity to initiate war at critical points,

Doran and Parsons (1980) instead return to the driving analogy and liken the initiation decision to oversteering a curve in the road. Such a mistake has the direct effect of deflecting a car’s path off the road, so that its new position is somewhere in the ditch. These verbal arguments

are used to persuade the reader that the likelihood of war should be highest at critical points, but there is no mathematical translation. Closer consideration of conflict dynamics might reveal subtleties heretofore unconsidered. Spiezio (1993), in an extension of Doran’s work from

Gathering Pieces of the Story 19 war initiations to militarized interstate disputes, demonstrates that the initiation of disputes is associated with Doran’s critical points less often than they would be by chance. Although this finding is consistent with Doran’s theoretical expectations, it does imply that power cycle theory is limited to explanations of very high levels of conflict alone. Why might

this be? A more careful explication of the relationship between power and conflict is in order.

Third, war initiation does not affect the relative power trajectory. Doran summarizes the impact of war on a nation’s power by emphasizing that “shifts in power on the cycle and the event of war itself are separable: abrupt change on the power cycle causes war, not the other way around” (1989b: 93). I argue instead that such a rigid characterization of power as the independent variable and conflict as the dependent variable is limiting and theoretically dull. The formalized power cycle model also does not directly specify an influence of power on war. The mathematical model designed in chapter 3 includes several interesting relationships between

power and conflict, so the independent/dependent variable distinction vanishes.

Last, the power cycle theory seems to present war as a unilateral activity. It is difficult not to wonder what potential rivals are doing when a nation reaches a critical point. Houweling and Siccama (1991) raise precisely this issue. They argue that critical points alone are necessary but not sufficient causes of war involvement. All war involvements among

dyads composed of the three or four top-ranking nations (or “contenders”) are preceded by acritical point, but not all critical points lead to war involvements. So, the authors search for an additional condition such that its joint occurrence with a critical point will be a sufficient for war. Not surprisingly, they settle on the power transition. Although a critical point together with a power transition still does not always produce war, the authors conclude that for the top three or four contenders “the presence of a critical point nation in a dyad that is experiencing a power transition is much more dangerous than a critical point in a nontransition dyad” (648). A greater amount of explanation is clearly gained by incorporating a dyadic story. Houweling and Siccama still stop short, however, of developing a fully dynamic model of this dyadic interaction.

Just as the relative power cycle theory does, the power-conflict model begins with elements of a basic power change model. This is done by carefully translating theoretical arguments into mathematical expressions. Additional contributions include the formal modeling of conflict behavior, specification of how conflict affects and is affected by power, and a dyadic approach to understanding conflict.

20 The Power-Conflict Story Dynamic Elements in Rational Choice Models

Further progress in developing a dynamic model of power and conflict in

a dyadic rivalry has been made in the rational choice literature. The most straightforward and common way that dynamic elements can be incorporated into rational choice models is by assuming that the dominant nation’s probability of success decreases over time while the challenging state’s rises (Kim and Morrow 1992; Morrow 1996; Powell 1996). The result is a simplified version of power growth. Simplifying assumptions are the business of formal modelers. They allow us to view the world in a more understandable way, perhaps by allowing us to find solutions to equations, to focus on Key relationships, to filter out extrane-

ous or secondary effects, to separate large problems into smaller and more manageable pieces, and so on. Just as they provide benefits they also come at a price. They narrow our focus, excluding potentially interesting features of the phenomenon of interest and limiting the kinds of questions we can ask. In balancing the costs and benefits of assumptions, the trick is for each researcher to identify key questions, choose the features that are most relevant, and use the tool that is best suited to answering those questions and capturing those features. With this in mind, I assess the usefulness of game-theoretic models of power-conflict dynamics.

Kim and Morrow model the probability of success for the challenger as a linear function of time (1992: 902). The challenging nation’s probability of success in a bilateral war with the dominant nation will be

the least before a power transition, a tossup at the point of transition, and the most following a power transition (Kim and Morrow 1992). Powell similarly assumes that the challenger’s utility of war and utility for peace are strictly increasing functions of time because the challenger’s probability of winning a war also increases over time (1996: 752,

759). In the incomplete information version of his model, Powell does add the possibility that the declining state’s utility for war eventually bottoms out and then rises by a small amount (758). Incorporating this expectation of “eventual improvement” produces a unique equilibrium that is otherwise unattainable. Thus, Powell explores one small way in which power levels might exhibit more diverse patterns of fluctuation. These simplified versions of changes in two rivals’ power level are perfectly reasonable assumptions for certain purposes. Game theorists’ models of power transitions are designed to examine the strategic decision making between the challenger and the dominant nation during a power transition. One question Powell (1996) ponders is the dilemma of the declining state: the longer it waits to stand firm, the less likely it is to

Gathering Pieces of the Story 21 win a war, but standing firm risks fighting an unnecessary war when the challenger’s demands are actually rather limited. He wants to determine when the declining state draws the line and refuses to yield to demands from the challenger. Kim and Morrow (1992) want to know how factors such as shifts in power (the probabilities of each rival winning a war), the risk propensity of the rivals, the expected costs of war, the rising state’s dissatisfaction with the status quo, and power equality affect the probability of war. Both of these games encompass only the short time period surrounding a classic Organski-style power transition.’ If the goal is to

examine decision making in this brief context, then a game-theoretic model with a simplifying assumption about the change in the dyadic power distribution over time makes good sense. Just as a line can meaningfully represent a small portion of a circle, the power levels of a challenger and dominant nation in the traditional PT story can be usefully approx1-

mated with monotonic (or virtually monotonic) trajectories. But the goal here is somewhat different, namely, to examine the causes and resulting patterns of a long-term process that tends to wash out the more myopic and immediate decisions. Power transitions are one manifestation of a complex process involving the evolution of national power, dyadic conflict behavior, and the feedback between the two. The broader goal of this book is to understand that process. Examining power levels and behavior near power transitions is merely one useful application of the power-conflict model, but the model is not a model of power transitions per se. As is discussed briefly below and in detail in chapter 4, the dynamic model does produce new and meaningful conclusions concerning power transitions themselves. Not only does this book’s modeling approach offer unique perspectives on power transitions; it allows for further analysis of nontransition issues concerning the power-conflict relationship such as determining when conflict is likely to decrease, how nations can preserve power levels while avoiding conflict, and the feasibility of maintaining a long-term peaceful balance. Given this focus, a differential equations model is better suited to capturing the essential features of the phenomenon. Power cycles in the life of nation-states, especially those that become major powers, frequently span centuries. During these longer time periods, many underlying and counteracting forces influence national power levels. In the extant game-theoretic models, national power (or the probability of success) is solely a function of time. This book’s model focuses on two other important forces—the positive effect of natural erowth and the costly effects of conflict with a rival. As a result of these two competing forces, the dynamic model produces power changes that are not always of the “one rival going up and one coming down” variety

22 The Power-Conflict Story (see chapter 4). Empirically, not all power transitions are characterized by this simple description (see chapter 5 and Wayman 1996). One central deduction from the dynamic model presented here is that there are three distinct types of power transitions, only one of which closely resembles the traditional one. The logic of differential equations allows me to build arguments about how a variable changes over time by considering the magnitude of changes, the speed of changes, the direction of changes, and the cause of changes. In turn, these arguments produce a more complex picture of power cycles and more nuanced types of power transitions. Rivalries between competing nations also often last for decades and

result in multiple disputes and wars (e.g., the United States and the Soviet Union, China and the Soviet Union, France and Germany, India and Pakistan, Israel and Syria). The ebb and flow of competition is not easily reflected in short-term and discrete decisions of whether or not to fight. It is better handled by considering the levels of conflict that two rival nations, X and Y, direct toward one another, c,, and c,,, as continu-

ous variables that encompass a broad range of both conflictual and cooperative behaviors. Their values might represent full-blown interstate war (high positive values), mobilization of troops (midrange positive values), verbal attacks (low positive values), statements of approval (low negative values), or unification (high negative values). Patterns in rising and declining conflict levels might be accounted for by escalation dynamics. Over time, opponents’ conflict behaviors toward one another tend to exhibit an action-reaction mechanism. Accordingly, one nation’s

hostile behavior leads to increases in the opponent’s, which in turn increases the first nation’s hostility level, and so on. These types of interdependencies over time, or feedback loops, can be readily captured using a system of differential equations. By using more than one equation, I can incorporate the idea that two or more variables can impact each other’s changes over time. Another type of feedback between variables involves the impact of

conflict on power distributions. In current game-theoretic models of power transitions (Kim and Morrow 1992; Morrow 1996; Powell 1996),

the decision of whether or not to go to war is the final outcome. This decision is primarily influenced by fluctuations in the probability of each side winning, that is, the dyadic power distribution.’ But the decision of

whether or not to go to war does not have an impact on subsequent power distributions. In the differential equations model, once the decision to go to war is generalized as a continuous variable representing conflict behavior each rival state’s conflict behavior becomes a component in the negative force operating on the other’s power level.

Gathering Pieces of the Story 23 It would be foolish to claim that game theory is completely unable to capture dynamic ideas. Many game theory texts discuss techniques for

incorporating dynamic elements into games (e.g., Kreps 1990; Ordeshook 1986; Morrow 1994). But differential equations can more elegantly capture ideas such as long-term processes, underlying causes of power growth and decay, the interdependencies among two states’ power levels and conflict behaviors, conflict behaviors short of war, and

escalation dynamics. More specifically, game theorists might have a harder time including several of these features within a single tractable game. Powell notes the tradeoffs inherent in using a game-theoretic model to address process questions. In order to focus on the timing of a significant decision, which in this context is the rising state’s decision to stop making further demands and the declining state’s decision to draw the line, the players’ set of options in simple timing games is extremely limited... . [T]he present formulation really only examines at most half the problem a declining state may face. In addition to being uncertain of the rising state’s minimal demands, the declining state may also fear that if it stands firm for defensive reasons, even after satisfying the rising state’s minimal demands, this action may be interpreted as a threat. (1996: 753)

Because he wants to investigate the timing of a decision, Powell sacrifices the variety of choices the players can make and the strategic essence of the game. Yet these are precisely the features that make game theory so useful. A model designed to suit the purposes of this book will necessarily be much different than one designed to understand the short-term workings of a traditional power transition. The most parsimonious method of dealing with those features is by using a system of differential equations. This allows one to explore not only how power and conflict behavior

evolve over time but how they are each a function of one another. Existing game-theoretic models say little about how power and conflict

behavior change over long periods of time and even less about how conflict behavior influences power levels.’ Given the BOP and PT literatures’ strong themes of changing power levels, shifting power relationships, and the importance of those changes in determining conflict, and given the proposition that conflict levels dynamically affect power distributions, I abide by Lave and March’s first rule of modeling social phe-

nomena: “Rule 1: Think ‘Process.’ A good model is almost always a statement about a process, and many bad models fail because they have

24 The Power-Conflict Story no sense of process” (1975: 40). Furthermore, I choose differential equations as a formal language with which to explicitly and rigorously record

my ideas due to their excellent ability to capture both the notion of change over time and the interdependent nature of Key variables, the two main ingredients of a process. The Power-Conflict Relationship

More explicit insight into what the dynamic model should look like is gained by considering the vast literature that examines the empirical relationships between power and conflict. ‘This section begins the consid-

eration of that literature. Its central task is to characterize the evidence that supports either BOP or PT theory. Throughout, I focus on studies that address dyadic power distributions and war. Treatments of systemic power distributions and the outbreak of war are presented to the extent that they inform the dyadic debate. Numerous attempts to settle the preponderance-transition debate by examining the real-world behaviors of nations have been made. Differences in time periods, operational definitions, universes of analyses, and statistical methods yield inconsistent conclusions among studies; some stud-

ies strongly support the notion that balances are peaceful (Ferris 1973; Siverson and Tennefoss 1984), while others clearly indicate that parity is dangerous (Organski and Kugler 1980; Houweling and Siccama 1988). A closer examination of the empirical work reveals that a single study often

supports both theories. In other words, there are mixed results because conclusions from different studies are contradictory and also because individual analyses yield evidence in support of both theories. An overview of Organski and Kugler’s seminal contribution to the empirical literature on power distributions and war demonstrates this point. Mixed Results in The War Ledger

In The War Ledger, Organski and Kugler (1980) perform empirical tests

of the two “power distribution models.”!° Their investigation begins with the division of the years of analysis, 1860-1975, into six test periods of approximately twenty years in length. The test periods are devised by

taking the initiation of a war as the termination year of a test period. Twenty years is subtracted from the date of war in order to determine the beginning of a test period. This is done for only four major power wars: the Franco Prussian War in 1870,"! the Russo Japanese War in 1904, World War I, and World War II. World War II and the Russo

Gathering Pieces of the Story 25

Japanese War fall into the same test period. The years 1914-18 and 1940-44 are excluded to eliminate years of fighting from the analysis. The remaining years are divided into three more twenty-year intervals.

The relevant nations in each test period are then paired into dyads whose power distributions for an entire period are assessed. As a first cut, Organski and Kugler begin with the entire group of major power nations, which form 126 dyads across the six test periods, or an average of 21 dyads per test period. Dyads are initially classified as

being equal or unequal during a test interval and as warring or not warring. Next, each equal dyad is classified as being simply equal or equal with one nation overtaking the other. This distinguishes cases in which a balance of power is stably equal from those in which it is not or when one nation is overtaking the other. The results appear in table 1. Equal and overtaking dyads have the highest risk of war, confirming that power transitions are indeed dangerous. Fixed equality, on the other hand, is always peaceful, lending support to BOP logic. Because both theories are supported, Organski and Kugler lament the “stubborn reluctance of the data to disconfirm one of the hypotheses” (1980: 51). They are thus motivated to further classify the dyads by type of nation. Major powers are divided into those on the periphery and those in the center. Of those in the center, some are classified as contenders. The group of contenders represents the transition school’s focus on the dominant nation and the few nations with the potential to challenge the dominant nation. It consists of either (1) the strongest nation in the major power system and any other nations possessing at least 80 percent of the strongest nation’s power share or (2), if no nation possesses 80 percent of the strongest nation’s power share, the top three nations. Only the results for contenders (see table 2) are more interesting than

those found for the entire group of major powers. The authors note TABLE 1. Organski and Kugler’s First Look at Power Distributions and War Power Distributions

No 81 11 15 (86.2%) (100% ) (71.0%) Yes 13 0 6

War Unequal Equal, No Overtaking Equal, and Overtaking

(13.8%) (29.0%)

Source: Information was taken from table 1.6 of The War Ledger (Organski and Kugler 1980: 50). Reprinted by permission of the University of Chicago Press.

Note: Dyadic power distributions from 1860 to 1975 are included. N = 126, Tau C = .05, not significant. Equality is defined as a difference no larger than 20 percent of the weaker’s power level.

26 The Power-Conflict Story that “If conflicts occur among contenders, they do so only if one of the contenders is in the process of passing the other” (Organski and Kugler 1980: 51). Table 2 additionally indicates that contender power transitions result in war only half of the time. Power transitions among con-

tenders are thus a necessary, but not sufficient, condition of war. An alternative way of characterizing this information would be to note the conditions for peace. Unequal contender dyads do a perfect job of avoiding war. Moreover, when major power dyads of any type are stably balanced, that is, they are equal and not overtaking, war never results. Thus BOP arguments cannot be dismissed. In sum, contender dyads have two conditions for peace —a stable balance and inequality.

A peaceful and stable balance is consistent with BOP expectations, while a peaceful inequality is consistent with PT expectations. Puzzles

Organski and Kugler’s findings therefore suggest that both the BOP and PT explanations are useful. Because the universe of analysis in this early

work is limited, one might be inclined to only cautiously accept this conclusion. Fortunately, later studies widen the scope of analysis by expanding the time frame, focusing on the broader major power classifi-

cation, or using a more encompassing definition for either the major power or the contender category. Regardless of the specific measure of national power, all of these replications produce the same kind of mixed results. Their contingency tables, like Organski and Kugler’s, clearly indicate that a stable dyadic balance is more likely to produce peace than is overtaking, regardless of the classification of dyads as major powers or as contenders (Geller 1992; Houweling and Siccama 1988; de Soysa, Oneal, and Park 1997). Should international relations theorists continue to pit BOP and PT logics against one another? TABLE 2. Contenders Only

No 4 6 5 Yes 0 0 5 Power Distributions

War Unequal Equal, No Overtaking Equal, and Overtaking

(100% ) (100% ) (50%) (50%)

Source: Information was taken from table 1.7 of The War Ledger (Organski and Kugler 1980: 52). Reprinted by permission of the University of Chicago Press. Note: N = 20, Tau C = .5, significance = .01.

Gathering Pieces of the Story 27 In The War Ledger, major power dyads show inequality to be less peaceful than fixed equality while contender dyads show both of the conditions to be equally war free. Similar uncertainty concerning the conflict tendencies of inequality appears in the replication studies. Some tables indicate that inequality is more peaceful than a steady balance (see Geller 1992: 278; and the portions of tables where power is measured using Correlates of War capabilities in de Soysa, Oneal, and

Park 1997: 521-24). Other tables indicate that inequality is more war prone (Houweling and Siccama 1988: 100-110; and portions of tables where power is measured as GDP in de Soysa, Oneal, and Park 1997: 521-24). Is inequality pacifying or antagonizing?

Further confusion is added by apparently inconsistent findings in several studies that move beyond contingency tables by using a continuous measure of the dyadic power relationship and analyzing its effects on the likelihood of war. Ferris finds that “the greater the disparity in the

power capabilities relationship between states, the greater the probability that the states will become involved in an intense conflict” (1973:

115-16). Geller (1993) and Maoz and Russett (1992, 1993) conclude instead that parity leads to conflict; Bremer (1992) and Lemke and Werner (1996) concur but find the relationship to be somewhat weak. Parity’s contribution to conflict is overshadowed by factors such as conti-

guity, lack of alliances, less advanced economies, and the absence of democracy (Bremer 1992) or by extraordinary military buildups (Lemke and Werner 1996). Some studies actually find that dyadic parity has no

bearing at all on war initiation (Kim 1989; Kim and Morrow 1992). Perhaps most unsettling is Bueno de Mesquita and Lalman’s finding that

their Hegemony variable (the difference between A’s power and B’s power) is inversely related to the probability of war, whereas their PowTrans variable (the difference divided by the sum) is directly related to the probability of war (1992: 212). Why should this more generalized approach produce such contradictory results?

Solving the puzzles of mixed results is the subject of the next section. It suggests that the solution lies in specifying the conditions under which each of the two theories makes sense. In other words, it stipulates different types of conflict behavior across various kinds of power relationships. A Solution: Three Regions

To date, straightforward statistical analyses that assess the relationship between power distributions and the outbreak of war, many of which are presented above, have proven largely unsatisfactory. Neither inequality

28 The Power-Conflict Story War

ns (i |

————— _ Balance of power

| — — — Power transition | |

Inequality

Fig.5. The BOP and PT explanations as step functions

nor equality is clearly associated with war proneness. The work of several authors suggests, albeit somewhat indirectly, that this failure is due to overly simple assumptions about the power variable and the nature of its relationship to war. These simplifications are conceptually distinct, but in practice they are indeed related. Traditional comparisons between the two standard power distribution theories typically present the independent variable, power, as dichotomous. Organski and Kugler (1980) and Houweling and Siccama (1988), for example, characterize the distribution of power as either equal or unequal. Using this simple dichotomy characterizes the BOP explanation by presenting the relationship between inequality and war as a step function that increases to the right; it similarly portrays the PT explanation as a step function that decreases to the right, as in figure 5. Imagine, instead, that the power distribution variable can take on other values as well. In order to get more explanatory mileage from the BOP and PT explanations, I introduce three distinct levels of inequality: approximate parity, a decisive advantage, and overwhelming preponderance. What is the distinction between approximate parity and disparity? Two rival nations tolerate a certain difference in power levels without considering their situation to be unbalanced. Such a condition is considered to be “approximate parity.” Inequality beyond that level is considered to be “disparity.” Ferris (1973: 76) settles on a weaker to stronger “disparity ratio” of about .7 to demarcate the point at which the power distribution reaches inequality.!* When the ratio crosses the disparity threshold, or as it becomes smaller than .7, “the number of war events

increases markedly” (76). Subsequent experimentation with dyadic power inequalities of different sizes (from .7 to .95) confirms the theoretical importance of distinguishing approximate parity from disparity

Gathering Pieces of the Story 29

(Houweling and Siccama 1988; Lemke and Werner 1996; de Soysa, Oneal, and Park 1997). Thus, BOP logic and overwhelming evidence from the empirical work discussed earlier should lead international conflict scholars to expect dyadic interaction under the condition of approximate parity to be quite pacific.

Disparity comes in two forms, a decisive advantage and overwhelming preponderance. Empirical evidence demonstrates that the movement from approximate parity to a decisive advantage for one rival, that is, the “equal and overtaking” scenario, is the most warprone condition. From the BOP perspective, the dangers of overtaking lie with the unchecked advantage of the more powerful nation. From the PT perspective, they lie with the challenger’s desire to assert its newly gained dominance over the formerly dominant nation. In both instances, it is the movement from rough equality to a noticeable advantage that is threatening. Each of these purportedly opposing theories tells us to expect change from approximate parity toward disparity to be highly conflictual.

Very large disparities should instead be quite peaceful. Weede (1976) claims that expectations for conflictual interaction under the condition of disparity are only reasonable up to a certain point, the point at

which one nation is Overwhelmingly preponderant. For contiguous Asian dyads in the 1950s and 1960s, Weede found that “war was much less frequent in the presence of overwhelming preponderance” (409).!°

The author suggests, and his findings support, the idea that a power advantage has a unique pacifying effect when that advantage is of the overwhelming, or ten-to-one, type. Inequality is therefore both conflictual and peaceful, depending on its magnitude. This helps to explain the inconsistencies regarding the “inequality” column of the contingency table analyses of the previous section. Therefore, PT logic governs conflict behavior in the region of overwhelming preponderance. Gross in-

equality ensures that the preponderant nation can dominate without being challenged.

A careful reanalysis of the information provided in the first two tables in Lemke and Werner’s (1996: 251-52) work supports the lowhigh-low pattern suggested by the power-conflict story. ‘Table 3 shows that a disparity of between .8 and .9 1s highly explosive, with between 16.7 and 27.3 percent of the dyads experiencing war (depending on how

national power is measured). It also shows that this rate drops off quickly for disparity levels below .8 and for levels at .9 or above.!4 Recognition of three distinct types of power distributions implies that the connection between inequality and conflict behavior is different un-

der each condition.

30 The Power-Conflict Story Systemic Contributions

Many researchers analyzing the effects of systemic power concentration

(i.e., the level of inequality in the way power is distributed across an entire group of nations) on the likelihood of war additionally recognize that inequality can actually take on an infinite number of values along an interval scale (Singer, Bremer, and Stuckey 1972; Bueno de Mesquita 1981; Bueno de Mesquita and Lalman 1988; Thompson 1983). Nonetheless, they fail to notice that the concentration-war relationship may not

be such that an increase in concentration always produces a greater likelihood of war (as specified by the BOP explanation) or that a decrease in concentration always leads to a lower probability of war (as expected by power transitionists). Levy criticizes this approach and suggests, as do Ferris (1973) and Weede (1976), that there are three distinct types of conflictual interaction, each exhibited by nations under a certain level of systemic power concentration:

Both balance of power and power preponderance formulations assume a linear (or at least monotonic) relationship between the [conTABLE 3. War Probabilities for Regional and Global Dyadic Rivalries

Probability of War Probability of War Using CINC to Using GDP to Measure

Measure Power Ratio Power Ratio

y78 p,, conflict equations (3.7.21) and (3.8.21) are combined with (3.3) and (3.4) to form a similar system. Under the second condition, when p, < p,,

(3.7.2u) and (3.8.2u) are used instead of (3.7.21) and (3.8.21). If the power relationship between two nations places them in the third condition, and p, > p,, then equations (3.7.31) and (3.8.31) form a system of equations along with (3.3) and (3.4). If the two nations are in the third condition and p, < p,, (3.7.3u) and (3.8.3u) replace (3.7.31) and (3.8.31) in the system of equations. The appropriate conflict equations are placed in the region(s) representing each condition in figure 11. I now turn to some fine tuning of the model. This is done so that an understanding of the model will be more complete and the “settings” can be used to generate results in chapter 4. Interpretation of Parameters

As I specified the model, I included a definition for each variable and

each parameter in the equations. The variables p,(t), p,(t), c(t), and c,,(t) were named and interpreted in a fairly intuitive way. The parameters, or Greek letters, on the other hand, were simply called constants. It is not uncommon for modelers to leave parameters unexplained, but a presentation of any model, and of the power-conflict model specifically, would be incomplete without a discussion of the meaning of the various parameters. Although the primary theoretical arguments of the model can be found in the previous sections, we may find supplementary information in the interpretation of the parameters. In addition to providing more information about how nations interact, the parameter meanings might give some insight about how specific parameter values should be chosen in order to perform the analytic simulations of the next chapter. Those meanings might also suggest methods for the empirical measurement of either the parameters themselves or the variables they modify. Dimension Analysis

In order to distill the meaning of each parameter, I consider the units in which each variable in the model is expressed. This technique is adapted

The Power-Conflict Story 69

from the dimension analysis presented by Mesterton-Gibbons (1989: 35-40). The underlying principle of this analysis is that the units on the right-hand side of an equation must be consistent with those on the lefthand side. For example, if I pay my babysitter six dollars per hour and she has worked 30 hours this past week, then I calculate what I owe her by multiplying her hourly rate by the total number of hours worked:

6hour dollars x 30 hours = 180 dollars. (3.9) I know that I owe her 180 dollars, and not, say, hours, because the units on the left-hand side reduce to dollars:

dollars x hours = dollars. (3.10) hour Further reasoning demonstrates that since the units on each side of an equality must be equal, the unknown units of any entity on either side of the equality can be determined provided that the units for all others are known. In this case, the unknowns are the parameters. Once a parameter’s units are determined, insight on what it represents can be gained. For example, if a parameter’s units are found to be miles per hour, it is understood to be a velocity. Because a parameter’s units are determined by referring to the units of all known variables, the first step is to establish the units of time (4), power (p, and p,), and conflict (c,, and c,,). To a certain extent, assigning units for each of these variables is a creative process. The units I assign to

them, especially to power and conflict, are solely for the purpose of coming to an understanding of the parameters. They should not be interpreted as absolute truths about the variables or their related concepts. Let us say, then, that the units of time will be years.!! While this is a straightforward assignment, those for the remaining variables are problematic. The two central variables of the dynamic model, power and conflict,

do not have standard units, so I will devise some for the purposes of analysis. Power will be gauged in terms of “dollars,” while conflict will be gauged in terms of “hits.” One should not try to make too much sense out of these. I do not mean to imply, for example, that we should measure the conflict that nation X directs toward nation Y by counting the number of times nation X physically slaps nation Y. Rather, I tried to choose generic

words that would roughly capture the notion of units relevant to power and conflict. Now that these units are in place, I can deduce the units associated with each parameter in the model.

70 The Power-Conflict Story In each of the power growth and decay equations ([3.3] and [3.4]), two parameters, a and £, appear. For brevity, I henceforth present only the unit analysis for equation (3.3). the parallel treatment of equation (3.4) would be redundant. The units of a are fairly easy to determine. First, I assess the units of the left-hand side of equation (3.3), that is, dp,/dt. Because it represents the change in national power over time, the units are dollars per year. In mathematical shorthand,

a = (3.11) d dollars

dt year

where the use of square brackets, [ |, indicates an expression’s units. The units of each term on the right-hand side; that is, both the a,p, and the B,p,c,, terms in equation (3.3), must also be dollars per year. So, for the a,p, term,

; 3.12 |— Ps | |= = fap, (3.12) dp

Substituting the known units into equation (3.12),

— = [a,] (dollars). (3.13) dollars

year

Solving for [a,],

[a] (3.14) Qa.)=— ——_. . “years

In equation (3.12), a, is multiplied by a power term, whose units are dollars, so a, must have the units l/years. The a, and a, parameters are thus easily recognized as instantaneous, or specific, growth rates. For every unit of power a nation has, it gains an additional a units of power.

At what specific value should a, and a, be set? That depends on how various values of a, and qa, translate into different predictions for national power growth. I assess this by momentarily assuming that no other factors affect changes in national power over time. In that case, equation (3.3) reduces to its earlier form as expressed in equation (3.1). The solution of equation (3.1) is:

p(t) = Ce, (3.15)

The Power-Conflict Story 71 Power

/, 3; /

yj |— 001

=

ae

a“

“

4

/

f

4

--- 6

Time

Fig. 12. Growth in national power according to three different growth rates

where C is aconstant. Arbitrarily setting the initial value of p,(t) at p,(0) = 10, and solving for C,

C = 10 (3.16)

and equation (3.15) becomes

p(t) = 10e. (3.17)

Substituting three different values for a, into equation (3.17) produces the three different curves in figure 12. All three exhibit exponential behavior and begin at the same level of national power, but each curve leads to different results. When a, = .6, the power level soon becomes much larger than it does when qa, = .3 or a, = .001. If a, = 1, the same type of pattern persists, although the slope of the curve is much steeper. Choosing specific values of a, and a, can now be divided into two steps. First, do I expect a, and a, to be equal or unequal? Second, are

there any clues in the literature that suggest reasonable levels for a, and a,? If one rival has a higher a value than the other, then it will be at a ereat advantage. As I demonstrated when writing equation (3.1), the a parameter reflects a nation’s natural abilities or internal tendencies for erowth. Unequal a values would therefore indicate that one nation’s

72 The Power-Conflict Story intrinsic qualities are greater than the other’s. Ceteris paribus, a disparity in specific growth rates will produce a disparity in power levels as well, as is shown in figure 12. The empirical investigations of this book focus exclusively on major power dyads. It is reasonable to expect that within this universe players have approximately equal natural growth abilities and power levels that place them in the same league with one another (see the section entitled “The Relevant System of Nations” in chapter 5 for a more lengthy discussion of this issue). In the analytic simulations of chapter 4, I will therefore assume that the two rivals have equal growth rates.

At what level should a, and a, be set? Ideally, either theoretical justification or empirical evidence for those levels should be found. To date, only one study approaches this issue, and the resulting information

is sparse. Doran and Parsons (1980) examine the behavior of major power growth over time, but because they are constrained by the need to fit a regression equation the mathematical structure they ultimately impose on the national data does not incorporate the instantaneous growth rate that is similar to those in both the population and the powerconflict models. This is unfortunate because they are clearly motivated by the population model’s incorporation of cyclical patterns. One curve which reflects this general dynamic is a logistic function

describing the growth of a population in a limited environment (Heiss, Knorr, and Morgenstern 1973; Kuznets 1966). The logistic growth curve was explored by biologist Raymond Pearl (1924) who deduced that human populations grow nearly exponentially until reaching an inflection point and then level off to approach an asymptote. Because the Pearl logistic curve models growth in the context of limited resources (homologous with growth of a major power’s share of relative capability in the international system), it provides a theoretically justified, readily applicable method of finding critical points in the growth of a nation’s capability. (Doran and Parsons 1980: 954).

The authors obviously believe that the proper model of a nation’s power

growth is similar to the general form of the population growth model introduced in equation (1.1) in chapter 1. Here, I rewrite it so that p, a nation’s power level, is the central variable:

dp HtPro

1 K (3.18)

The Power-Conflict Story 73 where: dp/dt is the change in the nation’s power level with respect to time, K is a positive constant reflecting the carrying capacity or maximum possible power level sustainable by resources, and A 1s the upper bound of the specific growth rate.

Instead of using this model, they devise an equation that produces similar curves but lacks a readily identifiable specific growth rate. This is done so that standard regression techniques can be used to estimate parameters of the equation (954). As a result, I am unable to use empirically derived values of A to estimate a value for the a parameters. I now explore a slightly less satisfying source of information.

Given that the population model originally motivated Doran and Parsons (1980) and myself to some extent, and given that a nation’s population is one factor I use to measure its power level in chapter 5, I will now turn to an empirical analysis of the population model itself.

Mesterton-Gibbons (1989) presents a careful assessment of U.S. population growth according to equation (3.18). In particular, he is concerned with establishing the value of A, the upper bound on the population’s specific growth rate. Using Bureau of the Census data on the U.S. population from 1850 to 1970, he estimates that A ~ .31 (2629). What, if anything, might this tell me about the specific growth rate for major powers in general? To begin, it seems reasonable to expect a to be less than one. A growth rate above 1.0 indicates that for every unit of power a nation generates more than one additional unit per time period. We do not witness such high growth rates even among the human population, where reproduction is relatively easy. I will assume that the difficulties involved in generating additional national power are at least as great as those involved in conception, pregnancy, and birth. So, for any major power, a < 1. Second, for lack of a better estimation, I will set a, and a, at a level similar to Mesterton-Gibbons’s estimation of the U.S. population’s up-

per boundary for the specific growth rate or A. In the end, the precise value is not critical, as all the curves in figure 12, regardless of the a value, exhibit the same pattern. Hence, I somewhat arbitrarily settle on a, = a, = .3 for the simulation setting. Most of the remaining parameters also have years in the units’ denominators and hence are rates as well. However, I am not aware of any other usage of those parameters by other researchers or theorists and therefore cannot use information exogenous to the model in this book to determine their values. Instead, I keep them all below or near

1.0, reasoning, as I did for a, that a rate well above 1.0 would be

74 The Power-Conflict Story extraordinary. A decay rate above 1.0 similarly indicates that for every unit of power a nation has it loses more than one unit per time period. A growth or decay rate well above 1.0 for conflict levels would likewise indicate that for every unit of conflict a nation directs it either increases or decreases its efforts by more than one unit. Except in dire circumstances, I do not expect this to be the case. The second parameter in equation (3.3) is B. Similar reasoning can

be used to establish its units. The units of the left-hand side of the equation must match the units of the term B,p,c,,:

— = [BP Cv. (3.19) d

Substituting the known units into equation (3.19),

dollars , years [B.|(dollars)(hits). (3.20)

= | 3.21

Solving for B, in equation (3.20),

B.I (hits)(years) ” (3.21) so the 6 parameters can be shown to have the form 1/hit-years. This is not as easy to interpret as the units of a. Let us rewrite these units as (1/hits)(1/years). We recognize the 1/years component as a rate.

It is modified by 1/hits, so we might think of B as being the rate of absorbing the cost of each hit from the opponent or the “conflict cost rate.” This fits nicely with the argument that subtracting the B,p,c,, and B,p,c,, terms represents the cost associated with responding to conflict from the opponent. Setting a value for B is also more problematic than it was for a. To

begin, it is difficult to estimate the impact of B on equation (3.2).!” Second, there are no available theoretical or empirical hints about what an appropriate value might be. Because it incorporates the “rate” notion, and setting arate above one seems to indicate extraordinary circumstances, I simply require it to be less than one. Again, somewhat arbitrarily, I set B, = .5 and B, = .1. Because the 6 parameters will play a crucial role in determining the occurrence of a special class of transitions and because they reflect the impact of issues, their meaning and values are more closely discussed in chapter 4’s simulation analysis. Now I move to the parameters appearing in the conflict equations.

The Power-Conflict Story 75 For simplicity, I will again work only with equations for nation X. In the base conflict equation, there is just one parameter, y,, which modifies the reaction term, p,c,,. The left-hand side of the conflict equation, dc,,/dt represents the change in conflict over time, so its units are expressed in hits/year. These units should parallel those on the right-hand side:

wean 7 LP (3.22) hits

= ! 3.23

Substituting the known units into equation (3.22) and solving for [y,],

LY (dollars)(years) - (3.23) That is, y, and y, must have units of the form 1/(dollar-years). We can think of these units as (1/dollars)(1/years). If we treat the second piece as arate, we can say that for every dollar a nation has it will react to the

opponent’s conflict at a certain rate. Thus, y is the “enabled rate of reaction.” It captures two aspects of national power that are difficult to

incorporate into a measurement tool—a nation’s ability to convert power resources into actions and its willingness to do so. Because y is a

rate, I set its value at less than one: y, = .2 and y, = .1. For the parameters appearing in the conditional conflict equations, I do not present a step by step unit analysis, as I did for a in equations (3.11) through (3.14). Instead, I focus on their meanings. Under the first condition, d* = |p,(¢) — p,(O|, there are two additional parameters for each nation’s conflict equation, x and e. The parameter e has the units of power, dollars, and simply serves as a mechanism for avoiding discontinuities in the second term on the right-hand side of equation (3.7.1). That term is a fraction, and if the denominator,

e + (d* — |p, — p,|), were ever zero its value would be undefined. Without the inclusion of € in the denominator, this would occur when the difference between p, and p, was precisely d*. Because I do not want € to play a large role in the model, I set it at a very low level: e = .OO1. The parameter x must have units of the form (dollars-hits)/year. We can again rewrite these units as (dollars)(hits/year). Recall that the term modified by « captures the weaker nation’s desire to avoid greater disparity and the stronger nation’s desire to promote movement toward disparity; SO we can say that a nation motivated by a current power disparity

will direct a certain number of hits per year toward its rival. Let us therefore call «x the “disparity-motivated conflict rate.” As another rate, K Is set below one: k, = .1 and k, = .15.

76 The Power-Conflict Story

For the second condition, d* < |p, — p,| < d, 1 will discuss the parameters appearing in the case in which p, > p,, as expressed in equations (3.7.21) and (3.8.21). Similar reasoning applies to the parallel parameters found in equations (3.7.2u) and (3.8.2u), where p, < p,. It

can be shown that a,’s units are hits/(dollar-years). These units can equivalently be expressed as (1/dollars)(hits/year). The term modified by o, represents the stronger nation’s attempt to consolidate its power.

We can therefore reason that for each dollar advantage held by the stronger nation it directs a certain number of hits toward the weaker nation each year. The parameter o, can be called the “rate of consolida-

tion.” The o parameters are thus understood to be rates and are set below one: o, = .2 and a, = .15. The weaker nation’s (i.e., nation Y’s) behavior under the second condition is influenced by the parameter w,. The term modified by a, was designed to capture the weaker nation’s attempts to prevent movement toward the third condition, under which the stronger nation’s ad-

vantage is guaranteed. Since w,’s units must be hits, this parameter represents the increased conflict the weaker nation is willing to perpe-

trate in order to prevent the stronger nation from securing such an overwhelming advantage. We can call w, a “commitment to prevention.” Because w, cannot be interpreted as a rate, I am left without much guidance for setting its value.!? Rather than being able to rely on infor-

mation from other scholars, I turn to the model itself for clues. In particular, I consider the relative importance of the two components on the right-hand side of equation (3.8.21), y,p,c,, and [@,/(p, — p,)|[d(p, —

p,)/dt]. Each one has a unique effect on the change in the amount of conflict nation Y directs at nation X, dc,,/dt. For example, while the first component always produces an increase in conflict, the second can produce a decrease in conflict if dp,/dt > dp,/dt. In a sense, the two pieces

compete with one another for governance of dc,,/dt. Should I expect them to have approximately equal ability to influence the left-hand side of equation (3.8.21) or that one outweighs the other? Recall that y,p,c,, is the component common to all three regions of conflictual interaction, as specified in the base conflict equation (3.8). This can be thought of as the primary ingredient in all equations describing nation Y’s conflictual behavior toward nation X, including (3.8.21). I therefore expect its effect to be stronger than that for the regional component, [w,/(p, — p,)|[d(p, — p,)/dt}:4

VP yCxy Sy Up. = Py) ; (3.24) Px ~ >Py( dt

The Power-Conflict Story T7

Substituting values for dp,/dt and dp,/dt from the right-hand sides of equations (3.3) and (3.4), Wy

VyP xy > ———( Px — BPyCyx — %Py + ByPCry). (3.25) Px ~ Py

Solving (3.25) for w,,

co,

p, and the difference between those two power levels 1s more than d* but less than d. The precise values of d* and d do not change the findings of this work as long

as d* < d, so in order to proceed with the simulations in chapter 4 the constants d* and d are arbitrarily set at d* = 7 and d = 15. Sop, — P, must be at least seven and less than 15. I want to set the difference close

to d* for the following reason. The inequality in (3.24) expresses the idea that the base conflict component of equation (3.8.21) is more impor-

tant than the weaker nation’s attempts to prevent the stronger from consolidating its advantage. This may no longer be realistic as the system

78 The Power-Conflict Story approaches d. As the stronger’s advantage becomes close to an overwhelming one, the weaker nation’s conflict behavior may be increasingly governed by its need for prevention. When the difference between p,and p, is close to 15, it would be reasonable to suspect that the [w,/(p, — p,)|\ld(p,. — p,)/dt] term may be larger than the y,p,c,, term. In other words, I expect the inequality in (3.26) to hold primarily when the difference between p, and p, is small. I will set the difference at eight. The midpoint of the range for possible initial values for p, and p, 1s 25, so p,, will be four units above 25 while p, will be four units below 25: p, = 29

and p, = 21. Information for determining the values ofc, and c,, can be found in the inequality in (3.26). Given the stipulation that @, is positive, the right-hand side of (3.26) must be positive as well. That is:

(YP Cry) (Dx 7 Py) > 0. (3.27) QP» — PyPyCy~ — AyPy + ByPxCry

Because the parameter values and the initial values for p,, p,, c,,, and c,, must be positive, and because (p, — p,) also must be positive, the numerator cannot be negative. Hence, the fraction can only be negative if the denominator 1s negative. In order to prevent this, the denominator must be larger than zero.

QP, — ByPCye — GPy + BP Cry > 0. (3.28) Solving for c,,,

CyB.Py p,, as expressed in equations (3.7.31) and (3.8.31). The parameters parallel to A, and 4,, that is, v, and

v, from equations (3.7.3u) and (3.8.3u), can be analyzed in a similar fashion. The parameter A, can be shown to have the units 1/(dollaryears). These units are the same as those for parameter y. Let us again think of these units as being (1/dollars)(1/years). We can now say that for every dollar a nation has beyond what its rival owns it can afford to lessen the number of hits it directs toward the opponent at a certain rate. The parameter 4, (and v,) can therefore be aptly named the “afforded rate of alleviation.” Parameter A, has the same units as A,, since equations (3.7.31) and (3.8.31) are of the same form. In the case of nation Y, however, we are concerned with the behavior of the weaker nation. We can say that for

every dollar a nation is behind its rival it resigns itself to defeat by lessening its conflict at a certain rate. The “resigned rate of surrender” is an appropriate name for A, (and »v,). Every other parameter that is interpreted as a rate is set below one.

The rationale for this is that a rate near or above one would indicate highly unusual circumstances. Perhaps such a situation exists in the third region. Here the stronger nation has established an overwhelming advantage, and it no longer needs to worry about consolidation. Any attempt

by the weaker nation to compete with the stronger would be in vain. Both would like to reduce their conflict levels as quickly as possible in order to avoid the retaliation effects of the rival’s conflictual behavior on national power. Because the passing of d*, an overwhelming advantage, does indeed represent an extreme case, choosing values near or greater

than one for the A and v parameters is warranted. For the sake of simplicity, | assume that A, = v, and A, = v,. This assumption represents the idea that when nation X is in the stronger position it will have an afforded rate of alleviation similar to that which would be held by nation Y if it were in the stronger position. And the resigned rate of surrender would be approximately the same, regardless of which rivals were in the weaker position: A, = 1.5, A, = .9, vu. = .9, and v, = 1.5.

80 The Power-Conflict Story From Alpha to Omega

By way of summarizing this lengthy discussion of the model’s parame-

ters, I explore two questions. First, what has been learned from the exercise? Second, what information has been used to determine the specific values of the model’s parameters? After answering these questions, I present all the parameters, their interpretations, and their values in table 5. The conceptual development of each of the parameters is the most useful contribution of this section of the book. The explanations of B as

a conflict cost rate, of o as a rate of consolidation, of A, and v, as afforded rates of alleviation, and so on are far more satisfying than the hand-waving statement that they are positive constants. They help make sense of the notion that power can be multiplied by conflict. For example, in equation (3.3) the second term, B,p,c,,, knowing the units for B,., 1/hit-years, allows us to multiply power by conflict and get dollars per year as the units. The advantages of conceptualizing these parameters goes beyond intellectual satisfaction. For example, the understanding of the y parameter gained here proves useful in choosing an indicator for national power later in chapter 5. One particularly lucid conceptual understanding 1s that all but two of the model’s parameters, € and w, are rates of action or reaction. In other

words, adjustments in those parameter values will primarily affect the speed at which nations’ behaviors occur. Very high parameter values might indicate that a nation’s particular form of government is highly efficient and can quickly respond to changes in an opponent’s levels of power and conflict. For example, if y,, nation X’s enabled rate of reaction, has a high value, then nation X would react quickly to conflict directed at it by nation Y by directing additional conflict toward nation Y.

In at least two instances, these conceptual understandings of the parameters provide clues about how their values might be empirically determined. In the case of the a parameters, the “specific growth rate” interpretation suggests that an analysis of national power data using a model similar to that used to understand population growth would yield a value for the parallel parameter, A. One could assume that a’s value will be similar to A’s. In the case of w, its units are the same as those for the conflict variables. ‘There are several usable, albeit imperfect, conflict

scales (some of which are discussed in chapter 6). A strategically designed analysis of a conflict data set for several major power nations in the weaker position might produce empirical values for w.

Currently, however, there is little empirical information on the value of most of the parameters appearing in the model. Instead, many

The Power-Conflict Story 81

other types of information have been used. The information for the a parameters is the most abundant, although it is still less than ideal. I drew on four sources to set the values for a, and a,. First, I analytically solved equation (3.1) to show the impact of different values of a on national power growth. Second, the empirical literatures on major power growth and population growth were consulted. Third, because a, and a, are rates, they were assumed to be less than one. Fourth, because the empirical analysis of this book deals with dyads consisting of major powers, rivals that could be seen as possessing roughly equal natural tendencies, a, was set equal to a,. Information of the first and second types was not available for the remaining parameters. The fourth type of information was not relevant.

Fortunately, most of them were determined to be rates as well. This allowed me to at least assume that those parameters should be less than unity unless unusual circumstances were demonstrated. Two rate parameters, A and v, were indeed assumed to be near or above one. This was because they are involved in the conflict behavior equations in the extreme condition of overwhelming preponderance. Only two parameters, € and w, were not found to be rates. The e

parameter reflects the units of power and is simply a mechanism for avoiding a discontinuity. Therefore, it was set very low. The w parameter

has the same units as the conflict variables. Its possible values were determined by examining the relative importance of different components of the model itself. In the next chapter, the particular parameter values established in this one are used as constant values in the simulations generated by the power-conflict model. Readers interested in a summary of all of those values can find them recorded in table 5.

A Synthesis

By synthesizing ideas from two seemingly opposed explanations of war,

a single (formal) model of power and conflict in dyadic rivalries was developed in this chapter. One integrative technique I employed was to design the conditional conflict equations ([{3.7.1] and [3.8.1], [3.7.21] and [3.8.21], [3.7.2u] and [3.8.2u], [3.7.31] and [3.8.3]], and [3.7.3u] and

[3.8.3u]). While each pair of equations shares it basic components, Y,PxCy, and y,p,c,,, with all other pairs, it also incorporates a unique component. The specialized piece changes as the value of the power differential, p, — p,, becomes larger or smaller. The principle underlying this “switching mechanism” is that changes in the conflictual behavior of

82 The Power-Conflict Story nation X and nation Y, dc,,/dt and dc,,/dt, are dependent on the power relationship between the two opponents. For example, suppose nations X and Y begin their rivalry when nation Y holds 15 units of national power and X holds less than one. Figure 11, in which each pair of conditional conflict equations is placed in the region of the p, versus p, plane in which it applies, indicates that equations (3.7.2u) and (3.8.2u) govern dc,,/dt and dc,,/dt at this point.

As the conflict equations for the upper condition and the universal power equations (3.3) and (3.4) generate subsequent values for p,, p,, C,,, and c,,, nation X’s power level grows while Y’s remains relatively constant. In other words, the system moves toward a smaller advantage TABLE 5. A Review of the Parameters in the Power-Conflict Model

Parameter Interpretation Assigned Value Those in the power equations: a, specific growth rate of X’s power 3 a, specific growth rate of Y’s power 3 Bb. conflict cost rate associated with X’s responses to conflict -) directed by Y

B, conflict cost rate associated with Y’s responses to conflict l directed by X Those defining regions of conflict behavior:

d* difference between two nations’ power levels denoting a 7 decisive advantage

d difference between two nations’ power levels denoting 15 an overwhelming preponderance Those in all conflict equations:

V, X’s enabled rate of reaction to Y’s conflict level 2 "Vy Y’s enabled rate of reaction to X’s conflict level wl

Those in condition 1 conflict equations:

K, X’s disparity-motivated conflict rate wl Ky Y’s disparity-motivated conflict rate 15

E mechanism for avoiding discontinuities .OO1 CO, when stronger, X’s rate of consolidation 2

Those in lower condition 2 conflict:

W, when weaker, Y’s commitment to prevention )

Those in upper condition 2 conflict:

0, when stronger, Y’s rate of consolidation 5

wo, when weaker, X’s commitment to prevention 6 Those in lower condition 3 conflict: A, when stronger, X’s afforded rate of alleviation 1.5

x, when weaker, Y’s resigned rate of surrender I v, when weaker, X’s resigned rate of surrender 29

Those in upper condition 3 conflict:

v, when stronger, Y’s afforded rate of alleviation 1.5

yo a Is ¥ x 12.5 a

The Power-Conflict Story 83

¢ _ 7 ne a

Py

20 .

a Py =p,td Py =p,+d “

17.5 a i Py = Px

o a a f = ” ¢ oe f rol “ a o“ 2 é a f ir Pa yo Pa Px

‘0 2 Py = Py~d

7.5 pete r

5 a Py = Px d

2.5 o YO

2.9 5 7.5 10 12.5 15 17.5 20

Fig. 13. A sample trajectory in the p, versus p, plane. The parameters are set at the values specified in table 5 in order to produce this sample trajectory

for nation Y. In figure 13, the initial power relationship 1s plotted at t = 0

and the evolution of that relationship is traced on the curved line, or “trajectory,” that twice passes across the d* boundary. When nation Y’s advantage diminishes to less than d* = 7 units just after ft ~ .21, equations (3.7.1) and (3.8.1) are substituted for (3.7.2u) and (3.8.2u). This trend persists until t ~ .98, when nation X’s power level reaches its maximum value and then begins to decline and nation Y’s power level gradually begins to lessen. Equations (3.3), (3.4), (3.7.1), and (3.8.1) continue to operate until the trajectory recrosses the line described by p, = p, + d* at t~ 1.92. Then equations (3.7.2u) and (3.8.2u) come back into play. They

84 The Power-Conflict Story operate in conjunction with equations (3.3) and (3.4) to ensure nation Y’s advantage by driving nation X’s power level to nothing at t ~ 1.97. Were the trajectory ever to cross into yet another region, the appropriate conflict equations would apply. After the integrative model was designed, I assessed the units of each of the model’s parameters in order to distill their meanings. From those meanings, I was able to make informed choices for the values of the parameters. It was necessary to determine their values in order to produce trajectories such as that in figure 13. In chapter 4, I analyze the model using a simulation procedure that generates a variety of trajectories. Each of these begins at a unique starting point and then traces the evolution over time of both power levels and conflictual behavior. The results of those simulations serve as deductions from my model and will be compared to the observable patterns in national power growth and dyadic conflict behavior in chapter 5.

CHAPTER 4 The Moral

Now that the model-building exercise is complete, I turn to an analysis

of the model itself. The purpose of this chapter is to elicit from the formal model certain deductions concerning power transitions, that is,

to reveal the moral of the story. In the next chapter, many of those deductions will be compared to empirical information. In particular, I am interested in whether transitions are peaceful or conflictual and in the types of power relationships that produce and result from transitions. Whether transitions are peaceful or conflictual is an interesting issue that arises from the BOP-PT debate. Proponents of the BOP explanation would expect transitions to be relatively peaceful (1.e., free from war) since they are thought to occur when rivals are approximately equal in power and stably balanced. Proponents of the PT explanation, on the other hand, would expect transitions to be relatively conflictual (i.e., war prone) since they signal the weaker nation’s challenge to the previously dominant nation. The hybrid model should provide expectations of its own concerning conflict behavior near power transitions. It should

additionally provide expectations concerning the patterns of power erowth and decay before and after a transition. Simulation Procedure

Expectations or conclusions can be determined from a system of differ-

ential equations in one of two fashions.! If the system can be solved analytically, the equations are rewritten so that they explicitly describe the behavior of p,, Py, C,,, and c,, as time passes. In other words, p,, p,, C,,, and c,, are the left-hand side of the equations as opposed to dp,/dt,

dp,/dt, dc,,/dt, and dc,,/dt. It, however, the system cannot be solved analytically, it must be solved numerically. Once a value is provided for each variable at a particular moment in time, the equations can gener-

ate all subsequent and previous values of p,, p,, C,,, and c,, through time. As misfortune would have it, the model proposed here requires a numerical solution. Such solutions are more time consuming and, if not 85

86 The Power-Conflict Story carefully done, can result in large errors in prediction. In their text on differential equations (1985), Edwards and Penney reassure the reader that properly executed numerical analysis is a viable alternative to analytic solutions.

Finding a particular solution of a differential equation is, in most applications, merely a means to an end. The goal is to be able to predict the shape of a column, the trajectory of a particle, or the modes of vibration of a mechanical system, with a certain degree of

accuracy. Numerical results are frequently just as useful to such ends. (401-2) Several steps are involved in the numerical solution. First, the pa-

rameter values are determined. Parameters are the constants represented by lower-case Greek letters throughout the equations. Next, several sets of initial values, or conditions, are selected. Several sets must

be used because rigor demands that conclusions not be drawn from special or unique cases. Since the general analytic solution cannot be determined, multiple instances must be used in order to make broadbased claims about the model’s expectations. Third, the equations are seeded with a set of initial conditions from which they determine subsequent and previous values of p,, p,, C,,, and c,,. The equations are then reseeded with each additional set of initial conditions, which provides several more runs or simulations. Last, the many sets of resulting paths for P,, Py» Cy, and c,, through time are examined in order to distill the commonalties held across all initial conditions. The selection of particular parameter values was completed in the

“Interpretation of Parameters” section of chapter 3. If a theoretical or analytic issue arises that suggests the need to adjust any particular parameter values after the initial investigation, this will be done. Throughout all baseline simulation runs, parameters are held constant at the following values:?

a, = 3 a, = 3

B= .5 Bb, = 1

y, = .2 WY = 1

K, = .1 K, = 15 w, = .6 oO, =. oO, = .2 o, = 15 A, = 1.5 A, = 9

v, = 9 v, = 1.5

e= .001

The Moral 87 Having settled on parameter values for the model, I now turn to the simulations themselves. As explained previously, I must decide on several sets of initial values for the four variables p,, p,, c,,, andc,,. In other words,

I must assign values to p,(0), p,(0), c,,(0), and c,.(0). These particular values provide the model with starting points in time. The sets of initial conditions satisfy two simple requirements. First, all must be positive and less than or equal to 50. That is, 0 < p,(0), p,(0), c,,(0), c,,(0) = 50. The purpose of this requirement is simply to limit the space under consideration. If such a limitation is not imposed, the initial values could take ona larger, unmanageable variety of values. This is a beginning step. If analysis suggests that interesting or unusual behavior is produced outside these boundaries, then they will be expanded. Second, all initial conditions will begin at a transition. Since transitions occur when two nations’ power levels intersect, this requirement can be expressed as p,(0) = p,(0). This will guarantee that the resulting conflict behaviors are concomitant with dyadic power transitions, the phenomenon I wish to investigate. A random number generator provides a variety of initial conditions meeting both of the stipulations specified earlier. * Now the equations are given a set of initial conditions from which both subsequent and previous values for each of the four variables can be determined.* In other words, each variable will have associated with it a time-series of values. These time-series can be plotted as trajectories, as in figures 14, 15, and 16. Each of these figures contains three graphs. The horizontal axis are indications of progress through time, while the vertical axes represent power levels, the difference in power levels,> and conflict levels, in that order, from top to bottom.® For each

set of initial conditions that the equations are given, a unique set of trajectories is generated. Hence, figures 14, 15, and 16 are different. The initial conditions that produced each set of trajectories can be read in the first and third graphs of each figure. The point at which a variable’s trajectory intersects the vertical axes indicates the value of that variable when ¢ = 0 or at the initial point in time. In the first graph of figure 15, for example, nation X’s and nation Y’s power levels cross the

vertical axis at p, = 10 and p, = 10.’ In the third graph of figure 15,

nation X directs 10 units of conflict toward nation Y at tf = 0. Nation Y reciprocates with six units of directed conflict at t = 0. The set of initial conditions that produced the unique trajectories in figure 15 are p,(0) = 10, p,(O) = 10, c,,(0) = 10, and c,,(0) = 6. For each variable, it is generally possible to generate values for an infinite number of points in both future and past time unless a discontinuity is reached. In figures 14, 15, and 16, the system of equations generates all variable values to the left and right of the vertical axis at t = 0.°

Px (0) = py (0) = 10, cxy (0) = 50, cyx (0) = 10

~‘Ny-‘ —— Py

20)

-a

am TTT TTT - 7

Time

—Q.125 —0.075 —Q.025 0.025

Power relationship

40 —— |px - Py!

30 corn aT ——= d= 15 20 10

\\ ~100 Oxy ~— Cyx —Q.125 —0,075 —0.025 0.025

Time

Conflict

~

“an ~— — —_ _ _ __ = —_ — —

—0.125 —Q.075 —0.025 0.025

Time

— 30)

— 100

Fig. 16. An example of a David and Goliath transition

The Moral 91 How, then, does one decide when to end a simulation? I employ the following rule. When either p,(¢) or p,(t) becomes zero, the simulation ends. J assume that a nation must possess some positive level of power in order to exist. Hence, the endpoints of the trajectories in figures 14, 15, and 16 correspond to the points in time — future or past — at which either

nation X or nation Y did not possess any power whatsoever. In other words, the model is not meaningful when p,(¢) or p,(¢) is nonpositive. I allow conflict levels to become negative, having reasoned that negative conflict could be interpreted as some type of concession or cooperative behavior. As was mentioned in chapter 3, Organski and Kugler see con-

flict in a similar fashion, arguing that “relations... , in reality, can range subtly in degree from full cooperation to armed conflict” (1980: 46-47).° Baseline Simulation Results: Deductions from the Model Approximately 50 different sets of initial conditions satisfying these crite-

ria were used to seed the dynamic model. Although each set of initial conditions produced a unique set of trajectories for p,, p,, C,,, and C,,, three distinct patterns of behavior emerged (the entire set of randomly generated initial conditions, along with the type of pattern each produces, is listed in appendix A). Each distinct pattern is represented by a single simulation result in figures 14, 15, and 16. For example, the graphs in figure 14 demonstrate the results of only a single simulation, but many other simulation results share similar trajectory behavior. So the trajectories in figure 14 can be thought of as a single, representative member of an entire family or class of like trajectories. The three distinct classes of behavior are distinguished primarily by the differences in how national power grows and decays before, during, and after a dyadic power transition. The first is characterized by the challenger’s obvious

inability to successfully complete a transition. All trajectories in the second class show the challenger rising to compete with the dominant nation for a period of time before returning to the weaker position. This group might also be characterized by saying that the formerly dominant nation finds a way to recover its lost position. The third class is characterized by a single, distinct transition in which the challenger succeeds. The first group of trajectories might be referred to as the bull and

gnat (B&G) deflection cases. They are represented by the graphs in figure 14 where p,(0) = p,(0) = 10, c,,(0) = 50, and c,,(0) = 10. In the first graph, the weaker nation’s power level, p,, approaches that of the

92 The Power-Conflict Story stronger but is deflected away at t = 0. Much as a gnat might annoyingly buzz in a bull’s ear and then fly off without inflicting much real harm, nation X only momentarily appears to threaten nation Y. Although the bull’s behavior does evolve from primarily cooperative to conflictual, its conflict level remains low in comparison with the gnat’s conflict behavior. In other words, the challenger (gnat) is consistently more conflictual than the dominant state (bull). Additionally, this disparity in conflict levels diminishes as the nations move forward in time or from left to right on the third graph of figure 14. Furthermore, two hostility peaks

arise. The first comes from the challenging gnat alone, prior to the deflection. A second comes from both the bull and gnat shortly afterward. Together, they persist in aggravating the hostile climate. Another interesting feature of all the B&G cases is that the difference between the values for conflict behavior, c,,(t) — c,,(¢), is large compared to that for the second type of simulation (depicted in fig. 15).

All B&G deflections occur when a trajectory in the p, versus p, plane reaches the line p, = p, (the line along which transitions occur) and

the slope of the line tangent to the curve at that time is unity, that 1s, when dp,(0)/dt = dp,(0)/dt. It can therefore be shown (by reference to equations [3.3] and [3.4]) that all deflection cases are characterized by the following stipulation on the initial conflict values:!°

Byc,(0) = Pye,,(0). (4.1) Since B, = .5 and 6, = .1, c,,(0) must be exactly five times larger than c,,(0) in order for deflection to take place. Equation (4.1) may be understood to mean that when the weaker nation’s (here nation X’s) power level becomes equivalent to the stronger’s (nation Y’s), the stronger must direct only one-fifth as much conflict at the challenger as the challenger directs at it in order to prevent the challenger from successfully achieving the advantage. A different pattern emerges in the second group of trajectories. In figure 15, a single simulation result demonstrates the typical behavior in this category. Here p,(0) = p,(0) = 10, c,,(0) = 10, and c,,(0) = 6." One nation’s power level rapidly rises to its peak and then declines. Its competitor, meanwhile, takes the more methodical path, maintaining a relatively even level of power and achieving the advantage in the end. An appropriate name for this group might be the tortoise and hare (T&H)

double transitions.!* In the fable of the tortoise and hare’s race, the boastful hare starts swiftly. Overconfident due to his large lead, the hare

naps midway through the race. From beginning to end, the tortoise plods along. When the hare wakes up, he sees the tortoise crossing

The Moral 93 the finish line. This scenario might remind us of the Soviet-American military rivalry during the Cold War. In the mid-1960s, the USSR began to experience a rapid rise in its capabilities vis-a-vis the United States.

American capabilities were maintained, and perhaps even declined slightly, at about this time. In 1992, however, the hare’s pace was broken

by the fall of the USSR, leaving the tortoise as the victor. Telling a strikingly similar story, some American policymakers and academics in the mid- to late 1980s predicted that the USSR would overextend its resources by pursuing a rapid armaments buildup that would debilitate the Soviet economy (more than the American buildup would harm the U.S. economy) and thereby render the United States the “winner” of the arms race (Collins 1986; Commission on Integrated Long-Term Strat-

egy 1988; Kennedy 1987). Unlike the first class of trajectories, I am unable to describe this class with a simple stipulation on the initial conditions, as in equation (4.1). What characterizes conflict behavior in the T&H class? As with the B&G cases, the challenger (this time the hare or nation X) is consistently more conflictual than the dominant state (the tortoise or nation Y). The conflict values also converge as the simulation progresses from left to right in the third graph of figure 15. Nation X is directing a great deal more conflict toward nation Y when t = —2 than Y is directing toward X. In fact, nation Y is directing a negative amount of conflict, or cooperation, toward nation X. As time passes, nation Y abandons its cooperative efforts and becomes increasingly conflictual. Nation X decreases its level of conflictual behavior until nation Y ceases cooperation at t ~ —.87. Afterward, they both consistently intensify their conflict levels. Last, the conflict levels are highest following the second transition. The first transition, when nation X surpasses nation Y, does not result in

a permanent advantage. Rather, the power position of the two rivals switches again when the second transition occurs. Why is the second transition the more conflictual one? Preceding the first transition, the hare (nation X) is highly conflictual while the tortoise (nation Y) is actually rather cooperative. The conflict prior to the first, or temporary, transition is high but one-sided. Between the first and second transitions, the tortoise, in order to compete with the hare, becomes less cooperative and then turns conflictual at t~ —.87. Up until the tortoise’s conversion to conflictual behavior, the hare’s level of conflict abates. After t ~ —.87, however, the amount of conflict it directs toward the tortoise rises again. So, from the point where the tortoise becomes hostile, both competitors constantly reinforce their conflictual behavior toward each other. This

continual increase persists until after the second transition, which is permanent. Not until well after the second transition do the c,, and c,,

94 The Power-Conflict Story trajectories level off and decline. The permanent transition, therefore, produces the highest joint conflict behavior of the entire simulation. In the third group of trajectories, a single successful transition occurs. This family of trajectories is represented in figure 16 by the behavior resulting when p,(0) = p,(O) = 10, c,,(0) = 10, and c,,(0) = 50. In the first graph, the decline of nation X’s power concurs with the ascent of

nation Y. Nation Y rises up to challenge and overtake the formerly dominant X, which experiences a prolonged period of decline. This group fits nicely with Organski and Kugler’s (1980) traditional version of a single, decisive, power transition. The David and Goliath (D&G) label is borrowed from the biblical story in which David, though only a boy,

fells the Philistine giant and eventually rules the kingdom of Israel.!° There is a distinct and rapid movement from lower condition 3 (where nation X, or Goliath, enjoys an overwhelming advantage) to upper condition 3 (where nation Y, or David, enjoys an overwhelming advantage).!4 This is different from the T&H group of trajectories in which the first transition is not a lasting one. The T&H dyads spend considerable time in the first condition or in lower condition 2 or 3 (where nation X, the hare, has an advantage) before experiencing a permanent transition,

which propels them into upper condition 2 or 3 (where nation Y, the tortoise, has the advantage). Again, I cannot determine a precise stipulation of the initial conditions that produce D&G transitions. Several general statements can be made about conflict behavior in the D&G group. Most notably, there is a mix between cases in which the challenger (David) is more conflictual and cases in which the dominant state (Goliath) is more conflictual. This prominently distinguishes the

D&G simulations from the B&G and T&H cases as well as from Organski and Kugler’s version of the biblical story. The two other simulation types and Organski and Kugler’s version of D&G predict the challenger to be more conflictual. But in many of the D&G simulations it is the dominant Goliath who is more aggressive. In the third graph of figure 16, the conflict levels converge as they move from the left to right. Beginning at tf = —.15, nation X is highly cooperative while nation Y is highly conflictual. As time progresses, each nation’s behavior becomes less extreme until nation X finally becomes conflictual at t~ —.06. From then until t = 0, nation X becomes increasingly conflictual while nation Y’s conflict behavior levels off. Although the small scale of the conflict

axis in the third graph makes it difficult to see, both nations slowly increase their conflictual behavior after the transition at ¢ = O until nation X’s power level becomes zero at t ~ .04. By the end of the simulation, the conflict behaviors appear to be coupled. Furthermore, they are highest following the transition. Although nation Y’s conflictual

The Moral 95 behavior is highest early on, it is rapidly decreasing at that time. It is increasing following the transition. Nation X’s corresponding behavior is cooperative when Y’s conflict is at its peak and is most conflictual follow-

ing the transition. Taken together, and accounting for the direction of change, nation X and nation Y can be considered to be most conflictual after the transition.

The Impact of Issues: Varying the Beta Parameters

In addition to defining the deflection cases, the 8 parameters play another special role. In chapter 2, their ability to symbolize the effect of important issues was highlighted. Here, I explore that effect by analyzing how changes in B values results in different types of power transitions. Large B values imply that a nation considers an issue to be very salient, perhaps because the stakes are high or tangible. For simplicity, I vary only £,. The results parallel those for B,. The logistics of this endeavor are straightforward. First, I use information from the classification of the deflection cases to establish critical values of B,. Simulation results for those values demonstrate their importance. Next, I run a set of simulations for each of the critical values. Each set is comprised of

different initial power conditions. Finally, I interpret the relevance of the results to provide substantive conclusions for the role of issues in dyadic rivalries. Equation (4.1) stipulates the condition that must hold in order for a deflection to take place. Earlier, it was used to determine the minimum level of conflict the dominant nation needs to direct at the challenger in

_ Sel) 1 B= Bo (4.2)

order to prevent a successful challenge. It can be rewritten to instead express the value of 8, at which a deflection occurs:

If, for example, c,,(0) = 10 and c,,(0) = 6, then a deflection occurs when B, = .167. This is true regardless of the magnitude of the power levels. As

long as p,(0O) = p,(0), c,,(0) = 10, and c,,(0) = 6, this exact value of 6, produces a deflection, with nation Y winning (See the first graph in fig. 19 in the “B, = .167” section). However, if B, is greater or less than .167, a

different picture emerges. Setting p,(0) = p,(0) = 10, c,,(0) = 10, and c,.(0) = 6 and varying £, to include values below, at, and above .167 produces figure 17. B, values below .167 yield single power transitions in

Py

35 Py=Pxt 13) Py=Pxt7 Py=Pe Py=Px-7 py =pr—15 20

13 Py 5 ge

al Px 10 20 30 40

Cyx ---- Bx — 02

30

Jeene By =.]

IB. = 167

20

wee fy = .5

10 » — By = A)

a iy -10 \ \° 20 ” ~10 ‘\ 4

%

‘,

—20 \

—30 ‘

Fig. 17. Varying the £, parameter

The Moral 97 which nation X is the ultimate winner (dashed trajectories). When 8, is above .167, Y wins, but in different ways (gray trajectories). In one case, where £B, = .5, it does so by playing the tortoise in a double transition. In another, where B, = .9, it rises to challenge a declining hegemon. Conflict behavior in all five of these simulations progresses to escalatory behavior,

with X or Y tapering off only toward the end. In general, X’s situation improves as B, gets smaller. When £, gets very large, even an overwhelming dominance cannot be maintained (see the B, = .9 trajectory). A closer look at each of these parameter values is useful. B, Values below .167

What conclusions can be drawn from the trajectories shown in figure 18? Using B, = .08, I vary the initial power condition along the p, = p, line. The smallest starting power value is two. The largest is 50, the maximum allowed in all simulations in this chapter. All trace back to an advantage for nation Y —from a small one in the case of p(0) = 2 to a large one in

the cases of p(0) = 30 and 50. Nation Y’s advantage then gradually declines for some time until it begins a precipitous fall just before reaching p, = p, Hence, a single transition is won by X, which ends up with anywhere from a negligible amount of power (in the p(0) = 2 case) to an overwhelming advantage (in the p(0) = 30 and 50 cases). Nation Y dies out except when the initial power level is very high (p(0) = 50). In this best-case scenario, Y’s extraordinary power insulates it from extinction,

but it must endure strong domination by X. Nation Y’s power does gradually increase after its fall but at such a slow rate that 1t would take eons before Y might hope to challenge X. So, when the underlying issue has very little importance to nation X, that nation is sure to come out ahead. Because X does not have much at stake, it is not vulnerable to Y’s attempts to diminish its power. But what if X has little control over how salient an underlying issue is? Suppose that

X 1s a State whose economy relies on fishing. Further suppose that X’s neighboring rival, Y, repeatedly attempts to seize fishing ports to which it has historic claims. We would expect 8, to be rather high. Changing that value is unrealistic because it would require large shifts in X’s economy. In this case, X’s best chance to win a power transition by insuring that B, < B,[c,,(0)|/[c,,.(0)] is to increase c,,(0), its conflict level at the time of transition. Rather than worrying about its own vulnerabilities, X can impose a

higher cost of competition on its rival. For example, if Y is performing

threatening naval maneuvers, X might consider firing in retaliation. When a nation has little control over the strength of an issue, it is not without options. Unfortunately, the remaining option of becoming more

Py

70 woe

vA Py = Prt

60 s. Py = Px Py = Py + 13 ‘ Py = Py —7

emai i ~. i } 20 } x i i j OL, | | 50

30 j ;YS |, ! aNeer 3 b, /

f

40 ieerts oe 4a 4 Dy = py - 15

10 20 30 40 50 60 70

—— pO) = 2

Cys —— p(0) = 15 30 ee s--- p(0) = 30

20 Pe --- p(0) = 50 10 rae a Fal

¥

ral

10 20) 30 40 50 60

_10 ~\. 7

.

—2() “SN XS

NN

YN

oN

Fig. 18. D&G trajectories for GB, = .08

Cxy

The Moral 99 conflictual is unnerving, costly, and dangerous. This conclusion provides formal support for Organski and Kugler’s (1980) theoretical claim that more dissatisfied states —those most affected by policy disagreements with the hegemon and therefore those with the highest 6 values — are the ones most likely to initiate war when undergoing a power transition with the hegemon. Empirical support for this claim can be found in Lemke and Werner (1996) and Kim (1991). The conflict behavior associated with low values of B, is character-

ized by a pattern of one-sided conflict followed by a joint increase in hostilities. When the initial power levels are high (30 and 50), the pattern continues with a sudden shift to diminished conflict by both sides. Both rivals benefit from lessening conflict because it relieves the drain on national power. In other words, the larger the magnitude of power at the point of transition the lower the rivals’ long-run cost of competition will be. This theme recurs across all B, values. B, = .167

Figure 19 presents simulation results that are clear-cut and easy to interpret. Because they all adhere to equation (4.1), they are all B&G cases. Nation Y has an early advantage and successfully fends off X’s pesky attempt to overtake it att = 0. Still, Y’s defense is costly because its power level declines. Ironically, the larger the bull’s pretransition advantage the bigger is its fall. In the B. < .167 cases, a large power advantage could insulate Y from complete demise. The insulation effect is less evident in the B, = .167 cases, however, because a large advantage does not lessen

the price of defending against an attempted transition. Still, Y would prefer B, = .167 because then it is guaranteed to win the transition. The conflict paths all move from an asymmetric situation in which Y cooperates while X is hostile to one in which both sides become increas-

ingly belligerent. Each experiences a small upsweep at the end, where X’s conflict level tapers off as its power vanishes. In the simulation involving the highest magnitude of power at the transition point | p(O) = 50], the upsweep is followed by a decrease in both states’ conflict behavior. This again serves to lessen the long-run cost of competition. B, Values above .167

When £B, > .167, the resulting trajectories exhibit a diverse set of patterns. For the two transitions of smallest power magnitude, p(0) = 2 and p(0) =15, nation Y successfully challenges X in a traditional D&G transition (see the first graph of fig. 20). Well before the transition, nation X

Py

col Ry Pe! 7 ~ Py = Px t 7 ~Y Py = Px 50

40 foehe"TF, .

ys

30 on —— p(0) =2 enn DP mmm (0) = 15 / : ------- p(0) = 30 20 a

10 J ---- p(0) = 50 10 20 30 40 50 60

40 a Cyx

Jt

} o’

20 eo

ral

er

*

“er

10 20 30 40 50 60 .*

>

\

y

~20 X Ny

Sy YN ON

‘. \,

Fig. 19. B&G trajectories for 6, = .167

Cxy

Px

Py

A Py = Pyt T Py > Px

~~ Fal iJ 7_foer

60 | nnn EPA OS fe ve 50 eee aa aa Py = Px- 15

AQ presen Ke —— p(o) -? 30 fn howe yon GP — (0) = 19

20 ~-== (0) = 30

LF De |

10 yh a ™ “ d* + 6, where 6 is currently unknown) yet remain very

Py }

J

+7 _ Py = Px

«0 A Py = Px + V5 an pT Tt tte eee - Py=Px-7 Var

40 eves et

' mes, ~ Dy = Px — 15

Se --- ae ..fo: —— pQ) = 2

cna aa aa — (0) = 15

wae = p(Q) = 30 Hr ~--- p(0) = 50

mA 20 40 D 60 ,

20 F*

05 awfr Cyx

t

View

WiZé

7

a

5 A 10 15 20 if

5 3s S.

3.

SS

N

~» xe

Ss

SA

.® —10 , ‘Ss ‘, *\

Fig.21. D&G and T&H trajectories for B, = .5

Oxy

104 The Power-Conflict Story sensitive on the issues underlying the rivalry are ultimately worse off than nations that are just as sensitive but have a smaller power discrepancy with their rivals. In the highest magnitude cases (p(0) = 50 in figs. 18 and 20 [and 21]), the rivals are more evenly matched because the power relationship never ventures deep into lower region 3. This leaves the formerly dominant state with a small but increasing level of power in the end, which is an improvement over the lower magnitude cases in which the dominant state faces ultimate extinction. In general, a nation should strive to force the stakes at issue below

the critical threshold because this improves its chances of winning a transition. When it is not possible to change how salient or tangible an issue 1s, alternatives do exist. If a challenging nation cannot do so, its best option is to greatly increase its conflict behavior near the transition point, thus ensuring its ability to inflict damage on the dominant state and win the transition. If an overwhelmingly dominant nation cannot adjust how vital its interest in the issue is, allowing the competitor to enjoy a similarly large power level is a viable long-run strategy for survival. These strategies both act as surrogates for controlling the impact of issues, but they are very different. One is aimed at obliterating the opponent while the other’s purpose is to share power with it. A Summary

At the end of chapter 1, I inquired about the patterns of growth in national power and dyadic conflict behavior that would produce and result from power transitions. Believing that the PT and BOP explanations of war initiation could provide some insight, I drew on them in order to design a formal, dynamic model. That model was intended to be a hybrid of the two original explanations, and it demonstrated how conflict behavior changes with varying conditions in the dyadic power relationship. In this chapter, the model, a system of first-order differential equations, was solved numerically using multiple simulations. The results of these simulations provided some answers to my initial question. The most obvious finding is that there are three different patterns in

national power growth that are related to power transitions. A B&G simulation comes close to producing a transition but ultimately does not. The T&H produces two transitions, only one of which is permanent in the

end. The D&G yields a single permanent transition. These three cases allow for a much more diverse world than either the BOP or PT explanation does separately. BOP theory focuses on the peaceful balancing of nontransitions or power relationships involving a difference of no more

The Moral 105 than d*. These are seen as inconsequential results of a balancing mecha-

nism and therefore do not produce wars. A decisive transition, or a movement past d*, transpires when the balancing mechanism breaks down, and this does indeed give rise to war. The relative severity of the approximate equality zone combined with the dangers of permanently passing d* are reminiscent of the middle and later portions of the T&H case. Conversely, power transitionists emphasize D&G transitions. They, too, depict the establishment of a permanent decisive advantage as precarious, but they add an overwhelming advantage, not equality, as the most tranquil scenario. Only by integrating BOP and PT logic can we

fully account for B&G, T&H, and D&G cases within one coherent model. Most importantly, this new classification simultaneously accounts for successful and unsuccessful challenges. Another unique result relates to the tendency of the conflict behavior of nations X and Y to become coupled as time progresses. In Statistics of Deadly Quarrels, Richardson offers a vignette demonstrating that policemen have reached a similar conclusion regarding conflictual behavior on the interpersonal level.

A police constable, on being asked how he decided who was the ageressor when he found a row beginning in the street, replied that street-quarrels commonly began with bickering, then shoving one another, then fighting; and that the best policy was to arrest both parties and to charge both with a breach of peace. (1960b: 17) Coupled dyadic conflict was seen, although to varying degrees, across all three classes of trajectories. The fact that c,, and c,, tend to converge is

important for two reasons. First, it suggests that the model might be refined to include only one conflict variable, c, which measures the total level of conflict experienced by the dyad at any given time. Diehl and Goertz (1994) appear to reach a similar conclusion, presenting a single continuous measure of the conflict level jointly experienced by both opponents in an enduring rivalry. Second, it implies that neither nation can clearly be considered the aggressor. In contrast, the PT explanation does portray one nation as the more aggressive by naming the challenger as the likely initiator of war. The BOP explanation, while naming the stronger as the likely initiator of war, does not exclude the possibility that the weaker may become highly conflictual in an attempt to restore balance.

A third conclusion shows that a pair of rival nations is most conflictual immediately following a power transition. Moreover, periods between the two transitions in the T&H cases are relatively peaceful. If, during this period, the |p, — p,| trajectory remains below d*, as it does in

106 The Power-Conflict Story figure 15, then it may be considered to be a relatively peaceful balance.

Peaceful balances are predicted by the BOP explanation. Hence, the conclusions concerning conflict behavior appear to be consistent with both the PT explanation’s predictions about conflictual transitions and the BOP explanation’s predictions about peaceful balances. It is also worth noting that the prescriptions for “winning a transition” may be somewhat counterintuitive. In several ways, the ultimate winner is not a fierce fighter in the typical sense. In the B&G case, the bull succeeds in fending off the gnat’s advances in power by returning only a fraction of the conflict that the gnat directs. If the bull already holds the advantage, it may make little sense to be more conflictual than is absolutely necessary to maintain that advantage. Conflict is costly in terms of national power, so using it sparingly is wise. In the T&H case, the ultimate winner has a slower initial growth rate than the ultimate TABLE 6. Baseline Deductions from the Power-Conflict Model General deductions G1: Three distinct patterns in national power growth are related to power transitions: B&G deflections, T&H double transitions, and D&G single transitions. G2: The ultimate winner is not necessarily a fierce fighter. G3: The rivals’ conflict levels become coupled over time. G4: Rival nations are most conflictual immediately following a power transition. Bull and gnat deductions B1: The challenger nation (gnat) is consistently more conflictual than the dominant nation (bull). B2: The difference between the rivals’ conflict levels diminishes over time. B3: There are two peaks in conflict behavior: (1) the challenger (gnat) alone is highly conflictual prior to the attempted transition and (2) both contestants are highly conflictual shortly after it. B4: The difference between the rivals’ conflict values is large compared to that for the T&H type of transition.

Tortoise and hare deductions T1: The challenger nation (hare) is consistently more conflictual than the dominant nation (tortoise). T2: The difference between the rivals’ conflict levels diminishes over time. T3: There are two peaks in conflictual behavior: (1) the challenger (hare) alone is highly conflictual preceding the first transition and (2) both contestants are highly conflictual immediately following the second transition. David and Goliath deductions D1: There is a mix between cases in which the challenger (David) is more conflictual and cases in which the dominant state (Goliath) is more conflictual. D2: The difference between the rivals’ conflict levels diminishes over time. D3: Both contestants are highly conflictual shortly after the power transition.

The Moral 107 loser. A slow but persistent rate of growth in national power is more effective than a fast but tiring one. Thus, D&G transitions in which Goliath is more conflictual but loses to the up and coming warrior David provide a final example of how winners can be relatively passive. All of the baseline conclusions deduced from the power-conflict model in this

chapter are summarized in table 6. These conclusions serve to inform the process of modeling national power growth and dyadic conflict, advise policymakers, and address substantive issues raised by the BOPPT debate. Last, the strength of the issues underlying a rivalry influences how well competitors fare in a power transition. In the best-case scenario, B, < B,[c,,(0)]/c,,(0)] and nation X will be sufficiently insulated from the impact of an issue and can easily succeed in challenging Y. If a state cannot reduce its issue sensitivity below that threshold, it must choose between two strategies — directing a much larger amount of conflict at its opponent in order to win a transition or avoiding the dangers of extinction by moving to a situation in which both itself and its rival hold a large

amount of national power. Because systematic data on issues are currently limited to typologies (e.g., territorial and fisheries disputes) and do not include measures of how strongly states are concerned with them (e.g., salience, tangibility, and vital importance), these conclusions are not tested in the next chapter. Still to Come: Testing the Deductions In the next chapter, I compare the theoretical conclusions of table 6 with empirical information on national power levels and dyadic conflict behavior. However, this process is not as straightforward as it is in other circumstances. Before proceeding with the empirical tests, I clarify the reason for using a less conventional method to investigate the historical record. In order to do so, I begin with an illustration of the more traditional approach. Consider the following word problem from a standard introductory textbook on calculus (Larson and Hostetler 1979: 177). The height of a ball at time ¢ is given by the equation h(t) = 96t — 16t’.

a) What was the initial velocity of the ball? b) How high did the ball go? c) What direction was the ball moving at time t = 4? d) Find the height of the ball at time ¢ = 1.

108 The Power-Conflict Story Using the equation h(t) = 96t — 16t* as a model of a ball’s behavior over time, one can deduce the position, speed, direction, and accelera-

tion of the ball at any point in time. Furthermore, these deductions could be easily checked against the measured speed, direction, and accel-

eration of a particular ball (or class of balls) whose behavior(s) one believes to be governed by the same equation. For example, it can be shown that at ¢, = 3 (seconds) the hypothetical ball should reach its maximum height of h,,.. = 144 (feet). Suppose that the behavior of a certain yellow tennis ball is thought to be governed by the h(t) = 96t — 16¢* equation. That tennis ball’s maximum height, A naxs and the amount of time it takes to get there, t ;,,» can both be measured using a yardstick

and a stopwatch. The tennis ball’s observed maximum height and the amount of time that elapsed while it got there can be compared to the parallel quantities predicted by the theoretical model. Furthermore, this test for a correspondence between the predicted and observed values of Aya andt, can be performed according to rigorous and well-established guidelines that indicate whether or not any discrepancies are statistically significant. If the differences between h,,... and A pox and between ¢, and

by are small enough that they are statistically insignificant, then the model would be confirmed as a useful representation of the tennis ball’s movement over time. Can this same method of analysis be used when comparing deductions from the power-conflict model with their empirical referents? If I

have a set of initial conditions and parameter values that can serve as substantively meaningful assumptions, I can use the equations in the power-conflict model to make highly precise predictions about the values Of P,, P,, C,,, and c,, at any particular point in time, just as I did in each simulation in this chapter. Moreover, I can answer the same kinds of questions that are asked in the word problem described earlier. For example, after running the simulation represented in figure 15, I can analyze the resulting theoretical time-series to find out when the hare (nation X) held the most power and at what point in time it did so. For the curious at heart, p, reached its maximum of approximately 19.25 at t ~ —.69. Similarly, the numerical analysis of the model yields the maximum and minimum values for each nation’s conflict behavior during a simulation. Suppose I believe that the national power levels and conflict behaviors of a particular dyad (or class of dyads) are governed by the equations

of the power-conflict model. Can I now measure each nation’s power level and conflict behavior at important points in time? The answer is a qualified yes. In chapter 5, I present indicators of national power and directed conflict, but they are imprecise measurement tools. So, when I

The Moral 109 say that France’s power level reached its peak of almost 22 percent of major power capabilities in 1841, I am not nearly as confident in that estimate as I would be in a standard yardstick’s measurement of the maximum height of a tennis ball. Alas, the international relations discipline is young and has not yet produced reliable stopwatches and yardsticks. While the model is capable of producing precise predictions, the best observation tools available are not capable of providing equally accurate measurements against which those predictions can be compared.

Rather than making point predictions whose precision of informa-

tion would be wasted and untestable, I instead compose alternative types of predictions, all of which are phrased in terms of broad-based patterns. Deduction T3 represents one type of conclusion that I will test in chapter 5. That statement relates the expectations for the timing of conflict peaks. In lieu of saying that the tortoise’s conflict level crests at a specific point in time, I say that it does so at two points in time that are positioned in relationship to the first and second power transitions. Scholars of BOP and PT similarly construct their timing of war expectations

with reference to certain power relationships (see chapter 2). I can therefore evaluate the evidence in terms of whether it is most consistent with my hybrid model, the BOP explanation alone, or the PT explanation alone. A second type of prediction made in this chapter is demonstrated by deduction G1, a statement about three general types of transitions. Its expectation is simply that all three exist in the world of major power rivalries. This expectation is different from that held by scholars wedded to the PT or BOP school because each group would predict only one

type of transition. If I discover B&G, T&H, and D&G transitions among the major powers, then there is more support for the powerconflict model than there is for either of the traditional versions of dyadic competition. A similar type of prediction is made in the field of particle physics. According to the Standard Model, scientists should expect to find top quarks as one of the six building blocks of matter. When researchers at the Fermi National Accelerator Laboratory discovered evidence of the top quark’s existence, they gained confidence in the superiority of the Standard Model as a useful characterization of the physical world (Browne 1995). Third, I make use of predictions that compare two important values. For example, deduction B1 specifies that the bull directs less conflict at the gnat than the gnat directs at it. Recall that by substituting B, = .5 and B, = .1 into equation (4.1) I was able to infer that c,,(0) must be exactly five times larger than c,,(0) in all deflection cases. Because I cannot depend on a precise measurement of two nations’ conflict values

110 The Power-Conflict Story at the deflection point or on that particular dyad’s B values, I am not able to assess the statistical significant of the difference between observed and predicted conflict values. I therefore generalize the prediction to the coarser statement listed as B1 in table 6. Calculating the mean ratio of conflict directed by a gnat to that directed by a bull, for instance, would produce a relatively inexact value. It would be nonsensi-

cal to then perform a statistical test geared to compare that value to a much more precise one. The underlying problem, then, is that indicators of national power levels and dyadic conflict behavior are less exact and reliable than would be necessary for meaningful statistical tests of point predictions. None-

theless, as long as they can be tested in some reasonably objective manner, less precise predictions can be useful inasmuch they either: (1) provide expectations that are distinct from parallel predictions by competing models or explanations or (2) provide novel expectations whose substance is not addressed by the competitors. Because the deductions in table 6 meet those criteria, the next chapter 1s devoted to testing their expectations against observable patterns in national power and dyadic conflict.

CHAPTER 5 Verifying the Story

Good stories might fascinate and intrigue us, but without an element of

truth they are merely forms of entertainment. This chapter tests the propositions derived from the numerical analysis of the power-conflict

model and listed in table 6. It begins with a brief discussion of the empirical context and the operationalization of national power and inter-

state conflict. As a test of hypothesis G1, it next identifies all power transitions occurring within the universe of analysis and attempts to classify them into the three types discovered in chapter 4—bull and enat, tortoise and hare, and David and Goliath. It continues with a systematic analysis of the remaining hypotheses, all of which concern dyadic conflict behavior. Finally, three historical case studies are conducted to supplement the systematic analysis. The Rivalry Context

The context of the power-conflict model 1s relevant because it is the background against which the story is told and because it determines the

universe of analysis that is used in any empirical test. In chapter 2, I elaborated on the basic assumption of rivalry or incompatible preferences that functions as the setting for the power-conflict story. Here I discuss how that is translated into selecting a group of interstate dyads that should exhibit the behavior patterns predicted by the formal model. A variety of methods of generating lists of competing interstate dyads are currently used in international relations research: geographic closeness measured by contiguity (e.g., Ferris 1973; Weede 1976), proximity (e.g., Maoz and Russett 1993), or regional military reach (Lemke 1995; Lemke and Werner 1996); enduring rivalries (Goertz and Diehl 1992a); and pairing states that hold a certain power status (e.g., Organski and Kugler 1980; de Soysa, Oneal, and Park 1997).! Drawing on the last method, pairs of major powers will serve as the set of meaningful dyads in this study. Major power nations, by virtue of their high-ranking capabilities and prestige, have played more significant 111

112 The Power-Conflict Story roles in international politics than other nations have. In their study of historical patterns in international conflict, Small and Singer reason that major powers “are involved in the bulk of the system’s diplomatic and martial activities” (1979: 65). One major power activity that is especially relevant to this book is vying for shares of world resources. In certain cases, this competition produces the special phenomenon called a power transition. Pairing each major power with every other is based on the

realist assumption that an important national actor is potentially in competition with any other actor of similar stature. Moreover, major power interactions “seem to set the tone for other nations’ interactions during the various historic periods” (65). Singer, Bremer, and Stuckey (1972) provide an intentionally “less than operational” description of the major power system membership from the Congress of Vienna until 1965. I consulted Gochman and Maoz’s (1984) and Hensel’s (1993) updated lists for current information on the power status of nations. This information is summarized in table 7. The decision to use major power dyads does not rule out generalizing to other types of competing dyads in the future. Measurement

This section is devoted to settling on indicators of national power and dyadic conflict behavior. A variety of issues are discussed. Alternative TABLE 7. Membership in the Major Power System, 1816-—Present

Member Years as a Major Power United Kingdom 1816—present France 1816-1940 1945—present

Prussia/Germany 1816-1918 1925-45 Austria-Hungary 1816-1918 Russia/Soviet Union 1816-1917

Italy 1860-1943 United States 1899—present 1922—present

Japan 1895-1945 China 1950—present Note: Neither Germany nor Japan has been restored to major power status, as both are lacking sufficient military capabilities.

Verifying the Story 113 measures are considered, and criteria for selection are established. Both of the chosen indicators have some weaknesses, and those are investigated and remedied. Building an Indicator of National Power

As a measure of national power, I have selected the Singer, Bremer, and Stuckey (1972) composite indicator of national capabilities (CINC). This indicator is built using information about the physical capabilities that a nation has at its disposal along three very general dimensions: military, industrial, and demographic.* Each of these dimensions is captured by two pieces of information. The military dimension includes the number of military personnel enlisted in a nation as well as the amount of money

expended on the military. Energy consumption (after 1885) plus iron production (1820-95) or steel production (1896-1985) reflect the industrial dimension.* The demographic dimension is composed of the total population and the population living in cities of 20,000 people or more.4

The first step in constructing the indicator is to compare a single nation’s capabilities on one aspect to the total of all major powers’ capabilities on that particular aspect. In other words, a nation’s system share of a single capability 1s assessed. For example, in order to compute

nation X’s CINC score, one starts with its share of the entire system’s military expenditures, % ME,, which is calculated in the following way:

% ME, = — 5 (5.1) ME.

ME, + & ME...

where: ME, is nation X’s military expenditures and ME, ners iS the sum of all other nations’ military expenditures. A similar calculation is done to determine nation X’s share of each of the

remaining five aspects of national power. This produces a value for YMP., %IS,, YNRG,, %UP,, and %TP,, that is, nation X’s share of military personnel, iron and steel usage, energy consumption, urban population, and total population. Because each of these percentages represents a portion of the whole, they range from zero to unity. Now it is possible to combine them into a single indicator of national power. Averaging the percentage shares for the six capability dimensions combines them into a single value. In other words, nation X’s CINC score, the indicator used to measure the variables p, and p, in the powerconflict model, is given by:

114 The Power-Conflict Story % ME. + %MP, + %IS, + %YNRG, + %UP, + %TP

CINC, = ——_—TYTY,_lwWJ—{\____—_.1.1— 6 *, (5.2) This procedure is followed for each nation in the system for each time period. Missing data were handled in two ways. In all cases but one, missing

information on any of the six capabilities indicators for any nation in a particular year caused that indicator to be removed from the analysis altogether. For example, if iron and steel information were unavailable for nation X, then equation (5.2) would become:

CINC. = YoOME, + %MP, + aw + %UP, + /oTPy (5.3) Similarly, the iron and steel component would be eliminated from the calculation of nation Y’s CINC score:

CINC, = a , (5.4) Y ME, + %MP, + %NRG, + %UP, + %TP

The same procedure also applies when information on two or more capabilities is missing. The case of missing data on all six capabilities for German-occupied France in 1943 was especially problematic. Use of the rule would produce

national power estimates of zero for all major power nations in 1943. Instead, I estimate the value for all six of France’s capability indicators by

looking at the pattern in each indicator’s behavior from 1932 to 1953. Each pattern is captured by using cubic splines, that 1s, by fitting a thirdorder polynomial between successive data points. The resulting interpolation function is then used to determine a reasonable value for a specific

indicator in 1943.5 Because data are also missing for French military personnel and expenditures in 1940 and 1942, and because an interpolating function covering those years must be devised to estimate values for 1943, I use that function to estimate values for 1940 and 1942 as well. My choice to use the interpolating function to make sensible approximations for the 1943 data should not be understood to imply that France was indeed France at that time. Rather, it is a prediction of what those values would have been if France had been unoccupied.°®

The most popular alternative indicator of national power is gross national product (GNP). Several researchers (e.g., Organski and Kugler 1980; Organski 1958; Hitch and McKean 1960; Russett 1965) prefer this

Verifying the Story 115 single capabilities indicator. Some of their arguments are based on the

series’ correlation with other measures of power such as the CINC score, diplomatic status, or perceived importance as judged by laypeople.’ If the correlation is high and positive, they reason, it is most efficient to use the simpler indicator. However, the reliability and avail-

ability of the six capabilities indicators composing the CINC scores seems to be better than that for GNP when we consider both the temporal and spatial universes that are covered. In particular, GNP data are typically available only for highly industrialized and capitalist societies.

This makes attaining reliable GNP information on a yearly basis for preindustrial nations, and for any nation before the 1930s, problematic (Rosen 1972: 171). Merritt and Zinnes (1989) consider the correlations

among indicators devised by Alcock and Newcombe (1970), Fuchs (1965), German (1960), Cline (1980), and the Singer, Bremer, and Stuckey CINC score. They conclude that there is a great deal of agreement among all of the indicators, particularly for the most powerful nations (Merritt and Zinnes 1989: 26). Power as a National Capacity

One necessary question is whether or not the composite indicator possesses “content validity” (Zeller and Carmines 1980: 78-79). In other words, does the measurement device reasonably capture the concept it is meant to represent? Specifically, I am interested in whether or not the CINC score makes sense when we consider power as a national capacity.

In designing the equations that determine how p, and p, behave over time, I argued that power is a national capacity and can therefore be accumulated and depleted. According to equations (3.3) and (3.4), accu-

mulation is driven by an internal force and depletion is the result of dyadic conflict. I would argue that the CINC indicator’s three dimensions of national

capacity, in turn, can be accumulated or depleted. What are some examples of growth and decline in each of these three areas? Increases in military capacity are the function of processes such as building and stockpiling weapons, training soldiers, and devising surprise combat strategies.

Military capacity is depleted when missiles are fired, lives are lost in battle, and strategies are divulged through implementation. Industrial capacity accumulates when the number of goods produced increases or when those goods are produced more efficiently, that is, in less time or with fewer resources. When bombing destroys factories or infrastructure,

that productive capacity is harmed. It is also harmed when available resources are diverted, say, for participation in international conflict, so

116 The Power-Conflict Story that production slows or stops altogether. Increases in demographic capacity can be reflected in a large and growing working population, which is typically capable of large-scale production while sustaining dependent segments of society. This capacity is threatened when significant portions of the productive population, civilian and military, die during war. Difficulties arise, however, if the accumulation and depletion char-

acteristics are imposed directly on the six indicators themselves. For example, it does not seem to make sense to talk about accumulating or depleting military expenditures or iron and steel production.® For that matter, how is energy usage stored? Instead, these indicators must be taken as proxies for concepts that do exhibit those characteristics. Kaplan warns:

The danger is that we succumb to what Coombs calls “operationism in reverse,” that is, “endowing the measures with all the meanings associated with the concept.” (1964: 199) This is not to say that we ought to ignore an indicator’s content validity,

but that we must not take it so far as to require measurements to be conceptually identical to their priors. Discrepancies between concepts and their indicators similarly exist in the physical sciences. Certain characteristics of time are not exhibited by a clock. Time is irreversible, but a clock can be turned back as far as we might like. In order to avoid the fallacy of “operationism in reverse,” Kaplan advises the researcher to check the indicator’s ability to predict values for some other empirical referent of the concept: Here the validity of a measurement is a matter of the success with which the measures obtained in particular cases allow us to predict the measures that would be arrived at by other procedures and in other contexts. The intelligence test, for example, 1s validated in the

degree to which the IQ score enables us to predict scholastic achievement, or performance in other problem-solving situations. (1964: 199)

Most notably, Organski and Kugler’s (1980) comparison of annual system shares of capabilities with system shares of GNP from 1870 to 1965 yields a correlation coefficient of .86. Their conclusion on the matter of indicator selection is that “so far as performance is concerned, there 1s no particular advantage in choosing one measure over the other” (38). Similarly, Taber’s (1989: 40) results from cross-correlation of the CINC score and GNP for 37 developed nations produces a Spearman’s r of

Verifying the Story 117 .893. As I have mentioned, the rationale for choosing the CINC score over some type of GNP measure lies primarily with the unavailability of the GNP data for more distant time periods. Hence, the CINC score is the best available indicator of national power. Adjusting the CINC Score

The use of a system of nations as a measurement tool raises an important issue concerning the power-conflict model. The equations presented in chapter 3 model a dyadic situation, that is, the power relationship and conflictual interactions between two rival nations. Built into the tech-

nique for measuring national power, however, is the comparison of a nation’s capabilities with those possessed by a group of many nations. Is there an inconsistency in measuring a strictly dyadic power distribution by referring to what are essentially the power levels of third parties? At first glance, 1t may seem that the inclusion of all major powers’ capabili-

ties in the indicator’s denominator is inappropriate. In their consideration of the CINC score, Organski and Kugler reason:

The measure that Singer and his colleagues have produced is a relative measure, in which the capabilities of one nation depend not

only on its own performance but also on that of the sample as a whole and on every other nation in the sample. When the relative power of one nation declines, one cannot determine whether this is because that particular nation is doing worse or whether the average growth of the sample as a whole is improving or, in the latter case, whether the overall improvement in the sample is due to a general increase in performance or to the increase in performance of one nation in particular. One cannot make a satisfactory deduction unless one goes back to the original data from which the percentage shares were computed. (1980: 36) Two solutions to this problem come to mind. First, one must ask if it

still might not be meaningful when one rival’s CINC score is pulled down by a sudden increase in a third party’s score while the other rival’s score is not affected. For example, following World War II, America’s

CINC score surged. Concomitantly, those of Germany and Italy declined while those of the Soviet Union and China rose.’ The latter two major powers weathered the storm much better than the former two and could therefore be said to have “won” any rivalry with the others. Second, the negative effects on a nation’s CINC score of a single nation experiencing temporarily extraordinary levels of capabilities are muted,

118 The Power-Conflict Story although not entirely eliminated, by estimating that nation’s power level in any particular year by averaging that year’s score with the previous and subsequent years’ scores.!° Any other sudden and unusual shocks to the CINC scores are also smoothed out in the averaging process. Hence, I adjust the indicator as follows:

CINC’ (1) = CINC(t — 1) + — (t) + CINC, (¢ + I) 5.5) Similar problems might be caused by the changing membership of the major power system (Kadera 1995; Kadera and Sorokin 1998). The number of nations whose material capabilities should be included in the denominator in equations like (5.1) does not remain constant. Table 7 indicates that there are 11 different times at which entries and exits of members to and from the major power system take place. In order to minimize their frequency, I include a nation in the analysis for all years in which data on its capabilities were available. The implication 1s that some nations are included in CINC score calculations even for years in which

they were not considered major powers (e.g., Japan after 1945, Italy after 1943, and the United States before 1899). This minimizes changes in the number of nations considered in the denominator to only four. The resulting five time periods are listed in table 8. Criteria for a Conflict Indicator

A good measurement tool will reflect qualities similar to those found in the concept it is attempting to gauge. Like the CINC score, the conflict indicator must possess content validity. In chapters 2 and 3, I presented several aspects of dyadic conflict behavior that reflected my conceptual TABLE 8. Major Powers Excluded in the Calculation of CINC Scores

Time Period Excluded Nations Number of Nations Included

1816-94 China 7 1895-1918 none 9 1919-45 Austria-Hungary 1946-51 Japan 68 Japan

Prussia/Germany Austria-Hungary

1952-85 Prussia/Germany 7 Austria-Hungary

Verifying the Story 119 understanding of that term. By way of review, a brief summary of the qualities used to characterize conflict in those chapters 1s given here: 1. Hostile behavior 2. Not necessarily physical or violent 3. Actor and target can be identified 4. War is indicated by a very high critical value of conflict 5. Can be thought of as the flexing of a nation’s power 6. The means by which nations oppose one another 7. Cooperation is its opposite The first four assumptions about conflict are fairly straightforward. One can easily picture the type of data that would be consistent with them. Any report of actions taken by one nation against another would qualify. For example, “France deploys troops to the German border” would be an example of a physically hostile action taken by France, the actor, against Germany, the target. Another example is given by the statement “German officials denounced the positioning of French troops on their border.” The verbal disdain shown by Germany toward France is certainly a hostile action, although it is not a physically violent one. Ona conflictual activity scale ranging from 1 to 10, the deployment of troops might rate a6 while the denouncement would be less, perhaps a3. If, however, the datum “German troops fire on French forces” were recorded, it would indicate a war-initiating action and would therefore rank much higher, possibly a 9 or 10. All three hypothetical data are both consistent with and illustrative of the first four assumptions concerning conflict. The remaining characteristics are slightly more complex. Numbers 5 and 6 summarize how conflict and power are related. Because these two assumptions concern the relationship between two variables, and not simply the nature of one variable, they are incorporated into the model itself. The portrayal of conflict as the flexing of national power is shorthand for saying that the more power a nation has at its disposal the more capable it is of taking increasingly conflictual action. In the base conflict equations, (3.7) and (3.8), the “flexing assumption” is demonstrated by the positive contribution of p, and p, to dc,,/dt and dc,,/dt, respectively. The sixth assumption, that conflict is the means by which nations oppose one another, translates into the idea that actors use conflict in order to

respond to unfavorable power relationships. It is integrated into the model in the second terms of the conditional conflict equations. For instance, in equation (3.8.21), representing nation Y’s response to a decisive (but not overwhelming) advantage held by the opponent, dc,,/dt is a function of, among other things, p, — p,. Nation Y’s conflict level

120 The Power-Conflict Story increases most rapidly when this difference is small or not yet close to an overwhelming advantage, d. Not only would it be difficult to incorporate these notions into a measurement of conflict, but it 1s unnecessary since the model already takes them into account.

The seventh characteristics is of yet another kind. It might be described as a deduction from the model as well as a prior assumption. Several simulations in chapter 4 produced conflict values less than zero, and the idea of cooperation as negative conflict was presented as a useful interpretation. It would therefore be practical to have an indicator possessing this quality. Conflict Indicator Options: COPDAB, WEIS, and MID

Three commonly accepted events data sets— Azar’s (1970, 1980) Con-

flict and Peace Data Bank (COPDAB), McClelland’s (1971) World Events Interaction Survey (WEIS), and the Correlates of War Project’s Militarized Interstate Disputes (MID) list (Gochman and Maoz 1984; Jones, Bremer, and Singer 1996) — present themselves as capable of demonstrating a reasonable number of the qualities numbered one through four and seven. Each begins with written records of international interactions similar to the sample statements given earlier. These are then converted into numerical data that specify the date on which an event took place, the national actor and target, and the level of conflict reached.

As a brief summary, the qualities each data set possesses, along with the greatest disadvantage of each, are listed in table 9. Although TABLE 9. Three Potentially Useful Conflict Data Sets

Data Set, Researcher(s), Characteristics

and Time Period Covered Demonstrated Greatest Disadvantage

COPDAB 1-4 and 7 Time period covered is short and

Azar 1980 does not overlap with many

1943-78 empirical power transitions. McClelland 1971 does not overlap with many 1966-79 empirical power transitions.

WEIS 1-4 and 78 Time period covered is short and MID 1, 2 (to some extent), © Low- to medium-level conflict

Gochman and Maoz 1984 3 (to some extent), and cooperative behavior are

1816-1976 and 4 excluded.

«McClelland originally intended for WEIS to be an unscaled set of categorical data. In other words, he did not want there to be any particular values, negative or positive, associated with events. However, efforts by Vincent (1983) and Goldstein (1992) have produced conflict-cooperation scales for WEIS. Thus, criteria numbers 4 and 7 are actually met through research completed by others.

Verifying the Story 121 both the COPDAB and WEIS data sets clearly meet more of the qualita-

tive criteria than does the MID list, neither covers a very large time period. More importantly, the time periods covered by COPDAB and WEIS overlap with very few empirical power transitions. Because the analysis of the power-conflict model presented in this book focuses on the nature of power transitions, it is essential that the conflict patterns considered in this chapter be observed near empirical power transitions. COPDAB covers the 1943 to 1978 period, which coincides with only three complete major power transitions, and WEIS’s 1966 to 1978 period only includes one.!! Because the MID data set spans more than a century, it serves as the basis for my examination of the empirical con-

flict record. The MID data set is not without its problems, however. Most importantly, it only reports instances that involve the threat, display, or use of military force. Criterion number 2, that conflict need not be physical or violent, is only vaguely met in the sense that threats to use force are verbal, albeit verbal promises of violence. The inclusion of cooperative behavior, criterion number 7, cannot possibly be met. These limitations and others will be discussed in the next section. First, however, I will briefly describe the list of disputes and the qualities that the data set does demonstrate. Gochman and Maoz offer the following concise description of their original collection: [W]e define a “militarized interstate dispute” as a set of interactions

between or among states involving threats to use military force, displays of military force, or actual uses of military force. To be included, these acts must be explicit, overt, nonaccidental, and government sanctioned. (1984: 587)

A dispute, then, 1s actually a series of substantively related interactions taking place over an extended period of time.!* Each of those interactions is characterized as one of 14 types, which are not ranked by inten-

sity (Jones, Bremer, and Singer 1996: 171). Those types are then grouped into three broad categories that are ranked according to their severity: the threat, display, or use of force. An example of one type of event falling into the first category would be the “threat by one state to use ships or troops to seal off the territory of another state so as to prevent either entry or exit” (Gochman and Maoz 1984: 588). An alert, or a “reported increase in the military readiness of a state’s regular armed forces, directed at another state” (588) would fall into the second general category. The use of force category includes actions such as the occupation of territory (i.e., the “use of military force by one state to

122 The Power-Conflict Story occupy the whole or part of another state’s territory for a period of at least 24 hours”) as well as war (“sustained military hostilities between the regular armed forces of two or more states, resulting in 1000 or more battle fatalities”) (589).!° To what extent do these data demonstrate qualities 1-4 and 7? The first, third, and fourth qualities are clearly demonstrated by MID interactions. Without a doubt, each type of interaction is antagonistic, so this measurement tool upholds the first assumption about conflict.

The third criterion is met by most, but not by all, of the 14 types of conflictual events. In their category descriptions and in the data set’s “originator” field, Gochman and Maoz (1984) typically specify that one State is the initiator and a second is the target. The first state 1s taking a particular action that is unquestionably directed at or against the other. There is one exception to this criterion. Gochman and Maoz define a clash as “military hostilities between the regular armed forces of two or more states that last for less than 24 hours and in which the initiator of the hostilities cannot be identified clearly” (589).'4 In terms of how conflictual each clashing or warring state is, though, Gochman and Maoz, much like the policeman in chapter 4, must consider both the initiator and the target as having reached the highest level possible. In the case of a clash, then,

both actors receive a four, as if both are equally culpable of initiation. Similarly, in the case of war, both participants receive a five. The operating assumption behind this view of physical combat is that both sides can exercise the option to back down and refuse to fight. Surrendering may not be an attractive alternative, but it exists nonetheless. Criterion number 4, that war is indicated by a very high critical value of conflict, is met by the MID list. The occurrence of war is distinguished

from other uses of force and receives a higher numerical value. Some concern might be expressed, however, over the allocation of a four to uses of force short of war and a five to incidences of war itself. While war has a higher conflict level, it is not a great deal higher than that attributed to other uses of force. Only one unit of conflict separates war from the use

of force, the same amount separating a threat from a display. This is because the MID scale is an ordinal one. So, while the occurrence of war is more severe than, say, a declaration of war, one cannot be sure of how much more severe it is. This problem will be addressed at length in the following section. Remedies for Weaknesses in the MID Data Set

In several ways, the MID data set proves disappointing for the purposes of this book. Two of its weaknesses have been briefly mentioned and will

Verifying the Story 123

be treated more thoroughly here. In particular, I discuss the consequences for assessing the empirical-theoretical link. Some additional difficulties, as well as potential remedies, are explored as well.

The greatest weakness of the dispute list is that the researchers establish a threshold of military force for inclusion of an event in the data set. This eliminates a large spectrum of interactions, from peaceful cooperation to nonmilitary conflict. A statement such as “High level diplomats from the United States traveled to Japan to negotiate a trade agreement” would not be reflected in a MID-based indicator of the conflictual behavior of the United States toward Japan. Ideally, such an event would be recorded as a negative level of conflict, that 1s, as cooperation. A statement representing low to medium levels of conflict would also be passed over by the MID measurement tool. For example, neither “France erected agricultural trade barriers designed to diminish the import of wines from Italy” nor “Britain’s minister of foreign affairs canceled her diplomatic trip to Moscow” would appear in the coded set of major power interactions. If they were included, they would be best represented by a positive conflict value below those assigned to events

involving military force. So instead of giving the entire picture of dyadic relations the MID data set, as its name clearly indicates, only portrays one extreme of those relations. One can think of this as being able to see only the tip of a conflict iceberg, with the tip being only the most conflictual portion. When empirical dyadic conflict behavior is plotted across time, one must imagine that the entire plot is shifted upward to lie above the horizontal axis representing the line along which conflict is zero and that there is potentially a good deal more of information below the horizontal axis.

This adjusted image reminds us that the lowest MID level, a threat to use force, actually represents a fairly high level of conflict when one considers the entire range of behaviors and that interstate cooperation is pursued at times even between opponents. Careful case studies partially remedy this problem and are implemented here. A more thorough solution would be to sift the historical record to collect the entire spectrum of cooperative and conflictual major power events for the pre-World War II period. Although this latter remedy is ideal, it is also quite costly and time consuming. At present, such a project is under way.!© When

the data set is complete, the empirical tests can be rerun using the updated information. The second problem is that the coding scheme uses only an ordi-

nal scale for assessing the severity of the conflictual behavior in a dispute. The authors assign the following values to each category of behavior:

124 The Power-Conflict Story

War 5

Threat to use force 2 Display of force 3 No military response (for target nations only) 1

Use of force short of war 4

For years in which neither disputant takes any military action against the other, no entry is made. This 1s functionally equivalent to recording a zero for each side of a dispute. Note that the assignment of a one to “no

military response” is misleading. If Germany threatens to use force against France and France does not respond, then Germany’s conflictual behavior 1s coded as a two while France’s behavior is coded as a one, even though France takes no action whatsoever. For the purposes of this investigation, it would be more appropriate for France’s entry to appear as an absolute zero to indicate a complete lack of military conflict on its

part. Therefore, I subtract one from each level’s value. That is, the assigned values become:

War 4 No military response (for target nations only) 0

Threat to use force 1 Display of force 2

Use of force short of war 3

Elsewhere I show that this revised MID scale closely replicates COPDAB’s ratio scale (Kadera 1995). How might the conflict scale be further adapted in order to account for frequent conflict? One simple recommendation would be to sum conflict values over small time periods. Beginning in 1880, I therefore ageregate all conflict behavior for each disputant in each year. As a consequence, wars might no longer appear as peaks in conflictual behav-

ior once events have been aggregated by year. Any combination of events adding up to five or more would be recorded as more severe than a single war. Since the revised measurement procedure approximates a

ratio scale, the relative severity of threats, displays, uses, and war are known and yearly aggregation is acceptable.

It is important to keep in mind that the measurement technique developed here is primarily meant for broadly based comparative purposes. The indicator should be able to produce information about general patterns in two opposing nations’ conflict levels over time. These can be compared to the theoretically derived patterns in c,,(¢) andc,,(¢), as can be seen in the third graphs of figures 14, 15, and 16. This comparison is not as

rigorous or objective as a standard statistical comparison of observed

Verifying the Story 125 values and expected values generated by a linear model. The particular values of data points are not relevant; rather, it is the overall pattern that is given attention. Moreover, the indicator is not meant to be universally valid; instead, it is offered as the best possible conflict measure, given a variety of constraints and concerns, for the purposes of this research. The Three-Stories Hypothesis The simulations in chapter 4 revealed three different patterns in the evolu-

tion of two rival nations’ power levels that are related to power transitions. In the B&G case, a power transition almost occurs but ultimately does not. Iwo power transitions are experienced in the T&H scenario, and only the second is a decisive one. The D&G pattern is characterized by a single permanent transition. Does the empirical record shown in figures 22 through 24 exhibit three similar patterns? Each figure depicts

—-— United States —_— United Kingdom

--- France =~. Prussia / Germany

0.4 eo Italy

CINC score oe Austria-Hungary

----—-- Russia / Soviet Union

0.35

OO ee _ 0.2 ee a eee a ye nok 0.3

ue’ ~~ we . ot O15 permSree menenWs NO”Zee a‘

0.1 ‘. f repens nal Senet ernment ” \

VY F < Teonee =ee ee = ee- one NT 0.05=pee plated moon 1820 1830 1840 1850 1860 1870 Fig. 22. Major power CINC scores, 1817-72

Year

126 The Power-Conflict Story

— United States —— United Kingdom

--- France

--- Prussia / Germany

0.4 — Italy 0.39 a China 03 oo Japan

CINC score —-—-- Austria-Hungary meores Russia / Soviet Union

0.25 0).2

0.15 aN ne . “- = “~ “ fm’ / ae ue o We wf WS

0.1 onrawas S~* Se er et:“ewe wo Pia

0.05 prremamearsereen ete mama ~ Te ag ee ™ essen’,

~ tai eat Cale ae | Year

1880 £890 1900 1910 1920 Fig. 23. Major power CINC scores, 1873-1928

one-third of the 1816-1984 time period, tracing the evolution of CINC scores for all major powers (or for those for which data exist) together. I visually examine these graphs for any power transition cases. Some difficulties arose in determining whether or not transitions occurred and in classifying them as a B&G, T&H, or D&G transition. When two nations’ power levels appeared to fit into the deflection category, I did not employ a particular standard in order to determine how closely the paths must come together in order for the scenario to be considered a deflection. In the simulations, of course, p, = p, at the actual point of deflection. Another question that arose was whether a pair of nations experiencing two transitions over a period of time should be classified twice in the D&G transition category or once in the T&H category. I am left, for now, without a rigorous rule for deciding whether a double crossing should be counted as two D&G transitions or one T&H transition. I rely instead on intuition, historical interpretation, and the

Verifying the Story 127 ——_ United States —— United Kingdom

_-- France

oo Prussia / Germany --—-- Austria—Hungary

—-=-= Italy

0.3 ete

CINC score _---—- Russia/ Soviet Union

0.4 —_—— China acon Japan

0.35

0.25 PON gine ccc 0.2 at ie / my fi E YN «NA 0.15 ; ‘.A!

a t a pecs.

- Year

0.177 ‘ pe~reet ~ en aeeenesesetenn, | ae

0.05 en RN Pain “ere eee eee. ™_ ‘ sneer meme cpmameememn to eee

1930 1940 1950 1960 1970 1980 Fig. 24. Major power CINC scores, 1929-84

advice of colleagues. It is especially reassuring to note that one other scholar, Frank Wayman (1996), has classified empirical power transitions among the major powers in a similar manner. He uses the categories of “single transitions” and “double transitions” in a manner consistent with

cases presented here as D&G and T&H transitions. In addition, Wayman’s “approach without transition” incident is much like a B&G case. Table 10 summarizes the results. Appendix B lists all power transitions by rivals, date, type, and winner. A total of 42 major power transitions were

found, with an ample number falling into each of the three categories. This supports G1’s expectation that all three types, not just D&Gs, occur. A brief comparison of the information in table 10 with the types of empirical power transitions found by Organski and Kugler (1980) and by

those replicating their work (e.g., de Soysa, Oneal, and Park 1997; Houweling and Siccama 1988; Kim 1991) is informative. For what types

128 The Power-Conflict Story

of transitions do these authors account? The “equal and overtaking” columns in their tables (see, e.g., tables 1, 2, and 3) capture the traditional D&G transition. For each dyad in each of The War Ledger’s six test periods from 1860 to 1975, “A country was seen to have passed another when the nation that was less powerful at the beginning grew more powerful than the other member before the period ended” (Organski and Kugler 1980: 49). Hence, the power relationship is examined

only at the beginning and end of each test period. If a T&H transition were to take place completely within a test period, it would be coded as “equal, no overtaking” since the tortoise both begins and ends ahead of

the hare. If a T&H transition begins in one test period and ends in another, it would be counted as two D&G transitions. For example, the Chinese-Soviet double transition from 1929 to 1941 (see fig. 24) begins in Organski and Kugler’s 1920-39 test period and ends in their 1945-55 test period and would therefore count as two instances in the “equal and overtaking” category. A B&G case would be included in the “equal, no overtaking” columns. The “equal, no overtaking” columns would thus contain B&G cases and some T&H cases as well as nontransitions cases (the stronger has no more than a 20 percent advantage over the weaker). The “equal, and overtaking” columns are comprised of D&G transitions and twice-counted T&H cases. Last, the “unequal” columns cover only instances of nontransitions in which one state has an advantage of 20 percent or more. Does it matter that the “equal, and no overtaking” and “equal, and overtaking” columns fail to make the same kinds of distinc-

tions among various power transition types as does table 10 and the power-conflict simulations? ‘The most obvious problem is that a simple frequency count of the different types cannot readily be made. TABLE 10. Empirical Power Transitions between Major Powers Total Number

1816-94 4 3 5 12 1895-1918 9 2 4 15 1919-45 5 4 2 11 1946-51 1952-85 00 00 13 13

Time Period Bulland Gnat Tortoise and Hare Davidand Goliath of Transitions

1816-1985 18 9 15 42 Note: I do not include transitions occurring when a nation enters the analysis with an average system share of capabilities higher than those of other nations. Approximately 10 simple transitions would be added if I did so. Nor do I include transitions occurring when a nation leaves the analysis. This would add two D&G transitions to the table. Also, the France and Russia/Soviet Union dyad is not included. I cannot readily decide in which category their transitions belong.

Verifying the Story 129 A more serious problem might arise if the cases that are misrepresented in these types of contingency tables fall into either the “war” or “no war” row in a biased manner. For instance, T&H cases that span two test periods and are counted twice in the “equal and overtaking column” will, according to the simulation results in chapter 4, tend to fall into the war/no war cells in the following way: the first transition (not

permanent) will be in the no war row (relatively low conflict) and the second (permanent) will be in the war cell (relatively high conflict). Such

a transition should instead be interpreted as one instance that produces high levels of conflict near its end. Counting them in this alternative way would improve support for the PT expectation that transitions are conflictual events. Given the simulation results telling us that they are conflictual phenomena, T&H transitions falling within a single test period would similarly fall into the war row. Because the coding rules do not count these cases as power transitions, they fail to lend the support they otherwise would to the PT theory. Hence, the oversimplified classification scheme of some extant PT research actually serves to detract from the viability of the theory. Lest it sound too much like the PT school of thought should be adopted in its entirety, I will point out that these types of tables regularly portray stable and balanced opponents as peaceful. Thus, BOP proponents could rightly claim that war rarely occurs when capabilities are balanced. That is, the “equal, no overtaking” and “no war” cells have far more entries than do the “equal, no overtaking” and “war” cells. These “equal, no overtaking” entries should be composed of B&G cases, T&H cases when the power difference remains too small to be detected, and nontransitions. It would be useful to separate them and search for the unique types of conflict behavior associated with each. It appears, then, that the distinctions drawn here are both more useful and less likely to bias investigations of the power-conflict relationship. A

closer look at the conflict behavior that occurs in the three types of transitions is now in order.

The Conflict Hypotheses

For each empirical power transition, I searched for MID data during the transition and for five years prior to and five years following the transition. Data were available for at least a portion of that time frame for 10 of the 18 B&G deflections, for five of the eight T&H cases, and for nine of the 14 traditional D&G transitions. Not only does the MID data set

provide conflict information for just 60 percent of the cases, but the

130 The Power-Conflict Story coverage of those cases is often sparse. There are 141 years leading up to, during, and following the 10 B&G deflections, but a MID occurs in only 34 of them, or less than 25 percent of the time. Only two of the 10,

or 20 percent, have at least five data points. Three of the 10, or 30 percent, have only one. There are 109 years near and during T&H cases,

and MIDs occur in 19 of them, or just 17.4 percent of the time. More than 70 percent of the T&H cases (five out of seven) have fewer than five data points. Little more than 18 percent, or 20 out of 109, of the years in the D&G transition setting overlap with the occurrence of MIDs. The Soviet-American case in 1972 is the only D&G case with five or more data points. Militarized disputes coincide with only one year in

four of these single cases, or slightly more than 44 percent of the time. MIDs occurring simultaneously with power transitions are sparse simply because militarized disputes are generally not common events. They represent the most extreme forms of conflict possible in an interstate rivalry, not the entire spectrum of conflictual actions. Because the data are limited in scope and coverage, I present a very simple data analysis, steering clear of high-powered inferential statistics that would be meaningless in this context. This analysis begins by looking at the first two hypotheses that concern conflict behavior for each of the three types of transitions. The B3, T3, and D3 deductions are together examined in the section on testing peaks, and B4 is then considered in the section on comparing B&G and T&H conflict patterns. I also follow up with a case study of each type of transition, using looser criteria for the time period analyzed and an assessment of dyadic conflict behavior. By way of summary, I consider the general hypotheses concerning conflict behavior. Bull and Gnat Conflict Patterns

Recall that the B&G story involves a weaker opponent (gnat) who directs a fair amount of conflict at the stronger opponent (bull) but never seriously threatens the stronger’s position. The hypothesis relating the opponents’ conflict behaviors during this deflection-like transition was:

BI: The challenger (gnat) is consistently more conflictual than the dominant nation (bull).

To compare the conflict levels of the two states in these 10 rivalries, I look at a variety of information: whether or not the gnat’s yearly sum of

hostility levels for militarized conflict directed toward the bull was greater than the bull’s yearly sum of behavior toward the gnat, whether

Verifying the Story 131 the sum of the gnat’s hostility levels for militarized conflict was greater than the bull’s for the entire time period, whether the frequency of the enat’s actions was higher than the bull’s, and whether or not the challenging gnat is consistently considered to be the initiator. Before summing the MID levels, I subtracted one from the score reported in the original data set. As was discussed earlier, this establishes a more meaningful zero point and also closely parallels Azar’s continuous COPDAB scale. The initiator’s identity is most meaningful when there are only a few acts and the rivals are not both listed as originators for one or more MIDs. For example, when the only dyadic conflict recorded is a single war and one rival was clearly the initiator, this can be used to represent the most

ageressive side. In other scenarios, such as those in which there are numerous events and both sides are listed as an originator for one or more MIDs, it is difficult to trace the origin of the conflict dynamics to one particular party. The results are reported in table 11. TABLE 11. Comparing Conflict Levels for Bulls and Gnats Conflict... > Conflict...

Each Most Number

Deflection Cases Year Years Total of Acts As Initiator@

1896-98 no no no no no

United Kingdom and Russia,

China and Russia, 1904-06 yes yes yes yes no, mixed China and Russia, 1911-14 no, yes yes yes no, mixed all but one France and Austria-Hungary, no no, no, no, yes Russia and United Kingdom, no yes, no, yes no, mixed

1915 1 year = = = 1917 2years > = 1 year < 1 year =

Russia and Germany, 1907-18 _ yes yes yes yes no, mixed Germany and France, 1925-32 yes yes yes yes no, mixed

United States and Germany, no, yes yes yes no, mixed

1938-41 all but one

Japan and France, 1939 no no no no no, mixed

France and Italy, 1941-45 no no, no, yes no, mixed 1 year > =

1 year < 2 years = aA “yes” entry means that the gnat is consistently listed as the originator. A “no” entry means that the bull is consistently listed as the initiator. A “no, mixed” entry means that both are simultaneously listed as the originator for one or more MIDs.

132 The Power-Conflict Story All but two of the 10 cases (the United Kingdom and Russia from 1896 to 1898) and Japan and France in 1939, show support for the Bl hypothesis. Several cases provide very strong support: China and Russia from 1904 to 1906, China and Russia from 1911 to 1914, Russia and Germany from 1907 to 1918, Germany and France from 1925 to 1932, and the United States and Germany from 1938 to 1941. The remaining cases

support the hypothesis but perhaps not quite as clearly. The France and Austria-Hungary case in 1915 supports B1, but only when the conflict initiator is considered. Because there is one dispute, the onset of war in 1914, the rivals are equally conflictual in all other categories. At first, the

Russia and U.K. case in 1917 might appear to yield mixed results. The United Kingdom (the challenging gnat) is not more conflictual in every year, but this does characterize two of the four years during which milita-

rized disputes overlap with the transition period. In one of the other years, Russia (the dominant bull) is more conflictual, and in the remaining year the two were equally conflictual. And, although the total conflict directed by each during the entire transition time is equal, the United Kingdom did instigate a higher number of acts. Overall then, the United Kingdom is slightly more conflictual than Russia. The same can be said of the France and Italy example from 1941 to 1945. There the two appear equally conflictual on all dimensions except when the number of acts category is considered. Then Italy (the gnat) emerges as more conflictual. While data limitations make conclusions somewhat tentative, there does appear to be fairly good support for the B1 hypothesis. A second hypothesis concerning the bull’s deflection of the gnat was derived from the power-conflict model: B2: The difference between the challenger’s (gnat’s) conflict level and the dominant nation’s (bull’s) conflict level diminishes over time.

This hypothesis is tested by subtracting the yearly sum of the bull’s conflict levels from that of the gnat and then plotting these differences over time. I simply looked for a negative slope coefficient when the difference in conflict levels was regressed onto time. Years in which there was a negative difference, or when the bull was more conflictual, were not included. I did not consider a goodness of fit measure because there are so few data points per case and the model does not predict a linear decline but a logistic one (see, e.g., fig. 25). The purpose of this analysis is simply to discover whether or not there is a decrease, not how well the data reflect a decrease with a particular functional form. The

analysis is done for all eight of the deflection cases that support B1, except the France and Austria-Hungary case in 1915, for which there is a single data point. ‘Table 12 reports the findings.

Verifying the Story 133 Conflict difference 80

60

20

—Q.2 —0). 1 0.1 0.2

Time

Fig. 25. Logistic decline in B&G conflict difference

Two cases, China and Russia from 1904 to 1906 and Russia and Germany from 1907 to 1918, do not demonstrate support for the B2 hypothesis. Still, the general picture from the other five cases is one of a declining difference. This supports the broader conclusion that dyadic conflict levels become coupled over time. A richer conflict history that includes nonmilitary forms of conflict would provide more conclusive findings. Tortoise and Hare Conflict Patterns

The moral of the tortoise and hare fable is that plodding wins the race. By fighting fiercely in the early stages, the hare surpasses the tortoise. But this strategy exhausts the hare, who ultimately loses. The powerconflict model predicts that in this scenario: TABLE 12. Patterns in Conflict Differences over Time: Bulls and Gnats

Deflection Cases over Time

Slope in Conflict Difference

China and Russia, 1904—06 positive China and Russia, 1911-14 negative Russia and United Kingdom, 1917 negative Russia and Germany, 1907-18 positive

Germany and France, 1925-32 negative

United States and Germany, 1938-41 negative

France and Italy, 1941-45 negative

134 The Power-Conflict Story T1: The hare is consistently more conflictual than the tortoise.

Empirical evidence concerning T1 resembles that for B1; the vast majority of the cases support the hypothesis. ‘The cases are summarized in table 13. The Franco-American example from the 1860s and the Sino-Soviet case from 1929 to 1941 solidly portray the hare as more conflictual. The first two cases, involving Germany as the hare during the World War II

era, show a similar picture of the hare, though only when the initiator category is considered. In the British-German example, Germany originates conflict by displaying force in 1938. Then in 1939 it starts its war with the United Kingdom. By 1941, the United Kingdom retaliates by initiating a use of force. Germany is therefore clearly the aggressor. In Germany’s competition with China, only Germany appears as an initiator. The Soviet-German case during World War II, however, is inconsistent with the expectations in T1. The Soviet Union, or tortoise, is generally more conflictual than Germany, the hare. In the case study section, I will consider a richer events record for the Soviet-Germany rivalry. That record will actually yield support for T1 (and T2). Table 13. Comparing Conflict Levels for Tortoises and Hares Conflict,,,. > Conflict, rise

Each Most Number

Double Cases Year Years Total of Acts As Initiator?

1861-66 yes yes yes yes yes 1929-41 3 years >

France and United States,

China and Soviet Union, no yes, yes yes no, mixed 2 years < 2 years =

United Kingdom and Germany, no no, no, no, yes

1933-44 1 year > = = 1 year < 1 year =

China and Germany, 1937-44 no no, no, no, yes

1 year > = = 1 year
2 years < 1 year =

4A “yes” entry means that the hare is consistently listed as the originator. A “no” entry means that

the tortoise is consistently listed as the initiator. A “no, mixed” entry means that both are simultaneously listed as the originator for one or more MIDs.

Verifying the Story 135 As in the other categories, the model’s second prediction for T&H scenarios concerned coupling: T2: The difference between the challenger’s (hare’s) conflict level and the dominant nation’s (tortoise’s) conflict level diminishes over time.

Table 14 records the results of regressing the difference onto time for each of the empirical T&H cases that supported T1, with one exception.

China and Germany from 1937 to 1945 cannot be assessed because elimination of the one year involving a negative difference leaves only one year remaining. One case, France and the United States from 1861 to 1866, demonstrates a positive slope because the hare becomes increasingly conflictual while the tortoise makes no aggressive moves at all. Negative slopes appear in the two remaining cases, so the weight of the evidence favors T2. David and Goliath Conflict Patterns

These transitions represent the biblical story used by Organski and Kugler (1980) to characterize their traditional version of power dynam-

ics. David, though only a boy, challenges and defeats the Philistine giant. The power-conflict model’s conclusions concerning D&G transitions sharply contrast Organski and Kugler’s prediction that David will be consistently more conflictual than Goliath. Many of the D&G simulations showed the declining giant as the more conflictual opponent. The deduction was:

DI: There should be a mix between cases in which David is more conflictual and cases in which Goliath is more conflictual.

Table 15 reports the relationship between the two rivals’ conflict levels.

In comparison with the challengers in the other two types of power transitions, the Davids are less likely to be the more conflictual rival. TABLE 14. Patterns in Conflict Differences over Time: Tortoises and Hares

Double Cases over Time

Slope in Conflict Difference

France and United States, 1861-66 positive China and Soviet Union, 1929-41 negative

United Kingdom and Germany, 1933-44 negative

136 The Power-Conflict Story

Goliath is clearly more conflictual in five of the nine cases: Prussia/ Germany and Austria-Hungary in 1861, Russia and the United Kingdom from 1896 to 1902, the United States and China in 1904, the USSR and the United States in 1972, and China and the United States in 1976. This means that in less than 45 percent (1.e., four out of nine) of the empirical D&G cases, David is the more aggressive, whereas the challenger (gnat or hare) is more conflictual in 80 percent of the B&G (eight out of 10)

and T&H (four out of five) cases. These results confirm the powerconflict model’s expectation that the traditional transitions should be accompanied by conflict behavior that sometimes portrays the challenger as more aggressive and sometimes portrays the dominant state as more aggressive. Coupling was the focus of the second D&G prediction: D2: The difference between the rivals’ conflict levels diminishes over time.

All the D&G cases except for the Soviet-American rivalry in the early 1970s have conflict values for two or fewer years. Looking for meaningTABLE 15. Comparing Conflict Levels for Davids and Goliaths Conflictp,yiq > Conflictg ian

Each Most Number

Single Cases Year Years Total of Acts As Initiator? Prussia/Germany and Austria- no no, no, no no, mixed Hungary, 1861 1 year = = Germany and France, 1885 yes yes yes yes no, mixed

United States and United yes yes yes yes no, mixed Kingdom, 1896—1900 Russia and United Kingdom, no no, no no no, mixed

1896-1902 1 year < 1 year >

United States and China, 1904 no no, no, no no, mixed 1 year = =

Japan and France, 1935 yes yes yes yes no, mixed Soviet Union and China, 1948 yes yes yes no no, mixed Soviet Union and United States, no no, no no no, mixed

1972 5 years < 2 years > 1 year =

China and United States, 1976 no no, no no no 2 years < 4A “yes” entry means that David is consistently listed as the initiator. A “no” entry means that Goliath is consistently listed as the initiator. A “no, mixed” entry means that both are simultaneously listed as the originator for one or more MIDs.

Verifying the Story 137 ful patterns over time is only possible for the U.S. and USSR case. There should be a trend toward a decreasing difference in conflict levels. Because either of the two rivals might ultimately be the more conflictual, I pay no attention to which side has the advantage. The regression of the conflict difference on time shows a negative relationship, so there does appear to be a coupling effect. Bull and Gnat versus Tortoise and Hare

One deduction, B4, calls for a comparison between the conflict patterns in B&G cases and those in T&H cases: B4: The difference between the challenger and dominant state’s conflict values seems large compared to that for the T&H type of transition. The empirical analysis was conducted by calculating the average positive yearly difference between challenger and dominant state conflict levels

for each type of transition and then performing a difference of means test. The mean difference in adjusted MID values for the tortoise and hare is approximately 2.11, while the mean difference for the bull and enat is about 2.78. Correcting for the small number of cases, the statistic t ~ 1.196, which is not large enough to imply that the means are statistically distinct at the .10 level. Although the B&G difference is larger, on

average, than the T&H difference, the size of the inequality is small enough that it occurs randomly more than 10 percent of the time.!’ While the relationship is in the predicted direction, it is not statistically significant. Peaks in Conflict Behavior

Limitations of the MID data set are especially problematic for testing predictions concerning when to expect conflict maxima. Hypotheses B3 and T3 specify two peaks in conflict behavior — one for the challenger alone prior to the transition and one for both the challenger and the dominant state following the transition. Hypothesis D3 specifies only a peak for both following the transition. Often, however, disputes do not occur across the entire period preceding, during, and following the transition. For instance, the British-German T&H transition runs from 1933 until 1944, so the test period would include years from 1928 to 1949. Determining whether or not Germany reaches a local conflict maximum prior to the first transition in 1933 is impossible because no militarized disputes occur until 1938. And establishing whether or not

138 The Power-Conflict Story both Germany and the United Kingdom reach local maxima following the second transition is hard because the last dispute in the test period occurs in 1941. Of the 21 empirical transitions for which there are some

data, just five have MID data both before and after the transition point. That means that MID data fully cover only 12.5 percent (five out of 40) of the empirical transitions.!* These five cases average fewer than five data points per competitor, so fitting a polynomial and looking for extremes is problematic. In addition, the challenger’s first peak 1s not characterized by the standard mathematical interpretation of a local maximum, where dc/dt = 0 and d’c/dt* < 0. Instead, it is distinguished by two other notable features. First, the challenger’s extraordinarily high conflict level coincides with its birth (where its power level is very small and positive), and then it declines sharply in the period immediately afterward. Second, during this time the dominant state actually exhibits high levels of cooperation, and this abates, too, until it turns into conflict. This somewhat unusual segment of the story occurs very early in the simulation, often two or more times further from the transition than the second peak 1s. Although this nascent period is beyond the scope of this investigation, its theme of nations springing forth during conflictual eras is historically familiar. !° Pakistan was born out of rivalry with India, and some would argue that the United States was reluctant to enter the world arena until events of World War I forced it to play its role as a major power. Because data concerning the cooperative behavior of the dominant state are currently unavailable for most of the major power era, and because the earliest stages of the challenger’s life span are beyond the time frame being investigated, an analysis of the first peak predicted by B3 and T3 is not performed in this book.

The later peak represents the simultaneous attainment of classic local maxima by both rivals:

Hesy Myx _ 9 (5.6) dt at and

22 Dery Pox — 9 (5.7) dt? dt?

However, because the average number of data points per rival is so small, fitting a polynomial and finding these maxima is troublesome.

Verifying the Story 139 Instead, I reason that jointly high conflict can be represented by years in which both rivals use military force against one another. This happened in 14 of the 21 transition periods. In only three of the 14, concurrent use of force uniquely manifests itself before the power transition. In one case, it occurs both before and after, in five it happens while the transition is under way, in four it appears after the transition has ended, and in one more case it happens both during and after the transition. Distributing these cases across time yields table 16, which indicates that the most

common time for the joint use of force is actually in the midst of a transition. Joint use after the transition is still more likely than before, however. Eliminating the single case that did not support the hypothesis concerning which rival is most conflictual produces table 17, which indicates that the joint use of force is most likely following the transition, as predicted by the power-conflict model. Although the pattern is not especially strong, it 1s evident.

Case Studies

Case studies are designed to give us richer information about international events. They are closer and more careful examinations of the historical context, the rivals’ relationship, and the interpretation of events. The context and chronology of one rivalry of each type is compared to the formal model’s characterizations. TABLE 16. Timing of Second Peak in Joint Use of Force Cases

Joint Use of Force before Joint Use of Force during Joint Use of Force after

the Power Transition the Power Transition the Power Transition

31 full full cases full cases halfcases case 15 half case 24 half cases N=14 TABLE 17. Timing of Second Peak in Joint Force Cases that Support T1, B1, or D1

Joint Use of Force before Joint Use of Force during Joint Use of Force after

the Power Transition the Power Transition the Power Transition

31 full full cases full cases halfcases case 14 half case 24 half cases N= 13

140 The Power-Conflict Story A B&G Transition: Germany and France, 1925-32

France’s national power begins to approach that of Germany and actually achieves parity twice, once in the 1924 to 1926 time period and once in the 1931 to 1932 time period (see figs. 23 and 24). I count this as a single B&G deflection because in the time between the two deflections the disparity remains quite small, never becoming larger than .019 percent. In 1933, France’s power falls off and Germany reasserts its dom1nance. This is very similar to the deflection simulation presented in figure 14. What some may find surprising is that, despite the German defeat in World War I and the return of the Alsace-Lorraine region to France, France was unable to surpass Germany’s strength in this interwar period. Its failure was not for lack of effort.

[T]his apprehension that France was not secure... drove Paris into seeking to prevent a revival of German power by all means possible: insisting on the full payment of reparations; maintaining

its own large and costly armed forces; endeavoring to turn the League of Nations into an organization dedicated to preserving the status quo; and resisting all suggestions that Germany be admitted to “arm up” to France’s level. (Kennedy 1987: 289)

France’s insistence on a large postwar army demonstrates its effort to bolster national power in order to compete with Germany. Hence, this analysis depicts France’s power rising from its wartime decline to meet Germany’s path. France’s other policy efforts can be interpreted as conflictual behavior clearly aimed at decreasing German power. This scenario is consistent with the MID data in supporting B1. One might wonder why France’s efforts failed to result in a success-

ful overtaking of Germany. As Kennedy points out, “Germany still possessed a much larger population than France and an iron-and-steel capacity which was around three times as big” (1987: 288). In other words, the failure came about because France did not possess enough power to put sufficient weight behind its efforts. How might this be interpreted in terms of the model’s power growth equations, (3.3) and (3.4)? Changing the subscripts to reflect the specific dyad in question, the equations become: d

~ = A¢PGo — BoP rfrc (5.8)

Cg, Germany can be said to be less sensitive to France’s conflictual efforts than France is to Germany’s efforts. One might alternatively say that France absorbs more of the conflict directed against it than does its German opponent. This provides a further explanation for why France did not make a strong comeback during the interwar period. Consistent with the model’s B2 coupling expectations, the disparity between the conflict behaviors of France and Germany diminishes. But expanding the time horizon forward brings results contrary to expectations; German conflict surpasses French conflict by World War II. From 1935 onward, German conflict dominates the rivalry. Germany’s sudden leap into highly conflictual activity vis-a-vis France is perplexing in light of the model. The bull needs to direct much less conflict than the gnat does in order to avoid being overtaken. In other words, Germany did

142 The Power-Conflict Story not need to expend resources on such high conflictual levels. By 1932, it had already used less conflictual strategies to easily deflect French attempts to surpass it. While Germany did indeed lessen its conflict level

during the deflection period itself, it failed to maintain that low level afterward. One way to make sense out of this unusual German behavior would

be to refer to the coming to power of one of history’s most unusual leaders in 1933. Idiosyncratic stories about a unique individual such as Hitler, however, are precisely what this book seeks to avoid. Instead, I refer to new elements in the power distribution following the deflection.

Not only did Germany avoid a transition, but it quickly dominated French strength. What motivated it to become so conflictual when its advantage was so great? Although it was not until 1939 that France and the United Kingdom formalized their military alliance against Germany, the seeds of this union were planted in the mid-1930s. Despite disagreement over the proper policies toward Germany following World War I, the United Kingdom and France found each other to be necessary allies given American isolationism and the reluctance of the United Kingdom to align with Communist Russia against Germany’s Fascist regime (Kennedy 1987: 317-18). Germany must have sensed that British appeasement policies were weakening in favor of joining French opposition to the pervasive German threat. As Germany witnessed early signs of an

opposing union, it reacted to a much larger force than French power alone. In essence, the alliance prolonged the transition by increasing the gap between Germany and its opponent. The Franco-British alliance against Germany therefore represents both additional conflict and increased capabilities on the part of France. It thus helps reconcile increased German aggression with Bl and B2, and it supports the second part of B3, which predicts that Germany and France would both be most conflictual following the transition. In terms

of the formal model, this historical case hints at the need to move beyond the simple dyadic scenario, taking into account allies and how they augment national power (as Kim [1989] argues) and thereby fuel conflictual actions by the opponent. It further suggests that the model needs to take into account the fact that alliances matter at least a few years before they are actually formalized. A T&H Transition: Germany and the soviet Union, 1938-44

Following World War I, German growth was phenomenal, allowing Hit-

ler to overtake his Soviet rival in 1939. Germany retained its power

Verifying the Story 143 advantage during World War II until its demise in 1945, at which point it was stripped of its major power status through defeat and occupation. Germany fits the stereotype of a hare, with its power level rapidly rising, peaking at p, ~ .219 in 1940, and then plummeting to practical nonexistence. The Soviet tortoise,”! on the other hand, increases its power level slowly in an almost cumbersome fashion. Nonetheless, it achieves the advantage in the end because it has not exhausted itself with an overly enthusiastic growth spurt in the beginning (see fig. 24). With regard to the conclusion stated in T1, the hare, Germany, should generally be more conflictual than the tortoise, the Soviet Union. In the MID data summarized in table 13, however, the two rivals are about equally conflictual. An undeniable and unvarying dominance of German conflict is not apparent unless we refer to the more detailed historical record. Consider, for instance, the 1938 crisis over Czechoslovakia. Germany used force short of war against the USSR in May of that

year. Although the Soviets responded with a display of force, their general strategy up to and perhaps even during the Munich crisis was one of appeasement. Ulam, for example, points out that out of fear of German aggression “Schulenburg from Moscow kept reassuring Berlin that the Soviet Union would not intervene with her army on behalf of Czechoslovakia” (1974: 256). Written accounts of events from 1938 to 1941 portray Germany as the clear aggressor and the USSR as a fretful and almost submissive player (Ulam 1974; Stoessinger 1993). During the 1939 to 1944 period, Stalin was not only cooperative with Hitler, he was

especially compliant, even when he became suspicious of what would later manifest itself as the Barbarossa attack in 1941.

On April 13, 1941,... Stalin threw his arm around the German military attaché, Colonel Krebs, at the Moscow railroad station, and said to him: “We will remain friends with you — through thick and thin!” Two days later, Stalin unconditionally accepted Hitler’s

proposals for the settlement of the border between the two countries. In addition, Stalin continued to supply grain, petroleum, manganese ore, and rubber to blockaded Germany. . . . In a supreme gesture of conciliation, Stalin closed the embassies of countries conquered by Hitler and proceeded to recognize the pro-Nazi government of Iraq. When, . . . under the weight of irrefutable facts coming in from

Soviet sources, Stalin began to realize in early 1941 that Hitler might indeed attack, this realization was too terrifying for him to bear. He then began to engage in appeasing Hitler. As Admiral N. G. Kuznetsov of the Soviet navy recounts: Stalin acted so as to

144 The Power-Conflict Story avoid giving Hitler the slightest pretext for an attack in order not to provoke a war. As a result, when German planes photographed our bases, we were told: Hold your fire! When German air intelligence agents were caught over Soviet fortifications and made to land at our airports, the order was: release them at once! (Stoessinger 1993: 45-46) This historic account is consistent with the power-conflict model’s portrayal of the tortoise (e.g., Stalin) as relatively cooperative early on and the hare as the most conflictual, ultimately providing support for T1. Closer observations also support T2’s conclusion that the rivals’ conflict levels became coupled over time. The Soviet Union’s behavior toward Germany after World War I, while typically less conflictual than German behavior toward it, did evolve from cooperation to conflict. In the Rapallo Treaty of 1922, the two parties agreed to extensive diplomatic and commercial relations, and they reaffirmed that agreement in the 1933 Treaty of Berlin. Nonetheless, Stalin’s suspicions of Hitler’s aggressive intentions grew (Ulam 1974: 149-67). Soviet appeasement in 1938 and its willingness to sign the Non Aggression Pact in 1939 signaled not purely cooperative efforts but a movement away from the friendly relationship that had characterized the period following Rapallo and toward a fearful and cautious policy vis-a-vis Germany. For Germany, the pact indicated a temporary willingness to relax hostilities. So, while the Soviet Union was slowly forging its mettle, Germany was on respite. This lull is consistent with the trough in the hare’s conflict curve from t ~ —1.5 tot ~ —.2 in the third graph of figure 15. Simultaneously, the tortoise’s behavior rises from cooperation to conflict in that theoretical simulation. Ultimately, in 1941, both fought openly in response to the German initiation of war, corresponding to the peak in both rivals’ conflict levels at t ~ .3 in figure 15 and

the second part of deduction T3. A D&G Transition: Russia and the United Kingdom, 1896-1902

The late nineteenth and early twentieth centuries can generally be characterized by the decline of the British hegemonic period. The relative strength of Russia (and the United States and Germany as well), on the other hand, grew rapidly (see fig. 23). The Anglo-Russian rivalry was particularly strong given Russia’s potential encroachment on British interests in the Far East, especially in India and China (Middleton 1947: 70-108). Russian power first surpassed British in 1896, but London’s

Verifying the Story 145 struggle with the Boers inflated its CINC score so that it reestablished a slight advantage (never more than .015 percent) from 1899 until 1902. Henceforth, the Russian power indicator dominated, with only a temporary lapse brought on by the Bolshevik Revolution. (The extra spurt in erowth from 1904 to 1905 is the result of the Russo-Japanese War. ) Since it specifies mixed behavior, D1 does not make a prediction that is testable with only one case. Therefore, only D2 and D3’s expectations are relevant here. First, I address D2’s coupling prediction. Do the Russian and British conflict levels converge over time, as did the simulated conflict trajectories in the Organski-style transition of figure 15? Early in the transition period, Russian conflict 1s not commensurate with British hostilities. Russia’s comparatively less aggressive behavior can be seen in its seemingly unusual decision, in 1895, to not respond to a display of force by its British opponent. Perhaps Russia felt it would be pointless to fight when it was about to overtake the United Kingdom anyway. Why bother spending resources in order to match the display when preponderance would be achieved without the expenditure? Hensel and Diehl’s (1994) empirical analysis of Latin American interstate disputes further suggests that nonresponses are much more likely when the initiator does not actually use force but merely threatens or displays its use, as is the case here. The expected matching (see Leng 1983), or Richardsonian action-reaction behavior, 1s established in 1898 when Russia occupied the Liaotung Peninsula in Manchuria. This action was immediately followed by the establishment of British rule in Wei-Hai-We1, directly across from the bay from the Russians (Middleton 1947: 79). Gochman and Maoz (1984) code these mobilizations of forces as actions taken against China. This certainly makes sense in light of the ensuing Boxer Rebellion, although the two events could just as easily be interpreted in the context of two rivals competing for domination of resources and markets in the Far East (Middleton 1947: 75-82). Overall, then, Russian and British conflict behavior surrounding their power transition from 1896 to 1902 does indeed appear to be coupled. Based on the model’s predictions, and as specified in the second

half of D3, this D&G transition should produce the highest levels of dyadic conflict after the challenger has surpassed the dominant nation. In the British-Russian case, the aggregate MID conflict scores for both rivals peak in 1895, just before Russia first achieves superiority. I have already pointed out that additional, uncoded aggressive behavior was exhibited by both rivals in 1898. This can also be said about the years 1902, when the Anglo-Japanese alliance was formed, and 1905, when that alliance was renewed for an additional 10 years. London’s decision

146 The Power-Conflict Story to promise support for Japan in the event of an attack not only signaled a vast change in British policy toward Russia, but it represented the absolutely worst point in dyadic relations. [The talks] concluded in a formal Anglo-Japanese treaty of alliance, whose publication caused a considerable diplomatic sensation. At

last, it appeared, England had decided to abandon her policy of avoiding foreign commitments and to follow the example of the other great Powers of Europe. The treaty was, of course, directed against Russia... . It appeared reasonable to believe that, whatever the consequences of the Anglo-Japanese treaty might be, the rift between Britain and the Dual Alliance [between Russia and France] had become complete. (Middleton 1947: 82)

One might reasonably conclude that this alliance and its reaffirmation following the Russo-Japanese War in 1905 were aggressive expressions of London’s rivalry with St. Petersburg. Accordingly, the peak in dyadic conflict behavior may well have been following the observed power transition that was completed by 1902. The Anglo-Japanese alliance could alternatively be interpreted not as British conflict directed toward Russia but as an attempt to restore the balance of power that had been disturbed by the Russian growth in national power. It seems sensible for the United Kingdom, upon the realization that Russia had surpassed it, to pursue alternative ways to check the opponent’s power. An alliance with Japan, whose interests were also threatened by expanding Russian influence in the Far East, would serve to counter this looming giant more effectively than the British could alone. In other words, the alliance was precipitated by the power transition, and once it was established the distribution of power was restored to approximate parity and Anglo-Russian conflict concomitantly lessened. In order for this second interpretation of the 1902 and 1905 events to map nicely onto my theoretical deductions, I had to draw once again on the role of alliance partners. Conclusion: Stories Well Told

Seeing the task laid out in this chapter to its end has been a long and somewhat painstaking process. After carefully designing indicators of national power and dyadic conflict, I examined their values for the major powers from 1816 to 1984. The same three types of power transitions that were predicted by Gl

Verifying the Story 147 were found in the historical record and were independently confirmed by Wayman’s (1996) work. Moreover, I demonstrated that the power transi-

tion typology developed in this book has certain advantages over that used by other researchers. In particular, the T&H and B&G stories are added to the traditional D&G story and the special conflictual consequences of each type of transition are more readily distinguished from one another. The next step was to determine whether or not those special consequences could actually be observed. Overall, they were. The powerconflict model performed well. The general hypothesis G2 stated that the ultimate winner would not necessarily be a fierce fighter. This ex-

pectation was met with support for Bl, T1, and Dl. The MID data indicated that the unsuccessful challengers (gnats and hares) were more

conflictual than the winning dominant states (bulls and tortoises). France’s aggressive interwar policies toward Germany and Germany’s thinly veiled hostility toward the Soviet Union preceding World War II add historically important confirmation to Bl and T1. The MID data also showed that successful challengers (Davids) were more conflictual only 44 percent of the time. Deduction G3 predicted a coupling effect for two rivals’ conflict levels and was well supported. In the investigations of B2, T2, and D2, the rivals’ MID levels clearly converged almost 73 percent of the time. The B&G case study of France and Germany supported B2; but when the time horizon was lengthened, this support weakened. The support was reestablished by considering the Franco-British alliance against Ger-

many and its role in the power-conflict relationship. Additionally, Soviet-German and British-Russian coupling was evident in the T&H and D&G case studies. The last general hypothesis, G4, told us to expect the highest degree of conflict following a transition. The subsidiary hypotheses, B3, T3, and D3, all specified two peaks in conflict behavior — one for the challenger

alone preceding the transition and one for both rivals following the transition. Because it indicated that both rivals would be conflictual, it was the latter type of peak that G4 emphasized as the most conflictual. Given the broader scope and additional difficulties associated with testing for the first type of peak, only the second type was considered here. The period following a transition was found to be slightly more prone to exhibit the joint use of force than were the pretransition and transition phases. This gives only weak to moderate support for G4. Given that the precision of the power indicator is not sufficient to make us confident of the exact timing of transitions, it 1s not surprising that the evidence for this hypothesis is less clear-cut.

148 The Power-Conflict Story

Hypothesis B4 garnered some support, although it 1s somewhat tenuous. The difference between the B&G conflict levels was indeed larger than the difference between those of the T&H but not by enough to claim that the relationship is statistically significant. Given the strong support for the first three general hypotheses and their subsidiary hypotheses, and considering that the direction of the relationships specified in G4 and B4 held up empirically, one can conclude that the primary predictions of the power-conflict model were well supported. Some additional findings from the case studies warrant attention. First, the usefulness of nonmilitarized forms of conflict spoke to the need

for a more comprehensive international events data set, which I am currently collecting. Next the model helped to explain France’s failure to overtake Germany in the B&G case study. In essence, France was not only weaker, but it was also more susceptible to conflict than was Germany. In other words, France’s conflict cost rate, B,, was greater than Germany’s, 8,. States in such a position might be well advised to consider methods of insulating themselves from the effects of a rival’s policies and

behavior. Last, despite my hesitance to explicitly include them in the formal model, alliances did seem to play a role in two of the case studies. It should nonetheless be noted that even without considering alliances the systematic investigations of the MID data did support the model’s predictions. If alliances are to be incorporated into the model later, the case studies give us some ideas about how to go about doing this. In both the France-Germany B&G case and the Russia—United Kingdom D&G case, one rival’s alliance with a third party had the effect of both altering the power distribution and demonstrating hostile intentions toward the second rival. Additionally, there may well be an interactive relationship

between power transitions and alliances, as evidenced by the AngloJapanese alliance against Russia, which was necessitated by Russia surpassing the United Kingdom. Alliances might also matter before they are officially formed. German conflict rose in response to early signs of an impending Franco-British alliance against it. Undoubtedly, measuring potential alliances would instead require the introduction of a different type of variable, one that captures how two nations draw closer together (or move further apart) over time. Just as the level of conflict changes over time, so do the bonds tying nations together (Lee, Muncaster, and Zinnes 1994; Lai 1999). In sum, the empirical work completed here has yielded some strong support for the power-conflict model, underscored the need for more diverse data on international conflict (and cooperation), and suggested some possible ways to incorporate alliances into the model itself.

CHAPTER 6 Epilogue

Rather than being concerned with the eventual fate of a main character, this chapter is devoted to the fate of the story itself. Here I must make

an objective judgment concerning the ability of the power-conflict model to help us understand the world of international relations. This is somewhat difficult because, as Lave and March caution, researchers run

the risk of “falling in love” with their models (1975: 60), much like fiction writers, who may find themselves particularly fond of a character in one of their novels. Toward the end of minimizing the impact of this

amorous tendency on the evaluation process, I propose two general criteria to be kept in mind throughout this chapter. Decisions to discard, retain, or modify the power-conflict model should be made by consider-

ing whether or not there is empirical support for the power-conflict model’s predictions and how well the model performs compared to the BOP and PT alternatives. These two assessment criteria are used to judge how well the powerconflict model answers important substantive questions —some of them generated by the BOP-PT debate and some revealed through analysis of the model. The questions of interest addressed here include: (1) what features characterize power transitions, (2) how do nations “win” a transition, (3) which rival is more responsible for aggression during a transition, and (4) when should we expect the highest conflict levels? These are precisely the issues addressed by the four general deductions listed in

table 14. As I evaluate the model, I also discuss the implications for whether or not, and how, it should be revised. Next, I review the potential incorporation of alliances and issues into the power-conflict model and the usefulness of the various characteristics of the extant model. Last, I speculate on the possibility of other analyses of the model. Three Stories, One Plot

The power-conflict model produced three versions of the evolution of a dyadic rivalry that undergoes a power transition: the bull and gnat 149

150 The Power-Conflict Story scenario, the tortoise and hare fable, and the classic tale of David and Goliath. Scholars of PT are concerned only with transitions that fit the prototype in figure 16, namely, D&G transitions. During this Organsk1style transition, Goliath’s power declines while David’s strength rises. Proponents of BOP do not directly address the mechanics of a power transition. They do, however, describe a situation consistent with the middle and last portions of a T&H case, with the approximate parity period being relatively peaceful. Confirmation of the existence of several types of power transitions was reported not only in table 10 and appendix B but in separate work by Wayman (1996). The significance of the three stories’ contribution includes the model’s ability to explain successful challenges as well as unsuccessful ones and its prediction of unique conflict behavior associated with each version of the transition story.

The PT theory, though largely motivated by the surges in Germany’s power that accompanied its precipitating roles in the world wars, does not explain why this determined challenger repeatedly failed. What distinguishes successful from unsuccessful challengers? Only the powerconflict model answers that question. Hitler’s ambitious and encompassing plans remind us of the hare’s overconfidence and exhausting pace in the T&H transition. In addition, Germany was probably too sensitive to its rivals’ conflict, that is, its conflict cost rate was too high. Factors that made f, high might have included Germany’s overextension from fight-

ing on many fronts and the vulnerability of its nationalist pride. As a general lesson, successful challengers are more evenly paced and less sensitive to their opponents’ conflictual efforts aimed at weakening their national capabilities. Each type of transition also demonstrated specialized conflict trajec-

tories leading up to, during, and following a transition. Most notably, Davids stood out as challengers who were not always the more aggressive (D1), contrary to Organski and Kugler’s prediction that they would be.

Hares and gnats, on the other hand, were consistently expected to be more conflictual rivals (Tl and B1). Additionally, the model predicted that the difference in the rivals’ conflict levels for T&H transitions would be smaller than for B&G transitions (B4). Empirical tests and case studies provided fairly good support for these conflict deductions.

The Meek Shall Inherit the Earth: Prescriptions for Winning a Transition

Proponents of BOP like Morgenthau would recommend that a statesman constantly strive to keep national power at a level at least as high as

Epilogue 151 that of the opponent. Conflict is one balancing mechanism available to both rivals, the potential loser as well as the potential winner. This idea was incorporated into the “cost of conflict” terms of the power equations ({3.3] and [3.4]). Scholars of PT would alternatively recommend that a statesman pursue national interests by maintaining a large, preferably overwhelming, advantage over the opponent. The method for doing so is found in the nation’s internal tendency for growth. This was integrated

into the formal model as the “natural growth” terms of the power equations. Aside from being general recommendations for how nations ought to pursue their interests vis-a-vis their rivals, neither gives advice specific to winning a power transition. Scholars of BOP probably do not address transitions because they see them as rare events, and transition

scholars do not recommend certain strategies for dealing with them because they see transitions as inevitable results of natural patterns in the growth and decay of national power. By integrating two explanations that do not have clear recommendations on this issue, I was able to generate three policy-relevant pieces of

advice. All can be subsumed under deduction G2’s statement that the “ultimate winner is not necessarily a fierce fighter.” The first stems from

the lesson of the hare. Nations that pursue rapid rates of growth can easily become exhausted. Hares gain an early lead but eventually lose the competition to the more evenly paced tortoise. Therefore, faster is not always better. A nation must distinguish rates of growth that allow it to compete from those that become burdensome.

The second piece of advice is derived from the B&G and T&H cases. According to B1 and T1, the bull and tortoise are the less conflictual rivals. Solid verification for these two hypotheses was presented in tables 11 and 13. Because high levels of conflictual behavior are unnecessary in order to maintain the advantage during a potential transition, it

makes good sense for the bull and tortoise to practice some restraint. Conflict, the reader may recall, comes at a price. It has a negative effect on power growth. Davids benefit from the last piece of advice in this section. These

young warriors do not automatically need to expend their power resources in order to successfully challenge giants. Deduction D1 tells us that Davids and Goliaths take turns being the more conflictual contestant. Davids, contrary to PT expectations, do not always exhibit more hostility. Even the traditional story found in 1 Sam. 17 portrays a less

ageressive David. The Philistines, not the Israelites, were the first to mobilize their army. It was also Goliath whose verbal provocations persisted for 40 days and who was better armed. David’s part was to throw one little, though amazingly successful, stone. In terms of the existing

152 The Power-Conflict Story

power distribution, one could say that David did indeed rise up and challenge Goliath. But in terms of the conflictual behavior, Goliath was the aggressor. Davids should resist being lured into unnecessary policies since they can win without them.

In sum, transition winners are often meeker than losers. Triumphant bulls and tortoises set a more even pace and are less impetuous than the opposing gnats and hares. They are also less belligerent. Victorious Davids are sometimes, though not always, less belligerent than Goliaths. This means that winners are not meek only in the classic PT cases in which Davids are more combative, a scenario that arose in less than 45 percent of the D&G cases examined in table 15. Culpability

Who is responsible for the unhappy consequences of international conflict? When two rivals expend extraordinary resources preparing for and waging war on one another, who is to blame? If the primary culprit can be identified with some regularity, we might be able to prevent the next offender’s actions or at least mitigate the effects. Throughout this book, the answer to the question of who is guilty has been that both rivals are. Proposed solutions to dyadic conflict must take that into account. Both the BOP and PT viewpoints identify one of the two competing nations as the initiating aggressor. For balance scholars, it is the stronger

nation, its power suddenly left unchecked, that will initiate war. For transition scholars, it is the challenger that will attack the previously dominant nation in order to claim its stake as the new hegemon. In contrast, deduction G2 predicts that over time two rivals’ conflict levels become coupled. As the rivalry progresses, pinpointing the protagonist is less tractable and of less importance. While the hybrid model allows us to say which nation is more conflictual at any given time during a simulation, this does not mean that we can always point to a single culprit. If the rivalry has endured over a long period, both are under suspicion. Under these circumstances, the action-reaction dynamics become en-

trenched and coupling more pronounced. Not only was evidence of coupled conflict found in the great majority of the cases examined in the

analyses of the specific deductions B2, T2, and D2 (eight out of 11 transition cases), but supplementary support is produced by Hensel and Diehl, who argue that it “takes two to tango” (1994). A likely culprit is the action-reaction dynamic, and this is where solutions should focus. Peacemakers might try to weaken the enabled rates of reaction, y,

and y,, that appear in the base conflict equations ([3.7] and [3.8]) by

Epilogue 153 making rivals less willing or able to use their power resources for conflictual purposes. Domestic and international penalties (or incentives) could be invoked. These might include costs to social programs, negative public reactions, erosion of government support, trade sanctions, suspension of diplomatic relations, and withholding military or financial assistance. Another strategy for weakening the y values would be to prolong the amount of time it takes for the rivals to respond to each other, perhaps by adding mediation as an intervening step. Because the action-reaction component of (3.7) and (3.8) includes interaction with the power variables, it should be noted that third parties that choose to intercede by adding to one rival’s power base only run the risk of escalating conflict.

In chapter 5, I suggested that the coupling finding might mean the model ought to be modified to reflect a single conflict variable instead of

two. Although coupled conflict was both predicted and rather consistently seen in the transition cases, I hesitate to conclude that c,, and c,, should be combined into a single variable. Why the reluctance? The simulations show conflict levels converging as time progresses, so that

they are not always similar in magnitude. Although they were not tested, the first components of the B3 and T3 peak predictions refer to a disparity between two rival nations’ conflict trajectories prior to power transitions. At the earliest stages of rivalry leading up to transition, this disparity is characterized by an asymmetric behavior pattern in which one rival cooperates while the other is hostile. Maintaining c,, and c,, as distinct variables leaves open the possibility of using the power-conflict model to understand how coupling evolves. Timing

Just as the BOP and PT explanations expect a particular nation to initiate war, they also predict that a certain power distribution will be associated with this initiation. The BOP explanation proposes that inequality is the most war-prone condition, while the PT explanation presents equality as the most dangerous situation in this respect. A synthesis of the specialized expectations for conflict behavior gave rise to some

general conclusions regarding conflict. Among them was G4, which states that “rival nations are most conflictual immediately following a power transition.” Among all the power-conflict hypotheses, G4 gained the least support. Although the case studies offered support for this hypothesis — unquestionably in the T&H case and when alliance effects were taken

154 The Power-Conflict Story

into account in the B&G and D&G cases—the more rigorous MID analysis gave only fragile support. Fortunately, there is still room for optimism. The nonmilitarized events data currently being collected promise a more detailed and aggregate picture of conflict that will in turn facilitate the location of classic local maxima in the empirical conflict trajectories. Finding these peaks will provide a better test of G4. Organski and Kugler’s empirical finding that wars occur after the transition (1980: 59-60)! further corroborates the posttransition conflict peak prediction. Given the potential for more meaningful tests, separate confirming evidence, and the solid support demonstrated for the model’s other hypotheses, G4 should be retained as a viable hypothesis. Because all three transition types exhibit the highest degree of conflict after the transition (the second component of B3, T3, and D3) and

because evidence verified this conclusion, one can conclude that PT theory’s prediction that transitions are conflict prone is valid. And because the T&H’s permanent transition is preceded by the relatively peaceful period during which the hare is resting while the tortoise is catching up one can conclude that the BOP theory’s prediction that balances are peaceful is also valid. The power-conflict model successfully integrates these seemingly opposed viewpoints by distinguishing among a negligible power differ-

ence (approximate parity), one that is noticeable yet not staggering (decisive advantage), and one that is extreme (overwhelming advantage). Consider, for instance, the BOP prediction of relatively peaceful parity. Such a situation characterizes those T&H simulations in which the |p, — Py| trajectory does not go beyond d*, a decisive advantage, between the two transitions.’ As is the case in figure 15, this in-between phase is associated with a trough in conflictual activity. At the same time, consider BOP scholars’ conclusion that when the d* threshold is crossed high levels of conflict should ensue. This setting is consistent with what happens after the second transition in all T&H simulations and with the posttransition phase of all B&G and D&G simulations, as in figures 14 and 16. In all cases, dyadic conflict does not peak until after the power difference is larger than a*. Next consider the transition scholars’ expectation that when one

nation’s power level overtakes that of another, high conflict levels should occur. This scenario, too, 1s parallel to that occurring when the PD. — Py trajectory in a D&G transition passes in succession from Goliath’s overwhelming advantage to Goliath’s decisive advantage, approximate parity, David’s decisive advantage, and then possibly David’s overwhelming advantage. The D&G transition simulation is different from the classic Organski and Kugler power transition in that it

Epilogue 155 specifies a high amount of conflict after the transition has been completed and only once the newly dominant nation has achieved a decisive advantage, d*. This is simultaneously consistent with the PT argument that transitions are conflictual periods and the BOP argument that once the inequality is no longer tolerable high levels of conflict Occur.

Last, consider the belief held by transitionists that gross inequality is a pacifying condition. Such was indeed the case for the more powerful nations that had an overwhelming advantage, d, at the beginning of all three representative simulations in figures 14, 15, and 16. Consequently, the first part of deductions B3 and T3 specify that only the challenging enat and hare are highly conflictual prior to the transition. The bull and tortoise, on the other hand, are actually cooperative. Hence, the power-conflict model 1s consistent with some aspects of

both explanations. In chapter 1, I indicated my intention was to integrate these two seemingly opposed explanations. Building the formal model as a hybrid of those competing explanations has allowed me to do

just that. Alliances

A striking feature of the historical case studies was the frequent role of alliances in helping to confirm the model’s deductions. Given that I reasoned at length in chapter 2 that the inclusion of alliances would be premature, how might I make sense of their regular appearance in chapter 5’s empirical investigation? Several authors (e.g., Ferris 1973; Siverson and Tennefoss 1984; Kim 1989) whose work was discussed in chapter 2 investigate the effect of alliance partners on distributions of power. Allies, they reason, augment a nation’s power and thereby alter the otherwise purely dyadic power distribution between that nation and its opponent. Motivated by the BOP-PT debate, these authors then ask what type of power distributions (equal or unequal) are more war prone when the contribution of alliances to those distributions are considered. The reader may recall that Kim’s answer to that question was that equal distributions are more likely to experience war. I chose not to include alliances in the formal model, however, because the theoretical justification for doing so was weak. On the one hand, Kim’s findings supported the PT school’s prediction that wars occur under the condition of equality and the BOP school’s supposition that alliances matter. On the other hand, they contradicted the PT scholars’ belief that alliances are not important and the

156 The Power-Conflict Story BOP group’s expectation that wars usually happen during inequality. Not only were Kim’s findings inconsistent with the both the PT and BOP viewpoints, but they were not given an alternative theoretical grounding either. The empirical analyses of chapter 5, however, suggested some ways

in which the role of alliances could be incorporated into the powerconflict story, thus providing it with a theoretical base. More importantly, chapter 5 revealed that the power augmentation role of alliances

is not the only important one. A nation tends to interpret an alliance between its primary rival and a third party as a form of conflict directed against it. When France and the United Kingdom joined forces against Germany prior to World War II, Germany reasonably interpreted this as a threat. Similarly, when the Anglo-Japanese alliance was formed in 1902, and later renewed in 1905, the purpose was to counter Russian

strength more effectively than either the United Kingdom or Japan could alone; and Russia clearly perceived both as hostile actions directed at it on the part of the other two nations. How might these two parts of the alliance story be incorporated into the existing model? The augmentation portion is problematic because the method by which two allies’ power levels, say p, and p., should be aggregated must be considered. It is unlikely that a simple additive function of the two levels, such as p, + p,, would be theoretically suffi-

cient. One possible alternative would be to rely on some weighting scheme that reflects how closely aligned those two nations are. This technique may be facilitated once the formalization of the second part of the story is more complete. Consideration of the way in which alliances might be perceived as

conflict directed at common enemies of alliance members led to the conclusion that a variable reflecting how closely two nations are aligned would be a useful addition to the formal model. In the case analysis for deduction D3, it was implied that the origins of the British-French alliance against Germany could be found in the mid-1930s. Extrapolating

from that case, I then suggested the introduction of a variable that “captures how two nations draw closer together (or move further apart) over time.” A very similar variable is the subject of work by Lai (1999) and Lee, Muncaster, and Zinnes (1994). Moreover, these authors are interested in the triadic nature of relationships. Drawing on Heider’s (1946, 1958) and Harary’s (1961, 1977) work, they reason that when a nation such as Germany perceives that its “enemy” has a “friend,” the friend of the enemy is deemed an enemy as well. Citing Liska (1962), Lai notes the relevance of relationships to conflict studies: “Alignment is a relationship between two countries. It occurs in the context of inter-

Epilogue 157 national conflict, and it is always formed against a third international actor” (1999: 3). I may be able to introduce a relationship variable in order to measure not only how aligned (or “friendly”) two nations are but how that association, as it grows stronger, 1s perceived as an increasing threat (or “enmity”) by their common enemy. As such a bond grows

stronger, the weight of the contribution an ally makes to a nation’s power level might also increase. The D&G case study further suggested that two rivals’ power relations might be one factor leading to changes in bonds with third parties:

db, . Py s+): (6.1) att (Pe

where b,, is nation X’s bond with a third party, K. More specific formalization of the relationships among relationship bonds, power, and conflict is an important task for future research.

Issues

Issues have been a recurring theme throughout this book. As a result of the discussion in chapter 2, it was decided that their contribution could be realized by incorporating them into the model’s conflict cost parameters, B, and B,. Higher B values represented how serious the issues at stake in a rivalry were. A sensitivity analysis of B, in chapter 4 yielded some interesting discoveries. Most clearly, a nation always prefers lower values of £; that is, it always wants to be less reactive to the issues driving the rivalry. This insulates a nation from the deleterious effects of its opponent’s conflictual actions. In particular, if B, is below the critical value given in equation (4.2), nation X is guaranteed to win the transition. How can a

nation control its B value and avoid power losses due to the cost of conflict? In the B&G case study of chapter 5, I examined the ratio of the conflict cost rates for Germany and France, 8,/B,. This ratio indicated

which nation was more susceptible to external conflict. Germany, the dominant nation, was much less sensitive to French conflict than France was to German conflict. Nations would be well advised to insulate themselves from uneven effects of conflict. Insulation might come in the form

of lessened or more equitable trade, reduced reliance on colonial or alliance ties, or construction of elaborate defense systems. Each of these strategies carries with it additional risks or costs. This conclusion gives

158 The Power-Conflict Story formal support to Keohane and Nye’s proposition that states that cannot respond to offset the deleterious effects of other states’ behaviors without deep costs are not just sensitive but vulnerable (1972, 1989). It also bolsters the claim that asymmetric interdependence, where one state is less dependent than the other, is a source of power (Hirschman 1945; Keohane and Nye 1989). Furthermore, the analysis of the B parameters revealed the intricate tradeoffs among power, conflict, and issues. For example, if states find it problematic to reduce their vulnerability to the issues at stake, increasing their aggressive behavior toward rivals is an option (though a costly one), lending reinforcement to Organski and Kugler’s hypothesis that the most dissatisfied states are the most likely to initiate war with the

hegemon (1980). Likewise, because they ultimately lead to joint cooperation, large initial power values serve to decrease long-run conflict values, ultimately saving states from extensive competition costs during protracted rivalries. The analysis of the B parameters produced novel information and demonstrated the great potential for understanding the role of issues through analyzing the model’s parameters. Useful Characteristics of the Power-Conflict Model

Insofar as the power-conflict model is able to produce better predictions

concerning transition types than is either the BOP or PT explanation alone, it is clear that the features of the model that allow it to do so should be retained. It is nearly impossible to isolate the individual characteristics of the model that are responsible for particular predictions, but one can nonetheless speak in general terms. A Formal Hybrid

Above all else, the model is a formal integration of two putatively competing sets of explanatory ideas. The BOP and PT theories were interwoven by considering both external and internal forces driving national power growth and decay. Equations (3.3) and (3.4) include the intrinsic progress of development as well as the deleterious effects of conflict with an opponent on a nation’s power. They were also integrated by designing the conditional conflict equations, in which increases or decreases in

the amount of conflict nation X directs toward nation Y are partially controlled by terms that change as a function of the quantity (p, — p,). Because the synthesis of those ideas was done mathematically, it was possible to tease out various deductions in a logically consistent

Epilogue 159 manner. One consequence was that the evolution of national power is not represented by an S-curve similar to that shown in figure 2. By implication, power transitions cannot be characterized by two intersecting Scurves, as in figure 3, but rather by three very different pairs of trajecto-

ries, aS shown in the first graphs of figures 14, 15, and 16. As was explained earlier, dyadic conflict is also much more diverse and complex than it is traditionally thought to be. In sum, fresh information relevant

to the BOP-PT debate resulted. Examples of this include the newly found B&G and T&H cases, the discovered importance of the ratio of the conflict cost rates, and the detection of Goliaths that are the more ageressive rivals. Dynamics

Dynamics provided the model with a complex richness that seems lacking in the extant literature. Dynamic properties allowed the model to generate pictures of how both power and conflict levels change over time. This is more satisfying than previous analyses, which were largely static. The conditional conflict behavior and the integration of the two dominant and completing explanations allowed me to characterize the power-conflict relationship as a complex and interdependent one. Not only do nations’ power levels naturally tend to grow, but they are hindered by external conflict weighted by the opponent’s strength. Similarly, conflict behavior has a perpetual action-reaction component in addition to being influenced by the current power distribution. The stories told by Morgenthau (1985), Organski (1958), and Organski and Kugler (1980) offer a rich view of national power growth,

dyadic conflict, and their interactions. If we ignore these stories, the world of rivalries seems static and boring. If we instead capture them in a general, dynamic model, the world of rivalries becomes the interesting place it is. Generalized Conflict

The generalized conflict variable allowed me to extrapolate from arguments about when war is most likely to statements about how a nation’s

conflict level might increase or decrease in response to other factors (e.g., the opponent’s level of conflict, that nation’s own power resources, and the distribution of power between the nation and the opponent). It also permitted me to observe how two nations’ conflictual actions might become increasingly coupled en route to or following a power transition. The empirical analysis fortunately confirmed the usefulness of examining

160 The Power-Conflict Story conflict at a variety of levels. There were persistent references to types of behavior other than those coded in Gochman and Maoz’s MID data set. Most obviously, lower levels of conflict were mentioned several times. In the analysis of the tortoise and hare case, I twice referred to the historical record concerning Soviet-German relations from 1939 to 1945, once to demonstrate that German conflictual behavior toward the USSR was

greater than Soviet conflict toward Germany and once to show that Germany’s conflict level had a small peak prior to the first transition. In addition, I twice referred to the formation of an alliance with a third party as a hostile, albeit nonmilitarized, action taken by one opponent against the other. Another type of behavior left uncoded by Gochman and Maoz but used in the case studies includes the higher levels of armed conflict that occur after a war has been initiated. Speculation on such behavior was made in the investigation of conflict peaks (G4). A third type of behavior excluded by the MID data set is cooperation . When examining the evidence concerning T3, which expected two conflict peaks separated by a trough in which one nation actually appeared cooperative, I was able to refer to the Soviet-German Non Aggression Pact. Together with other recorded appeasement efforts on the part of Stalin, the pact seemed to

indicate dyadic behavior that was asymmetric, with Germany acting rather aggressively while the Soviet Union practiced unanswered attempts at cooperation. Two important conclusions can be reached at this point. First, there is a Serious need to expand the coverage of our events data sets. Militarized conflict alone does not represent the subtle variety of actions that states can take vis-a-vis one another. Data sets that do cover cooperative

and conflictual events at many levels, such as COPDAB, need to be temporally expanded so that meaningful analyses can be done for longer time periods. The need to gather information about the events that take place during wars is also suggested. With few exceptions (e.g., Massoud 1995), little work has been done in this area. Second, the repeated reference to nonmilitarized behavior indicates that two modeling decisions were good ones. Some of the work reviewed

in chapter 2 (e.g., Garnham 1976b; Geller 1996; Bueno de Mesquita, Morrow, and Zorick 1997), as well as Organski and Kugler’s disappointment over not being able to use a finer-grained indicator of conflict in the empirical analysis of The War Ledger, suggested that the dichotomous war variable needed to be restructured as a more general conflict variable. My own predisposition on the issue being quite similar, I devised the conflict variables in chapter 3, c,, and c,,, to cover a variety of behavior. Not only did I choose to let them become infinitely positive,

Epilogue 161 but I also allowed them to become infinitely negative. The decision to

allow c,, and c,, to take on negative values during the simulation of chapter 4 was based on the idea that such values might represent cooperation. Because the case studies of chapter 5 revealed the importance of conflictual behavior that was both more and less intense than war, as well as cooperative behavior, those two decisions appear to have been justified. Sequels to the Power-Conflict Story: A Future Research Agenda

Having discussed the performance of the model and its potential modifi-

cations, I now briefly sketch out a plan for how research in this area should proceed. The first sequel to this book will involve a systematic analysis of new events data that reflect a more diverse range of state behaviors. The goal will be to provide a more rigorous analysis of the power-conflict model’s deductions, especially those concerning when to expect peaks in conflict (B3, T3, and D3). The fruits produced by the three fine-grained historical cases in chapter 5 will not be lost in this endeavor, since they

underscored the need for additional data. Later, I would like to reinitiate the storytelling process, this time focusing on how interstate bonds factor into the original dyadic story. Afterward, it would be inter-

esting to use the original or revised model (or both) to answer new questions such as: (1) when or under what conditions does conflict decrease, (2) how might a nation preserve some minimal level of national power and simultaneously avoid high levels of conflict with its opponent, and (3) what types of situations give rise to power transitions? Now that the reader has gotten a glimpse of how the power-conflict

story might generate answers to new questions, I feel compelled to conclude this book with the traditional closing for all good stories. The End

Blank Page

Appendixes

Blank Page

APPENDIX A Simulation Results

The following is a list of randomly generated initial conditions. Each produced either a B&G deflection, a T&H double transition, ora D&G single transition and is given as a set of three values. The first represents

both the p,(0) and p,(0) values, which are set as equal. The second is c,,(0), and the third is c,,(0). Initial Conditions Producing Deflections

{48, 50, 10} {26, 47, 10} {10, 50, 10} Initial Conditions Producing Single Transitions

{17, 17, 43} {41, 20, 32} 146, 4, 26} {31, 16, 37} {31, 26, 19} {29, 38, 35} {40, 14, SO} {29, 11, 33} {40, 2, 10} {12, 28, 28} {41, 45, 36} {24, 42, 46}

{22, 24, 18} {7, 32, 40} 13, 1, 49} {29, 2, 48} {10, 26, 20} {2, 42, SO} {37, 34, 42} 144, 1, 15} {7, 21, 46}

{1, 2, 13} {7, 9, 30} {18, 25, 49} {20, 4, 8} 19, 34, 30} {3, 1, 47} {12, 22, 14} {43, 16, 14} 18, 34, 46}

{10, 1, 5} 19, 38, 37} {48, 36, 44} {10, 10, 50} {20, 45, 4} Initial Conditions Producing Tortoise and Hare Transitions

{20, 34, 13} {29, 47, 35} {28, 50, 23} {47, 40, 13} {30, 27, 15} {42, 8, 3}

{19, 20, 6} {44, 16, 8} {45, 35, 13} {47, 25, 17} {26, 30, 19} {6, 31, 13}

{5, 43, 18} {14, 15, 8} {17, 45, 23} {10, 10, 6} 165

APPENDIX B

Power Transitions Among the Major Powers

Bull and Gnat Cases

Dates “Winner” “Loser”

Approximate

1821 United Kingdom Russia

1847 Austria-Hungary United States 1870-78 United States Prussia-Germany

1888-94 Russia Germany 1896 United States Germany 1896-98 United Kingdom Russia 1896-1904 United States Russia

1903-6 Austria-Hungary Japan

1904-6 1911-14China China Russia Russia

1915 France Austria-Hungary 1917 Russia United Kingdom 1907-18 Russia Germany 1925-32 Germany France 1938-41 United States Germany 1939 United States Soviet Union

1939 Japan France 1941-45 France Italy

166

Tortoise and Hare Cases

Dates “Winner” “Loser”

Approximate

1860-67 1861-66 Russia France United United States States 1867-81 United States Russia

1896-18 Italy Japan 1929-41 China Soviet Union 1933-45 United Kingdom Germany

1937-45 China Germany 1938-45 Soviet Union Germany David and Goliath Cases

Dates “Winner” “Loser”

Approximate

1834 United States Prussia-Germany 1859 United States Austria-Hungary 1861 Prussia-Germany Austria-Hungary

1879 United States France

1885 Germany France 1896-1900 United States United Kingdom 1902 Russia States UnitedChina Kingdom 1904 United 1921 JapanFrance Italy 1935 Japan 1948 Soviet Union China 1960 JapanUnion United Kingdom 1972 Soviet United States

1976 China United States

167

Blank Page

Notes

Chapter 1 1. Throughout this book, I refer to the sixth edition of Morgenthau’s Politics among Nations (1985), a revision of the fifth edition done by Kenneth Thomp-

son following Morgenthau’s death. I chose to cite Morgenthau alone, as the portions of the book to which I refer have not been altered by Thompson, whose substantial revisions are primarily found in sections on “human rights, détente, and the nuclear problem.” (Thompson’s preface to Morgenthau 1985: vi). Read-

ers interested in the historical placement of Morgenthau’s balance of power ideas might note that the first edition was published in 1948. 2. Emphasis added.

Chapter 2 1. Emphasis added. 2. Yet one of the greatest difficulties with balance of power theory is that its central concept has too many meanings. Haas (1953), for example, found at least eight different common uses of the phrase “balance of power.” Zinnes (1967) similarly reveals numerous understandings of the concept. The balance of power explanation of war, it seems, uses just one of those meanings, namely, “stability and peace in a concert of power” (Doughterty and Pfalzgraff 1981: 24). 3. This is because Organski (1958) sees economic, political, and social devel-

opment as closely related, but not equivalent to, power in the international system since power is a relative concept. Ultimately, however, only the growth of power is operationalized, so the discrepancy becomes irrelevant. 4. Less powerful nations are only of interest later in their development. 5. Examples of these contingency tables can be found later in this chapter in tables 2.1 and 2.2. 6. Emphasis in original. 7. In Kim and Morrow’s (1992) data analysis, the time period for assessing shifts in dyadic power relations is ten years. Powell (1996) does not perform an empirical test, but the historical case (Britain and Germany in the 1930s) on which he draws to develop the formal model clearly demonstrates an elapsed period of no more than ten years. 8. The utility of winning and losing is held constant (e.g., the winner gets all 169

170 Notes to Pages 23-42 of the disputed territory and the loser gets none minus the cost of fighting, regardless of the point in time in the game).

9. Wolfson, Puri, and Martelli (1992) develop a dyadic, rational choice model that accounts for changes in both the power distribution (i.e., the degree of “hegemony”) and conflict (1.e., its inverse, the “probability of peace”) (124). These variables are related to one another via the role each plays in both rivals’ utility functions. While Wolfson, Puri, and Martelli’s model provides good evidence of the usefulness of dynamic rational choice models for capturing dyadic interaction, its incorporation of economic constraints makes for a much different substantive focus. 10. Organski and Kugler (1980: 17-19) also discuss the “collective security” model, but the empirical analysis pays little attention to this set of arguments. 11. Because a lack of data prevented the first test period from beginning in 1850, it begins in 1860 and runs to 1880.

12. Ferris actually reports a stronger to weaker disparity ratio of 1.45. Its inverse is .6897, or about .7. 13. Emphasis added. 14. Tables 1 and 2 in Lemke and Werner (1996) give the “static power parity” results for rival dyads. Subsequent tables give the results for “dynamic power parity,” but these are more difficult to assess both because they give information about movement and because that information is only about movement from inequality to equality and leaves out information about movement in the opposite direction. 15. Similarly, deterrence scholars have proposed that our expectations concerning when war will be initiated and by whom depends on the disparity between the costs nation 7 can impose on nation / and the costs j can impose on i. These expectations are different under different conditions (Kugler and Zagare 1992).

16. There are 469 disputes and 238 randomly chosen cases in which no dispute occurred. 17. See chapter 5 for a more elaborate discussion of the COPDAB data set. 18. The attributes of the COPDAB and WEIS data set are more thoroughly discussed in chapter 5. 19. Another way to put this is that knowing something about conflict will tell you nothing about cooperation levels and knowing something about cooperation will tell you nothing about conflict levels. 20. Friendship-enmity would be a more accurate label for this dimension. 21. Oddly enough, Organski and Kugler argue at length that it is the changes in alliances commitments that lead to high threat perceptions. “[W]e should stress again that it is not simply the degree of tightness or discreteness in the alliance system but the shifts in these arrangements that are critical” (1980: 39). One wonders how protagonists in an environment of change would perceive that alliances are permanent. 22. For one rival, each ally’s potential contribution is captured by multiplying its capability score by the probability that it will join that rival in a bilateral war

Notes to Pages 43-61 171 with the other rival. The sum of the products for all allies is added to the rival’s own capabilities score. The probability that a certain ally will join the rival is determined by considering how similar that ally’s policies are to the rival’s (Kim 1989, 1992, 1996; Kim and Morrow 1992). 23. The short term decision-making under uncertain conditions that characterizes alliance formation and its relationship with international conflict is already the subject of an impressive literature. Aside from research on the trade-offs between arms and alliances, which is mentioned above, there have also been investigations of the deterrent effect of alliances (Huth 1988; Huth and Russett 1984), how nations choose sides during war (Altfeld and Bueno de Mesquita

1979), and the conflictual behavior of allies toward one another (Bueno de Mesquita 1981).

24. Because the empirical investigations in chapter 5 consider only majormajor dyads, I only present Siverson and Tennefoss’ results for that type of case. 25. These issue categories are not necessarily exhaustive or mutually exclusive. 26. An interesting and important exception is McLaughlin and Prins’s study

(1999) addressing the importance of fisheries, maritime boundaries, and resources of the sea in disputes between democracies since 1946. 27. Similarly, Gartzke (1997) argues that the similarity of preferences is responsible for peaceful interactions among democracies. 28. Morrow (1986) and Morgan (1994) offer excellent discussions of how to model issue salience in a rational choice approach to international relations.

Chapter 3 1. There has been a slight effort to capture dynamics by including a test for whether growth rates are equal or unequal (e.g., Organski and Kugler 1980; Houweling and Siccama 1988). Nonetheless, this approach is still limited to a test of certain variable values at a single point in time. 2. Of course, another way to do this would be for a rival to increase its own

power level. Here, however, I am interested only in the forces that cause a nation’s power level to decay. 3. This is not to say that conflict is the only means by which nations act. It is simply the means that is important in this model. 4. In the empirical work of chapter 6, a specific value is associated with war.

5. The idea that conflictual behavior depletes national power is captured indirectly in the second term of the power equations. The depletion can be thought of as an expenditure made in order to respond to the opponent’s conflict. The response, an action in itself, is conflictual in nature and requires the expenditure of power reserves.

6. This is reminiscent of Most and Starr’s “pretheoretic” framework for understanding international politics, which is based on the organizing concepts of “opportunity” and “willingness.” (1989: 23-46) Opportunity is seen as the “possibilities that are available within any environment,” and willingness is seen

172 Notes to Pages 61-87 as the “choice (and process of choice) that is related to the selection of some behavioral option from a range of alternatives” (23). 7. Iintentionally leave this possibility out of my discussion because it essentially refers to the case in which the basic assumptions about what drives nations

no longer hold. I am concerned here only with the scenario in which these assumptions do hold. 8. Alternatively, we might be interested in d as a certain critical proportion between p, and p,. 9. This prevents the denominator from becoming zero. 10. Emphasis added. 11. This does not mean that the time scale appearing in the simulation trajectories of chapter 4 can be interpreted as years. 12. This is so because equation (3.2) cannot be solved by integrating each side since two variables other than p, (i.e., p, and c,,) appear on the right-hand side. Holding those variables constant makes little sense since their dynamic features are intrinsic to the model. 13. Later I will suggest that w might be empirically measured.

14. Again, at some later point in this research it would be interesting to investigate how the model behaves when the opposite assumption is made. If the prevention effect is stronger than the action-reaction effect, a different pattern might emerge.

Chapter 4 1. I also use the words predictions, implications, hypotheses, and deductions to refer to conclusions that can be reached by logically or mathematically analyzing a formal model.

2. The constants d* and d are arbitrarily set at 7 and 15, respectively. The choice of their particular values does not change the findings of this work as long

as d* < d. 3. This procedure was done using the Mathematica software developed by Wolfram (1991).

4. Note that starting at t = 0 and generating four subsequent values and three previous values, for example, yields the same results as beginning at t = —3 and generating seven subsequent values. 5. The second graph shows the system’s movement through the three conditions defined by the magnitude of the difference between p, and p,.

6. The reader may notice that the presentation style in these figures differs from that in figure 13. I have switched gears here in order to trace the behavior of the conflict trajectories that were absent in figure 13.

7. Note that p,(0) = p,(0) is a requirement for all initial conditions, as specified previously.

8. A system of differential equations can generate previous values of variables if —?t 1s substituted for ¢ in all equations.

Notes to Pages 91-116 173 9. Likewise, Richardson reasons that a “suitable name for negative preparedness for war seems to be ‘co-operation’” (1960a: 19). 10. Because p,(0) = p,(0) in all simulations, and because a, = a,, the a,p, and a,p, terms are subtracted from each side of the equation.

11. These are the same initial conditions used to generate the trajectory in figure 13. In other words, the trajectory in figure 13 contains the same information as do the trajectories in the first and second graphs of figure 15. 12. The confinement of the power relationship to approximate parity (condition 1) during the hare’s peak is not generalizable to all tortoise and hare simulations. In some, the power relationship passes d* during the peak, and in others it even passes d during the peak. Even when that happens, the time between the two transitions 1s relatively peaceful.

13. Organski and Kugler use the labels David and Goliath to refer to their predictions of winners and losers in wars of all types, not just those resulting from power transitions (1980: 64-103). 14. The system, of course, goes through lower condition 2, condition 1, and upper condition 2 before reaching upper condition 3. 15. But this does not occur in all T&H simulations. Chapter 5 1. Tucker (1997) has designed a computer program that will pair each member of the interstate system with every other member for every year from 1816 to 1994. This universe of analysis provides a useful null comparison group.

2. Morgenthau (1985) gives a great deal of emphasis to one particular element of power, namely, “capabilities.” This includes entities such as military capabilities, industrial productivity, and population, the standard Correlates of War indicators for national power. Organski and Kugler’s (1980) notion of “power resources” closely resembles that of capabilities. Both conceptualizations portray capabilities or resources as closely and directly related to power. 3. These figures are given in coal ton equivalents.

4. These six capability measures have fairly low levels of specificity. For example, total population taps into the notion of labor as a basic resource and

urban population captures one large sector of that resource, namely, the nonagrarian one. 5. This procedure is completed by using the interpolation function in Mathematica (Wolfram 1991). 6. An additional criticism of the French wartime data is that the installation of the Vichy government and the partial German occupation may also indicate the nonexistence of the French state. 7. Organski and Kugler (1980: 37-38) actually investigate the correlation between a nation’s system share of GNP and its CINC score. 8. For that matter, increasing military expenditures might signal a decline in

174 Notes to Pages 117-43 national power, rather than an increase, if expenditures constitute increasing usage of available resources. In this case, power resources have been “spent” and are unavailable for other purposes. Capturing this phenomenon would require that I use the ratio of military expenditures to GNP, GDP, or the government budget. Because the availability of economic data is sketchy, as mentioned earlier, I cannot use such a ratio as an indicator of power. 9. See figure 24, which is presented later in this chapter.

10. I also considered averaging over five years;t— 2,t—1,¢,t+1,andr+ 2. This would not work because one of the time periods in table 8, 1946-51, is only six years long. Because endpoints in the time period must be eliminated, there would only be data for the years 1948 and 1949.

11. Four T&H transitions and one B&G end in 1945, but all begin before the COPDAB starting date of 1943. An extension of the Correlates of War data set on shares of major power would likely demonstrate that the single SovietAmerican transition in 1972 is followed in the early 1990s by a second transition

in which the Soviet Union clearly loses. Because the two single transitions together are actually of the T&H sort, this transition is not completed before the COPDAB data set ends in 1978. This leaves only three D&G transitions in the 1943-78 time period: the Sino-Soviet transition in 1948, the BritishJapanese transition of 1960, and the Sino-American transition of 1976.

12. However, the researchers “do not report repeated exchanges within a given hostility level, nor are there de-escalatory actions in the data set” (Jones, Bremer, and Singer 1996: 172). 13. A “minimum of 100 battle fatalities or 1000 troops in active combat is required before a state is considered to be a participant in a war” (Gochman and Maoz 1984: 589).

14. Emphasis added. 15. Even the COPDAB Project, which measures the full range from cooperative to conflictual activity, produces a data set greatly biased toward the reporting of conflictual events (Azar 1975). Perhaps the greatest reason for this is the reliance on news sources for raw data. Here one finds more of a tendency to report that something has occurred than to report that all is well. Conflict makes a better story. 16. The National Science Foundation has provided me with funding for this collection effort (Grant SBR 98-06123). 17. Including the cases of equal levels of conflict results in an average T&H difference of 1.46, an average B&G difference of 1.79, and ¢-value of .58. 18. All of the cases for which there is sufficient peak information also support the relevant first hypothesis regarding which rival is more conflictual.

19. Expanding the time period to 10 or more years before and after the transition does little to improve coverage by the conflictual data. In fact, it decreases the percentage of rivalry-years in which MIDs occurred.

20. It is difficult to get a precise estimate for c,,(0) and c,(0) because the t = 0 point falls at 1927, a year for which there are no MID data. 21. My apologies to those who prefer a bear analogy.

Notes to Page 154 175 Chapter 6 1. This contradicts their theoretical prediction that wars should occur just before the power transition. 2. Recall from chapter 4 that not all tortoise and hare trajectories are confined in this manner.

Blank Page

References

Alcock, N., and A. Newcombe. 1970. “The Perception of National Power.” Journal of Conflict Resolution 14:335-43. Altfeld, M., and B. Bueno de Mesquita. 1979. “Choosing Sides in War.” [nternational Studies Quarterly 23:87-112. Axelrod, R. 1984. The Evolution of Cooperation. New York: Basic Books.

Azar, E. 1970. “Analysis of International Events.” Peace Research Reviews 4:1-113. Azar, E. 1975. “Ten Issues in Events Research.” In E. Azar and J. Ben-Dak, eds., Theory and Practice of Events Research, 1-17. New York: Gordon and Breach. Azar, E. 1980. “The Conflict and Peace Data Bank (COPDAB) Project.” Journal of Conflict Resolution 24:143-52. Azar, E., and T. Havener. 1976. “Discontinuities in the Symbolic Environment: A Problem in Scaling.” International Interactions 2:231—46.

Bennett, S. 1996. “Security, Bargaining, and the End of Interstate Rivalry.” International Studies Quarterly 40:157-84.

Boyce, W., and R. DiPrima. 1986. Elementary Differential Equations and Boundary Value Problems. 4th ed. New York: Wiley.

Bremer, S. 1992. “Dangerous Dyads: Interstate War, 1816-1965.” Journal of Conflict Resolution 36:309-41. Bremer, S. 1993. “Democracy and Militarized Interstate Conflict, 1816-1995.” International Interactions 18:231-49.

Browne, M. 1995. “Physicists Track Down an Elusive Atomic Particle.” New York Times, March 3, Al, A13. Bueno de Mesquita, B. 1975. “Measuring Systemic Polarity.” Journal of Conflict Resolution 19:187-216. Bueno de Mesquita, B. 1981. “Risk, Power Distributions, and the Likelihood of War.” International Studies Quarterly 25:541-68. Bueno de Mesquita, B. 1989. “The Contribution of Expected-Utility Theory to the Study of International Conflict.” In M. Midlarsky, ed., The Handbook of War Studies, 143-69. Boston: Unwin Hyman. Reprint, Ann Arbor: University of Michigan Press, 1993. Bueno de Mesquita, B., and D. Lalman. 1988. “Empirical Support for Systemic and Dyadic Explanations of International Conflict.” World Politics 41:1—20. Bueno de Mesquita, B., and D. Lalman. 1992. War and Reason: Domestic and International Imperatives. New Haven: Yale University Press. 177

178 References Bueno de Mesquita, B., J. Morrow, and E. Zorick. 1997. American Political Science Review 91:15-27.

Bull, H. 1969. “International Theory: The Case for a Classical Approach.” In K. Knorr and J. Rosenau, eds., Contending Approaches to International Politics, 20-38. Princeton: Princeton University Press. Carlson, L. 1995. “A Theory of Escalation and International Conflict.” Journal of Conflict Resolution 39:511-34. Carr, E. H. 1974. The Twenty Years’ Crisis, 1919-1939. New York: Harper and Row.

Claude, I. L. 1962. Power and International Relations. New York: Random House. Cline, R. 1980. World Power Trends and U.S. Foreign Policy for the 1980s. Boulder: Westview Press. Collins, R. 1986. Weberian Sociological Theory. New York: Cambridge University Press. Commission on Integrated Long-Term Strategy. 1988. Discriminate Deterrence. Washington, D.C.: Government Printing Office.

Coombs, C. H. 1953. “Theory and Methods of Social Measurement.” In L. Festinger and D. Katz, eds., Research Methods in the Behavioral Sciences, 471-535. New York: Dryden. de Soysa, I., J. Oneal, and Y. Park. 1997. “Testing Power-Transition Theory Using Alternative Measures of National Capabilities.” Journal of Conflict Resolution 41:509-28. Diehl, P. 1985. “Arms Races to War: Testing Some Empirical Linkages.” Sociological Quarterly 26:331-49. Diehl, P. 1992. “What Are They Fighting For? The Importance of Issues in International Conflict Research.” Journal of Peace Research 29:333-44. Diehl, P., and G. Goertz. 1994. “Linking Rivalries: Connecting International Conflict across Space.” Paper presented at the Annual Meeting of the International Studies Association, Washington, D.C., March 29—April 1. Dixon, W. 1993. “Democracy and the Management of International Conflict.” Journal of Conflict Resolution 37:42-68. Dixon, W. 1994. “Democracy and the Peaceful Settlement of International Conflict.” American Political Science Review 88:14-32.

Doran, C. 1971. The Politics of Assimilation: Hegemony and Its Aftermath. Baltimore: Johns Hopkins University Press. Doran, C. 1989a. “Systemic Disequilibrium, Foreign Policy Role, and the Power Cycle: Challenges for Research Design.” Journal of Conflict Resolution 33:371-401.

Doran, C. 1989b. “Power Cycle Theory of Systems Structure and Stability: Commonalities and Complementarities.” In M. Midlarsky, ed., The Handbook of War Studies, 83-110. Boston: Unwin Hyman.

Doran, C., and W. Parsons. 1980. “War and the Cycle of Relative Power.” American Political Science Review 74:947-65. Dougherty, J., and R. Pfaltzgraff Jr. 1981. Contending Theories of International Relations. New York: Harper and Row.

References 179 Edwards, C. H., and D. Penney. 1985. Elementary Differential Equations with Applications. Englewood Cliffs, N.J.: Prentice-Hall. Ferris, W. 1973. The Power Capabilities of Nation-States. Toronto: Lexington. Fuchs, W. 1965. Formeln zur Macht: Prognosen tiber Volker, Wirtschaft, Potentiale. Stuttgart: Deutsche Verlags-Anstalt. Galtung, J. 1964. “A Structural Theory of Aggression.” Journal of Peace Research 1:95-119. Garnham, D. 1976a. “Dyadic International War, 1816-1965: The Role of Power Parity and Geographical Proximity.” Western Political Quarterly, 29:231-42. Garnham, D. 1976b. “Power Parity and Lethal International Violence, 19691973.” Journal of Conflict Resolution 20:379-94. Gartzke, E. 1997. “The Logic of the Democratic Peace.” Ph.D. diss., University of Iowa. Geller, D. 1992. “Capability Concentration, Power Transition, and War.” /nternational Interactions 17:269-84. Geller, D. 1993. “Power Differentials and War in Rival Dyads.” International Studies Quarterly 37:173-93. Geller, D. 1996. “Relative Power, Rationality, and International Conflict.” In J. Kugler and D. Lemke, eds., Parity and War: Evaluations and Extensions of the War Ledger, 127-43. Ann Arbor: University of Michigan Press. George, A., and R. Smoke. 1974. Deterrence in American Foreign Policy. New York: Columbia University Press. German, F. Clifford. 1960. “A Tentative Evaluation of World Power.” Journal of Conflict Resolution 4:138-44.

Gochman, C. 1990. “Capability-Driven Disputes.” In C. Gochman and A. Sabrosky, eds., Prisoners of War? 141-59. Lexington, Mass.: Lexington.

Gochman, C., and R. Leng. 1983. “Realpolitik and the Road to War: An Analysis of Attributes and Behavior.” International Studies Quarterly 27: 97-120. Gochman, C., and Z. Maoz. 1984. “Miulitarized Interstate Disputes, 1816— 1976.” Journal of Conflict Resolution 28:585-616. Goertz, G., and P. Diehl. 1992a. “The Empirical Importance of Enduring Rivalries.” International Interactions 18:151-63. Goertz, G., and P. Diehl. 1992b. Territorial Changes and International Conflict. London: Routledge.

Goldstein, J. 1992. “A Conflict-Cooperation Scale for WEIS Events Data.” Journal of Conflict Resolution 36:369-85.

Goldstein, J. 1995. “Great-Power Cooperation under Conditions of Limited Reciprocity: From Empirical to Formal Analysis.” International Studies Quarterly 39:453-77. Haas, E. 1953. “The Balance of Power: Prescription, Concept, or Propaganda?” World Politics 5:442-77. Harary, F. 1961. “A Structural Analysis of the Situation in the Middle East in 1956.” Journal of Conflict Resolution 5:167-78. Harary, F 1977. “Graphing Conflict in International Relations.” Papers of the Peace Science Society (International) 27:1-9.

180 References Hart, J. 1976. “Three Approaches to the Measurement of Power in International Relations.” [International Organization 30:289-305. Heider, F. 1946. “Attitudes and Cognitive Organization.” Journal of Psychology 21:107-12. Heider, F. 1958. The Psychology of Interpersonal Relations. New York: Wiley.

Heiss, K., K. Knorr, and O. Morgenstern. 1973. Long Term Projections of Power: Political, Economic, and Military Forecasting. Cambridge, Mass.: Ballinger.

Hensel, P. 1993. “COW Militarized Interstate Dispute Data Set: Description of Variables.” University of Illinois at Champaign-Urbana. Mimeo. Hensel, P. 1996. “Charting a Course to Conflict: Territorial Issues and Militarized Interstate Disputes, 1816-1992.” Conflict Management and Peace Science 15, no. 1: 43-73. Hensel, P., and P. Diehl. 1994. “It Takes Two to Tango: Nonmilitarized Response in Interstate Disputes.” Journal of Conflict Resolution 38:479-506.

Hensel, P, and W. Reed. 1997. “Introducing the Issue Correlates of War (ICOW) Project: Territorial Claims and Militarized Interstate Disputes.” Paper presented at the Annual Meeting of the International Studies Association, Toronto, Canada, March 18-22. Hewitt, J., and J. Wilkenfeld. 1996. “Democracies in International Conflict.” International Interactions 22:123—42. Hirschman, A. 1945. National Power and the Structure of Foreign Trade. Berkeley: University of California Press. Hitch, C., and R. McKean. 1960. The Economics of Defense in the Nuclear Age. Cambridge, Mass: Harvard University Press. Houweling, H., and J. Siccama. 1988. “Power Transitions as a Cause of War.” Journal of Conflict Resolution 32:87-102. Houweling, H., and J. Siccama. 1991. “Power Transitions and Critical Points

as Predictors of Great Power War.” Journal of Conflict Resolution 35: 642-58.

Huth, P. 1988. “Extended Deterrence and the Outbreak of War.” American Political Science Review 82:423-43.

Huth, P 1996. Standing Your Ground: Territorial Disputes and International Conflict. Ann Arbor: University of Michigan Press.

Huth, P., and B. Russett. 1984. “What Makes Deterrence Work? Cases from 1900 to 1980.” World Politics 36:496—526.

Jones, D., S. Bremer, and J. D. Singer. 1996. “Militarized Interstate Disputes, 1816-1992: Rationale, Coding Rules, and Empirical Patterns.” Conflict Management and Peace Science 15:163-213. Kadera, K. 1990. “Capability Distribution and Major Power War: Choosing an Explanation.” Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago, April 5-7.

Kadera, K. 1995. “Power Growth and Decay and Conflict Behavior in Dyadic Rivalries: A Dynamic Model.” Ph.D. diss., University of Illinois at UrbanaChampaign.

References 181 Kadera, K., and G. Sorokin. 1998. “Measuring Capabilities in Regional and Global Rivalries.” University of Iowa. Manuscript. Kaplan, A. 1964. The Conduct of Inquiry: Methodology for Behavioral Science. San Francisco: Chandler. Kaplan, M. 1957. System and Process in International Politics. New York: Wiley. Kennedy, P. 1987. The Rise and Fall of the Great Powers. New York: Random House. Keohane, R., and J. Nye Jr. 1972. Transnational Relations and World Politics. Cambridge: Harvard University Press. Keohane, R., and J. Nye Jr. 1989. Power and Interdependence: World Politics in Transition. 2d ed. Boston: Little, Brown. Kim, W. 1989. “Power, Alliance, and Major Wars, 1816-1975.” Journal of Conflict Resolution 33:255-73. Kim, W. 1991. “Alliance Transitions and Great Power War.” American Journal of Political Science 35:833—50.

Kim, W. 1992. “Power Transitions and Great Power War from Westphalia to Waterloo.” World Politics 45:153-72.

Kim, W. 1996. “Power Parity, Alliance, and War from 1648 to 1975.” In J. Kugler and D. Lemke, eds., Parity and War: Evaluations and Extensions of the War Ledger, 93-105. Ann Arbor: University of Michigan Press. Kim, W., and J. Morrow. 1992. “When Do Power Shifts Lead to War?” American Journal of Political Science 36:896—922. Kissinger, H. 1979. The White House Years. Boston: Little, Brown. Kissinger, H. 1994. Diplomacy. New York: Simon and Schuster.

Kocs, S. 1995. “Territorial Disputes and Interstate War, 1945-1987.” Journal of Politics 57, no. 1: 159-75. Kreps, D. 1990. Game Theory and Economic Modelling. Oxford: Clarendon. Kugler, J., and A. F K. Organski. 1989. “The Power Transition: A Retrospective and Prospective Evaluation.” In M. Midlarsky, ed., The Handbook of War Studies, 171-94. Boston: Unwin Hyman.

Kugler, J., and F Zagare. 1992. “The Long-Term Stability of Deterrence.” International Interactions 15:255-78.

Kuznets, S. 1966. Modern Economic Growth. New Haven: Yale University Press. Lai, D. 1999. “Alignment, Structural Balance, and International Conflict in the

Middle East, 1948-1978.” Paper presented at the Annual Meeting of the International Studies Association, Washington, D.C., February 16-20. Larson, R., and R. Hostetler. 1979. Calculus with Analytic Geometry. 2d ed. Lexington, Mass.: Heath. Lave, C., and J. March. 1975. An Introduction to Models in the Social Sciences. New York: Harper and Row. Lee, S., R. Muncaster, and D. Zinnes. 1994. “The Friend of My Enemy is My Enemy: Modeling Triadic Internation Relationships.” Synthese 100:333-58. Lemke, D. 1995. “The Tyranny of Distance: Redefining Relevant Dyads.” [nternational Interactions 21:23-38.

182 References Lemke, D., and J. Kugler. 1996. “The Evolution of the Power Transition Perspective.” In J. Kugler and D. Lemke, eds., Parity and War: Evaluations and Extensions of the War Ledger, 3-33. Ann Arbor: University of Michigan Press.

Lemke, D., and S. Werner. 1996. “Power Parity, Commitment to Change, and War.” International Studies Quarterly 40:235—-60. Leng, R. 1983. “When Will They Ever Learn? Coercive Bargaining in Recurrent Crises.” Journal of Conflict Resolution 27:379-419. Levy, J. 1987. “Declining Power and the Preventive Motivation for War.” World Politics 40:82-107. Levy, J. 1989. “The Causes of War: A Review of Theories and Evidence.” In P. Tetlock et al., eds., Behavior, Society, and Nuclear War, 209-333. New York: Oxford University Press. Liska, G. 1962. Nations in Alliance: The Limits of Interdependence. Baltimore: Johns Hopkins University Press.

Majeski, S., and S. Fricks. 1995. “Conflict and Cooperation in International Relations.” Journal of Conflict Resolution 39:622—45. Mansbach, R., and J. Vasquez. 1981. In Search of Theory: A New Paradigm for Global Politics. New York: Cambridge University Press. Mansfield, E. 1992. “The Concentration of Capabilities and the Onset of War.” Journal of Conflict Resolution 36:3-24.

Maoz, Z., and B. Russett. 1992. “Alliance, Wealth, Contiguity, and Political Stability: Is the Lack of Conflict among Democracies a Statistical Artifact?” International Interactions 17:245-67. Maoz, Z., and B. Russett. 1993. “Normative and Structural Causes of Democratic Peace, 1946-1986.” American Political Science Review 87:624-—38. Massoud, T. 1995. “Anatomy of Wars: Military and Diplomatic Dimensions of 20 Interstate Wars.” Paper presented at the Annual Meeting of the Midwest American Political Science Association, Chicago, April 6-8. McClelland, C. 1971. “The Management and Analysis of International Event Data.” University of Southern California at Los Angeles. Mimeo. McLaughlin, S., and B. Prins. 1999. “Beyond Territorial Contiguity: Issues at Stake in Democratic Militarized Interstate Disputes.” [International Studies Quarterly 43:169-83. Merritt, R., and D. Zinnes. 1989. “Alternative Indexes of National Power.” In

R. Stoll and M. Ward, eds., Power and World Politics, 11-24. Boulder: Lynne Rienner. Mesterton-Gibbons, M. 1989. A Concrete Approach to Mathematical Modelling. Redwood City, Calif.: Addison-Wesley. Middleton, K. W. B. 1947. Britain and Russia: An Historical Essay. London: Hutchinson. Morgan, T. C. 1994. Untying the Knot of War. Ann Arbor: University of Michigan Press. Morgenthau, H. 1948. Politics among Nations. New York: Knopf. Morgenthau, H. 1985. Politics among Nations. 6th ed. New York: Knopf.

References 183 Morrow, J. 1986. “A Spatial of International Conflict.” American Political Science Review 80:1131-50.

Morrow, J. 1991. “Alliances and Asymmetry: An Alternative to the Capability Aggregation Model of Alliances.” American Journal of Political Science 35:904—33.

Morrow, J. 1993. “Arms versus Allies: Tradeoffs in the Search for Security.” International Organization 47:207-33. Morrow, J. 1994. Game Theory for Political Scientists. Princeton: Princeton University Press. Morrow, J. 1996. “The Logic of Overtaking.” In J. Kugler and D. Lemke, eds., Parity and War: Evaluations and Extensions of the War Ledger, 313-30. Ann Arbor: University of Michigan Press.

Most, B., and R. Siverson. 1987. “Substituting Arms and Alliances, 18701914: An Exploration in Comparative Foreign Policy.” In C. F. Hermann, C. W. Kegley Jr., and J. N. Rosenau, eds., New Directions in the Study of Foreign Policy, 131-57. Boston: Unwin Hyman. Most, B., and H. Starr. 1989. Inquiry, Logic, and International Politics. Columbia: University of South Carolina Press. Olson, M., Jr., and R. Zeckhauser. 1966. “An Economic Theory of Alliances.” Review of Economics and Statistics 48:266-—79.

Oneal, J., and H. C. Whatley. 1996. “The Effect of Alliance Membership on National Defense Burdens, 1953-88: A Test of Mancur Olson’s Theory of Collective Action.” /nternational Interactions 22:105—22. Ordeshook, P. 1986. Game Theory and Political Theory. Cambridge: Cambridge University Press. Organski, A. F. K. 1958. World Politics New York: Knopf. Organski, A. F K., and J. Kugler. 1980. The War Ledger. Chicago: University of Chicago Press. Pearl, R. 1924. Studies in Human Biology. Baltimore: Johns Hopkins University Press. Polachek, S. 1994. “Peace Economics: A Trade Theory Perspective.” Peace Economics, Peace Science and Public Policy, 1, no. 2: 12-15.

Powell, R. 1996. “Uncertainty, Shifting Power, and Appeasement.” American Political Science Review 90:749-64.

Pudaite, P., and G. Hower. 1989. “National Capability and Conflict Outcome: An Application of Indicator Building in the Social Sciences.” In R. Stoll

and D. Ward, eds., Power in World Politics, 79-96. Boulder: Lynne Rienner.

Reuveny, R., and H. Kang. 1996. “International Conflict and Cooperation: Splicing COPDAB and WEIS Series.” /nternational Studies Quarterly 40: 281-306. Richardson, L. F. 1960a. Arms and Insecurity. Chicago: Homewood.

Richardson, L. F. 1960b. Statistics of Deadly Quarrels. Chicago: Quadrangle. Rosen, S. 1972. “War, Power and the Willingness to Suffer.” In B. Russett, ed., Peace, War, and Numbers, 167-83. Beverly Hills: Sage.

184 References Rosenau, J. 1971. The Scientific Study of Foreign Policy. New York: Free Press. Russett, B. 1965. Trends in World Politics. New York: Macmillan. Schampel, J. 1993. “Change in Material Capabilities and the Onset of War: A Dyadic Approach.” International Studies Quarterly 37:395—408. Singer, J. D. 1980. “Accounting for International War.” Annual Review of Sociology 6:349-67. Singer, J. D., S. Bremer, and J. Stuckey. 1972. “Capability Distribution, Uncertainty, and Major Power War, 1820-1965.” In B. Russett, ed., Peace, War, and Numbers, 19-48. Beverly Hills: Sage. Singer, J. D., and M. Small. 1972. The Wages of War, 1816-1965: A Statistical Handbook. New York: Wiley. Siverson, R., and M. Sullivan. 1983. “The Distribution of Power and the Onset of War.” Journal of Conflict Resolution 27:473—94. Siverson, R., and M. R. Tennefoss. 1984. “Power, Alliance, and the Escalation of International Conflict, 1815-1965.” American Political Science Review 78:1057-69. Small, M., and J. D. Singer. 1979. “Conflict in the International System, 18161977: Historical Trends and Policy Futures.” In J. D. Singer, ed., Explaining War, 57-82. Beverly Hills: Sage.

Snyder, G. 1961. Deterrence and Defense. Princeton: Princeton University Press. Sorokin, G. 1994. “Arms, Alliances, and Security Tradeoffs in Enduring Rivalries.” International Studies Quarterly 38:421—46. Spiezio, K. E. 1993. “Power Cycle Theory and State Involvement in Militarized Interstate Disputes, 1816-1976.” Conflict Management and Peace Science 13:87-100. Stoessinger, J. 1993. Why Nations Go to War. 6th ed. New York: St. Martin’s. Taber, C. 1989. “Power Indexes in the Third World.” In R. Stoll and D. Ward, eds., Power in World Politics, 29-48. Boulder: Lynne Riennertr. Thompson, W. 1983. “Uneven Economic Growth, Systemic Challenges, and Global Wars.” International Studies Quarterly 27:341-55.

Tucker, R. 1997. “Dyad-Hard: The Interstate Dyad-Year Dataset Creator.” Political Methodologist 8:28-29. Ulam, A. 1974. Expansion and Coexistence: Soviet Foreign Policy, 1917-73. 2d ed. New York: Holt, Rinehart and Winston. Vasquez, J. 1993. The War Puzzle. Cambridge: Cambridge University Press. Vasquez, J. 1996. “Distinguishing Rivals That Go to War from Those That Do

Not: A Quantitative Comparitive Case Study of the Two Paths to War.” International Studies Quarterly 40:531-58.

Vasquez, J., and R. Mansbach. 1984. “The Role of Issues in Global Cooperation and Conflict.” British Journal of Political Science 14:128—42.

Vincent, J. 1983. “WEIS vs. COPDAB: Correspondence Problems.” /nternational Studies Quarterly 27:161-68. Waltz, K. 1979. Theory of International Politics. Reading, Mass.: AddisonWesley.

References 185 Ward, M. D. 1981. “Seasonality, Reaction, Expectation, Adaptation, and Memory in Cooperative and Conflictual Foreign Policy Behavior.” International Interactions 8:229-45. Ward, M. D. 1982. “Cooperation and Conflict in Foreign Policy Behavior.” International Studies Quarterly 26:87-126. Wayman, F. 1989. “Power Shfits and War.” Paper presented at the Annual Meeting of the International Studies Association, London, March 29—April 1.

Wayman, F. 1996. “Power Shifts and the Onset of War.” In J. Kugler and D. Lemke, eds., Parity and War: Evaluations and Extensions of the War Ledger, 145-62. Ann Arbor: University of Michigan Press. Wayman, F., and J. D. Singer. 1990. “Evolution and Directions for Improvement in Correlates of War Project Methodologies.” In J. D. Singer and P. Diehl, eds., Measuring the Correlates of War, 1-20. Ann Arbor: University of Michigan Press. Weede, E. 1976. “Overwhelming Preponderance as a Pacifying Condition among Contiguous Asian Dyads, 1950-1969.” Journal of Conflict Resolution 20: 395-411.

Werner, S., and J. Kugler. 1996. “Power Transitions and Military Buildups: Resolving the Relationship between Arms Buildups and War.” in J. Kugler, ed., Parity and War: Evaluations and Extensions of the War Ledger, 187210. Ann Arbor: University of Michigan Press. Wolfram, Stephen. 1991. Mathematica: A System for Doing Mathematics by Computer. 2d ed. Redwood City, Calif.: Addison-Wesley.

Wolfson, M., A. Puri, and M. Martelli. 1992. “The Nonlinear Dynamics of International Conflict.” Journal of Conflict Resolution 36:119-49. Zeller, R., and E. Carmines. 1980. Measurement in the Social Sciences. London: Cambridge University Press.

Zinnes, D. 1967. “An Analytical Study of the Balance of Power Theories.” Journal of Peace Research 4:270-88.

Zinnes, D. 1976. “The Problem of Cumulation.” In J. N. Rosenau, ed., Jn Search of Global Patterns, 161-66. New York: Free Press.

Blank Page

Index

Action-reaction mechanisms, 22, 39, 136-37, 140-42, 147-52, 154-55,

40, 51, 60-61, 145, 152-53, 159 157, 159 Afforded rate of alleviation, 67, 79-

81, 86 Capabilities, 11-12, 16, 27, 42-43,

Aggressor, 105, 134, 143, 152 46, 93, 111, 113-18, 129, 142,

Alliance commitments, 41—42 150

Alliances, 9, 12, 27, 38, 40-46, 53, Carrying capacity (K), 2-3, 73 142, 145-49, 153, 155-57, 160 Challenger, 13-15, 20-21, 29, 35, 56,

alliance system, 42, 45-46, 53 65, 91-95, 105, 130, 132, 135-38, Anglo-Japanese alliance, 145-46, 145, 147, 152

148, 156 Check and balance, 12, 58

British-French alliance against China, 22, 117, 128, 132-36, 144-45 Germany, 142, 147-48, 156 Cold War, 10, 47-48, 93

Alsace-Lorraine region, 140 Commitment to prevention, 66, 76—

Appeasement, 142-44, 160 81, 86

Approximate parity, 10, 28-29, 33- Composite indicator of national capa35, 45, 47, 51-52, 146, 150, 154. bilities (CINC), 113-18, 126, 145

See also Parity Concert of Europe, 47

Armaments levels, 37, 43 Conditional conflict equations, 75,

Arms races, 6, 36-37, 60, 93 81-82, 119, 158 Austria-Hungary, 43, 132, 136 Conflict

as a balancing mechanism, 12, 63,

Balance of power (BOP) theory, 1, 5, 105, 151 7-12, 23-26, 28-31, 33-34, 40- continuum, 38-39 42, 44, 46, 50, 52, 55-61, 85, coupling, 56, 67, 94, 102, 105, 133, 104-6, 109, 129, 146, 149-56, 135-37, 141, 144-45, 147, 152-

158-59 53, 159

Barbarossa, 143 definition, 60-61

Bolshevik Revolution, 145 dynamics, 18, 20, 46, 51-52, 131 BOP-PT debate, 34, 46, 50, 52, 85, escalation, 22—23, 35-36, 44, 49,

104, 107, 109, 129, 149-50, 152- 97, 102

53, 158-59 negative values (cooperation), 11, Bosnia, 49—50 35-40, 52, 59, 91, 93, 102, 119Boxer Rebellion, 145 20, 123, 138, 144, 148, 158, Bull and gnat (B&G), 6, 91-95, 99, 160-61 102, 104-6, 109-11, 125-32, peaks, 109, 160 187

188 Index Conflict and Peace Data Bank Germany (Prussia), 10, 22, 24, 47, (COPDAB), 36, 38, 120-21, 124, 49-50, 114-15, 117, 119, 124,

131, 160 132-38, 140-44, 147-48, 150,

Conflict cost rate, 50, 54, 58-59, 70, 156-57, 160

74, 77-78, 80, 86, 92, 95, 97, 99, Gross domestic product (GDP), 27 102, 106, 109-10, 140-41, 148, Gross national product (GNP), 58,

150, 157-59 114-17

Consolidation, 63, 67, 76, 79-80. See

also Rate of consolidation Hegemony, 27, 31, 102

Contenders, 19, 25-26 Hitler, 142-44, 150

Content validity, 115, 118

Correlates of War (COW), 12, 16, 27, Independent and dependent vari-

43, 120 ables, 5, 19, 28, 35, 53, 59

Critical points, 16-19, 72 India, 22, 53, 138, 144

Curve fitting, 18 Inequality-war relationship, 31 Cyclical patterns, 11, 16, 18, 22,51, 72 Inflection points, 16

Initial conditions, 77, 86—87, 91, 93-—

David and Goliath (D&G), 94, 97, 24,108 99, 102, 104-7, 109, 111, 125-30, Internal capabilities, 43, 46 135-36, 144-45, 147-48, 150-52, International system, 7, 10-11, 35,

154, 157, 159 41,56, 72, 117

Decisive advantage (d*), 28-29, 33- Inverted U-shape, 31, 33-34 34, 44-45, 51-53, 61-64, 66, 75— ‘Israel, 22, 94 77, 79, 83, 102, 105, 154-55 Issues, 21, AO, 42, 46-51, 53-54, 74,

Defense coefficient, 7 95, 104, 106, 112, 149, 157-58

Democratic peace, 36 salience, 48, 50, 54, 95, 97, 104, 106

Deterrence, 11, 31 tangibility, 48, 106

Dimension analysis, 68—79 territory, 46, 48, 50, 54 Disparity-motivated conflict rate, 64, vital interests, 48

75, 86 Italy, 117-18, 123, 132

Dissatisfaction, 13, 21, 67, 99, 158 Japan, 24, 58, 118, 123, 132, 145-46,

Enabled rate of reaction, 45—46, 53, 148, 156 60, 63-64, 66-67, 75-81, 86, Kashmir, 53 152-53

Enduring rivalries, 105, 111 League of Nations, 140 Exponential growth or decay, 71—72 Level of analysis, 11-12, 30-31, 42 Liaotung Peninsula, 145

Feedback loops, 21-22, 39 Linearity and linear models, 6, 17, 20, France, 22, 24, 49-50, 109, 114, 119, 30-31, 38, 125, 132 123-24, 132, 134-35, 140-42, Logics-in-use, 55-56

146-48, 156-57 Love race, 37

Game-theoretic models, 20—23, 33, Major powers, 15, 18, 21, 25-26, 41,

42. See also Rational choice 44—45, 73, 81, 109, 111-13, 117-

theory 18, 126-27, 146

Index 189 Militarized interstate disputes shifts, 15, 19, 21, 35, 51 (MIDs), 19, 120-24, 129-31, Power-conflict relationship, 6—7, 10-—

137-38, 140, 143, 145, 147-48, 11, 16, 19, 21, 34-35, 40, 51-52,

154, 160 129, 147, 159

Minor powers, 44—45 Power cycle theory, 16, 18-19, 21, 53

Misperception, 18 Power relationship or distribution, 1,

Munich crisis, 143 5-8, 10-15, 21-24, 27-30, 33-

36, 40—48, 51-53, 55-56, 61, 63-

National capacity, 57, 115 68, 75-79, 81-83, 85, 102, 104— Natural growth, 3, 5, 21, 57, 72, 151 5, 109, 117, 119, 128, 142, 146, Negative armaments, 36-37. See also 148, 152-55, 158-59

Love race Power transition (PT) theory, 10-14,

Non Aggression Pact, 144, 160 21, 23-24, 26, 28-29, 31, 33-35,

Numerical solutions, 9, 36, 68, 72—74, 40-42, 44-46, 52, 85, 104-7, 109, 77, 81, 84-87, 91-95, 97, 99, 102, 129, 149-56, 158-59 104, 108, 120, 125-26, 128-29, Power transitions, 1,2, 4-5, 7-10, 13135, 138, 140-41, 144, 152-55, 16, 19-23, 25-26, 30, 35, 44, 48,

161 55-57, 59, 65, 67, 74, 85, 87, 9195, 97,99, 102, 104-6, 109, 111-

Opportunity and willingness, 61, 66, 12,121, 125-30, 135-36, 138-39,

67 145-51, 153-55, 159, 161

Overwhelming preponderance (d ), winning, 93, 97, 104, 106, 127, 147,

28-29, 34, 44-45, 51-52, 62-64, 150-52 66-67, 76-79, 81, 94, 97, 102, Preferences, 39, 47, 49-50, 111 105, 120, 154-55. See also Perma- Preventive motivation for conflict,

nent advantage 63-65, 76, 78

Probabilistic or statistical models, 6,

Pakistan, 22, 53, 138 13-14, 18, 20-22, 24, 27, 30, 32, Para bellum adage, 12 42—43, 105, 110, 124, 130, 137 Parameters, 45—46, 50—54, 68-71, Puzzles, 11, 26-27, 40, 51 73-81, 84, 86-87, 95, 97, 108,

141, 157-58 Rapallo Treaty, 144

Parity, 10, 15, 24, 27-29, 31, 34-35, Rate of consolidation, 66, 76, 80, 86 44—45, 47, 51-52, 63, 65, 140, Rational choice theory, 11, 14, 20,

146, 150, 154 49-51

Peace, 7, 10-12, 20-21, 24—27, 29, Realism, 10, 46 35-37, 40-41, 44-45, 47, 85, Resigned rate of surrender, 67, 79,

104-6, 120, 123, 129, 150, 154 81, 86 Peaceful balance, 21, 106 Richardson, L. F., 6—7, 36-37, 39,

Peloponnesian War, 47 59-60, 105

Permanent advantage, 93 Population models, 1—3, 16, 72—73 S-curve, 1-5, 12-13, 159

Power Simulations. See Numerical solutions definition, 57-58 Specific or instantaneous growth rate,

relative nature, 16-19, 117 2—4, 58-59, 70-74, 77-78, 80resources, 75, 113, 151, 153, 159 81, 86, 140-41

190 Index Stages of development, 3, 13, 16 United Nations, 49

Stalin, 143-44, 160 United States, 10, 22, 47-48, 58, 73, Strategic decision making, 20, 51 93, 117-18, 123, 130, 132, 134Submission and dominance, 3, 10, 29, 38, 142-45, 147, 160

53, 63, 66, 97, 140, 143 U-shaped curve, 31, 34

Syria, 22 USSR (Russia), 10, 22, 24, 47-48, 93, 117, 128, 130, 132-34, 136-37,

Territory, 121-22. See also Issues, ter- 142-48, 156, 160 ritory Tit-for-tat, 45

Tortoise and hare (T&H), 92-94, 97, War, timing of, 13-14, 33, 48, 109,

102, 104-6, 108-9, 111, 125-30, 159 133-37, 142-44, 147-48, 150-55, War initiation, 9, 12-14, 17-19, 27,

159-60 31, 36, 41-42, 46, 52, 59-60, 65,

Treaty of Berlin, 144 104-5, 144

World Events Interaction Survey

United Kingdom (Britain), 10, 47, (WEIS), 38, 120-21 123, 132, 134, 136-38, 142, 144- World Wars I and II, 10, 24-25, 117,

48, 156 123, 134, 138, 140-44, 147, 156