UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Testing the anarchy tenet : an empirical analysis of the anarchy-cooperation and anarchy-conflict relationships Wong, Kelvin Richard 1997

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1997-251888.pdf [ 9.93MB ]
Metadata
JSON: 831-1.0088081.json
JSON-LD: 831-1.0088081-ld.json
RDF/XML (Pretty): 831-1.0088081-rdf.xml
RDF/JSON: 831-1.0088081-rdf.json
Turtle: 831-1.0088081-turtle.txt
N-Triples: 831-1.0088081-rdf-ntriples.txt
Original Record: 831-1.0088081-source.json
Full Text
831-1.0088081-fulltext.txt
Citation
831-1.0088081.ris

Full Text

TESTING THE ANARCHY TENET: AN EMPIRICAL ANALYSIS OF THE ANARCHY-COOPERATION AND ANARCHY-CONFLICT RELATIONSHIPS by KELVIN RICHARD WONG B.S., Portland State University, 1987 M.S., Portland State University, 1990 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY i n THE FACULTY OF GRADUATE STUDIES Department of P o l i t i c a l Science We accept t h i s thesis as conforming to th{= frequ^r^d standard THE UNIVERSITY OF BRITISH COLUMBIA October 1997 ® Kelvin Richard Wong 1997 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree. that the Library shall make it freely available for reference and study, I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of \*Lbt<J ^OU^CSL. The University of British Columbia Vancouver, Canada Date 0% l-Mv DE-6 (2/88) 11 TESTING THE ANARCHY TENET: AN EMPIRICAL ANALYSIS OF THE ANARCHY-COOPERATION AND ANARCHY-CONFLICT RELATIONSHIPS ABSTRACT This research tests the v a l i d i t y and accuracy of the hypothesis which asserts that changes i n l e v e l s of anarchy are p o s i t i v e l y associated with changes i n l e v e l s of c o n f l i c t and negatively associated with changes on cooperation. A model of anarchy as a variable i s constructed with neorealist assumptions about the character of states and environment i n which they e x i s t . The model u t i l i z e s two variables, the s i m i l a r i t y of states' ends and the s i m i l a r i t y of states' means, to construct a variable which measures anarchy. The synthesis of these two variables y i e l d s the anarchy variable. After developing the t h e o r e t i c a l framework which j u s t i f i e s the use of the s i m i l a r i t y of ends and the s i m i l a r i t y of means as measures for anarchy, anarchy's rel a t i o n s h i p to cooperation and c o n f l i c t i s tested. The hypothesis i s tested with p a r a l l e l analyses of data from the Behavioral Correlates of War and International C r i s i s Behavior data sets. Using nonparametric tests of sig n i f i c a n c e and association, the research concludes that the test hypothesis i s not accurate and must be rejected. Results show that hypothesized cooperation and c o n f l i c t r e lationships to anarchy are not v a l i d or accurate, and that d i f f e r e n t types of cooperation and c o n f l i c t have d i f f e r e n t r elationships to anarchy. These re s u l t s provide empirical r e s u l t s which undermine neorealist assertions about anarchy and i t s consequences, as well as theories of i n t e r - s t a t e cooperation whose foundation i s b u i l d on the idea that an erosion of anarchy w i l l increase i n t e r - s t a t e cooperation. Aside from the empirical analyses and res u l t s , an important feature of the research i s the development of anarchy as a measurable empirical variable. TESTING THE ANARCHY TENET: AN EMPIRICAL ANALYSIS OF THE ANARCHY-COOPERATION AND ANARCHY-CONFLICT RELATIONSHIPS i v TABLE OF CONTENTS Abstract i i Table of Contents i v L i s t of Tables v i i L i s t of Figures x Acknowledgement x i i CHAPTER 1 Introduction 1 CHAPTER 2 Previous Research 14 Bodin and Hinsley on Sovereignty . . 16 The Modern Mainstream View 33 D e f i n i t i o n and C e n t r a l i t y of Sovereignty . . . 34 Anarchy's Ef f e c t s 39 Erosion of Sovereignty 42 The C r i t i c s 47 C r i t i c s on the E f f e c t of Anarchy 4'8 C r i t i c s on the Conceptual Fuzziness of Anarchy 50 Conclusion 52 CHAPTER 3 Modelling Anarchy . . . . . . 53 Defining Anarchy 57 Step 1: counter-position of anarchy and government i n terms of fun c t i o n a l ( d i s ) s i m i l a r i t y 57 Step 2: Counter-position i n terms of rule making and rule enforcing 59 Step 3: Rule making and Rule enforcing in terms of s i m i l a r i t i e s of ends and means . . 62 Integrating the steps: l i n k i n g anarchy to s i m i l a r i t i e s of ends and^means 65 Anarchy as a Variable 68 S p e c i a l i z a t i o n Argument 71 S i m i l a r i t y of ends and anarchy 73 S i m i l a r i t y of means and anarchy. 76 Interactive components and anarchy 77 S p e c i a l i z a t i o n 83 End-Drive-Means Argument 87 V End-Drive-Means Argument . . . . ' . . . . .. . . 87 Ranking Anarchy . . 89 Summation . . 93 CHAPTER 4 Methodology . . . . . . . . .96 Data. . . . . . . . . 96 BCOW C o n f l i c t Variables. . . . . . . . . . . . 96 ICB C o n f l i c t Variables . 100 BCOW Cooperation Variables 102 ICB Cooperation Variables. . . . . . . . . . .104 BCOW's Anarchy Measure . 107 ICB's Anarchy Measure . 113 S t a t i s t i c a l Tests: Dancers and Wallflowers . . 115 Goodman-Kruskal Gamma . . . . . . . . . . ' ; . 116 Kruskal-Wallis H-Test 119 Kendall C o e f f i c i e n t of Concordance . . . . .. 120 Wallflowers 120 Test Application and Projected Results . . . . 124 Matched-Pair Analysis . . . . 125 Aggregated Analysis 126 Significance Level . . . . . . . . . . . . . 127 Projected Results . . . . . . . . . . . . . . 128 CHAPTER 5 Analysis of Bcow Data . . . . . . . 131 Findings Concerning the S i m i l a r i t y of Means Hypothesis . . . . . 134 Si m i l a r i t y of Means: Summary of Results . . . 141 Findings Concerning the S i m i l a r i t y of Ends Hypothesis . 142 Si m i l a r i t y of Ends: Summary of Results . . . 148 Findings Concerning the Anarchy Hypothesis . . 149 Anarchy C o n f l i c t Relationships . . . . . . . 150 H-test Results . . . . . . . . . . . . . . . 150 Difference i n Accurate and Erroneous Results i n Actual and Random Dist r i b u t i o n s 154 Goodman-Kruskal Gamma Results. . . . . . . . 158 Anarchy-Conflict Summary . 159 Anarchy Cooperation Relationships . . . . . 161 H-test Results. . . . . . . . . . . . . . . 161 Difference i n Accurate and Erroneous Results i n Actual and Random Dis t r i b u t i o n s . 164 Goodman-Kruskal Gamma Results 166 Anarchy-Cooperation Summary . . . . . . . . 167 Summary of Results . . . 169 Si m i l a r i t y of Means Hypothesis . 169 Si m i l a r i t y of Ends Hypothesis . . . . . . . 170 Anarchy Hypothesis . . . . . . 171 v i CHAPTER 6 Analysis of ICB Data . •• 173 Data Preparation . . . . . . . . . . . . . . . 173 Hypothesis Testing . . . . . 175 Findings Concerning the S i m i l a r i t y of Means Hypothesis 175 S i m i l a r i t y of Means: Summary of Results . . 181 Findings Concerning the S i m i l a r i t y of Ends Hypothesis . . . . . . . . . . . . .. 183 S i m i l a r i t y of Ends: Summary of Results . . . 188 Findings Concerning the Anarchy Hypothesis. . 189 Anarchy C o n f l i c t Relationships . . . . . . . 189 H-test and Kendall Results. . . . . . . . . 189 Difference i n Accurate and Erroneous Results i n Actual and Random Dist r i b u t i o n s . . . . 195 Goodman-Kruskal Gamma Results . . 198 Anarchy-Conflict Summary. 199 Anarchy Cooperation Relationships .. 201 H-test Results . . . . . . . . 201 Difference i n Accurate and Erroneous Results i n Actual And Random Dis t r i b u t i o n s . 206 Goodman-Kruskal Gamma Results . 208 Anarchy-Cooperation Summary . . 209 Summary of Results . . . . 210 Si m i l a r i t y of Means Hypothesis . ... . . . . . 210 Si m i l a r i t y of Ends Hypothesis 211 Anarchy Hypothesis . . . . . . . . 212 CHAPTER 7 Conclusion. 214 The S i m i l a r i t y of Means, C o n f l i c t , and Cooperation: V a l i d i t y pf Accuracy of the S i m i l a r i t y of Means Hypothesis 215 The S i m i l a r i t y of Ends, C o n f l i c t , and Cooperation: V a l i d i t y of Accuracy of the S i m i l a r i t y of Ends Hypothesis. . . . . . . 220 Anarchy, C o n f l i c t , and Cooperation: V a l i d i t y and Accuracy of H3 225 Anarchy-Cooperation results . . . . . . . . 226 Anarchy-Conflict 231 Discussion . . . . . . . . . . . . . . . . . 234 Bibliography . 243 Appendix A: Calculation Method of Proportional S i m i l a r i t y of Attributes Score. . 253 Appendix B Calculation of Kendall Co e f f i c i e n t s of Concordance . . . 266 v i i Tables Table Page Table 5.1 Projected and Actual Findings: S i m i l a r i t y of Means and Cooperation and C o n f l i c t . . . . . . . . . . . . . . . . 129 Table 5.2 Goodman-Kruskal gamma s i m i l a r i t y of means-cooperation & c o n f l i c t variables 131 Table 5.3 S i m i l a r i t y of Means Frequency of Errors i n Actual and Random Distributions . . . . . 134 Table 5.4 Hypothesized and Actual Findings: S i m i l a r i t y of Ends and Cooperation and C o n f l i c t .138 Table 5.5 Goodman-Kruskal gamma s i m i l a r i t y of ends-cooperation & c o n f l i c t variables . . . . 139 Table 5.6 S i m i l a r i t y of Means Frequency of Errors i n Actual and Random Distributions 141 Table 5.7 Hypothesized and Actual Findings: Anarchy and Cd n f l i c t 145 Table 5.8 Projected and Actual Findings: Anarchy and C o n f l i c t , Tri-chotomous Form . . . . . . 147 Table 5.9 Frequency of Errors i n Actual and Random D i s t r i b u t i o n 149 Table 5.10 Anarchy-Confliet: Goodman-Kruskal Gamma Results 152 Table 5.11 Projected and Actual Findings: Anarchy and Cooperation Scores . . . . . . . 156 Table 5.12 Projected and Actual Findings: Anarchy and Cooperation tri-chotomous form 158 Table 5.13 Frequency of Errors i n Actual and Random Distributions . 159 V l l l Table 5.14 Anarchy-Cooperation Relationship: Goodman-Kruskal Gamma Results . 161 Table 6.1 Projected and Actual Average C o n f l i c t and Cooperation Scores and Significance by Level of S i m i l a r i t y of Means . . .... . . 172 Table 6.2 Gamma Results S i m i l a r i t y of Means and C o n f l i c t and Cooperation 174 Table 6.3 Frequency of Errors i n Actual and Random Distr i b u t i o n s S i m i l a r i t y of Means . . . . . . 176 Table 6.4 Hypothesized and Actual Average C o n f l i c t and Cooperation Scores and Significance by Level of S i m i l a r i t y of Ends . . . . . . . 181 Table 6.5 S i m i l a r i t y of Ends and C o n f l i c t and Cooperation Goodman-Kruskal gamma results 182 Table 6.6 Frequency of Errors i n Actual and Random Distributions S i m i l a r i t y of Ends . . . 183 Table 6.5 Projected, Actual Average, and Actual Ordinal C o n f l i c t Scores by Level of Anarchy . . . . . . . . . . . . 187 Table 6.8 Kendall Coefficient of Concordance 190 Table 6.9 Projected, Actual Average, and Actual Ordinal C o n f l i c t and Cooperation Variables Scores by Level of Anarchy i n Tri-chotomous Form . . . . . . . . . . . . . 191 Table 6.10 Kendall Coefficient of Concordance Among Anarchy and C o n f l i c t Variables, Tri-chotomous . . . . . . . . . . . . . . . . 192 Table 6.11 Difference of Actual and Random Distributions . . . . . . 194 Table 6.12 Summary of gamma c o e f f i c i e n t s 196 Table 6.13 Projected, Actual Average, and Actual Ordinal Cooperation Variables Scores by Level of Anarchy, Quad-chotomous Form . . . .201 Table 6.14 Projected, Actual Average, and Actual Ordinal Cooperation Variables Scores by Level of Anarchy, Tri-chotomous Form . . . . 203 I X Table 6.15 Frequency and Magnitude of Errors 205 Table 6.15 Summary of anarchy-cooperation gamma c o e f f i c i e n t s . . . . . . . 207 Table 7.1 BCOW and ICB Results Adherence to the Five C r i t e r i a of the S i m i l a r i t y of Means Hypothesis . ...... • • • • . . . . . 216 Table 7.2 S i m i l a r i t y of Means Hypothesis H-test and gamma Significance Results . . . . 219 Table 7.3 BCOW and ICB Results Adherence to the Five C r i t e r i a of the S i m i l a r i t y of Ends Hypothesis . 221 Table 7.4 S i m i l a r i t y of Ends Hypothesis H-test and gamma Significance Results . . . . . . . 224 Table 7.5 BCOW and ICB Results Adherence to the Five C r i t e r i a of the Cooperation-Anarchy Hypothesis . . 227 Table 7.6 Cooperation-Anarchy Relationships H-test and gamma Significance Results i n each coding format . . . . . . . . . . . . . 229 Table 7.7 Summary of Results Anarchy-Conflict . . . . 231 Table 7.7 Summary of Results Anarchy-Cooperation . . . 233 Table A1.2 PSA score calculations for n=3, n=4 and n=5 actors 232 X Figures Figure page Figure 3.1 Pure Anarchy to Pure Government Continuum . . . 68 Figure 3.2 Levels of Anarchy and S i m i l a r i t i e s of Ends and Means 75 Figure 3.3 The Model of Anarchy and Rank Ordering . Levels of Anarchy 88 Figure 4.1 General Hypothesizes . . . . . . . . . . . 93 Figure 4.2 Empirical Hypotheses 94 Figure 4.3 Operational, BCOW, ICB Proxy Variables Anarchy, Cooperation, and C o n f l i c t . . . . 97 Figure 4.4 ICB C o n f l i c t Variables' Codings 101 Figure 4.5 Recoding of C r i s i s Management Technique Into Levels of Cooperation . . . 105 Figure 4.6 Recoding of Timing of Violence Variable . 106 Figure 4.7 Anarchy BCOW Measure: Origi n a l and Recoded Values . . . . . . . . . . . . 110 Figure 4.8 Interest Recodes: Generation of High S i m i l a r i t y (HS) and Low Si m i l a r i t y (LS) Scores 112 Figure 4.9 The Model of Anarchy. Levels of Anarchy . . . . . . . . . . . . . 113 Figure A l . l N d i s t r i b u t i o n s f o r n Actors when n=3,4,5 . . . . . . . . . . . . . . . 256 Figure A1.2 Calculation of Anarchy for n=3, n=4, and n=5 actors using Ends and Means PSA scores . 259 X I Figure A1.3 Anarchy, S i m i l a r i t y Of Ends (ESIM) and Means (MSIM) PSA Recodings .... . . . .263 x i i Acknowledgements I would not have started t h i s t h esis, nor the M.S. thesis, without Gary Scott, who f i r s t taught me about IR. Kim Holley read several versions of more than a few chapters. I suspect that few have benefitted from a better colleague. C h r i s t i n a Blumel read every word of every version, and sustained the next version with just the r i g h t kind of comment and bon mot— stern, sensitive, and humorous a l i k e . Ken Conca, Stephan C u f f e l , Kathy Doherty, and Michael Merlingen read portions of the thesis and provided valuable and extensive comments. Others helped with c r i t i c a l support. Chris, Kenny, Joe and P r i s c i l l a , many thanks. My supervisory committee provided c r i t i c a l help f o r which I am g r a t e f u l . As thesis supervisor Mike Wallace's s t y l e was just the right blend of autonomy and d i s c i p l i n e to keep me going and on track. And, as readers one could not f i n d a better combination than Brian Job and Kal H o l s t i . My thanks to each of you. Needles to say, the remaining errors are mine. L a s t l y and most importantly, I must recognize and acknowledge my wife whose requited a f f e c t i o n i s as important as anything else. 1 CHAPTER 1 INTRODUCTION If there i s a single tenet of international r e l a t i o n s theory i t i s that the anarchical structure of the international system both impedes cooperation and fosters c o n f l i c t between and among states. This tenet, which I w i l l r e f e r to as the anarchy tenet, i s accepted i n one way or another by leading scholars such as Axelrod and Keohane, B u l l , Buzan, Jones and L i t t l e , Keohane, Keohane and Nye, Grieco, H o l s t i , Oye, Ruggie, Snidal, Waltz, Watson, and Zacher. 1 Each concurs that anarchy i s an important, i f not 1 Robert Axelrod and Robert Keohane, "Achieving cooperation under anarchy, strategies and i n s t i t u t i o n s , " World Politics 38, no.1 (1985); Headley B u l l , The Anarchical Society: A Study of Order in World Politics (New York:Columbia, 1977); Barry Buzan, Charles Jones, and Richard L i t t l e , The Logic of Anarchy: Neorealism to Structural Realism {Hew York:Columbia, 1993); Joseph M. Greico, "Anarchy and the l i m i t s of cooperation," International Organization 42, no.3 (1988); idem, "Anarchy and the l i m i t s of cooperation: a r e a l i s t c r i t i q u e of the newest l i b e r a l i n s t i t u t i o n a l i s m , " i n Controversies in International Relations Theory: Realism and the Neoliberal Challenge, ed. Charles W. Kegley (New York:St. Martin's Press, 1995); K.J.Holsti, The Dividing Discipline (Boston:Allen & Unwin, 1985); Robert 0. Keohane, After Hegemony: Cooperation and Discord in the World Political 2 the most important, obstruction to increased cooperation as well as a cause of increased c o n f l i c t between and among states. While these scholars may disagree to the extent and the ways i n which anarchy affects cooperation and c o n f l i c t , each contends that anarchy s i g n i f i c a n t l y a f f e c t s these aspects of states' r e l a t i o n s . These scholars would also not dispute that the re l a t i o n s h i p between anarchy and the behavior of states i s one where states' relationships are more c o n f l i c t i v e and less cooperative when the international system i s anarchically organized or more anarchically organized, and less c o n f l i c t i v e and more cooperative when the system i s not anarchically organized or less anarchically organized. To varying degrees the anarchy tenet i s also accepted by Economy (Princeton N.J.:Princeton, 1984); Robert 0. Keohane and Joseph S. Nye, Power and Interdependence: World Politics in Transition (Boston:Little Brown, 1977); Kenneth Oye, "Explaining cooperation under anarchy: hypotheses and strategies," World Politics 38, no.1 (1985); John G. Ruggie, "International regimes, transactions and change: embedded l i b e r a l i s m i n the post-war economic order," International Organization 36, no.2 (1982); Duncan Snidal, "Relative gains and the pattern of international cooperation," (American Political Science Review 85, no.3 (1991); Kenneth N. Waltz, Theory of International Politics (Reading MA:Addison-Wesley, 1979); idem, Man, the State, and War: A Theoretical Analysis (New York:Columbia, 1956); Adam Watson, The Evolution of International Society: A Comparative Historical Analysis (New York:Routledge, 1992); Mark Zacher, "The decaying p i l l a r s of the Westphalian temple: implications f o r international order and governance" i n Governance Without Government: Order and Change in World Politics, ed. James N. Rosenau and Otto-Ernst Czempiel (New York:Cambridge, 1992). 3 Balance of Power, Regime, International Society, Power Transition, Power Parity, and Hegemonic S t a b i l i t y theories of international r e l a t i o n s . The acceptance of the anarchytenet requires concomitant acceptance of three propositions about i n t e r -state r e l a t i o n s . F i r s t , the i n t e r - s t a t e system i s best characterized as an anarchical system, a system without government, where there exists no authority higher than the state. Second, as a result of anarchy, states are themselves the only legitimate and authoritative a r b i t e r s for c o n f l i c t resolution. Third, because of anarchy the tendency i s for c o n f l i c t to abound and for cooperation to wane. The acceptance of the anarchy tenet, and by extension these three propositions, can be translated into the acceptance of the v a l i d i t y and accuracy of the hypotheses that there i s a p o s i t i v e r e l a t i o n s h i p between anarchy and c o n f l i c t and a negative relationship between anarchy and cooperation. These two hypotheses can be s p e c i f i e d i n the more formal terms of the following four anarchy tenet hypotheses: H i . when anarchy i s high, c o n f l i c t i s high; H i i . when anarchy i s low, c o n f l i c t i s low; H i i i . when anarchy i s high, cooperation i s low; H i v . when anarchy i s low, cooperation i s high. 4 Yet, despite the widespread acceptance of the anarchy-tenet and, by extension, the tenet's four hypotheses, d i r e c t inquiry about the anarchy, cooperation, and c o n f l i c t r e l a t i o n s h i p has been l i m i t e d to th e o r e t i c a l research. 2 No empirical research d i r e c t l y addresses the relationships between anarchy and cooperation, and anarchy and c o n f l i c t . Furthermore, and more b a s i c a l l y , while Deutsch and Kaplan, 3 for example, have tangentially treated anarchy as variable, anarchy as variable has not been a focus of inquiry. Quite the opposite, anarchy i s widely accepted as a constant. 4 As such, the consequences of anarchy contained i n the anarchy tenet hypotheses constitute assumptions, since i n the absence of evidence relationships can only be assumed. The primary purpose of t h i s research i s to examine the v a l i d i t y and accuracy of the anarchy tenet by t e s t i n g the four derivative hypotheses. In order to test the hypotheses, anarchy must be developed as a measurable varia b l e . Such a concept of anarchy needs to be developed 2 B u l l , Anarchical Society; Buzan, L i t t l e and Jones, Logic of Anarchy; Waltz, Theory. 3 Deutsch W. Deutsch, Political Community in the North Atlantic Area (Princeton NJ:Princeton University Press, 1957) ; and Morton A. Kaplan, System and Process in International Politics (New YorkrJohn Wiley and Sons, 1963) . 4 Axelrod and Keohane. "Achieving cooperation under anarchy; " Oye, "Explaining cooperation under anarchy,-" Waltz, Theory . 5 since, as we w i l l see i n chapter 2, i t has been accepted as s t a t i c and constant. The development of anarchy as a variable w i l l be accomplished i n chapter 3, where the model of anarchy as a measurable variable i s presented. The general hypothesis tested i s that the l e v e l of anarchy i s related to the le v e l s of states' c o n f l i c t i v e and cooperative behaviors. The general hypothesis i s based on and derived from three assumptions about international p o l i t i c s , none of which i s novel, and a l l of which are consistent with neorealist assumptions of international p o l i t i c s . Assumption {i} i s that the important actor which inhabits the international p o l i t i c a l milieu i s the state. Assumption {ii} i s that the state i s , has been, and always w i l l be an egoist; that i t i s an actor which acts on i t s own behalf, pursues ends which s a t i s f y i t s e l f , engages i n behavior which i s self-serving, subordinates the int e r e s t s of a l l other actors to i t s own, and seeks to dominate other states. Assumption { i i i } i s that, i n the milieu of international p o l i t i c s , the anarchy of the international system, i s a primary determinant of the range of choices available to states. While these assumptions may be challenged, they are made because the hypotheses being tested are e s s e n t i a l l y neorealist hypotheses. As such i t i s appropriate to make assumptions consistent with the hypotheses. From these assumptions the general hypothesis i s b u i l t . From Assumption {ii}, the state as egoist assumption, i t i s hypothesized that as anarchy increases the l i k e l i h o o d f o r cooperation decreases. The l o g i c i s as follows. Since states are egoists, as anarchy increases so increases the incentive to be a defector and the l i k e l i h o o d that partners w i l l defect. 5 These incentives increase because both the cost of partners defecting and the benefit of being a defector increase. These increases i n turn render partners or p o t e n t i a l partners both less t r u s t i n g and less trustworthy. 6 The benefit of being a defector, the cost of partners defecting, and the decreased t r u s t i n g and trustworthiness of states contribute to a reduced l i k e l i h o o d fo r cooperation. From assumption { i i i } , the anarchy menu assumption, I hypothesize that i f there i s change i n the l e v e l or type of anarchy, the range of choices available to actors w i l l also change. .And, from assumption {i} I w i l l focus on states. Aside from te s t i n g a widely accepted and previously untested tenet, an important feature of the research i s the development of anarchy as an empirical variable. Two 5 The use of the term "defect" i s intended to convey the same meaning as i t s usage i n terms of Prisoner's Dilemma. 6 The argument about t r u s t i n g and trustworthiness was suggested by Anatol Rapoport i n The Strategy of Conscience (New York:Schocken Books, 1969). 7 arguments are i m p l i c i t i n developing anarchy beyond the mainstream negative absence of government s p e c i f i c a t i o n . F i r s t , anarchy i s an important variable for analyses of international p o l i t i c s , but has been mis-specified, or at least under-specified, as a constant. 7 Second, i f anarchy i s defined and treated as variable, analyses of the. propensities of states to engage i n c o n f l i c t i v e and cooperative behaviors can be s i g n i f i c a n t l y improved. My d i s s a t i s f a c t i o n with the mainstream concept anarchy i s not unique. Milner, Ashley, Wendt, Onuf and Klink, and Buzan, Jones and L i t t l e , 8 have i n one way or another argued that the proposition that anarchy i s a constant needs to be reformulated or developed i f i t i s retained as, i n Robert 7 The mis-specification claim i s not unique. Alexander Wendt makes a sim i l a r claim when he argues that Waltz's three part structure i s under s p e c i f i e d . See "Anarchy i s What States Make of i t : the Social Construction of Power P o l i t i c s , " International Organization 46, no.2 (1992) ,- and John Ruggie made a sim i l a r point i n his c r i t i c i s m of Waltz's theory and i t s i n a b i l i t y of Waltz to account for the s h i f t from the feudal-medieval system to the modern international system. See "Continuity and Transformation i n the World P o l i t y : Toward a Neorealist Synthesis," i n Neorealism and it Critics., ed. Robert Keohane (New York:Columbia University Press, 1986) . 8 Helen Milner, "The assumption of anarchy i n international relations theory: a c r i t i q u e " Review of International Studies 17 (1991); Ashley "Problematique;" Wendt "Anarchy i s what states make of i t ; " Nicholas Greenwood Onuf and Frank F. Klink "Anarchy, authority, and ru l e , " International Studies Quarterly 33, no.2 (1989); and Buzan, Jones and L i t t l e , Logic of Anarchy. . 8 North's term, the "master variable" of international r e l a t i o n s . 9 The l o g i c behind my disagreement with the mainstream s p e c i f i c a t i o n of anarchy as simply the absence of government i s that anarchy i s more than just the absence of government. As we accept d i f f e r e n t ideas about, measurement of, and kinds of government, so we should about anarchy. If anarchy i s the opposite of and/or the absence of government, then v a r i a t i o n i n government r e f l e c t s v a r i a t i o n i n anarchy. If one does not reject that anarchy i s and can be treated as variable, then the anarchy-cooperation and anarchy-conflict relationships, as asserted i n the anarchy tenet, can be evaluated and tested. The development and reformulation of anarchy as variable and r e j e c t i o n of anarchy as constant i s thus a required and a priori task to the assessment of the four anarchy tenet hypotheses. I quarrel with neither the consequences of anarchy asserted i n the four anarchy tenet hypotheses gua assumptions nor the argument that anarchy i s of great importance to international relations theory. However, with truant evidence I do question the empirical v a l i d i t y and accuracy of the asserted relationships. I also question the acceptance of analyses that are weakened by an imprecise and 9 Robert North used the term "master variable" to describe variables i n f i r s t order assumptions i n h i s discussion at the 1994 Annual Meeting of the American P o l i t i c a l Science Association, Washington D.C., 1994. 9 approximate s p e c i f i c a t i o n of anarchy. The problem with the mainstream s p e c i f i c a t i o n of anarchy i s that there i s very l i t t l e s p e c i f i c a t i o n . Anarchy i s simply s p e c i f i e d as.the absence of government, and treated as the constant value of a dichotomous nominal variable. In other words, anarchy i s one value i n the dichotomy of "anarchy" and "hierarchy". 1 0 As such, i t i s constant, and provides l i t t l e a n a l y t i c a l benefit beyond an elementary h e u r i s t i c device. This view, which I challenge, i m p l i c i t l y asserts that degrees of anarchy do not exist, cannot be measured, are not i n t e r e s t i n g or useful, and can therefore be treated as a constant. My dispute with th i s s p e c i f i c a t i o n of anarchy i s not with the counter-position of anarchy and government, but d i s s a t i s f a c t i o n with accepting the counter-position as the definition. As we w i l l see i n the next chapter, most scholars tend to accept the mainstream counter-position s p e c i f i c a t i o n of anarchy, and then proceed to.discuss the consequences of anarchy without further c l a r i f i c a t i o n . The problem with accepting the mainstream d e f i n i t i o n of anarchy i s that i t seems both unreasonable and a n a l y t i c a l l y perilous to discuss the consequences of anarchy as a tenet 1 0 Barry Buzan characterizes anarchy as a negative condition i n People, States, and Fear: The National Security-Problem in International Relations (Chapel H i l l , N.C.:University of North Carolina Press, 1983), 94. 10 of international relations without f i r s t having spec i f i e d , understood, and conceived of anarchy as p r e c i s e l y as possible. If anarchy i s r e s t r i c t e d to one value i n the dichotomy, where anarchy i s the absence of government, i t seems that analyses which make central use of such an under s p e c i f i e d concept are of li m i t e d h e u r i s t i c value. The conception of anarchy developed i n chapter 3 and made operational i n chapter 4 w i l l introduce enough s p e c i f i c i t y to remedy these shortcomings. Anarchy w i l l be sp e c i f i e d i n consistent theore t i c a l and empirical terms. This conception of anarchy conceives of anarchy as an ordinal variable. It i s congruent with the mainstream's conception of anarchy as the absence of government, but does not leave the matter there. This part of the research w i l l , as p r e c i s e l y as possible, respond to the following questions: What i s anarchy? What do di f f e r e n t types of anarchy look l i k e ? And what i s the relationship between d i f f e r e n t types of anarchy and the cooperative and c o n f l i c t i v e behaviors of states? 1 1 In chapter 3 a model of anarchy as a non-constant, non-dichotomous variable i s b u i l t with two indicators, the 1 1 Implicit i n these questions i s the argument that anarchy i s dynamic and may be an important and neglected explanatory variable for states' cooperative and c o n f l i c t i v e behaviors, and vice versa. This may seem l i k e common sense, and i t i s . However, i t i s neglected common sense when anarchy i s s p e c i f i e d as constant and when the anarchy tenet i s accepted i n the absence of evidence. 11 s i m i l a r i t y of states' ends and the s i m i l a r i t y of states' means. The s i m i l a r i t y of ends and means are conceptually formulated as dichotomous ordinal variables with values of "high s i m i l a r i t y " and "low s i m i l a r i t y " . The s i m i l a r i t y of means i s a measure of the s i m i l a r i t y of tools or instruments used by actors to secure goals and objectives. The s i m i l a r i t y of ends i s a measure of the s i m i l a r i t y of actors' goals or objectives. The relevance of these two variables to anarchy i s established i n a three step process which (1) counter-poses anarchy and government and casts the d i f f e r e n t i a t i o n i n terms of unit functional s i m i l a r i t y , 1 2 (2) restates and expands anarchy qua functional d i f f e r e n t i a t i o n i n terms of rule making and rule enforcing, and (3) l i n k s differences i n rule making and rule enforcing to s i m i l a r i t i e s of ends and means. Since the s i m i l a r i t y variables are dichotomous, t h e i r combination y i e l d s four permutations. 1 3 Each 1 2 The concept of unit functional s i m i l a r i t y u t i l i z e d here i s consistent with that advanced by Waltz i n Theory; by John Ruggie i n "International regimes, transactions and change: embedded l i b e r a l i s m i n the post-war economic order," International Organization 36, no.2 (1986); and by Buzan, Jones, and L i t t l e , i n Logic of Anarchy. 1 3 The ( d i s ) s i m i l a r i t y of ends and ( d i s ) s i m i l a r i t y of means has a minimum of four d i f f e r e n t configurations. Where HS=similar and LS=dissimilar the possible range of configurations are: (1) E=LS, M=LS; (2) E=LS, M=HS; (3). E=HS, M=LS; (4) E=HS, M=HS. 12 permutation represents a unique configuration of the s i m i l a r i t i e s of ends and means, and each also constitutes a unique l e v e l of anarchy. Then two arguments about the relat i o n s h i p between l e v e l s of anarchy and configurations of the s i m i l a r i t i e s of ends and means are made. One i s about s p e c i a l i z a t i o n 1 4 and the other i s about the r e l a t i v e importance of ends and means for rule making and rule enforcing on anarchy. Once an operational construct of anarchy as variable i s b u i l t , the four anarchy tenet hypotheses can be examined and asserted claims about the relationship between anarchy and states' cooperative and c o n f l i c t i v e behaviors can be assessed for empirical v a l i d i t y and accuracy. This i s done i n chapters 5, 6, and 7. Chapter 5 tests the relationships asserted i n the four hypotheses using data from Behavioral Correlates Of War (BCOW) collected by Russell Leng. Chapter 6 does the same using data from International Crisis Behavior (ICB) data set c o l l e c t e d by Michael Brecher and Jonathan Wilkenfeld. And chapter 7 i s a comparative analysis of the BCOW and ICB findings. 1 4 For other a r t i c u l a t i o n s of th i s argument see Waltz Theory. 127-128; Christopher Layne, "The unipolar i l l u s s i o n : why new great powers w i l l r i s e , " i n Perils of Anarchy: Contemporary Realism and International Security, ed. Michael E. Brown, Sean M. Lynn-Jones, and Steve E. M i l l e r (Cambridge:MIT Press 1995), 140-141. 13 The ICB and BCOW data sets are selected to conduct empirical analyses because they contain data required to test the hypotheses. Measures of cooperation and c o n f l i c t , as well as s i m i l a r i t y of ends and means used to measure anarchy are r e a d i l y available i n them. The BCOW and ICB data sets have the additional advantage of focusing on international c r i s e s . Using international c r i s e s data i s an advantage because cooperative and c o n f l i c t i v e behaviors are more c l e a r l y i d e n t i f i a b l e i n cri s e s than i n other types of interac t i o n s . In chapter 4 the hypothesis t e s t i n g method i s outlined. A f t e r introducing test variables and s t a t i s t i c s , empirical expectations derived from the general hypotheses are generated so that the empirical findings can be compared to hypothesized r e s u l t s . This w i l l allow assessment of the v a l i d i t y and accuracy of the anarchy hypotheses. The methodology outlined i n chapter 4 w i l l allow the research, i n chapters 5 , 6 , and 7 , to contribute more precise conclusions about the anarchy tenet, the general hypothesis, and the relationship between anarchy, c o n f l i c t , and cooperation. Having introduced the research, l e t me turn to the next chapter, a discussion of how anarchy has been treated i n previous research. 14 CHAPTER 2 PREVIOUS RESEARCH The focus of th i s chapter i s on selections from two bodies of l i t e r a t u r e which suggest i n d i r e c t l y that anarchy i s variable. One addresses int e r n a l or domestic sovereignty, and the other addresses external or systemic anarchy. The purpose of thi s review i s to demonstrate that (1) concepts of sovereignty and anarchy are intimately related, (2) sovereignty theorists i m p l i c i t l y accept that sovereignty i s variable, (3) arid that while anarchy has been accepted as a constant, the v a r i a b i l i t y of anarchy i s i m p l i c i t i n the arguments of a number of contemporary international relations theorists. I w i l l f i r s t address sovereignty, how i t relates to anarchy, and why the claim that sovereignty varies i s important for the claim that.anarchy varies. Then, under the rubric of the modern mainstream view, I w i l l turn to consider a number of contemporary voices to demonstrate that while anarchy i s generally accepted as constant, there are unacknowledged and i m p l i c i t suggestions that anarchy i s 15 v a r i a b l e . 1 5 This part i s intended to demonstrate that while v a r i a b i l i t y of anarchy has been an i m p l i c i t argument, i t has not been the focus of d i r e c t attention. Those i n t h i s group i m p l i c i t l y assert that anarchy i s variable when they argue that anarchy has eroded, and that with t h i s erosion state behavior has changed. 1 6 1 5 This discussion emphasizes contributions made a f t e r Kaplan's i n System; John David Singer's, "The l e v e l s of analysis problem" i n The International System: Theoretical Essays, ed. Klaus Knorr and James N. Rosenau (New York:Free Press, 1961) and Kenneth N. Walt's Man, the State, and War: A Theoretical Analysis (New York:Columbia, 1956) f i r s t addressed systemic analyses. The development of the concept of systemic analysis i s important because i t i s the foundation on which concepts of anarchy as a variable i n i t s own right could be b u i l t . The notion that c h a r a c t e r i s t i c s of a c o l l e c t i v i t y wield influences independent of i n d i v i d u a l members could not be f u l l y developed i n the absence of systemic analysis. As such, anarchy cannot be adequately addressed i n the absence.of systemic analysis because the idea of anarchy and i t s consequences are systemic. While e a r l i e r analysts such as Rousseau and Hobbes, for example, posited what might be considered as s t r u c t u r a l or systemic constraints and consequences, t h e i r concerns were prim a r i l y about sovereignty and how i t relates to domestic and i n t e r n a l a f f a i r s rather than external ones. Furthermore, they located the source of c o n f l i c t i n flawed human nature and/or the structures of domestic r e l a t i o n s . 1 6 Robert 0. Keohane and Joseph S. Nye, Power and Interdependence: World Politics in Transition (Boston:Little Brown, 1977) ; Stephen D. Krasner, "Structural causes and regime consequences," i n International Regimes (Ithica New York:Cornell, 1983); John Ruggie, ed., Multilateralism Matters The Theory and Praxis of an Institutional Form (New York:Columbia, 1994). 16 F i n a l l y I w i l l discuss others who are also d i s s a t i s f i e d with the t r a d i t i o n a l concept of anarchy, as well as those who dispute the anarchy tenet. This discussion i s intended to feature others who are d i s s a t i s f i e d with the mainstream concept of anarchy. Yet, with one exception, no attempt has been made to develop anarchy as a measurable v a r i a b l e . 1 7 In part, the conceptual and empirical developments of anarchy i n Chapters 3 and 4, respectively, respond to shortcomings of the ef f e c t s of anarchy made by these c r i t i c s . 1 8 Bodin and Hinsley on Sovereignty This consideration of sovereignty i s not (and i s not intended to be) a complete review of the sovereignty l i t e r a t u r e . While a number of sovereignty theorists w i l l be considered, the focus wil-l be on Jean Bodin 1 9 and F.H. 1 7 Buzan, L i t t l e and Jones, Logic of Anarchy. 1 8 Ashley, Problematique; Charles Lipson, "International cooperation i n economic and security a f f a i r s , " World Politics 37, no.l (1984); Milner, Assumption of Anarchy; Nicholas Greenwood Onuf, World of Our Making: Rules and Rule in Social Theory and International Relations (Columbia:South Carolina, 1989); and Wendt, "Anarchy i s what states make of i t " . 1 9 Jean Bodin i s credited by Torbjorn L. Knutsen as the one who made "the seminal c l a r i f i c a t i o n " of sovereignty. Torbjorn L. Knutsen, A History of International Relations 17 Hinsley. 2 0 The purpose of t h i s consideration i s to i l l u s t r a t e that the idea that sovereignty i s variable has been present since Bodin made his " l a s t i n g d e f i n i t i o n of sovereignty. 1 , 2 1 Focusing oh sovereignty i s important because sovereignty and anarchy are c l o s e l y related. It i s widely accepted that the concepts of anarchy and sovereignty are intimately connected. 2 2 The linkage between sovereignty and anarchy runs along these l i n e s . An actor which claims sovereignty for i t s e l f i s bound i n l o g i c to accept the sovereignty of other actors outside the area of i t s own Theory (New York:Manchester, 1992), 58. 2 0 Jean Bodin, J?epuJbligue [1583] , i n On Sovereignty trans, and ed. by J u l i a n H. Franklin (New York:Cambridge, 1992); and E.F. Hinsley, .Sovereignty, 2nd ed. (New York:Cambridge, 1986) . 2 1 Knutsen, A History of International Relations Theory, 58. 2 2 Hinsley, On Sovereignty, 158; Barry Buzan, People, States, and Fear: The National Security Problem in International Relations (Chapel H i l l , N.C.:University of North Carolina Press, 1983), 94; Arend Lijphart, "The structure of the the o r e t i c a l revolution i n international r e l a t i o n s , " International Studies Quarterly 18 no.l (1974): 44; Charles W. Kegley J r . and Gregory A. Raymond, When Trust Breaks Down: Alliance Norms and World Politics (Columbia:University of South Carolina, 1990), 3. 18 claim. 2 3 This point echoes F.H. Hinsley, who writes that a state which claims sovereignty "claims to be free of l i m i t and control within i t s community i s bound i n l o g i c to concede the same freedom to other states i n t h e i r s . " 2 4 This world of co-existing sovereigns i s nothing more nor less than anarchy. Anarchy cannot exist i n the absence of sovereign actors and sovereign actors only ex i s t i n anarchy. As Hinsley has pointed out, the very notion that states are sovereign has the simultaneous concomitant that states ex i s t i n anarchy. Sovereignty and anarchy are then, Hinsley reasons, "the inward and. outward expressions, the obverse and reverse sides, of the same idea." 2 5 The relevance of sovereignty for anarchy i s that each i s the sine qua non of the other. The importance of the claim that sovereignty varies to the claim that anarchy varies i s the l o g i c a l dependence of each on the other. As such, ideas about internal sovereignty are c r u c i a l f or analyses of anarchy. Given th i s l i n k , i f sovereignty i s variable then an important step has been made for the claim that anarchy must also be variable. 2 3 Martin Wight, Power Politics (New York:Penguin, 1986),35. 2 4 Hinsley Sovereignty, 158. 2 5 Hinsley, Sovereignty, 158. 19 Bodin's and Hinsley's analyses of sovereignty are important for the analysis of anarchy i n at least three ways. F i r s t , as we w i l l see, both posit sovereigns with variable c h a r a c t e r i s t i c s . This d i r e c t claim that sovereigns are variable i s , as noted above, important to the claim that anarchy i s variable. Second, t h e i r d e f i n i t i o n s of sovereign actors i s important for our d e f i n i t i o n of anarchy i f anarchy i s accepted as the absence of a sovereign. Third, t h e i r analyses of sovereignty address the problematigue of rule making and rule enforcing when contending rule makers and rule enforcers e x i s t . 2 6 As such, relations among contending rule making and rule enforcing actors constitute what we consider to be anarchy. Furthermore, the ideas Bodin emphasized and the problems he associated with l i m i t e d sovereignty are important since they address a s i m i l a r condition to that which exists between and among states insofar as no actor i s the rule maker and enforcer. This p o l i t y and the inter-state system are si m i l a r because both are forms of anarchy. Bodin conceived of the sovereign as an actor with supreme rights within a bounded purview. His sovereign was i n d i v i s i b l e and supreme but not absolute. Hinsley's concept of sovereignty i s similar. For Hinsley, sovereignty exists when "there i s f i n a l and absolute authority i n the p o l i t i c a l 2 6 Contending sources such as the Pope and Emperor, and John and the Magna Carta Barons. 20 community ... and no final and absolute authority exists elsewhere."27 However, Hinsley makes the additional claim that the sovereign must also be separate from i t s community. He writes that [A] community and i t s government [sovereign] must be s u f f i c i e n t l y d i s t i n c t , as they are when the government i s i n the form of the state, before the concept of sovereignty i s relevant. 2 8 Thus for Bodin and Hinsley sovereignty exists when there i s only one rule maker and rule enforcer. Both agree on the issue of i n d i v i s i b i l i t y , insofar as they both agree that there can only exist one sovereign at a single time i n the same t e r r i t o r y . Both also recognize the supremacy of the sovereign, and both acknowledge that sovereigns can have bounded purviews, and as such are not s t r i c t l y absolute. In t h i s way each accepts that an actor without absolute authority and/or c a p a b i l i t y could s t i l l be sovereign. 2 9 Both Bodin's and Hinsley's sovereign was l i m i t e d insofar as the sovereign was not absolute. Hinsley i s quite cl e a r on t h i s matter. He argues that the proposition that a "state has experienced a decline i n i t s freedom of action" 2 7 Hinsley, Sovereignty, 26. 2 8 Hinsley, Sovereignty, 21. 2 9 In a p a r a l l e l argument i s made by Daniel Philpott i n "Sovereignty: An Introduction and Brief History," Journal of International Affairs 48, no.2 (1995). 21 and i s no longer sovereign confuses "the possession by the state of freedom to act as i t chooses . . . w i t h the absence over and above the state of a superior authority. 1 1 3 0 Hence, fo r Hinsley, an actor's i n a b i l i t y to engage i n p a r t i c u l a r actions does not necessarily a f f e c t i t s sovereignty, so long as the actor took the form of a state and no higher authority existed. In t h i s scheme there can exi s t sovereign actors with d i f f e r e n t ranges of c a p a b i l i t y to enforce and make rules. Sovereignty was for Hinsley what i t was for Bodin i n so far as both viewed sovereignty as i n d i v i s i b l e and supreme but not absolute. 3 1 One important difference between Bodin and Hinsley i s that Hinsley required a sovereign to be a state. But, Hinsley l i v e d i n a world made up of states and sovereign actors, while Bodin did not. As such, i t should not be surpr i s i n g that Bodin was exclusively concerned with 3 0 Hinsley, Sovereignty, 226. 3 1 The d i s t i n c t i o n between supreme and absolute i s a d i s t i n c t i o n between d i f f e r e n t l y ranged purviews. Absolute indicates supreme i n a purview bounded only by time and t e r r i t o r y , while supreme indicates additional boundaries. For example, the sovereign United States e x p l i c i t l y declares i n the B i l l of Rights boundaries excluded to i t s sovereign purview. France, likewise l i m i t s i t s range of sovereign purview regarding economic l a t i t u d e by i t s association with the European Union. (See Philpott, "Sovereignty: An Introduction and Brief History".) Thus, d i f f e r e n t sovereigns may have d i f f e r e n t purviews and s t i l l be supreme within t h e i r respective purviews. As such, while both may be supreme, neither i s necessarily absolute. 22 est a b l i s h i n g i n t e r n a l sovereignty 3 2 while Hinsley's concerns were about both inte r n a l and external dimensions of sovereignty. According to many of the sovereignty analysts c i t e d below, 3 3 i n and p r i o r to Bodin's time no sovereign had been absolute and singular. No single actor had been sovereign because the r i g h t s to rule making and rule enforcing were derived from d i f f e r e n t a u thorities. From the polis to the emergence of the absolutist state the right to supreme authoritative rule making and rule enforcing was e i t h e r concurrently shared by d i f f e r e n t actors or so embedded i n s o c i a l and community r i t e s that no single actor could be properly defined as the sovereign. 3 4 3 2 Bodin, according to Hinsley, refused to write about the consequences of sovereignty between sovereigns. Hinsley, Sovereignty, 181. 3 3 Antony Black, Political Thought In Europe 1250-1450 (New York:Cambridge, 1992); David Held, Political Theory and the Modern State: Essays on State, Power, and Democracy (Stanford, CA:Stanford University Press, 1989); Bertrand de Jouvenel, Sovereignty and Inquiry into Political Good, trans. J.F. Huntington (Chicago:University of Chicago Press, 1957) ; V.G. Kierman, State and Society in Europe 1550-1650 (New York:St. Martin's Press, 1980); R.R Palmer and Joel Colton "The France of Louis XIV- a b r i e f perspective," i n The Absolutism of Louis XIV-- The End of Anarchy or the Beginning of Tyranny?, ed. Brian Tierney, Donald Kagan, and L. Pearce Williams (New York:Random House, 1967) . 3 4 John A. Hall and G. John Ikenbeery characterize pre-modern states as "puny Leviathans" who could not control or a l t e r t h e i r relationship to society. See The State 23 For example, i n the polls, according to Hinsley, the l i n k s between the government structure and kinship and t r i b a l society were so c l o s e l y woven that the two could not be separated. 3 5 In Rome the right to rule was derived from the populus Romanus and implemented by those t i t l e d imperium of which there could exist any number simultaneously. 3 6 And, i n Medieval times not only was there the dual authority of the Pope and Emperor, but every other person had a superior and subordinate, and a l l society was "a great chain of d uties." 3 7 De Jouvenel asserts that [U]nder these conditions, command was never sovereign ... never was i t e n t i t l e d to a l t e r the content of the obligation which f e l l on the i n f e r i o r , or to whittle away the content of the right retained by him. 3 8 It was i n t h i s general context of no true sovereign and the s p e c i f i c context of Bodin's France, which was fraught with r e l i g i o u s and p o l i t i c a l c o n f l i c t and war between (Minneapolis:University of Minnesota Press, 1989), 23. . 3 5 Hinsley, Sovereignty, 28. 3 6 Hinsley, Sovereignty, 37. 3 7 Augustin Thierry, as c i t e d i n Bertrand de Jouvenel, Sovereignty and Inquiry into Political Good, trans. J.F. Huntington (Chicago:University of Chicago, 1959), 171. 3 8 de Jouvenel, Sovereignty and Inquiry into Political Good, 172. 24 Huguenots and Catholics, that Bodin developed the ideas i n the Republique. In fact, i n seventeenth century France sovereignty was shared between King and parliaments. Parliaments were responsible to uphold laws "which the king could not overstep." 3 9 Furthermore, at the time three hundred or so d i f f e r e n t systems of law were i n force within France. 4 0 The problems Bodin i d e n t i f i e d of his time, as a consequence of the anarchical environment gua the absence of a sovereign, are a l l too f a m i l i a r . In the r e l i g i o u s and p o l i t i c a l c o n f l i c t between the Huguenots and Catholics the Crown had neither the a b i l i t y nor the accepted authority to compel resolution. The remedy was his concept of the absolutist^ a l b e i t l i m i t e d , 4 1 sovereign against whom 3 9 R.R. Palmer and Joel Colton, "The France of Louis XIV- A B r i e f Perspective," i n The Absolutism of Louis XIV--The End of Anarchy or the Beginning of Tyranny?, ed. Brian Tierney, Donald Kagan, and L. Pearce Williams (New York:Random House, 1967),' 4. 4 0 Palmer and Colton, "The France of Louis XIV- A B r i e f Perspective", 4. 4 1 The interpretation of l i m i t a t i o n on sovereignty i s not accepted without objection. For example Joseph A. C a m i l l e r i claims that Bodin's po s i t i o n asserted sovereignty as unlimited. Joseph A. Ca m i l l e r i ("Rethinking Sovereignty i n a Shrinking, Fragmented World" i n Contending Sovereignties: Redefining Political Community, ed. R.B.J. Walker and Saul H. Mendlowitz (Boulder, CO:Lynn Reinner, 1990), 16. J.H. Burns makes a si m i l a r claim when he att r i b u t e s to Bodinian sovereignty the q u a l i t y of being "absolute, perpetual, and i n d i v i s i b l e . " See "The idea of Absolutism," i n Absolutism in Seventeenth-Century Europe, 25 existed no competitor as the sole source and enforcer of rul e . The -Republigue's central thesis i s , according to Joseph A. C a m i l l e r i , that a central p o l i t i c a l authority should rule so that order and security could be restored. 4 2 And Hobbes unabashedly viewed the absence of a sovereign as the f a c t o r which permitted eternal c o n f l i c t . Bodin constructed a concept of sovereignty i n which the sovereign was i n d i v i s i b l e and supreme but not absolute. The sovereign was an i n d i v i s i b l e sovereign i n the sense that no other sovereign existed, and supreme i n the sense that no higher p o l i t i c a l authority existed. The sovereign was not absolute because he was bound by both divine and secular constraints. Thus, while Bodin claimed sovereignty to be both i n d i v i s i b l e and supreme, he recognized i t s l i m i t a t i o n s . Bodin c l e a r l y posits l i m i t s of the sovereign when he argued that [I]f the prince, then, does not have the power to over step the bounds of natural law, which has been established by God ... he w i l l also not be able to take another's property without just and reasonable cause. 4 3 ed. John M i l l e r (London:Macmi11am, 1990), 27. 4 2 C a m i l l e r i , "Rethinking Sovereignty i n a Shrinking, Fragmented World", 16. 4 3 Jean Bodin, Republique, [1583] , i n On Sovereignty, ed. and trans., J u l i a n H. Franklin (New York:Cambridge, 1992), 39. Hence forth referred to as i?epuJbligue. Pagination i s Franklin's. 26 The rights of the sovereign were not only l i m i t e d by divine r e s t r i c t i o n , but also by secular authorities, i n the form of courts, Parlement, and lords. These r e s t r i c t i o n s are c l e a r i n Bodin's claim that "our kings are required by the ordinances and decrees of the court to r i d t h e i r properties that have f a l l e n to them by unjust and unreasonable means" so that "mense lords may lose nothing of t h e i r r i g h t s . " 4 4 Not only did Bodin recognize these l i m i t a t i o n s on i n t e l l e c t u a l and analytic bases, but his successful action to counter Henry I l l ' s new taxation i n 1576 c l e a r l y demonstrates his b e l i e f s on the secular l i m i t s Of sovereignty. 4 5 Thus rather than claiming absolute and Unlimited sovereignty, Bodin's argument about the i n d i v i s i b i l i t y of sovereignty only pertains to a s p e c i f i c and l i m i t e d range of issues, actions, and purviews. Absoluteness of supremacy, for Bodin was only v a l i d .within a s p e c i f i e d range. As such Bodin i s , as Franklin points out, " i m p l i c i t l y working with a concept of l i m i t e d supremacy. A king's authority, accordingly, could be sovereign yet less than absolute." 4 6 It i s less than absolute i n the sense that beyond i t s purview the sovereign was no more, and i n d i v i s i b l e i n the 4 4 Bodin, Republique, 41. 4 5 Franklin, Oh Sovereignty , x - x i . 4 6 Franklin, On Sovereignty, x - x i . 27 sense that within the space the sovereign was both singular and supreme. Burns, who i s c i t e d above as accepting Bodinian sovereignty as absolute and i n d i v i s i b l e , recognizes that the Bodinian sovereign's power was "to be used within bounds." 4 7 And, V.G. Kiernan's hi s t o r y of state-society r e l a t i o n s also found Bodin's concept of sovereignty to be l i m i t e d and not unrestrained. 4 8 David Held, as well, recognizes l i m i t a t i o n s on the Bodinian sovereign. While Held notes, as does Kiernan and Burns, the unconditional and absolute q u a l i t i e s of Bodin's sovereign, Held also recognizes that Bodin's sovereign was restrained by "the fundamental or customary righ t s and laws of the p o l i t i c a l community," which Held d i f f e r e n t i a t e s from natural or divine r i g h t s . 4 9 Thus, l i m i t a t i o n s placed upon the sovereign i s conceded even by those who uphold Bodin's claim about the absoluteness of the sovereign. The notion of a li m i t e d sovereign may seem to be contradictory. The contradiction stems from the equation of unlimited and sovereign, an equation which i s , I think, i n 4 7 Burns, "The Idea of Sovereignty", 28. 4 8 V.G. Kiernan, State and Society in Europe 1550-1650 (New York:St. Martin's Press, 1980), 96. 4 9 David Held, Political Theory and the Modern State: Essays on State, Power, and Democracy (Stanford, CA:Stanford University Press, 1989), 220. 28 error. Sovereignty i s supreme, not absolute: "a sovereign need not be sovereign over a l l matters. 1 , 5 0 It i s clear from the c i t e d passages that both Bodin and Hinsley postulated l i m i t a t i o n s of the sovereign, a l b e i t d i f f e r e n t ones. Bodin posited l i m i t s on the authority of the sovereign, while Hinsley posited l i m i t s on the c a p a b i l i t y of the sovereign. Bodin c l e a r l y posited l i m i t a t i o n s of the sovereign authority, as indicated i n his words "our kings are required by the ordinances and decrees of the court" and "mense lords may lose nothing of t h e i r r i g h t s . " 5 1 And Hinsley also c l e a r l y posited that l i m i t a t i o n s of sovereign c a p a b i l i t y do not necessarily constitute the demise of the sovereign, an assertion which, according to Hinsley, confuses "sovereignty with the possession by the state of freedom to act as i t chooses instead of with the absence over and above the states of a superior authority." 5 2 Even Bishop Jacques Benigne Bossuet, an ardent supporter of the divine rights of kings recognized that the king, as absolute sovereign, i s bound by secular law, a law which "does not only include 5 0 Philpott, "Sovereignty: an introduction and b r i e f h i story," 357. 5 1 Bodin, Republique, 41. 5 2 Hinsley, Sovereignty, 226. 29 r e l i g i o n , but the law of the realm" which the king i s "subject to as much as any other." 5 3 The issues of d i v i s i b i l i t y , absoluteness, and supremacy of sovereignty are d i f f e r e n t issues. The claim that sovereignty i s i n d i v i s i b l e i s not the same claim that sovereignty i s absolute. Sovereignty i s i n d i v i s i b l e i n the sense that there cannot be more than one sovereign. Morgenthau was eloquent on t h i s point when he claimed that "[T]he simple truth i s that a divided sovereign i s l o g i c a l l y absurd and p o l i t i c a l l y unfeasible." 5 4 But, by i n d i v i s i b l e Morgenthau meant that sovereignty cannot be shared, and that there can only be one sovereign. These claims are d i f f e r e n t from the claim that sovereigns must have the same purviews of rule making and rule enforcing. Morgenthau's contention that both Cuba and the United States are sovereign, but that Cuban sovereignty rests on the consent of the United States, 5 5 indicates his recognition that d i f f e r e n t 5 3 Jacques Benigne Bossuet, P o l i t i g u e tiree des propre paroles de l'Ecriture Sainte, [1870], i n The Absolutism of Louis XIV, trans. L. Pearce Williams, ed. Brian Tierney, Donald Kagan, and L. Pearce Williams (New York:Random House), 16. 5 4 Hans Morgenthau, Politics Among Nations. : The Struggle for Power and Peace, 2nd ed. (New York:Knopf, 1954), 330. 5 5 Morgenthau, Politics Among Nations, 324. 30 sovereigns have d i f f e r e n t ranges for v a l i d rule making and rule enforcing. Sharing sovereignty i s an issue regrading the d i v i s i b i l i t y of sovereignty, while absoluteness of sovereignty i s an issue regarding the range of v a l i d rule making and rule enforcing. A sovereign i s absolute i n so far as i t has no l i m i t s placed upon i t s purview aside from some temporal and spacial ones, an idea which even Hobbes seems to discount when he claimed the right of a l l to defend t h e i r l i v e s against the sovereign; as such, the sovereign was bound to not v i o l a t e t h i s right, and was therefore not absolute. 5 6 Supremacy i s an issue regarding the absence of contention within a bounded purview. A sovereign i s supreme i n so f a r as i t has no r i v a l s i n i t s purview. The theme of sovereignty as i n d i v i s i b l e , in-absolute, and supreme, as developed by Bodin and a r t i c u l a t e d by Hinsley, contains the p r i n c i p l e that sovereignty i s variable. Sovereignty i s variable by vi r t u e of the v a r i a b i l i t y i n the purview of sovereigns. If v a r i a t i o n i n the purview of sovereigns i s possible, then so i s v a r i a t i o n i n sovereignty. Given the relationship between anarchy and sovereignty, i f sovereignty can vary, then so can anarchy. The idea that sovereignty and anarchy are variable i s also contained i n the thoughts of Emile Cruce (1623) . His 5 6 Burns makes t h i s point i n "The Idea of Absolutism, 11 41.. 31 idea that some aspects of rule making and rule enforcing be made more h i e r a r c h i c a l i m p l i c i t l y asserts that both anarchy and sovereignty vary. The relevance of these sovereignty analysts i s that t h e i r ideas about sovereignty suggest that i t i s variable, and given the anarchy-sovereignty linkage, that anarchy i s also variable. The modern post-systemic analysts b u i l t on these concepts. The l a t t e r accepted t h e i r predecessors' view that anarchy i s an important cause of or contributor to c o n f l i c t , just as they accepted the notion that a r e s t r a i n t on anarchy would contribute to more peace and cooperation. In much the same way they viewed an erosion of sovereignty as the key which could provide an escape from war and other c o n f l i c t . Yet, they do not d i r e c t l y conceive of anarchy as variable. For example, while Morgenthau does not d i r e c t l y speak of anarchy, the concept i s c l e a r l y present i n his rationale for power seeking. But, t h i s should not be sur p r i s i n g given his emphasis on human nature exemplified i n his statement that p o l i t i c s " i s governed by objective laws that have t h e i r roots i n human nature." 5 7 Scholars' p r i o r to the introduction of systemic analyses were, as Waltz notes, unable to conceive f u l l y of anarchy and i t s importance i n the absence of the idea that 5 7 Morgenthau, Politics Among Nations, 4 . 32 systemic factors influenced behavior. 5 8 While Hobbes and Rousseau hinted at these factors, they did so i n d i r e c t l y . However, t h i s f a i l u r e does not discount the important contributions of Bodin, Hinsley and others f o r our analysis of anarchy, which concludes that sovereignty, and therefore anarchy, varies. One purpose of our examination of sovereignty has been to show that sovereignty and anarchy are intimately related. They are, as Hinsley pointed out, the reverse side of the same thing. This notion w i l l help when an operational d e f i n i t i o n of anarchy i s offered i n Chapter 3. A second point of importance i s that from Bodin onward the idea that sovereignty varied, insofar as the range of sovereign purview varied, has been present. 5 9 Given the close and interdependent relationship between sovereignty and anarchy, 5 8 Kenneth Waltz. "Realist Thought and Neorealist Theory," i n Controversies in International Relations Theory: Realism and the Neoliberal Challenge, ed. Charles W. Kegley J r . (New York:St. Martin's, 1995), 78. 5 9 In a recent a r t i c l e Janice E. Thompson claims some v a r i a b i l i t y of sovereignty when she acknowledges that "there may be degrees of sovereignty." She also attributes the acceptance of sovereignty as variable to l i b e r a l interdependence theorists such as Ruggie i n MutHateralism, and others such as Ashley i n "Problematique." See Janice E. Thompson "State Sovereignty i n International Relations: Bridging the Gap Between Theory and Empirical Research," International Studies Quarterly 39, no.2 (1995): 216-18, 227. 33 the idea that sovereignty varied e x p l i c i t l y indicates that . anarchy i s variable. The Modern Mainstream View I now turn to an examination of the modern mainstream view of anarchy. As we w i l l see below, modern analysts of international relations generally accept three points about anarchy and international r e l a t i o n s . F i r s t , anarchy i s :• accepted as nothing more than the absence of government and i s constant. Second, anarchy i s of central importance for international p o l i t i c s and the primary element of international p o l i t i c s . Third, anarchy causes c o n f l i c t and impedes cooperation. Even though these three points are generally accepted, c r i t i c s of them ex i s t . The discussions which follows focuses on how anarchy has been conceived of, de-fined, upheld, critiqued, and under-specifled as the absence of government and mis-s p e c i f i e d as a constant. Some c r i t i c s point to the conceptual fuzziness of anarchy. Others challenge the assertion that anarchy necessarily impedes cooperation.and fosters c o n f l i c t . F i n a l l y , a t h i r d voice asserts that anarchy has been eroded, and thus i n d i r e c t l y questions anarchy as a constant by virt u e of suggesting that anarchy has changed. 34 Definition and Centrality of Anarchy Oran Young's d e f i n i t i o n of anarchy as the absence of government, can stand for a host of others: [A]bove a l l , I want to emphasize that the international p o l i t y i s an anarchically organized p o l i t i c a l system. L i t e r a l l y anarchy refers to the absence of a r u l e r . More generally, p o l i t i c a l anarchy i s the condition of any p o l i t y that i s lacking formal i n s t i t u t i o n s or government at the system l e v e l , and that i s highly decentralized with the respect to the d i s t r i b u t i o n of authority and power.60 This d e f i n i t i o n of anarchy i s close to that given by Martin Wight when he f i r s t acknowledged that anarchy i s more than the absence of government. But, a f t e r t h i s acknowledgement he f e l l back on the absence of government d e f i n i t i o n . Wight hints at a more complex notion of anarchy, one that I w i l l pursue at a l a t e r point, when he claims that anarchy i s "a, m u l t i p l i c i t y of powers without government."61 B u l l makes the si m i l a r point i n his d e f i n i t i o n of anarchy as "the absence of central authority 6 0 Oran Young, "Anarchy and Social Choice: Reflections on the International P o l i t y , " World Politics 30, no.2 (1978) :242 . 6 1 Wight, Power Politics, 101. 35 commanding overwhelming force and the legitimate use of i t But, too many t a c i t l y accept the mainstream d e f i n i t i o n , as d i d Wight four pages a f t e r his dual condition d e f i n i t i o n c i t e d above, when he posits that " i f anarchy i s understood to mean the absence of common government... ."" This error, which ignores both considerations of power as well as rule making and rule enforcing c a p a b i l i t i e s , i s p r e c i s e l y the type of acceptance of anarchy which ignores important elements of anarchy or which glosses over them. Thus i n spite of the acknowledgement of a more complex notion of anarchy which includes both the absence of government and asymmetric d i s t r i b u t i o n of force hinted a t , 6 4 many simply focus on the absence of government i f not by an act of commission, then by acts of omission. To remedy t h i s shortcoming the d e f i n i t i o n of anarchy that w i l l be employed i n Chapters 3. and 4 w i l l emphasize both the c a p a b i l i t y and central authority of rule making and rule enforcing. 6 2 B u l l , The Anarchical Society, 62. 6 3 Wight, Power Politics, 105. 6 4 B u l l , Anarchical Society; Kegley and Raymond, When Trust Breaks Down; Young, "Anarchy and s o c i a l choice,-" and Wight Power Politics. 3 6 It i s commonly accepted that anarchy occupies a central place i n international relations theory. Richard W. Mansbach and John A. Vasquez claim that "[S]ince the b i r t h of the modern nation-state . . .a single paradigm has held sway," one which assumes "an anarchic environment." 6 5 Robert G i l p i n assumes that "the fundamental nature of international relations has not changed over the millennia. . . . [That i t i s the relationship] between independent actors i n a state of anarchy." 6 6 Torbjorn L. Knutsen, i n a h i s t o r i c a l review of the theory of international r e l a t i o n s , writes that "[T]heories about international society are d i s t i n c t from other p o l i t i c a l theories by being preoccupied with human behavior i n an anarchical society."67 Stephen Krasner writes that the contributors to his edited International Regimes accept that states "act i n anarchic environment." 6 8 And K. J. H o l s t i argues that "there has 6 5 Richard W. Mansbach and John A. Vasquez, In Search of Theory: A New Paradigm for Global Politics (New York:Columbia, 1981), 3. 6 6 Robert G i l p i n , War and Change in World Politics (New York:Cambridge, 1981), 6. 6 7 Knutsen, A History of International Relations Theory, 3. 6 8 Stephen Krasner "Structural Causes and Regime Consequences," i n .International Regimes (Ithica New York:Cornell, 1983), 2. 37 been fundamental consensus [that] states operate i n a system characterized by anarchy." 6 9 Most scholars of international r e l a t i o n s have accepted that anarchy i s a constant, either by d e f i n i t i o n or by default. Examples of e x p l i c i t claims can be found i n Robert Keohane's and Robert Axelrod's assertion that "[A]narchy, defined as lack of common government remains constant" 7 0 and Kenneth Oye's that "[N]ations dwell i n perpetual anarchy. " 7 1 The currently popular game theoretic approach also treats anarchy as a constant and as the absence of government by the use of Prisoners' Dilemma as an analogy for the international system. Robert Axelrod's claim c i t e d above and Robert Jervis assessment i s that "the research on cooperation under anarchy assumes the actors are i n P[risoner's] D [ilemma] " 7 2 are but two examples. Those who emphasize other variables by default accept anarchy as constant. They point to the v a r i a b i l i t i e s i n 6 9 K . J . H o l s t i , The Dividing Discipline (Boston: A l l e n & Unwin:, 1985), 10. 70 Axelrod and Keohane, "Achieving Cooperation Under " 226. Anarchy, 71 Oye, "Explaining Cooperation Under Anarchy, " 1. Emphasis added. 7 2 Robert Jervis, "Realism, Game Theory, and Cooperation," World Politics, 40, no.l (1988):329. 38 hegemony,73 the p o l a r i t y of the system and/or the d i s t r i b u t i o n of c a p a b i l i t i e s among states, 7 4 interdependencies between states, 7 5 regimes, 7 6 and epistemic communities 7 7 as accounting for v a r i a b i l i t y i n the r e l a t i o n s between or among states. For example, G i l p i n and Waltz 7 8 argue that a v a r i e t y of configurations of organization, or orders, ex i s t i n the international system. Thus, i t would seem that they do not treat anarchy as a constant. However, upon closer examination one finds that i t i s hegemony (Gilpin) and p o l a r i t y (Waltz) which are treated as variable, and not anarchy, which both treat as constant. 7 3 Keohane, A f t e r Hegemony; Adam Watson, The Evolution of International Society:- A Comparative Historical Analysis (New York:Routledge, 1992). 7 4 Kaplan, Systems; G i l p i n , War and Change; Waltz Theory. 7 5 Keohane and Nye, Interdependence. 7 6 Krasner, "Structural causes and regime consequences." 7 7 Peter M. Hass, "Epistemic communities and . international environmental protection," -International Organization 46, no.2 (1992). 78 War and Change and Theory, respectively. 39 Anarchy's Effects The e f f e c t s of anarchy on actors i s also generally accepted. Many argue that because of the absence of government, actors cannot r e s t r a i n themselves from c o n f l i c t . 7 9 The i n a b i l i t y of actors to r e s t r a i n themselves from c o n f l i c t stems from both the absence of government and, depending on the perspective, i n d i v i d u a l , s o c i a l , or systemic forces. 8 0 For example, Robert J e r v i s asserts that [T]he lack of an international sovereign not only permits wars to occur, but also makes i t d i f f i c u l t for states s a t i s f i e d with the status quo to a r r i v e at a goal that they recognize as being i n t h e i r common i n t e r e s t s . 8 1 Other studies which focus on patterns of cooperation and c o n f l i c t concur with J e r v i s ' assessment. In the game theory approach the introduction of variables such as the shadow of the future and side payments, and the argument about t h e i r e f f e c t on the environment i n which actors 7 9 See for examples, Je r v i s , "Cooperation under the security dilemma;" and Waltz Theory. 8 0 These d i f f e r e n t perspectives are highlighted by Morgenthau, Politics Among Nations; Mercer, "Anarchy, s e l f help, r e l a t i v e gains;" and Waltz Theory, respectively. 8 1 Robert Jervis, "Cooperation Under the Security Dilemma," World Politics, 30, no.2 (1978):167. 40 intera c t (which results i n increased cooperation and decreased c o n f l i c t ) i s tantamount to making the argument that the erosion of anarchy promotes cooperation and de f l e c t s c o n f l i c t . 8 2 For Robert Keohane and Joseph Nye, the mutual dependencies between and among states erodes anarchy and i t s constraint on cooperation. 8 3 Edward Morse asserts that the change i n the state during the i n d u s t r i a l and technological revolutions made viol e n t c o n f l i c t less l i k e l y . 8 4 Krasner and colleagues view the presence of international regimes as forces which curb state propensity fo r c o n f l i c t and fosters cooperation. 8 5 Ruggie and contributors to Multilateralism Matters view the presence of m u l t i l a t e r a l i n s t i t u t i o n a l i s m i n the same l i g h t as Regime 8 2 See for example: Duncan Snidal "Coordination versus Prisoners' Dilemma: Implications for International Cooperation and Regimes," American Political Science Review 79, no.4 (1985); idem, "The Game Theory of International P o l i t i c s " World Politics 38, no.1 (1985); Robert Axelrod, The Evolution of Cooperation (New York:Basic Books, 1984); Harrison Wagner "The Theory of Games and the Problem of International Cooperation," American Political Science Review 77, no.2 (1983). 8 3 Robert 0. Keohane and Joseph S. Nye, Power and Interdependence: World Politics in Transition (Boston:Little Brown, 1977). 8 4 Edward L. Morse, Modernization and the Transformation of International Relations (New York:The Free Press, 1976). 8 5 Krasner, International Regimes. 41 t h e o r i s t s . 8 6 Haas sees the increased influence of those with both speci a l i z e d and s c i e n t i f i c t r aining, which he c a l l s epistemic communities, as a contributor to increased cooperation. 8 7 While each of these scholars' emphasis may be on a d i f f e r e n t variable, each addresses the erosion of anarchy. Thus, while they locate explanatory variables i n d i f f e r e n t places, 8 8 each i n his own way posits that as anarchy erodes cooperation increases and c o n f l i c t decreases. Each also accepts the mainstream notion of anarchy, does not d i r e c t l y addresses anarchy as a variable, and leaves unquestioned the e f f e c t of anarchy on state behavior held i n the mainstream, view. Other assaults on mainstream anarchy can be found i n a diverse body of l i t e r a t u r e which asserts consequences of and conditions for the erosion of anarchy. This l i t e r a t u r e at 8 6 Ruggie, Multilateralism Matters. 8 7 Hass, "Epistemic communities and international environmental protection." 8 8 The d i f f e r e n t explanations are found i n asymmetric d i s t r i b u t i o n of power or c a p a b i l i t i e s found i n n-po l a r i t y and hegemonic s t a b i l i t y theories (Waltz, Theory; and Keohane, A f t e r Hegemony) , the regularization of some interactions seen i n regime, i n s t i t u t i o n a l , and m u l t i l a t e r a l theories (Krasner, Regimes; and Ruggie, Multilateralism Matters), Morse's state development theory, and Haas' epistemic community theory. 42 l e a s t t a c i t l y argues that anarchy i s variable. A lineage of th i s genre can be found i n the s p e c i a l i z a t i o n (neo)functionalist, interdependence, regime, i n s t i t u t i o n a l i s e (neo)liberal, and mul t i l a t e r a l i s m l i t e r a t u r e s . E s s e n t i a l l y t h i s group of researchers p o s i t that anarchy can be, and i s at times, eroded. As a consequence of the erosion of anarchy, and presumable sovereignty, the ef f e c t s of anarchy are mitigated, and states are free to pursue goals precluded by anarchy. While within the lineage d i f f e r e n t variables are (under)specified, the commonality i s that erosion of sovereignty i s possible, and that the consequence of such an erosion i s the waning of c o n f l i c t and the prospering of cooperation. Erosion of anarchy While many arguments, t a c i t l y imply that anarchy i s variable by vi r t u e of i t erosion, v some have made more d i r e c t suggestions that anarchy i s variable. However, the research on anarchy as a variable i s underdeveloped, at best. 8 9 8 9 When anarchy i s thought of i n terms of rule making and rule enforcing, l e v e l s of anarchy are not as novel as i t might f i r s t seem. For example, Deutsch's concepts of (non)integrated and (un)amalgamated communities suggest four categories of systems which also r e f l e c t l e v e l s of anarchy. From most to least anarchical they are: non-integrated and none amalgamated, labeled the state system; integrated and non-amalgamated, labeled p l u r a l i s t i c security community,-non- integrated and amalgamated, labeled empire; and 43 Some deny that anarchy i s the defining c h a r a c t e r i s t i c of the international system. For example, contributors to Volker Rittenberger's edited volume reject "that the inte r n a t i o n a l system only admits two e s s e n t i a l l y opposed structures: anarchy and world state . " 9 0 He writes for himself and contributors that "we challenge . . . that the inte r n a t i o n a l system i s best conceived as an anarchical society." 9 12 Rather, they suggest that a conception of regulated anarchy i s better. 9 2 Mark W. Zacher and Richard A. Matthew make a s i m i l a r claim. They claim that the Lib e r a l t r a d i t i o n of integrated and amalgamated, labeled the state. For a concise discussion of these concepts see Donald J. Puchala "Integration Theory and the Study of International Relations," i n From National Development to Global Community: Essays in Honor of Karl W. Deutsch, ed. Richard L. Merritt and Bruce M. Russett (London:George, Allen, and Unwin, 1981). Kaplan, as well, hints at le v e l s of anarchy, or at least types of anarchies, i n his conception of the di f f e r e n t systems he discusses i n System. 9 0 Volker Rittenberger and Michael Ziirn, ed. Tnternational Regimes in East-West Policies (London:Printer Publishers, 1990), 2. 9 1 Rittenberger and Ziirn, International Regimes in East-West Policies, 2. 9 2 Likewise, Daniel Dudney, voicing a d i s s a t i s f a c t i o n s i m i l a r to Rittenberger, argues that the dichotomous anarchy-hierarchy schema needs the t h i r d structure type of negarchy. "Binding Powers, Bound States: the Logic of Geopolitics and Negarchy" (Paper presented and the 35th Annual International Studies Association Meeting, Washington D.C., March, 1994). 44 international relations does not accept the anarchy tenet propositions. They assert that the second defining c h a r a c t e r i s t i c of the l i b e r a l t r a d i t i o n holds that "the organization of the international system i s not captured by the term "anarchy" since i t has many important elements of governance embedded i n i t . " 9 3 However, Zacher and Matthew, as l i b e r a l s , accept that the f i r s t defining c h a r a c t e r i s t i c of the l i b e r a l t r a d i t i o n i s that "states are currently the primary actors i n the international system. 1 , 9 4 It seems that one cannot both accept the state as the primary actor and reject the anarchic organization of the international system, since both the state and anarchy are interdependent. And, i n spite of t h e i r claim, Andrew Moravcsik 9 5 makes no mention of Zacher and Matthew's f i r s t two defining c h a r a c t e r i s t i c s of libe r a l i s m , and i n fact, makes the counter claim--reminiscent of E r i c Nordlinger's 9 3 Mark W. Zacher and Richard A. Matthew, "Liberal International Theory: Common Threads, Divergent Strands" (Paper Presented at 88th annual American P o l i t i c a l Science Association Conference, 1992), 9. 9 4 Zacher and Matthew, "Common Threads, Divergent Strands," 9 . 9 5 Andrew Moravcsik, "Liberalism and International Relations Theory," CFIA Working Paper (Cambridge:CFIA Harvard University, 1992). 45 argument--that i t i s not states, but individuals, which are "the fundamental actors i n world p o l i t i c s " . 9 6 Thus i t seems that there i s support for the notion that anarchy varies despite the fact that no one d i r e c t l y makes the claim. Recent works by K.J. H o l s t i , P h i l l i p G. Cerny, and Mark W. Zacher, 9 7 i n addition to the e a r l i e r c i t e d works of Wendt, Onuf and Klink, and Ruggie, 9 8 imply that anarchy i s variable. K. J . H o l s t i implies d i f f e r e n t forms of anarchy and that anarchy i s therefore not constant when he uses the terms "anarchies." 9 9 Cerny claims that "the post-Cold War [international p o l i t i c a l ] structure i s a new form of quasi-9 6 E r i c Nordlinger, The Autonomy of the Democratic State (Cambridge:Harvard, 1981), 2. 9 7 P h i l l i p G. Cerny, " P l u r i l a t e r a l i s m : s t r u c t u r a l d i f f e r e n t i a t i o n and functional c o n f l i c t i n the post-cold war world order," Millennium 22, no.l (1993); K.J. H o l s t i , "Government without governance: polyarchy i n the nineteenth century european international p o l i t i c s , " i n Governance Without Government: Order and Change in World Politics, ed. James N. Rosenau and Otto-Ernst Czempiel (New York: Cambridge, 1992); and Mark W. Zacher, "The decaying p i l l a r s of the Westphalian temple: implications for international order and governance," i n Governance Without Government, ed. Rosenau and Czempiel (New York:Cambridge, 1992). 9 8 Wendt, "Anarchy i s what states make of i t " ; Onuf and Klink, "Anarachy, authority, and rule,-" and Ruggie "Continuity and transformation". H o l s t i , "Government without Governance," 30. 46 anarchy." 1 0 0 And, Zacher implies that anarchy i s not a constant when he writes that "the international system i s moving from the high l e v e l of anarchy that previously existed to one i n which reasonably important regimes e x i s t . " 1 0 1 The recognition of d i f f e r e n t i a t e d anarchy i s also found i n Lynn M i l l e r ' s analysis. M i l l e r , as Marcus, notes that p r i o r to The Peace of Westphalia, Europe was anarchically organized. M i l l e r asserts that the European system i n t h i s period "was decentralized i f any ever has been; i t s chief c h a r a c t e r i s t i c seems to have been anarchy, i . e . the absence of central r u l e . " 1 0 2 Recognition of the v a r i a b i l i t y of anarchy i s found i n M i l l e r ' s claim that "the anarchy of the middle ages was not the theore t i c a l equivalent of anarchy i n the modern world." 1 0 3 If pre- and post- Westphalian Europe are characterized by anarchy, but anarchy i s d i f f e r e n t i n each period, then i t seems that anarchy must have changed, and i s , as such, variable. 1 0 0 Cerny, " P l u r i l a t e r a l i s m : Structural D i f f e r e n t i a t i o n and Functional C o n f l i c t , " 48. 1 0 1 Zacher, "The Decaying P i l l a r s , " 61. 1 0 2 Lynn H. M i l l e r , Global Order: Values and Power in International Politics (Boulder CO:Westview 1985), 20. M i l l e r , Global Order, 20. 47 Perhaps the most d i r e c t attempt to examine anarchy as a variable can be found i n work by Buzan i n 1983 and 1993. Buzan argued that anarchy may be best represented as a spectrum which ranges between mature and.immature anarchies. 1 0 4 However Buzan's notion of anarchy conflates anarchy and i t s effects, rather than i s o l a t i n g anarchy on i t s own, when he defines the leve l s of anarchy by l e v e l s of chaos and so c i e t y . 1 0 5 In his 1993 portion of The Logic of Anarchy Buzan presents a more refined and precise idea of anarchy as a variable. Anarchy i n the 1993 formulation i s divorced from i t s effects, and i s more of a variable i n i t s own r i g h t . Buzan accomplishes t h i s by introducing un i t c h a r a c t e r i s t i c s into the concept of anarchy, and i n doing so, also introduces v a r i a b i l i t y into anarchy. Further, i n t h i s formulation anarchy i s also p o t e n t i a l l y d i f f e r e n t i a t e d by what he c a l l s "sectoral differences." The Critics The previous section outlined some of the accepted wisdom about anarchy. However, c r i t i c s of the conventional wisdom e x i s t . Sounds of d i s s a t i s f a c t i o n about conceptual c l a r i t y and voices of dissent can be heard. Some of these 1 0 4 Buzan, People, States, Fear, 96. 1 0 5 Buzan, People States, Fear, 96. 48 voices question the u t i l i t y of anarchy and seek to have i t exorcised, while others seek to re-specify i t and i t s consequences. The following section i s intended to i l l u s t r a t e that the anarchy tenet i s not unanimously accepted. C r i t i c s on the effects of anarchy Nicholas Greenwood Onuf has "grave doubts about the claim that anarchy i s the central and defining feature of international r e l a t i o n s . " 1 0 6 Alexander Wendt argues that anarchy i s misunderstood as a structural variable, and that i f the c h a r a c t e r i s t i c s t r a d i t i o n a l l y attributed to the anarchical condition are present, they are present "due to process, not structure." 1 0 7 The arguments set f o r t h by Wendt and Onuf have been labeled as c o n s t r u c t i o n i s t . 1 0 8 The essential proposition which buttresses the constructionist p o s i t i o n i s that the self-help and egoist character of actors i n anarchy are s o c i a l l y constructed, and 1 0 6 Nicholas Greenwood Onuf, World of Our Making: Rules and Rule in Social Theory and International Relations (Columbia:South Carolina), 14. 1 0 7 Wendt, "Anarchy i s what States Make of I t " , 394. 1 0 8 Mercer, "Anarchy, s e l f help, r e l a t i v e gains," uses t h i s l a b e l . 49 that they are therefore not necessary consequences of anarchy. In the proposition i t i s the competitive s e l f - h e l p and e g o t i s t i c a l i d e n t i t y of actors which creates what the mainstream views as necessary consequences of anarchy. Therefore, the constructionists argue that change i n the i d e n t i t y of the actors can also change the consequences of anarchy. 1 0 9 Richard Ashley, i n the same vein, argues that accepting the "fact of anarchy" and the premises which follow from i t "might be exposed as a r b i t r a r y and r h e t o r i c a l rather than unproblematic. " 1 1 0 Patrick Morgan, i n a d i f f e r e n t kind of argument, reaches a s i m i l a r conclusion to the constructionist. Morgan argues that i f anarchy i s the absence of forces which can prevent c o n f l i c t , and i s as such an impediment to cooperation, not only i s anarchy being treated as the equivalent to insecurity, but anarchy i t s e l f cannot be held to account for c o n f l i c t i v e behavior. It cannot, Morgan 1 0 9 Ted Hopf has argued that the constructionist p o s i t i o n so severely undermines Realism that "[WJe [can] no longer unproblematically accept the idea that anarchy ... has determinate implications" for state action i n "Domesticizing Anarchy: State Identity(s) and Interstate Security" (Paper Presented at the 36th International Studies Association Conference, Chicago Feb. 1995), 1. For an argument to the contrary see Mercer "Anarchy, Self Help, Relative Gains". Ashley, "Problematique," 227. 50 argues, be held a cause since i t i s f a c i l i t a t i v e rather then causal. Morgan writes that, anarchy i s f a c i l i t a t i v e - - the root cause [of c o n f l i c t ] i s the p o t e n t i a l l y l e t h a l c o n f l i c t s among states, which, as anarchy leaves states on t h e i r own, makes each open to attack and therefore insecure. 1 1 1 Morgan reasons that i f insecurity among states could be reduced inse c u r i t y would no longer a f f e c t state behavior and would no longer be an impediment to cooperation even i f anarchy were present. 1 1 2 C r i t i c s on the conceptual fuzziness of anarchy Another c r i t i q u e of the anarchy tenet i s based on the conceptual fuzziness of what anarchy i s or may be. Helen Milner and Charles Lipson are two who have made t h i s point. Helen Milner writes that " c l a r i f i c a t i o n of t h i s central concept i n international relations [(anarchy)] i s important 1 1 1 Patrick Morgan, "Multilateralism and security: prospects i n Europe," i n Multilateralism Matter: The Theory and Praxis of an Institutional Form, ed. John Gerald Ruggie (New York:Columbia, 1993), 337. 1 1 2 Morgan, "Multilateralism and security, " 337. 51 since such a key term should not be used without knowing what i s meant by i t . " 1 1 3 Charles Lipson i s , i n a s i m i l a r vein, d i s s a t i s f i e d with the concept of anarchy. He concludes that [T]he idea of anarchy i s , i n a sense, the Rosetta stone of international r e l a t i o n s : a h e u r i s t i c device for decoding i t s basic grammar and syntax. But what was once a blinding insight-- profound and evocative-- has o s s i f i e d and so become blin d i n g and i n the other sense of the word--l i m i t i n g and obscuring. 1 1 4 Both Milner and Lipson are, i n part, d i s s a t i s f i e d with the acceptance of a negatively defined and undifferentiated concept of anarchy.. If anarchy i s the absence of government, and vice versa, then i t seems that d i f f e r e n t i a t i o n of government suggests d i f f e r e n t i a t i o n i n anarchy. In th i s l i g h t i t i s in t e r e s t i n g to note, while "government" has been categorized according to and measured by numerous schemes,115 anarchy has not. 1 1 3 Helen Milner, "The Assumption of Anarchy, " 68. 1 1 4 Charles Lipson, "International Cooperation i n Economic and Security A f f a i r s , " World Politics 37, no.l (1984) -.22. 1 1 5 Democratic, authoritarian, s o c i a l i s t , communist, republican, parliamentary, and so forth. 52 Conclusion Three notes are important to make at t h i s point. F i r s t , anarchy and sovereignty have been accepted as the obverse of one another; the absence of one signals the presence of the other, and vice-versa. Second, while anarchy has been accepted as a constant, there ex i s t s a foundation for questioning t h i s acceptance. Given t h i s foundation, an exploration of the notion that anarchy i s variable beyond that found i n the reviews of both modern and past analyses presented i n thi s chapter i s appropriate. Third, the notion of anarchy as actors who have both the task and a b i l i t y to make and enforce rules, suggested by Wight, B u l l , and Young, i n the e a r l i e r quoted passages, leads d i r e c t l y into the idea developed i n the next chapter: that a s p e c i f i c a t i o n of anarchy can be achieved by using the two elements these authors allude to, namely the goals and c a p a b i l i t y to make and enforce rules. In the next chapter goals and c a p a b i l i t i e s are developed i n terms of ends and means, respectively. Means i s an ind i c a t i o n of actors' c a p a b i l i t y while ends i s a r e f l e c t i o n of actors' goals. 53 CHAPTER 3 BUILDING THE ANARCHY MODEL This chapter presents a conceptual model to d i f f e r e n t i a t e and measure d i f f e r e n t l e v e l s of anarchy. The model w i l l use inputs which are independent from cooperation or c o n f l i c t inputs, since i t w i l l generate a measure of anarchy to test the association between anarchy and c o n f l i c t and cooperation. The proposed model i s but one way to model and measure anarchy. Other avenues which emphasize d i f f e r e n t measures for, or concepts of anarchy, may be equally v a l i d . However, given the dearth of attempts to measure anarchy t h e i r evaluation i s not yet possible. The concept of anarchy advanced i n the model d i f f e r s from the mainstream notion of anarchy i n two important ways. F i r s t , the model sp e c i f i e s anarchy beyond the absence of government. Second, the model asserts that anarchy i s neither constant nor ubiquitously s i m i l a r . In other words, not only does anarchy vary, but d i f f e r e n t l e v e l s of anarchy may exi s t concurrently. 1 1 6 1 1 6 For example, d i f f e r e n t l e v e l s of anarchy co-exist between or among in t e r and i n t r a a l l i a n c e partners. Regarding the difference between i n t e r - a l l i a n c e partners, 54 The model uses two inputs to generate a measure of anarchy. Both inputs are actor based variables. One input variable i s the s i m i l a r i t y of means. The s i m i l a r i t y of means variable i s a measure of the s i m i l a r i t y of tools or instruments used by actors to secure goals and objectives. The other input variable i s the s i m i l a r i t y of ends. The s i m i l a r i t y of ends i s a measure of the s i m i l a r i t y of actors' goals or objectives. Each input variable i s conceived of as a dichotomous ordinal variable with values of "high s i m i l a r i t y " and "low s i m i l a r i t y " . The output from the model, which i s derived from the configuration of the dichotomous ends and means variables, generates a range of four permutations. Each permutation represents a unique configuration of the s i m i l a r i t i e s of ends and means, which together, i t w i l l be argued, represent four lev e l s of anarchy. 1 1 7 David A. Lake has argued that d i f f e r e n t l e v e l s of anarchy existed between the counter-poised U.S. and Soviet l e d a l l i a n c e s ("Anarchy, hierarchy and the v a r i e t y of international r e l a t i o n s , " International Organization 50, no.l (1996) . 1 1 7 A c l a r i f y i n g note about ends should be made p r i o r to further discussion. Not a l l ends are appropriate to use i n the model. The model requires that ends are directed to scarce and incompatible things, or at least that actors perceive them to so be. Ends sought must be incompatible and scarce because, i f they are not, then rule making and rule enforcing regarding these types of ends i s meaningless. If ends seek things which are compatible then rules about end seeking are i r r e l e v a n t . Rules are i r r e l e v a n t 55 Conceptually the linkage between anarchy and the configuration of ends and means begins with a three step process which s p e c i f i e s the anarchy-government dichotomy i n terms of the s i m i l a r i t i e s of ends and means. While taking these steps results i n a d i f f e r e n t i a t i o n of anarchy and government where each i s s t a t i c , the l o g i c u t i l i z e d to make each step w i l l f a c i l i t a t e modelling anarchy as a v a r i a b l e . The f i r s t step i s to counter-pose anarchy and government and cast the counter-position i n terms of unit because there i s no need for a rule: i n anarchy i f ends are compatible actors w i l l behave as they w i l l whether a rule i s present or not. See Arthur A.Stein, Why Nations Cooperated: Circumstance and Choice in International Relations (Ithica:Cornell University Press, 1990), 30; Keohane After Hegemony, 51; and idem, International Institutions and State Power: Essays in International Relations Theories (Boulder CO:Westview, 1989), 159. , Each makes the same point i n d i f f e r e n t ways than i s made here. If what i s considered to be a rule i s present, then as a rule i t i s meaningless i n so f a r as actors' behavior i s independent of the ru l e . In t h i s case the rule i s an a r t i f a c t of behavior rather than a s t i p u l a t i o n or regulation for behavior. For discussion see Arthur A.Stein, "Coordination and collaboration: regimes i n an anarchic world," i n International Regimes, ed. Stephen Krasner (Ithica, New York:Cornell, 1983), 117; Keohane After Hegemony, 51; and idem, International Institutions and State Power, 159. Ends sought must be scarce. When actors seek goods which are p l e n t i f u l actors have no interest i n rules regarding these types of goods, and no rule w i l l be present. Should a rule be perceived to be present then, as above, the rule i s an a r t i f a c t of rather than a s t i p u l a t i o n or regulation for behavior. 56 functional s i m i l a r i t y . 1 1 8 The second step restates and expands the counter-position of anarchy and government i n terms of rule making and rule enforcing. That i s , the difference between anarchy and government i s cast i n terms of rule making and rule enforcing. And i n the t h i r d step, differences i n rule making and rule enforcing are linked to differences i n the s i m i l a r i t i e s of ends and means. At t h i s point i n the development of the model, anarchy i s discussed as a variable with four d i f f e r e n t i d e n t i f i a b l e values. Once these steps are taken, two arguments about the rela t i o n s h i p between anarchy and the configuration of the s i m i l a r i t i e s of ends and means are made. One i s about s p e c i a l i z a t i o n of actors and anarchy 1 1 9 and the other i s about the r e l a t i v e importance of ends and means for rule making and rule enforcing on anarchy. These two arguments j u s t i f y the rankings of the four values of anarchy. The chapter w i l l end with a detailed discussion of why one configuration of s i m i l a r i t y of ends and means i s more/less anarchical than another. While t h i s discussion i s 1 1 8 See Buzan, L i t t l e , and Jones, Logic of Anarchy; Ruggie, "Continuity and transformation i n the world polity,-" and Waltz, Theory, for others who use the concept of unit functional s i m i l a r i t y . 1 1 9 See Waltz, Theory, 127-128; and Layne, "The unipolar i l l u s s i o n , " 140-141, for other a r t i c u l a t i o n s of t h i s argument. 57 li m i t e d to a two-actor case, i t s l o g i c can be generalizable to n-actor cases. 1 2 0 Defining Anarchy This section outlines the three step linkage of anarchy to the configuration of the s i m i l a r i t i e s of ends and means. The f i r s t step i s to counter-pose anarchy arid government i n terms of actors' functional s i m i l a r i t y . Then, i n the second step anarchy i s conceptualized i n terms of rule making and rule enforcement. And, f i n a l l y , the t h i r d step l i n k s rule making and rule enforcing to the s i m i l a r i t y of ends and the s i m i l a r i t y of means. step 1: counter-position of anarchy and government in terms of functional (dis)similarity The mainstream counter-position of anarchy and government serves as our s t a r t i n g point. The notion of counter-posing anarchy and government, or conceiving the two as polar opposites i s common.121 Each i s , as Hinsley has 1 2 0 The expansion to an n-actor case i s made i n the following chapter since i t i s a question of method. 1 2 1 See for examples: B u l l , Anarchical Society, 46; Hinsley, Sovereignty, 158; Kegley and Raymond, When Trust Breaks Down, 4; and Waltz, Theory, 93. 58 noted, the opposite of the other. 1 2 2 Waltz d i f f e r e n t i a t e s anarchic and governmental environments by the functional (dis) s i m i l a r i t y of u n i t s . 1 2 3 Government i s composed of f u n c t i o n a l l y d i s s i m i l a r actors, while anarchy i s composed of f u n c t i o n a l l y s i m i l a r actors. Waltz t e l l s us the functional s i m i l a r i t y d i f f e r e n t i a t i o n i s v a l i d because "[h]ierarchy e n t a i l s r e l a t i o n s of super- and subordination ... which implies d i f f e r e n t i a t i o n . 1 , 1 2 4 Employing the functional ( d i s ) s i m i l a r i t y counter-p o s i t i o n as a proxy for the anarchy-government counter-p o s i t i o n i s v a l i d because there i s an i n e x t r i c a b l e l i n k between anarchy and functional d i s - s i m i l a r i t y and government and functional s i m i l a r i t y . The i n e x t r i c a b l e connection between fu n c t i o n a l l y s i m i l a r actors and actors i n anarchy i s that actors i n anarchy must be considered f u n c t i o n a l l y s i m i l a r because each i s a t i t u l a r decision maker and faces the same problems of making and enforcing rules for t h e i r subjects and keeping other actors from unwanted e f f e c t s to 1 2 2 Hinsley, Sovereignty, 158. 1 2 3 Waltz, Theory. 1 2 4 Waltz, Theory, 93. Waltz does not delve into t h i s difference beyond i t s usefulness to him, which i s to discard the functional s i m i l a r i t y variable i n analyses of international p o l i t i c s . 59 t h e i r i n t e r n a l realm. 1 2 5 As Waltz noted, government implies hierarchy which e x p l i c i t l y denotes actors with f u n c t i o n a l l y d i s s i m i l a r subordinate and supervisory r o l e s . The u t i l i t y of making the d i s t i n c t i o n in terms of functional ( d i s ) s i m i l a r i t y i s that i t provides a single non-negative (one i s the absence of the other) standard to d i s t i n g u i s h between anarchy and government. Given the equivalence of the two counter-positions, e s t a b l i s h i n g that actors are functionally s i m i l a r i s enough to e s t a b l i s h that the environment i s anarchic and that actors who are f u n c t i o n a l l y d i s s i m i l a r exist under government. Step 2: counter-position in terms of rule making and rule enforcing A more precise d i s t i n c t i o n between anarchy and government i s b u i l t on the functional ( d i s ) s i m i l a r i t y standard discussed i n step 1. Precision i s obtained by counter-posing anarchy and government i n terms of rule making and rule enforcing: government, the condition with 1 2 5 Both Stephen D. Krasner ("State Power and the Structure of International Trade," World Politics 28, no. 3 (1976), 316-317) and K.J. H o l s t i ("The Comparative Analysis of Foreign Policy: Some Notes on the P i t f a l l s and Paths to Theory," i n The Political Economy of Foreign Policy in South East Asia, ed. David Warfel and Bruce Burton (New York:St. Martin's Press), 14-15), claim that a l l states have four basic interests or goals. 60 one rule maker and enforcer, i s counter-posed to anarchy, the condition with multiple rule makers and enforcers. An anarchic environment requires that actors be fun c t i o n a l l y s i m i l a r to each other in-so-far as the actors are t h e i r own rule makers and rule enforcers. This i s necessarily the case because i f an actor were not i t s own rule maker and rule enforcer, another would be i t s rule maker and rule enforcer. An actor who i s not i t s own rule maker and rule enforcer i s subject to another actor's rule making and enforcing and would be subsumed into that actor's realm of rule making and rule enforcing. This environment i s not anarchic since i t involves actors with subordinate and supervisory roles. At t h i s point the following d i f f e r e n t i a t i o n between anarchy and government can be made i n terms of rule making and rule enforcing. I f government is the presence of an entity with ultimate authority of decision and sole legitimacy of use of force, whose, task is to regulate other actors' behavior and settle their disputes, and anarchy is the absence of such an entity, then anarchy is the condition where each actor f u l f i l l s the task of behavior regulation and dispute settlement on its own, and where every actor retains ultimate authority for itself and need not recognize any other actors' rule making and rule enforcing. In other words, government is the condition where there is one rule maker and rule enforcer, and anarchy is the condition when 61 there are multiple rule makers and enforcers.126 Casting the d i s t i n c t i o n between anarchy and government i n these 1 2 6 It has been argued that i n anarchy actors a priori r e j e c t the acceptance of other actors' of rule making and rule enforcing. Because actors attempt to r e t a i n ultimate authority of rule making and rule enforcing f o r themselves does not denote that they are a l l equally successful. Thus confusion may arise about anarchy and actors' success or f a i l u r e i n t h e i r pursuit of rule making and rule enforcing. The successes and f a i l u r e s which actors experience i n t h e i r rule making and rule enforcing endeavor does not ef f e c t the presence or absence of anarchy. Some actors may have more of an a b i l i t y to make and enforce rules than other actors. This does not denote that the condition of anarchy i s not present. Successes and f a i l u r e s of rule making and rule enforcing attempts a f f e c t the l e v e l of anarchy but not the presence or absence of anarchy. The condition of anarchy w i l l be present so long as actors do not in v o l u n t a r i l y defer or de-emphasize rule making and rule enforcing and are intolerant of other actors' rule making and rule enforcing v i s - a - v i s themselves. In other words, the condition of anarchy w i l l be present so long as rule making and rule enforcing dominate actors' ends or i s at least primus inter pares among ends. P a r t i c u l a r successes and f a i l u r e s which actors' experience i n t h i s endeavor does not affe c t the presence or absence of anarchy. Actors' i n a b i l i t y to make and,enforce p a r t i c u l a r rules for themselves does not denote that the condition of anarchy i s not present. It does however indicate a d i f f e r e n t l e v e l of anarchy than when actors are successful i n t h e i r rule making and rule enforcing endeavors. Each l e v e l of anarchy i s anarchy because each involves actors who want to ret a i n for themselves, even i f they f a i l to, ultimate authority and f u l f i l l the task of behavior regulation and dispute settlement on t h e i r own. Actors i n anarchy, whatever the l e v e l of anarchy, do not a priori recognize the legitimacy of any other actor's rule making and rule enforcing as legitimate v i s - a - v i s themselves. In anarchy actors a priori reject the legitimacy and authoritativeness of exogenous rule making and rule enforcing. 62 terms provides the support to l i n k rule making and enforcing to s i m i l a r i t i e s of ends and means. Step 3: Rule making and rule enforcing in terms of the similarities of ends and means In step 2 the importance of the difference between anarchy and government was made i n terms of rule making and rule enforcing. In anarchy each actor rejects and has the c a p a b i l i t y to reject others as rule makers and rule enforcers, retains the right to u n i l a t e r a l l y abrogate agreements, and champions i t s own rule making and rule enforcing, while i n government actors accept rule making and rule enforcing of another actor. As a result, i n anarchy there are many rule makers and rule enforcers while i n government there i s but one. Because actors i n anarchy seek to act as t h e i r own rule makers and rule enforcers no single rule maker and rule enforcer can e x i s t . Put another way, i n anarchy each actor has the same end to act as i t s own rule maker and rule enforcer, while i n government t h i s end i s uniquely held by one actor known as the government. In government, the end to be the rule maker and rule enforcer i s precluded to a l l but one. Consequently we can conclude that i n anarchy a l l actors necessarily have rule making and rule enforcing as t h e i r dominant end. 63 The desire of actors to be t h e i r own rule maker and rule enforcer a f f e c t s other ends as well. When actors have the s i m i l a r end to be the rule maker and rule enforcer other substantive ends w i l l also be si m i l a r . Substantive ends w i l l be s i m i l a r because only those ends which contribute to rule making and rule enforcing w i l l be maintained on a constant basis. Ends which impede rule making and rule enforcing w i l l be discarded. When substantive ends are similar, actors w i l l be reluctant to accept another actor's rule making and rule enforcing regarding the end. In thi s case, actors w i l l increase t h e i r own rule making or rule enforcing c a p a b i l i t y or place more emphasis on i t ; i f they do not they r i s k that another w i l l make and enforce rules regarding the substantive end. Hence the tendency w i l l be for actors with rule making and rule enforcing as an end to have s i m i l a r substantive ends as well., This dynamic i s supported by a second one. When actors have sim i l a r substantive ends, they w i l l emphasize t h e i r own rule making and rule enforcing. This common emphasis on rule making and enforcing r e s u l t s i n an increase i n the s i m i l a r i t y of actors' ends. The two di f f e r e n t dynamics act to reinforce each other, and re s u l t i n a homogenizing ef f e c t i n actors ends. 1 2 7 1 2 7 Conversely, when ends are di f f e r e n t , not only i s the dynamic just discussed not present, but i f one actor has the end of rule making and rule enforcing the other actor 64 When actors are egoists, as we have assumed, t h i s tendency i s certain. If ends are substantive, and i f actors have s i m i l a r ends, fragmented rule making and enforcing w i l l ensue. Thus, i t can be concluded that i f actors seek to be t h e i r own rule makers and rule enforcers, not only w i l l t h i s end be similar, but substantive ends w i l l also be s i m i l a r . The relationship between the s i m i l a r i t y of means and rule making and rule enforcing i s more straightforward than the r e l a t i o n s h i p between the s i m i l a r i t y of ends and anarchy. To f u l f i l l i t s rule making and rule enforcing role an actor must possess the c a p a b i l i t y to reject other actors rules at the same time i t makes and enforces rules. This requires of the actor the a b i l i t y to make and enforce rules i n the event i t i s opposed, which i n turn demands that the actor has c a p a b i l i t i e s dis-proportionally higher, and therefore d i s -s i m i l a r enough, r e l a t i v e to other actors, to make and enforce rules i n the face- of opposition. In other words, the rule making and rule enforcing actor must have greater c a p a b i l i t i e s , or more means, to make and enforce opposed rules: the rule making and enforcing actor, having more means than other actors, therefore has means d i f f e r e n t from other actors. cannot have the same end i f ends are d i f f e r e n t . A d d itionally, when ends are di f f e r e n t an actor with the end of rule making and rule enforcing w i l l face less opposition to rule making and rule enforcing since other actors are not contending the rule making and enforcing r o l e . 65 This argument i s akin to those advanced by hegemonic s t a b i l i t y and power p a r i t y t h e o r i s t s . 1 2 8 When actors have s i m i l a r means the a b i l i t y of an actor to behave as the rule making and rule enforcing i s diminished since opposition to i t i s , as a re s u l t of s i m i l a r means, enough to keep the rules from being enforceable. The a b i l i t y of a single actor to act as the rule maker and enforcer i s diminished because other actors, who have si m i l a r means, have the a b i l i t y to r e s i s t the rule making and rule enforcing i f they wish to. With these conditions rule making and enforcing i s fragmented. Even i f , under these conditions, less fragmented rule making and rule enforcing emerges i t w i l l be on an ad hoc basis and dependent on the consent of other actors. Integrating the steps: linking anarchy to similarities of ends and means The three steps taken, (1) the counter-position of anarchy and government where one i s the absence or reverse of the other, (2) the re-conceptualization of anarchy i n terms of rule making and rule enforcing, and (3) the 1 2 8 For a concise review of hegemonic s t a b i l i t y and power p a r i t y theorists see Jack S. Levy "The causes of war: a Review of theories and evidence," i n Behavior, Society, and Nuclear War, Volume One, ed. P h i l i p Tetlock et. a l . (New York:Oxford, 1989) . 66 connection of rule making and rule enforcing to the s i m i l a r i t i e s of ends and means, esta b l i s h the linkage between anarchy and the s i m i l a r i t i e s of ends and means., This linkage provides a point-of-launch to model anarchy based on the s i m i l a r i t i e s of ends and means. Three points about anarchy and government can be drawn from the above discussions: 1. anarchy and government are the obverse of the same thing, that the presence of one i s the absence of the other (step 1); 2. i n anarchy each actor has the si m i l a r end to be t h e i r own rule maker and rule enforcer, and as a resu l t t h e i r substantive ends w i l l also be sim i l a r (steps 2 and 3); 3. i n government one actor has unique and di f f e r e n t means from other actors (step 3). We can deduce two additional points from the three just made. Since that which makes for anarchy makes f o r the absence of government and vice versa (point 1) and since actors i n anarchy have similar ends (point 2) we can deduce that 4. i n government actors have d i f f e r e n t ends. We can also deduce that since i n government one actor has d i s - s i m i l a r means from other actors (3) and since anarchy and government are obverse (1) that 67 5. i n anarchy actors have s i m i l a r means. We are now i n a pos i t i o n to make the following four claims about anarchy and the s i m i l a r i t i e s of ends and means: i . i n anarchy actors have s i m i l a r ends (restated from point 2); i i . i n anarchy actors have s i m i l a r means (re-stated from point 5, deduced from points 1 and 3); i i i . i n government actors have d i f f e r e n t ends (re-stated from point 4, deduced from points 1 and 2); i v . i n government actors have d i f f e r e n t means (restated from point 3). These four claims can be integrated into two statements. F i r s t , that anarchy i s the condition when actors have si m i l a r ends and sim i l a r means. Second, government i s the condition when actors have d i f f e r e n t ends and d i f f e r e n t means. These two statements are commensurate with each other. They are statements of the same thing from d i f f e r e n t perspectives. These aspects of the presence and absence of anarchy are i l l u s t r a t e d i n Figure 3.1. Having offered a d e f i n i t i o n of anarchy, defined what anarchy i s and i s not i n terms of s i m i l a r i t y of ends and means and i n terms of rule making and rule enforcing, l e t me now turn to the discussion of how anarchy varies, and how i t s variance can be measured. 68 Figure 3.1 Pure Anarchy to Pure Government Continuum Anarchy-Government Continuum anarchy pole government pole anarchy range government range multiple rule makers and enforcers I I one rule maker and enforcer ends & means similar I I ends & means different Anarchy as a Variable The f i r s t step i n developing anarchy as a variable i s to d i s t i n g u i s h between anarchy being constantly present and anarchy being a constant. It i s an error to a t t r i b u t e to anarchy an i d e n t i c a l l y homogeneous qual i t y because i t has been a durable feature of international r e l a t i o n s . Making such an error confuses the continuous presence of anarchy with the uniformity of anarchy. I do not intend to suggest 69 that anarchy i s not and has not been present f o r the duration of inte r s t a t e r e l a t i o n s . However, I w i l l contend that there are d i f f e r e n t l e v e l s and types of anarchy, and that anarchy must be conceived of and treated as variable and not constant. 1 2 9 The four claims about anarchy and the s i m i l a r i t i e s of ends and means developed i n the previous section w i l l provide an avenue to i d e n t i f y and then measure d i f f e r e n t l e v e l s of anarchy. These claims y i e l d a s t a t i c model for the presence and absence of anarchy. In the s t a t i c model the presence of anarchy was associated with actors who have s i m i l a r ends and means, while the presence of government (absence of anarchy) was associated with actors who have d i f f e r e n t ends and means. In a model of anarchy where anarchy varies d i f f e r e n t l e v e l s of anarchy occur with d i f f e r e n t l e v e l s of s i m i l a r i t y of ends and/or means. More precisely, d i f f e r e n t l e v e l s of anarchy occur with d i f f e r e n t configurations of the s i m i l a r i t i e s of ends and means. Since the s i m i l a r i t i e s of ends and means are dichotomous variables, there are four possible permutations for the configuration of the s i m i l a r i t i e s of ends and means: 1 2 9 As noted i n the previous chapter, others such as Deutsch North Atlantic; Kaplan System; Buzan, People, States, Fear; and Buzan L i t t l e and Jones, Logic of Anarchy, have t a c i t l y treated anarchy as variable. 70 (1) both are of high s i m i l a r i t y ; (2) both are of low s i m i l a r i t y ; (3) ends are of high s i m i l a r i t y and means are of low s i m i l a r i t y ; (4) ends are of low s i m i l a r i t y and means are of high s i m i l a r i t y . The four configurations of s i m i l a r i t i e s of ends and means are statements for four d i f f e r e n t l e v e l s of anarchy. Each l e v e l i s unique and d i f f e r e n t i a t e d from other l e v e l s . At t h i s point i n the development of the model anarchy i s a variable with d i f f e r e n t values. To complete the model the four configurations need to be ranked i n magnitude of anarchy. Two guidelines w i l l be established to guide these rankings. F i r s t , a c r i t e r i o n to rank the two configurations when both ends and means are si m i l a r and when both are d i s s i m i l a r (1 and 2 from above), and then a second c r i t e r i p n to assess t h e two mixed configurations (3 and 4 from above) w i l l be established. Once these two c r i t e r i a are established the model w i l l be capable of assessing the rank order magnitude of the four configurations outlined above. The f i r s t c r i t e r i o n i s developed i n the next section under the. rubric of sp e c i a l i z a t i o n , and i s followed by development of the second c r i t e r i o n under the rubric of ends-drive-means. 71 Specialization Argument Ranking the l e v e l of anarchy of configuration (1) , when both ends and means are high s i m i l a r i t y , and configuration (2), when both are low s i m i l a r i t y , i s based on the relationships between the configurations and rule making and rule enforcing. An assessment of these relationships, which i s an extension of step 3 i n section 2, begins with an examination of the individual relationship of each component which makes up the configurations of rule making and rule enforcing, and then turns to t h e i r i n t e r a c t i v e e f f e c t s . A f t e r establishing t h i s ground work, the findings are generalized i n the s p e c i a l i z a t i o n section. Similarity of ends and anarchy The proposition being advanced i s that when anarchy i s higher, ends are of higher s i m i l a r i t y , and when anarchy i s lower, ends are of lower s i m i l a r i t y . At higher l e v e l s of anarchy ends are more si m i l a r because actors place more p r i o r i t y on rule making and rule enforcing. As discussed above i n section 2, when actors place more p r i o r i t y on rule making and rule enforcing substantive ends tend to become common, which reinforces the p r i o r i t y of rule making and rule enforcing, and results i n homogenizing ends. A concomitant of the homogenizing dynamic i s an increase i n 72 the l e v e l of anarchy. Anarchy increases as a r e s u l t of actors propensity to increase t h e i r own rule making and rule enforcing r o l e . Consider a d i f f e r e n t l i n e of reasoning which leads to the same conclusion. When ends are of higher s i m i l a r i t y the l i k e l i h o o d that a single rule maker and rule enforcer w i l l or can emerge i s reduced. If a single actor has the end of rule making and rule enforcing, other actors w i l l have the same end of rule making and rule enforcing. Not only i s th i s true by d e f i n i t i o n , but the b e l i e f that other actors w i l l abuse rule making and rule enforcing i f they are l e f t to make and enforce rules w i l l result i n opposition and resistance to the emergence of a rule maker and rule enforcer other than themselves. 1 3 0 As the l e v e l of anarchy changes from higher to lower l e v e l s , actors increasingly d i v e r s i f y t h e i r ends insofar as they are less s i m i l a r . Ends d i f f e r e n t i a t e because actors no longer put utmost p r i o r i t y on t h e i r e f f o r t s to make and enforce rules. It i s the increased opposition to the emergence of a single rule maker and rule enforcer when ends are of higher s i m i l a r i t y and the reduced concern with rule making and rule enforcing when ends are of lower s i m i l a r i t y 130 This b e l i e f i s held because we have assumed actors to be egoists. This i s sim i l a r to the r e l a t i v e gains argument made by Joseph M. Grieco i n "Anarchy and the l i m i t s of cooperation i n s t i t u t i o n a l i s m . " 73 which makes the former more anarchical than the l a t t e r . 1 3 1 The rela t i o n s h i p between the s i m i l a r i t y of ends and anarchy i s i l l u s t r a t e d below i n Table 3.2. Similarity of means and anarchy The relationship between means and anarchy i s more straightforward than between ends and anarchy. The l o g i c which underlies the relationship between the s i m i l a r i t y of 1 3 1 There may seem to be a contradiction between what makes l e v e l s of anarchies si m i l a r and that which d i f f e r e n t i a t e s them. How can i t be that i n anarchy actors reserve the rights of sovereignty, while at lower l e v e l s of anarchy centralized rule making and enforcing are s t i l l possible? The condition of anarchy, where actors reserve the r ights of sovereignty and r e t a i n for themselves license to make and enforce rules, does not preclude the p o s s i b i l i t y that actors w i l l agree to l e t others f u l f i l l these tasks e s p e c i a l l y on a temporary and ad hoc basis with their consent. To assert that actors cannot give such consent and s t i l l remain i n anarchy assumes two points. F i r s t i t assumes that sovereign actors do not have the l a t i t u d e to abstain or r e l i n q u i s h part of t h e i r duties and rights as sovereign actors. Second, i t assumes actors' i n a b i l i t y to reclaim t h e i r abstained or temporarily relinquished r i g h t s . Just because actors allow another entity, whether i t be another state or a c o l l e c t i o n of them, to make and enforce a r u l e does not denote that these actors have given up t h e i r rights to do so. The h i s t o r i c a l record provides ample evidence of such occurrences, such as the Concert of Europe and the c o a l i t i o n of states i n the Gulf War, f o r example. See K.J. H o l s t i , "Governance Without Government," for a discussion of the r i s e and f a l l of the Concert of Europe. The condition of anarchy remains at both high and low l e v e l s of anarchy because actors reserve for themselves the righ t s to rule making and rule enforcing. 74 means and the l e v e l of anarchy i s based on the c a p a b i l i t y of actors to make and enforce rules. This argument i s s i m i l a r to that made by the hegemonic s t a b i l i t y t h e o r i s t s . As noted e a r l i e r , when one actor has preponderant c a p a b i l i t y i t has the a b i l i t y to make and enforce rules. This a b i l i t y i s precluded for those with i n f e r i o r c a p a b i l i t i e s . Those with i n f e r i o r c a p a b i l i t i e s cannot enforce rules because they do not possess the means to do so. Those with preponderant c a p a b i l i t y w i l l make and enforce rules because actors are egoists. When means are of lower s i m i l a r i t y the l e v e l of anarchy i s lower because one actor has disproportional means, and therefore has the a b i l i t y to act as a rule maker and rule enforcer. On the other hand when means are of higher s i m i l a r i t y , actors have very si m i l a r c a p a b i l i t i e s , and i t i s more d i f f i c u l t f or any p a r t i c u l a r actor to play the role of a rule maker and rule enfor.cer. It i s more d i f f i c u l t f or a rule maker and rule enforcer to emerge because opposition to i t w i l l l i k e l y r e s u l t i n i t s demise by vi r t u e of actors' equal means. The inverse r e l a t i o n between, s i m i l a r i t y of means and level s of anarchy i s due to two factors. F i r s t , the equality of c a p a b i l i t i e s when means are highly s i m i l a r r e s u l t i n the e f f i c a c y of opposition to the demise of rule makers and enforcers. Second, actors oppose and r e s i s t the emergence of a single rule maker and enforcer because they are e g o t i s t i c a l . The relationship between s i m i l a r i t y of 75 means and anarchy i s i l l u s t r a t e d i n Table 3.2. Figure 3.2 Levels of Anarchy and S i m i l a r i t i e s of Ends and Means Ends Higher S i m i l a r i t y Higher Means Higher S i m i l a r i t y LEVELS OF ANARCHY Lower S i m i l a r i t y Lower Lower S i m i l a r i t y Interactive components and anarchy In configurations when ends and means are both high s i m i l a r i t y or low s i m i l a r i t y the aff e c t s of the high s i m i l a r i t y or low s i m i l a r i t y act to reinforce one another. In the two high s i m i l a r i t y configuration, the tendency of actors with high s i m i l a r i t y of ends to reject others' rule making and rule enforcing combined with t h e i r a b i l i t y to reje c t i t , as a resu l t of the high s i m i l a r i t y of means, 76 r e s u l t s i n an environment where rule making and rule enforcing i s more fragmented and where the l e v e l of anarchy i s higher. Conversely, i n the low s i m i l a r i t y configuration the r e s t r i c t i v e dynamics for rule making and rule enforcing are l e s s potent. In t h i s environment actors are both more w i l l i n g to accept rule making and rule enforcing and l e s s able to r e s i s t rule making and rule enforcing. As a r e s u l t , rule making and rule enforcing i s less fragmented, and the l e v e l of anarchy i s lower. Aside from claiming the high s i m i l a r i t y configuration i s more anarchical than the low s i m i l a r i t y configuration, the claim that the high s i m i l a r i t y configuration i s the most anarchical and that the low s i m i l a r i t y configuration i s least anarchical can be sustained. The high s i m i l a r i t y configuration i s most anarchical because there i s no component which supports rule making and rule enforcing. Both the ends and means components of high s i m i l a r i t y foster and support each others' tendency td reject rule making and rule enforcing. Conversely, i n the low s i m i l a r i t y configuration both ends and means components are more accepting of rule making and rule enforcing, and each supports the other's tendency. These mutually r e i n f o r c i n g tendencies away from rule making and rule enforcing make the two high s i m i l a r i t y configuration the most anarchical, and 77 the absence of the tendencies i n the two low similarity-configuration the least anarchical. Specialization S p e c i a l i z a t i o n occurs with the d i v i s i o n of what was once similar, whole, or single, into that which i s d i f f e r e n t , p a r t i c u l a r , or multiple, where the d i v i s i o n creates d i f f e r e n t roles f o r actors, who perform the 'new' tasks, and where these new task f u l f i l l e r s have d i f f e r e n t ends. 1 3 2 S p e c i a l i z a t i o n involves the r e d e f i n i t i o n of ends and means from general and ubiquitous to s p e c i f i c and unique. More s p e c i a l i z a t i o n indicates a decrease i n the s i m i l a r i t y of actors' ends and means, and les s s p e c i a l i z a t i o n indicates an increase i n the s i m i l a r i t y of actors' ends and means. The dynamic of change i n s i m i l a r i t i e s of ends and means can be thought of i n terms of actor s p e c i a l i z a t i o n . Higher s i m i l a r i t y of ends and means involves less actor s p e c i a l i z a t i o n and lower s i m i l a r i t y of ends and means involves more s p e c i a l i z a t i o n . Using the linkage between s p e c i a l i z a t i o n and the s i m i l a r i t i e s of ends and means, the 1 3 2 Buzan, L i t t l e , and Jones define f u n c t i o n a l l y s i m i l a r units as actors who perform the whole range of functions while fu n c t i o n a l l y d i f f e r e n t i a t e d units perform a l i m i t e d range of function, Logic of Anarchy, 38. 78 re l a t i o n s h i p between s p e c i a l i z a t i o n and anarchy can be s p e c i f i e d as inverse. At a higher l e v e l of anarchy actors are less s p e c i a l i z e d and at a lower l e v e l of anarchy actors are more s p e c i a l i z e d . 1 3 3 S p e c i a l i z a t i o n by actors at lower l e v e l s of anarchy i s possible because actors re l i n q u i s h part of t h e i r r o l e as rule maker and rule enforcer. 1 3 4 Actors r e l i n q u i s h part of th i s i n so f a r as t h e i r role as rule maker and rule enforcer i s less dominant. Because of t h i s abdication actors can have increased differences i n interests, tasks, goals, and c a p a b i l i t i e s . 1 3 5 In the areas of abdicated rule making and rule enforcing, actors can and w i l l d i f f e r e n t i a t e t h e i r ends and means. Actors can d i f f e r e n t i a t e ends and means because they no longer have the dominant imperative to contend as the rule maker and rule enforcer (which requires the generalization of ends and means). Actors w i l l d i v e r s i f y 1 3 3 This l o g i c i s . s i m i l a r to the "sameness" dynamic discussed by Waltz Theory, 127; and Layne, "The unipolar i l l u s i o n , " 140-141. 1 3 4 This l i n e of argument has been informed by Waltz Theory, and Buzan, L i t t l e , and Jones, Logic of Anarchy. 1 3 5 An example of such an abdication i s the E.C. member state's abdication of some economic r i g h t s . See Philpott, " P l u r i l a t e r a l i s m , s tructural d i f f e r e n t i a t i o n and functional c o n f l i c t . " 79 t h e i r ends and means i n order to pursue t h e i r role as egoists. Increased d i f f e r e n t i a t i o n of interests, tasks, goals, and c a p a b i l i t i e s w i l l occur when the l e v e l of anarchy-decreases because actors do not, cannot, or w i l l not seek to act as the supreme rule maker and rule enforcer. When actors are "free" to ignore rule making and rule enforcing i n p a r t i c u l a r areas, actors can pursue more sp e c i a l i z e d i n t e r e s t s and goals. F u l l e xploitation of opportunities require that rule making and rule enforcing be abdicated or eroded, and that tasks and c a p a b i l i t i e s be s p e c i a l i z e d i f e f f i c a c y i s to be maximized. And, because actors are egoists, they w i l l pursue interests and goals which are most s e l f s a t i s f y i n g , which i n turn w i l l r e s u l t i n increased s p e c i a l i z a t i o n i n areas of abdication. Hence the tendency w i l l be for actors to d i v e r s i f y t h e i r interests and goals.. This tendency w i l l i n turn d i v e r s i f y tasks and c a p a b i l i t i e s i n areas where actors have abdicated or eroded supreme authority. Crudely put, a d i v i s i o n of labor w i l l occur. S p e c i a l i z a t i o n w i l l occur because some actors w i l l concentrate t h e i r e f f o r t s i n one area, and others, who wish to avoid the erosion of s e l f i n t e r e s t s brought about by competition and threats to the e f f i c a c y of t h e i r endeavor, w i l l concentrate i n other areas. This dynamic process thus fosters increasing s p e c i a l i z a t i o n of actors' tasks, goals, interests, and d i f f e r e n t i a t i o n of 80 c a p a b i l i t i e s as the l e v e l of anarchy decreases and impedes s p e c i a l i z a t i o n as the l e v e l of anarchy increases. The relationship between s p e c i a l i z a t i o n and anarchy provides us with t h e o r e t i c a l l y grounded empirical expectations of the relationship between l e v e l s of anarchy and the s i m i l a r i t y of ends and means. As the l e v e l of anarchy increases the s i m i l a r i t y of actors' ends and means increases and the l e v e l of s p e c i a l i z a t i o n decreases. As the l e v e l of anarchy decreases the s i m i l a r i t y of actors ends and means decreases and the l e v e l of s p e c i a l i z a t i o n increases. 1 3 6 An example of the relationship between s p e c i a l i z a t i o n and the s i m i l a r i t i e s of ends and means, and rule making and rule enforcing can be found i n the "development" from the Imperial Tutor or the Single Room Schoolhouse teacher to contemporary teachers who s p e c i a l i z e at various grade l e v e l s and i n d i f f e r e n t d i s c i p l i n e s . 1 3 7 Teachers i n d i f f e r e n t 1 3 6 At lower leve l s of anarchy subjects d i v e r s i f y t h e i r ends. While they may s t i l l r e t a i n as an end rule making and rule enforcing, i t no longer predominates other ends. Other ends such as development, trade, and welfare, for example, may gain prominence over t h e i r previous status. While ends among subjects may d i f f e r e n t i a t e , the condition of anarchy remains as long as each subject retains f o r i t s e l f the right to make and enforce rules. 1 3 7 These examples are meant to describe s p e c i a l i z a t i o n . The use of the individual l e v e l of analysis does not indicate a v i o l a t i o n of the state as the main actor and subject asserted e a r l i e r . 81 d i s c i p l i n e s and at d i f f e r e n t l e v e l s have d i f f e r e n t ends. Math teachers seek to impart mathematical knowledge, language teachers impart reading and writing s k i l l s , while elementary school teachers imprint basic and general knowledge and uni v e r s i t y professors introduce, and sometimes impart, a n a l y t i c a l c a p a b i l i t i e s and s p e c i f i c knowledge i n sub-disciplines. Teachers also have d i f f e r e n t means. Geographers have maps, chemists test tubes, and anatomists cadavers. States, l i k e educators, s p e c i a l i z e when they emphasize ce r t a i n interests and c a p a b i l i t i e s . For example, when the United States decided to renew development and deployment of cruise missiles i n the early 1970s, i t did so with two sp e c i a l i z e d i n t e r e s t s . The f i r s t was to acquire a weapon system outside the purview of the Strategic Arms Limitation Talks (SALT). The second was to b u i l d a weapon system which would contribute to the maintenance of a force structure balance i n the event of unrequited Soviet deployment of an Anti B a l l i s t i c M i s s i l e (ABM) system. 1 3 8 These s p e c i a l i z e d i n t e r e s t s lead the United States to develop and deploy cruise m i s s i l e s . 1 3 9 1 3 8 For the U.S. pos i t i o n i n ABM negotiations see chapter 4 i n Steve Weber, Cooperation and Discord in U.S.-Soviet Arms Control (Princeton, N.J.:Princeton, 1991). 1 3 9 Because cruise missiles have very small radar cross-section and tr a v e l at altit u d e s under 100 feet they are extremely d i f f i c u l t to intercept and are less vulnerable 82 In t h i s example the U.S. sought increased c e n t r a l i z e d rule making and rule enforcing as manifested i n SALT and ABM agreements. The U.S.'s speci a l i z e d means was the cruise m i s s i l e and the Soviets spe c i a l i z e d means was the ABM system. The Soviets sought ABM c a p a b i l i t i e s while the U.S. sought cruise m i s s i l e c a p a b i l i t i e s . As an i l l u s t r a t i o n , we have a case with low s i m i l a r i t y of means and ends which resulted i n exogenous rule making and rule enforcing i n the form of SALT and ABM agreements. 1 4 0 The s p e c i a l i z a t i o n argument provides a measure f o r anarchy for the high and low s i m i l a r i t y configuration. Anarchy i s highest when both ends and means are of high s i m i l a r i t y , and anarchy i s lowest when both ends and means are of low s i m i l a r i t y . It t e l l s us nothing about the mixed configurations of high and low s i m i l a r i t i e s . That i s , i t to an ABM system. Additionally, due to i t s slower subsonic speed the cruise m i s s i l e i s s t r a t e g i c a l l y a second s t r i k e weapon and i s therefore a s t a b i l i z i n g system. See Rose E. Gottemoeller "Land Attack Cruise M i s s i l e s " (Adelphi Papers #226 Winter 1987/88) for more on cruise m i s s i l e s . 1 4 0 Paul Schroder, using a 19th century example, argued along s i m i l a r l i n e s . He argues that i n the Concert of Europe "states and t h e i r leaders ... used the device of functional d i f f e r e n t i a t i o n to remove problem areas ... by recognizing and sanctioning p a r t i c u l a r roles and functions fo r p a r t i c u l a r states." Paul Schroder, "Anarchy i n the 19th Century,"in Perils of Anarchy: Contemporary Realism and International Security, ed. Michael E. Brown, Sean M. Lynn-Jones, and Steve E. M i l l e r (Cambridge:MIT Press 1995), 440-441. 83 does not elucidate the r e l a t i v e importance of ends and means v i s - a - v i s the l e v e l of anarchy. In order to make claims about the r e l a t i v e importance of ends and means the ends-drives-means argument w i l l prove to be u s e f u l . 1 4 1 The Ends-Drives-Means Argument The argument which j u s t i f i e s ranking the l e v e l of anarchy of the mixed configurations of the s i m i l a r i t i e s of ends and means i s the ends-drives-means argument. The argument responds to the question, what i s more important.in changes between the two middle lev e l s of anarchy, ends or means? Put i n terms of rule making and rule enforcing, are actors' means or actors' ends the more important influence on changes between a no-rule condition and a condition where rules exists at the d i s c r e t i o n and with the consent of actors? The ends-drives-means argument asserts that ends exert a more important influence on the establishment and maintenance of rules than means and that ends are therefore more important determinants of l e v e l s of anarchy than means. 1 4 1 I would note that I am not making the claim that at high l e v e l s of anarchy there i s no s p e c i a l i z a t i o n or that low l e v e l s of anarchy there i s no generalization. I am making the weaker claim that the amount of s p e c i a l i z a t i o n at high l e v e l s of anarchy i s smaller and the amount of generalization i s greater, and that at low l e v e l s of anarchy s p e c i a l i z a t i o n i s greater and generalization i s smaller. 84 There are two supporting elements of the ends-drives-means argument. The f i r s t i s that actors' ends are more important than means fo r both the establishment and maintenance of rules. The second i s that ends a f f e c t means more than means af f e c t ends. I w i l l address each of these elements i n order. In order to demonstrate that ends are more important than means l e t us think of l e v e l s of anarchy as the amount of rules or rule governed behavior: the higher the l e v e l of anarchy the less the amount of rule governed behavior, and the lower the l e v e l of anarchy the greater the amount of rule governed behavior. 1 4 2 In terms of anarchy, i n Figure 3.1 anarchy ranges between the pure anarchy at the anarchy pole and q u a l i f i e d anarchy just before the government range. Pure anarchy i s the condition where there i s an absence of rules, while lower q u a l i f i e d l e v e l s of anarchy are conditions where rules may exists but actors r e t a i n the right to abrogate them. To make thi s argument we need to consider the role of ends and means and the purpose and source of rules. Rules emerge and are maintained to increase p r e d i c t a b i l i t y and to 1 4 2 This i s consistent with the d e f i n i t i o n of anarchy advanced e a r l i e r . 85 increase a consistent order. 1 4 3 As such, actors must have an i n t e r e s t to create and maintain them. In the absence of such an intere s t rules w i l l not exist or w i l l not be consistent even i f actors have the c a p a b i l i t y to enforce them. Take for example rules which govern vehicles and pedestrians. Rules exist for vehicles which do not exis t for pedestrians because there i s an interest i n rules f o r the former and no such interest i n rules for the l a t t e r . Many vehicular rules, such as which side of the road vehicles occupy, t r a f f i c signs and signals, and so forth, e x i s t and are maintained due to drivers' interest, and not the c a p a b i l i t y that drivers or the pol i c e have to enforce them. Pedestrians have no rule about which side of a sidewalk they occupy because there exists no intere s t i n esta b l i s h i n g or maintaining such a rule. For both pedestrians and vehicles a rule i s present because there exists an interest i n having a rule, and a rule i s absent because there i s no interest i n having a rule. The fact that authorities undoubtedly have the capacity to enforce vehicular and pedestrian rules does not a priori a f f e c t the presence or absence of a rule . 1 4 3 B u l l , Anarchical Society, 53; Lynn H. M i l l e r , Global Order: Values and Power in International Politics (Boulder CO:Westview 1985), 4. 86 At a basic and fundamental l e v e l rules emerge from actors' interests and not from t h e i r c a p a b i l i t y for rule enforcement. While c a p a b i l i t y i s an important factor i n the a p p l i c a t i o n of the rule, an antecedent inter e s t for a rule a p p l i c a t i o n i s more important than the c a p a b i l i t y to enforce a r u l e . Thus, the r e l a t i v e importance of ends and means v i s - a - v i s changes i n rules between higher and lower l e v e l s of anarchy (or between higher and lower l e v e l s of rule making and rule enforcing) must place p r i o r i t y on ends rather than means. Means are not, and have not been excluded; they are of secondary importance to rule creation and maintenance. 1 4 4 The preceding discussion suggests that rules emerge and are maintained because actors have an in t e r e s t to create and maintain them. This leads to the second element of the end-drives -means argument, that ends influence means more than means do ends. In the a c t i v i t y of rule making and rule enforcing, one can conclude, as we have, that an i n t e r e s t for rule making and rule enforcing i s more important, f o r the a c t i v i t y , than the c a p a b i l i t y to make and enforce rules. Interest i s more important i n the sense that i t i s i n t e r e s t which guides c a p a b i l i t y . Interest dictates how c a p a b i l i t y i s used and what i t does. From t h i s perspective, i n t e r e s t 1 4 4 In the model means are not excluded. They form an i n t e g r a l part of the model. However, means are subordinate to ends i n the c a l c u l a t i o n of the l e v e l of anarchy. 87 influences c a p a b i l i t y more than c a p a b i l i t y does i n t e r e s t . If one thinks of c a p a b i l i t y as means and inte r e s t s as ends, as we have, the conclusion i s that ends influence means more than means influence ends. While c a p a b i l i t y i s an important factor for opposed rule making and enforcing, there must be an antecedent in t e r e s t to use ca p a b i l i t y . Capability i n the absence of inte r e s t i s moot. I do not suggest that c a p a b i l i t y i s unimportant, as i t s incl u s i o n i n the model indicates. I only claim that c a p a b i l i t y i s subordinate to interest, as far as measuring anarchy i s concerned. Ranking Anarchy To t h i s point I have reasoned that the l e v e l of anarchy i s highest when ends and means are of high s i m i l a r i t y , and lowest when ends and means are of low s i m i l a r i t y ( s p e c i a l i z a t i o n argument). I have also reasoned that anarchy i s higher when ends are of high s i m i l a r i t y and means are of low s i m i l a r i t y than when ends are of low s i m i l a r i t y and means are of high s i m i l a r i t y (ends-drives-means argument). Since the value of the anarchy i s based on the combined values of the s i m i l a r i t i e s of ends and means which are dichotomous ordinal variables, the range of values for the anarchy variable i s 1 to 4. When the two arguments are 88 synthesized, the resu l t i s a single framework to measure anarchy. The s p e c i a l i z a t i o n argument provides f o r the ranking of the highest and lowest l e v e l s of anarchy and the ends-drives-means argument for the middle two l e v e l s . The framework sp e c i f i e s the following ranking f o r configurations of s i m i l a r i t i e s of ends and means i n descending order of anarchic magnitude: anarchy=4 ends are of high s i m i l a r i t y and means are of high s i m i l a r i t y ; anarchy=3 ends are of high s i m i l a r i t y and means of low similarity,-anarchy=2 ends are of low s i m i l a r i t y and means of high s i m i l a r i t y ; anarchy=l ends are of low s i m i l a r i t y and means are of low s i m i l a r i t y . These rankings can also be expressed i n the two-by-two matrix i l l u s t r a t e d i n Figure 3.3. Figure 3.3 The Model of Anarchy and Rank Ordering Levels of Anarchy Ends Ends High Low Si m i l a r i t y S i m i l a r i t y Means High {4} {2} Similarity- highest medium low Means Low {3} {1} Si m i l a r i t y medium high lowest 89 To t h i s point I have explained how and why ends and means and changes i n ends and means are related to l e v e l s of anarchy. I have concluded that the anarchy scores represents, i n descending order, l e v e l s of anarchy, and I have explained why one anarchy score i s more or less anarchical than another. Note that t h i s ranking of l e v e l s of anarchy has been developed, defined, and argued separate from and independent of notions of c o n f l i c t and cooperation. As noted e a r l i e r , t h i s was necessary because I w i l l argue that l e v e l s of anarchy and level s of cooperation and c o n f l i c t are strongly i n t e r r e l a t e d . While the above discussion i s lim i t e d to a two actor case, n-actor cases can also be evaluated using the s p e c i a l i z a t i o n and ends-drives-means arguments. Summation This chapter established the conceptual base of the model. It presented a d e f i n i t i o n and model of anarchy which y i e l d s ordinal measures for level s of anarchy. Anarchy was defined i n terms of rule making and rule enforcing. Then a l i n k between rule making and rule enforcing and ( d i s ) s i m i l a r i t y of ends and means was made. This l i n k provided for a measure of anarchy based on s i m i l a r i t i e s of ends and means. F i n a l l y rule making and rule enforcing propensities were related to four configurations of 90 ( d i s ) s i m i l a r i t y of ends and means, and an anarchy variable with four ordinal l e v e l s was constructed. Systemic variables are indicators of systems. A System has been defined as a "sets of elements standing i n i n t e r a c t i o n , " 1 4 5 "a set of variables so related . . . that describe behavioral r e g u l a r i t i e s , 1 , 1 4 6 and a "structure and i n t e r a c t i n g u n i t s . " 1 4 7 In his recent book (1997) on in t e r n a t i o n a l systems and t h e i r e f f e c t s , J e r v i s defines a systems as "a set of units or elements ... interconnected so that changes i n some elements or t h e i r r e l a t i o n s produce changes i n other parts of the system." 1 4 8 These d e f i n i t i o n s suggests that a systemic variable must be exogenous to any single actor, since i t i s the r e l a t i o n between or among rather than an actor's c h a r a c t e r i s t i c which defines the a system. The measure of anarchy developed f u l f i l l s both these two conditions. It i s both a measure of r e l a t i o n and i t i s exogenous to any single actor. F i r s t , the anarchy variable i s a measure of r e l a t i o n a l p o s i t i o n of actors which i s based 1 4 5 Ludwig vow Bertalanffy, "General Systems Theory, " General Systems 1 (1956):3. 1 4 6 Kaplan, System, 4. 1 4 7 Waltz, Theory, 79. 1 4 8 Robert Je r v i s , System Effects: Complexity in Political and Social Life (Princeton, N.J.: Princeton University Press, 1997), 6. 91 on some aggregation of actor c h a r a c t e r i s t i c s . More prec i s e l y , i t i s an in d i c a t i o n of the configuration of the s i m i l a r i t i e s of c a p a b i l i t y and int e r e s t s . Second, the anarchy variable i s exogenous to any single actor. That i s , no actor can alone determine i t s values. Let me address t h i s i n greater d e t a i l , beginning with the construction and operation of the anarchy variable. F i r s t , following Waltz and Kaplan 1 4 9 and many other system theorists, the operation anarchy i s based on r e l a t i o n a l aspects derived from an aggregation of actor c h a r a c t e r i s t i c s . As with balance of power and p o l a r i t y theories of international relations, which are t r a d i t i o n a l l y taken to be a analyses based on systemic v a r i a b l e s , 1 5 0 the power of actors r e l a t i v e to other actor(s) along with the inter e s t s of each actor r e l a t i v e to other actors are the foundation of the anarchy variable. Second, while individual actor c h a r a c t e r i s t i c s c l e a r l y contribute to the composition of anarchy, no single actor can determine i t s value. This i s a d i r e c t r e s u l t of the r e l a t i o n aspect of the measure which occurs when actor c h a r a c t e r i s t i c s are aggregated. While t h i s formulation and s p e c i f i c a t i o n of the r e l a t i o n i s mutually (although perhaps not equally) determined by the actors i n the system, and 1 4 9 Waltz, Theory; and Kaplan, System. 1 5 0 Levy, "The causes of war: a review." 92 while an actor can influence the value of t h i s systemic variable, none can alone determine i t s value. The anarchy variable developed here f u l f i l l s these conditions. Not only i s the measure of anarchy consistent with the idea of a systemic variable established i n the other research noted above, but i t i s also systemic because i t i s based on and derived from d i s t r i b u t i o n s , and because these d i s t r i b u t i o n s are exogenous to any single actor. Having established anarchy as variable and l e v e l s of anarchy, the next chapter addresses how the variable can be made operational and used i n empirical analyses. 93 CHAPTER 4 METHODOLOGY This chapter outlines the methodology employed to test the hypotheses. The general hypotheses are (1) as the l e v e l of anarchy increases the l e v e l of cooperation decreases and the l e v e l of c o n f l i c t increases, and (2) that as the l e v e l of anarchy decreases the l e v e l of cooperation increases and the l e v e l c o n f l i c t decreases. Formally, t h i s i s noted i n figure 4.1. Figure 4.1 General Hypothesizes general hypothesis 1: that as anarchy increases, cooperation between or among states decreases and c o n f l i c t between or among states increases general hypothesis 2. that as anarchy decreases, cooperation between or among states increases and c o n f l i c t between or among states decreases 94 Note that the two general hypothesis are restatements of the four derivative hypotheses expressed i n chapter 1. From these four hypotheses three empirical hypotheses are derived. The lineage from the anarchy tenet to the empirical hypotheses i s as follows. F i r s t , the general hypotheses were derived from anarchy tenet. Then, according to the measure of anarchy developed i n chapter 3, s i m i l a r i t i e s of ends and means replaced anarchy i n the hypotheses. F i n a l l y , the general hypotheses were then d i s -aggregated into constituent parts, which yielded the s i x empirical hypotheses. The empirical hypotheses are noted i n below i n Figure 4.2. The three empirical hypotheses are i n Figure 4.2 Empirical Hypotheses HI. that for any states as the means of the states become more s i m i l a r the l e v e l of c o n f l i c t w i l l be higher and the l e v e l of cooperation w i l l be lower than when the means of the states are less similar; H2. that for any states as the ends of the states become more s i m i l a r the l e v e l of c o n f l i c t w i l l be higher and the l e v e l of cooperation w i l l be lower than when the ends of the states are less similar; H3. that for any states as the means of the states become more s i m i l a r and as the ends of the states become more si m i l a r the l e v e l of c o n f l i c t w i l l be higher than i f only ends or means are simi l a r . 95 a testable form which allow an assessment of both the v a l i d i t y and accuracy of the anarchy tenet. Two d i f f e r e n t types of relationships can support the hypotheses. One i s a di r e c t relationship, and the other i s a concordant relationship. A di r e c t r e l a t i o n s h i p i s when le v e l s of s i m i l a r i t y of ends, s i m i l a r i t y of means, and anarchy concur with lev e l s of cooperation and c o n f l i c t . A concordant relationship i s when le v e l s of cooperation and c o n f l i c t are equal to or larger than l e v e l s of s i m i l a r i t y of ends, s i m i l a r i t y of means, and anarchy. Two p a r a l l e l analyses w i l l test the v a l i d i t y and accuracy of the hypotheses. Each analysis w i l l follow the same methodology and w i l l use the same test s t a t i s t i c s , but w i l l use d i f f e r e n t data. One analysis w i l l use variables from the Behavioral Correlates of War (BCOW) data set as reported i n Russell Leng's International Crisis Behavior and the other w i l l use variables from International Crisis Behavior (ICB) data set coll e c t e d by Michael Brecher and Jonathan Wilkenfeld. The remainder of the chapter i s divided into three sections. Section 1 addresses the data: how variables are made operational and what they measure. Section 2 presents the s t a t i s t i c a l tests which assess the hypotheses. And section 3 presents how tests are applied, and expected . res u l t s of each test. 96 Data This section outlines the variables used to test the hypotheses. The order of discussion f i r s t addresses the c o n f l i c t variables, then the cooperation variables, and f i n a l l y the anarchy variable. BCOW data i s addressed p r i o r to ICB data. The s i m i l a r i t y of ends and s i m i l a r i t y of means are discussed when anarchy i s addressed. Figure 4.3 i l l u s t r a t e s these variables. BCOW Conflict Variables Three BCOW variables are used to measure c o n f l i c t . They are the h o s t i l i t y magnitude, h o s t i l i t y escalation, and c o n f l i c t index variables. H o s t i l i t y magnitude represents the "overall l e v e l of h o s t i l i t y , 1 , 1 5 1 and h o s t i l i t y e scalation represents "the d i r e c t i o n and magnitude of the slope and mix of cooperation and c o n f l i c t . " 1 5 2 Both variables are measured on a six point scale which ranges from +3 to -3, where +3 indicates the maximum l e v e l of h o s t i l i t y and -3 represent the maximum l e v e l of 1 5 1 Russell Leng, Interstate Crisis Behavior 1816-1980; Realism Versus Reciprocity (New York:Cambridge, 1993), 69. 1 5 2 Leng, Interstate Crisis Behavior, 70. 97 Figure 4.3 Operational, BCOW, ICB Proxy Variables Anarchy, Cooperation, and Conflict Variable BCOW Variable ICB Variable Ends Interest Means Capability Anarchy Interest & Capability Status Conflict (1) Mean Magnitude of Escalation (2) Mean Rate of Escalation (3) C o n f l i c t Index Cooperation (1) Reciprocity Direction (2) Reciprocity Distance Issues Power Status Issues & Power (1) Intensity of Violence (2) Importance of Violence (3) Overall Violence (4) C o n f l i c t Index (1) C r i s i s Management Techniques (2) Timing of Violence (3) Cooperation Index (3) Cooperation Index 98 cooperation. 1 5 3 The higher the value of h o s t i l i t y magnitude and h o s t i l i t y escalation, the more c o n f l i c t i v e the r e l a t i o n s h i p . Conversely, the smaller the value of the score the more cooperative the relationship. The benefit of using both h o s t i l i t y magnitude and h o s t i l i t y escalation i s that each r e f l e c t s d i f f e r e n t aspects of c o n f l i c t . H o s t i l i t y magnitude 1 5 4 r e f l e c t s how c o n f l i c t i v e the r e l a t i o n s h i p i s , while the rate of e s c a l a t i o n 1 5 5 r e f l e c t s how quickly states increased or decreased h o s t i l i t y l e v e l s . A higher escalation rate indicates both faster and larger changes i n l e v e l s of c o n f l i c t than a smaller escalation rate. A lower score indicates the reverse, smaller and slower changes i n c o n f l i c t . The sign of the h o s t i l i t y escalation score indicates whether the sides are escalating ( i f the sign i s 1 5 3 Leng, Interstate Crisis Behavior, 70. 1 5 4 H o s t i l i t y magnitude i s obtained by taking the weekly h o s t i l i t y scores ( h o s t i l i t y scores during a seven day period) for each side i n a c r i s i s , summing them, and then taking the mean value of each side's h o s t i l i t y score for the duration of the c r i s i s and adding them together. 1 5 5 H o s t i l i t y escalation scores are obtained by c a l c u l a t i n g the difference between weekly h o s t i l i t y scores of each side,- that i s one side's h o s t i l i t y score at one time point i s subtracted from i t s h o s t i l i t y score at the previous time point. The c a l c u l a t i o n i s , i n Leng's words "from one seven day i n t e r v a l to the next" (Leng, Interstate Crisis Behavior, 70). He then takes the mean value of each side's differenced h o s t i l i t y scores for each side and f i n a l l y sums the two mean values to obtain the rate of escalation score. 99 positive) or de-escalating (when the sign i s negative). A h o s t i l i t y escalation score of +4.8, for example, indicates that a high l e v e l of escalation. A h o s t i l i t y score of -.71 indicates a smaller l e v e l of de-escalation. The c o n f l i c t index variable i s a general in d i c a t o r the l e v e l of c o n f l i c t . It i s calculated by summing the h o s t i l i t y magnitude and h o s t i l i t y e s c a l a t i o n . 1 5 6 Since the c o n f l i c t index variable i s the sum of h o s t i l i t y magnitude and h o s t i l i t y escalation, i t r e f l e c t s both the l e v e l of c o n f l i c t and speed at which i t was achieved. Taking both of these elements together into account provides f o r a more accurate o v e r a l l measure of c o n f l i c t than i f either were used alone. One consequence of combining h o s t i l i t y magnitude and h o s t i l i t y escalation into an index variable i s that such a combination equates rate and magnitude i n the c a l c u l a t i o n of the c o n f l i c t index variable. Thus, a high c o n f l i c t index score could indicate either a higher rate of escalation and a lower magnitude of h o s t i l i t i e s or a lower rate of escalation and a high magnitude of h o s t i l i t y . The i n a b i l i t y of the c o n f l i c t variable to d i f f e r e n t i a t e between these two cases i s not as unreasonable as i t may. seem to be since i t does not a l t e r the c o n f l i c t variable's a b i l i t y to assess d i f f e r e n t l e v e l s of c o n f l i c t . 1 5 6 Leng himself uses t h i s c a l c u l a t i o n and index indicator. 100 Even though the c o n f l i c t index variable does not d i f f e r e n t i a t e between cases where escalation i s high and magnitude i s low and cases when escalation i s low and magnitude i s high, i t can determine higher and lower l e v e l s of c o n f l i c t . In the high escalation-low magnitude case c o n f l i c t increases quickly but reaches a lower magnitude. In t h i s case the increase i n h o s t i l i t y follows a pattern s i m i l a r to a step-level-function, that i s an quick increase i n the l e v e l of c o n f l i c t and then the l e v e l i n g off of the rate of increase to zero. In the low escalation-high magnitude case the rate of c o n f l i c t growth would be constant at a lower rate.of change, but the eventual magnitude of h o s t i l i t y would be higher than i n the high escalation-low magnitude case. ICB conflict variables To measure the l e v e l of c o n f l i c t three variables from the ICB data set are employed: the i n t e n s i t y of violence, the extent of violence, and the importance of violence. Each variable measures a d i f f e r e n t facet of c o n f l i c t . The extent of violence variable measures the "the extent of violence i n an international c r i s i s as a whole." 1 5 7 The 1 5 7 Jonathan Wilkenfeld and Michael Brecher, "Code Book for ICB1, System Level Data" (College Park,MD:University of Maryland, 1989, photocopied), 45. 101 i n t e n s i t y of violence variable measure the i n t e n s i t y of violence i n a c r i s i s , and the importance of violence measures the importance of violence i n a c r i s i s . Each variable i s measured on a four point ordinal scale where the higher the score the more violence, the more important the violence, and the more intense the violence. The higher the score the more c o n f l i c t i v e i s the behavior, and the lower the score the less c o n f l i c t i v e the behavior. The coding of these variables are l i s t e d below i n Table 4.4 Figure 4.4 ICB Conflict Variables' Codings Intensity of Violence and Extent of Violence 1 no violence 2 minor clashes 3 serious clashes 4 f u l l scale war Importance/Centrality of Violence as a C r i s i s Management Technique 1 no violence 2 minor Role 3 important 4 pre-eminent 102 In addition to these variables a c o n f l i c t index v a r i a b l e w i l l also be included. This index variable i s the sum of the constituent c o n f l i c t variables divided by three. BCOW Cooperation Variables Three variables from the BCOW data set are used to measure cooperation. They are the r e c i p r o c i t y distance and re c i p r o c i t y d i r e c t i o n variables, and t h e i r sum, the cooperation index variable. As with the BCOW c o n f l i c t variables, the r e c i p r o c i t y variables' are calculated on the same six-point scale. Leng conceives of r e c i p r o c i t y as being elements of both distance and di r e c t i o n . The r e c i p r o c i t y distance scores indicates the distance (arithmetic difference) between the two sides h o s t i l i t y scores. 1 5 8 The r e c i p r o c i t y d i r e c t i o n 1 5 8 To obtain the r e c i p r o c i t y distance score Leng takes the difference between the two sides h o s t i l i t y scores at each time point. Then Leng calculates the sum of differences between the two sides' weekly h o s t i l i t y scores over the course of the c r i s i s , and takes the mean value of th i s sum. This i s a d i f f e r e n t procedure from that used to obtain the h o s t i l i t y escalation score. The h o s t i l i t y e scalation score i s calculated by taking the average of the differences between one side's h o s t i l i t y score at one week i n t e r v a l s . The difference taken i n the r e c i p r o c i t y distance score i s between one side's escalation score and the other side's escalation score i n the same week. The larger the value of the r e c i p r o c i t y distance score the greater the distance between the two sides h o s t i l i t y scores, thereby i n d i c a t i n g a lower l e v e l of r e c i p r o c i t y . For further 103 score indicates the degree to which the two sides are moving toward or away from one another. The r e c i p r o c i t y d i r e c t i o n score i s an in d i c a t i o n of the degree to which the two sides' l e v e l s of h o s t i l i t y are moving away from or toward one another. 1 5 9 High r e c i p r o c i t y i s when there i s a small difference between h o s t i l i t y scores where the h o s t i l i t y scores have the same dir e c t i o n , i n the case where they are both increasing or decreasing. 1 6 0 A low l e v e l of discussion see Leng, Interstate Crisis Behavior, 71. 1 5 9 Leng obtains the r e c i p r o c i t y d i r e c t i o n score by taking the difference of h o s t i l i t y scores for each side from one time point to another. The difference of each side's h o s t i l i t y score between t n and t n. x i s calculated. This indicates the congruence of change between one side and the other i n the week. These scores are then translated into absolute values. Leng then takes absolute values to negate the canceling e f f e c t of summing d i f f e r e n t signed values. Then, the difference between the absolute values of each side i s calculated for each i n t e r v a l time point and the sum of the difference i s taken. F i n a l l y , the mean value of the difference scores i s taken and standardized i n order to y i e l d a single measure of the concurrence of each sides changes i n h o s t i l i t y . A higher value of the r e c i p r o c i t y d i r e c t i o n variable indicates that the differenced h o s t i l i t y scores are further from one another and that one side i s , r e l a t i v e to the other side, increasing i t s l e v e l of h o s t i l i t y . A smaller value of the r e c i p r o c i t y d i r e c t i o n variable indicates that the differenced h o s t i l i t y scores are closer to one another and that the two sides h o s t i l i t y scores are approaching one another thereby i n d i c a t i n g a greater l e v e l of r e c i p r o c i t y . For further discussion see Leng, Interstate Crisis Behavior, 72-73. 1 6 0 Leng, Interstate Crisis Behavior, 71 104 r e c i p r o c i t y i s when there i s a large difference and the d i r e c t i o n i s not the same. ICB cooperation variables Two ICB variables and an index variable are be u t i l i z e d . Each variable i s from the ICB "system data," and r e f l e c t s cooperation i n the c r i s i s as a whole. The constituent variables are the c r i s i s management techniques and the timing of violence variables, and the c o n f l i c t index variable i s t h e i r average. The c r i s i s management techniques variable measures the most c o n f l i c t i v e c r i s i s management technique used by a c r i s i s actor. There are eight values for c r i s i s management techniques which range from (1) negotiations to (2) adjudication or a r b i t r a t i o n , (3) mediation, (4) multiple non-violent, (5) non-military, (6) non-violent m i l i t a r y , (7) multiple including violent, and (8) vio l e n t c r i s i s management techniques. These eight categories have been collapsed into four c l a s s i f i c a t i o n s which r e f l e c t four l e v e l s of cooperative behavior. Negotiation, adjudication/arbitration, and mediation are c l a s s i f i e d as most cooperative. Multiple non-violent and non-military are c l a s s i f i e d as medium-high cooperative, while non-violent m i l i t a r y i s c l a s s i f i e d medium-low cooperative. The f i n a l two categories which include violence are c l a s s i f i e d as 105 le a s t cooperative. Ori g i n a l and recoded values are l i s t e d i n Table 4.5. Figure 4.5 Recoding of Crisis Management Techniques Into Levels of Cooperation o r i g i n a l coding (1) negotiations (2) adjudication or a r b i t r a t i o n (3) mediation (4) multiple non-violent (5) non-military (6) non-violent m i l i t a r y (7) multiple including violent (8) v i o l e n t c r i s i s management techniques recoding most cooperative techniques medium high cooperative techniques medium low cooperative techniques least cooperative techniques The second variable which measures cooperation i s the timing of violence variable. In i t s o r i g i n a l form the timing of violence variable i d e n t i f i e s the chronological sequence i n which violence occurred i f i t did. It has four values which range from (1) no violence to (2) violence p r i o r to the c r i s i s , (3) violence triggered the c r i s i s , and (4) violence c r i s i s i n i t i a t i o n . These values have been recoded into a variable which indicates the l e v e l of 106 cooperation, with values which range from more cooperative to l e s s cooperative. The l o g i c i s that the longer i t took for violence to occur, the more cooperative the behavior of c r i s i s actors. The most cooperative c r i s i s i s one i n which no violence occurred, followed i n reverse chronological order of the occurrence of violence: violence a f t e r c r i s i s i n i t i a t i o n , violence triggered c r i s i s , violence p r i o r to c r i s i s . Figure 4.6 Recoding of Timing of Violence Variable Original Codes l.no violence <<-2. violence p r i o r to c r i s i s <— 3. violence triggered c r i s i s <-4.violence a f t e r c r i s i s {= to c r i s i s ' Recoded Values ->>l.no Violence =} 2.violence a f t e r c r i s i s ->3 .violence triggered c r i s i s -> 4.violence p r i o r The rationale for these recordings i s that the longer actors took to resort to violence the longer actors waited to use violence. This delay constitutes more cooperative behavior than i f actors acted with violence e a r l i e r , which indicate less cooperative behavior. The recoded values are 107 based on the notion that, prima facie, the longer i t takes violence to occur i n a c r i s i s , the more cooperative the behavior of the participants. Both the recoded timing of violence and c r i s i s management techniques indicate l e v e l s of cooperation. As with the BCOW cooperation variables, the lower the score the more cooperative behavior and the larger the score the less cooperative the behavior. BCOW's anarchy measure Using BCOW data, anarchy i s measured by assessing the ( d i s ) s i m i l a r i t y of actors' interests and c a p a b i l i t i e s . Both the i n t e r e s t and c a p a b i l i t y variables are tri-chotomous ordinal variables. The interest variable's values range from v i t a l security interests, to non-vital security int e r e s t s , and other tangible but non-security in t e r e s t s and i s coded into the following ordinal categories: l = v i t a l security; 2=non-vital security; 3= other tangible non-s e c u r i t y . 1 6 1 The ca p a b i l i t y variable's values ranges from preponderant power, to equal and i n f e r i o r power and i s coded into the following ordinal c l a s s i f i c a t i o n : l=state has preponderant power, eminent m i l i t a r y success; 2=equal power, Leng, Interstate Crisis Behavior, 47-48 108 c o s t l y uncertain outcome; 3=much less power, sure defeat. 1 6 2 The BCOW interest and c a p a b i l i t y variables are proxies fo r the ends and means, respectively, discussed i n chapter 3. The interests variable r e f l e c t s what states desire, and i d e n t i f i e s the type end a state pursues: v i t a l s ecurity end, non-vital security end, or a tangible non-securing end. The BCOW c a p a b i l i t y variable measures each side's perception of i t s m i l i t a r y c a p a b i l i t y r e l a t i v e to the other side's, and the l i k e l i h o o d for successful u t i l i z a t i o n of m i l i t a r y force. The interest and ca p a b i l i t y variables are tri-chotomous ordinal variables i n t h e i r BCOW forms. They need to be modified i n order to f i t the dichotomous form of the s i m i l a r i t y of ends and s i m i l a r i t y of means variables, respectively, with values of 'high s i m i l a r i t y ' and 'low s i m i l a r i t y ' . 1 6 3 The interest variable i s recoded from l = v i t a l security, 2=non-vital security, 3= other tangible non-security to l = v i t a l security interest and 2=non v i t a l security/non security i n t e r e s t s . The decision to collapse non-vital security and non-security interests into a single category 1 6 2 Leng, Interstate Crisis Behavior, 47-48. 1 6 3 S t r i c t l y , from a l o g i c a l point, t h i s i s not a necessary step i n the case of the interest variable; however, for the c a p a b i l i t y variable i t i s necessary. The reason why the collapsing manipulation was performed i s discussed further below. 109 was made f o r two reasons. F i r s t , v i t a l security i n t e r e s t s are d i s t i n c t from non-vital security and other tangible security by v i r t u e of the increased value states place on v i t a l security i n t e r e s t s . Non-vital security and other tangible non-security are i n a more s i m i l a r value category than v i t a l security since neither i s v i t a l . The second reason to collapse the variable i n t h i s manner i s that the boundary between non-vital security and tangible non-security interests i s more porous than the boundary between v i t a l security and non-vital security. Non-vital security and other tangible non-security interests are d i f f i c u l t , i f not impossible, to untangle and d i f f e r e n t i a t e , since the value which states place on non-vital security and other tangible interests may be the same. The rendering of the tri-chotomous c a p a b i l i t y variable i n t o a dichotomous means variable i s straight forward. Here the values of 'state has preponderant power & eminent m i l i t a r y success' and 'state has much less power & sure defeat' are placed into the same category. Placing these two values i n the same category i s appropriate because each indicates the same thing about perceptions of difference between two actors: that one actor has preponderant c a p a b i l i t y and the other i n f e r i o r c a p a b i l i t y . If one has preponderant capability, the other necessarily has i n f e r i o r c a p a b i l i t y . 110 The recoding of 'state has preponderant power & eminent m i l i t a r y success' and 'state has much less power & sure defeat' into a single value indicates means of low s i m i l a r i t y , while the value of 'equal power, uncertain recoded form the c a p a b i l i t i e s variable e s s e n t i a l l y captures notions of high and low s i m i l a r i t y of means. Put another way, the o r i g i n a l BCOW ca p a b i l i t y variable values of 1 and 3 are collapsed together since they indicate that means are of low s i m i l a r i t y , since one state has more c a p a b i l i t y and the other less c a p a b i l i t y . outcome' indicates means of high s i m i l a r i t y . In t h i s Figure 4 .7 Anarchy BCOW Measure: Original and Recoded Values ORIGINAL INDICATOR VALUE RECODED INDICATOR VALUE Interest 1= v i t a l security 2= non-vital security 3= other tangible non-security. 1=1 v i t a l security 2,3=2 not v i t a l security Capability 1= state has pre-ponderant power 2= equal power, uncertain outcome; 3= much less power, sure defeat 1,3=1 means are low s i m i l a r i t y eminent m i l i t a r y success; 2=2 means are high s i m i l a r i t y Once the interest and ca p a b i l i t y variables are recoded they can be used to generate l e v e l s of s i m i l a r i t y values for I l l the ends and means components of the anarchy model. The c a p a b i l i t y variable t r a n s l a t i o n into s i m i l a r i t y of means i s straight forward. Since c a p a b i l i t y measures the ( d i s ) s i m i l a r i t y between two actors, the same thing s i m i l a r i t y of means measures, i f c a p a b i l i t y i s 1 then means i s low s i m i l a r i t y , and i f c a p a b i l i t y i s 2, then means i s high s i m i l a r i t y . We know t h i s to be the case since the recoded c a p a b i l i t y value of 1 i s when one side has preponderant power and the other side has i n f e r i o r power. This recoded 1 therefore i s e s s e n t i a l l y when there i s a low s i m i l a r i t y of c a p a b i l i t i e s . . The recoded c a p a b i l i t y value of 2 i s when both states have equal power, and thus represents c a p a b i l i t i e s of high s i m i l a r i t y . 1 6 4 Transforming the interest variable into ends i s more complex, since interest i s not an indicator of ( d i s ) s i m i l a r i t y , but an attr i b u t e of a single actor. To ar r i v e at an indicator of, (dis) s i m i l a r i t y we must compare actors' i n t e r e s t s . If actors' interests are the same, both have the same interest, and then the s i m i l a r i t y of ends i s coded as high s i m i l a r i t y . If actors' i n t e r e s t are di f f e r e n t , then ends value i s coded as low s i m i l a r i t y . This coding i s i l l u s t r a t e d i n Figure 4.8. 1 6 4 If there i s a disagreement between each sides' perception of the others c a p a b i l i t y the case i s coded as low s i m i l a r i t y . 112 Figure 4.8 Interest Recodes: Generation of High Similarity (HS) and Low Similarity (LS) Scores actor (a) Interest 1 2 actor (b) HS LS LS HS When both c a p a b i l i t y and interest variable have been recoded into s i m i l a r i t y of means and s i m i l a r i t y of ends variables, respectively, they can be used as inputs to the anarchy model, and the l e v e l of anarchy, given the configuration of s i m i l a r i t i e s of ends and means, can be calculated. This process i s i l l u s t r a t e d below i n Figure 4 . 9 113 Figure 4.9 The Model of Anarchy: Levels of Anarchy Ends Ends High Low S i m i l a r i t y S i m i l a r i t y Means . High {4} {2} S i m i l a r i t y Means Low {3} {1} S i m i l a r i t y ICB's anarchy measure The ICB power status and the issue variables are used as proxies for means and ends, respectively, i n the construction of the anarchy variable. The power status variable i s a measure of and actor's power r e l a t i v e to other actors' power. Its values range from small power to middle, great and superpower. According to the p r i n c i p a l researchers, "the power status of a state was determined by i t s c a p a b i l i t y within the international system i n which i t s 114 c r i s i s occurred. 1 , 1 6 5 Therefore, as they point out, an actor might be a smaller power i n the international system as a whole, but a major power i n the subsystem which the c r i s i s occurred, as i s the case for Egypt and Israel i n 19 6 7.166 As such, the power status variable i s a suitable i n d i c a t o r of means.167 The issues variable categorizes the p r i n c i p l e issues areas for actors into f i v e c l a s s i f i c a t i o n s : m i l i t a r y -security, p o l i t i c a l - d i p l o m a t i c , economic-development, c u l t u r a l status, and other. These issue area c l a s s i f i c a t i o n s are a d i r e c t indicator of an actors' p r i n c i p a l i n t e r e s t . As such they are appropriate indicators of actors' ends. To generate s i m i l a r i t y of ends and s i m i l a r i t y of means, in d i v i d u a l actors ends (or means) attributes are compared to other actors ends (or means) attrib u t e s . If the proportion of s i m i l a r attributes i s .larger than or equal to 0.5 then 1 6 5 Jonathan Wilkenfeld, et. a l . , Crisis in the Twentieth Century, Volume II (New York:Permagon, 1989), 12-13. 1 6 6 Wilkenfeld, Crisis in the Twentieth Century, Vol. II, 12-13. 1 6 7 The system sensitive measure does not pose a problem for our purposes so long as the system or subsystem remain constant for a l l actors i n a c r i s i s , as i s the case i n the ICB data. See Wilkenfeld et. a l . , Crisis in the Twentieth Century Vol. II, 12-13 for further discussion. 115 ends (or means) i s coded as high s i m i l a r i t y and i f the proportion i s less than 0.5 i t i s coded as low s i m i l a r i t y . E s s e n t i a l l y t h i s process i s the same use i n the two actor two att r i b u t e BCOW cases. 1 6 8 However, the increase i n the number of actors and attributes to as many as f i v e i n ICB requires a more formal method to assess the l e v e l s of s i m i l a r i t y of ends and means. A det a i l e d description of the process can be found i n Appendix A. Statistical Tests: Dancers and Wallflowers This section discusses the three tests s t a t i s t i c s , the Goodman-Kruskal gamma, the Kruskal-Wallis H-test, and Kendall's Coefficient of Concordance, why they are selected, and why other tests are rejected. Each test y i e l d s a nonparametric s t a t i s t i c , and each i s appropriate f o r ordinal data. 1 6 9 : 1 6 8 While BCOW reports three attributes for intere s t and c a p a b i l i t y they were collapsed into two at t r i b u t e variables with r e l a t i v e ease. 1 6 9 Gibbons reports that while parametric s t a t i s t i c s are more e f f i c i e n t than nonparametric s t a t i s t i c s by 4%. For further discussion see Jean Dickinson Gibbons, Nonparametric Statistics: An Introduction (Newberry Park, CA:Sage, 1993), 3. Non-parametric s t a t i s t i c s are u t i l i z e d since assumptions of parametric s t a t i s t i c s are not met by the ordinal l e v e l data i n both data sets. As such, nonparametric s t a t i s t i c s are most appropriate. To u t i l i z e parametric methods would attach assumptions to the measurement of anarchy I am not w i l l i n g to make, and which are not necessary to make, given 116 Goodman-Kruskal gamma The gamma c o e f f i c i e n t i s a resul t of an examination of the concordance or discordance of two variables' values. The examination compares how consistently one variable's value i s ranked equal to and higher or lower than the other variable's value. This i s a bi - v a r i a t e pair-wise method of analysis which y i e l d s information about the accuracy and v a l i d i t y of hypotheses which project that one variable's values w i l l be equal to and higher (or lower) than another variable's values. Gamma's values range from -1.0 to 1.0. A gamma value of 1.0 (or-1.0) indicates that one variable's value i s equal t o 1 7 0 or higher (or lower) than the other variable's value i n every pair-wise case. Gamma i s 1.0 (or -1.0) when the percentage of cases where one variable i s equal to and higher (or lower) than the second variable i s 100%. Gamma i s 0.0 when .the percentage of cases where one variable i s equal to (or lower) than the second variable i s 50%. Gamma provides a more focused perspective than the Kendall test. Rather than examining association among average scores, as i s done with the Kendall test, gamma examines matched-pair data. Unlike Kendall's W, gamma can the a v a i l a b i l i t y of nonparametric methods. 170 This case i s also c a l l e d a t i e . 117 be interpreted as a proportional reduction of error s t a t i s t i c (PRE) where knowledge of one variable's value reduces the error i n predicting the other variable values by the value of gamma. Champion asserts that "[0]ne of the most useful measures of association between ordinal l e v e l variables i s gamma.11171 Gamma i s also the most appropriate s t a t i s t i c to use when data i s ordinal and when there exists a large number of t i e s . One drawback of using gamma i s that i t i s sens i t i v e to the way a variable i s categorized. 1 7 2 This means, according to Agresti and Finley, that " d i f f e r e n t researchers could reach d i f f e r e n t conclusions i f they studied the same variables but used d i f f e r e n t categorizations." 1 7 3 To o f f s e t the eff e c t categorization has on gamma, gamma c o e f f i c i e n t s i n three d i f f e r e n t categorization forms, quad-chotomous, tri-chotomous,, and dichotomous, and the average gamma value of the gamma values which have s t a t i s t i c a l l y s i g n i f i c a n t p values, are calculated. This procedure w i l l 1 7 1 Dean J. Champion, Basic Statistics for Social Research (Scranton:Chandler, 1970), 219. 1 7 2 Alan Agresti and Barbara Finley, S t a t i s t i c a l Methods of the Social Sciences, 2nd ed. (San Francisco:Dellen, 1986), 226. 1 7 3 Agresti and Finley, Statistical Methods of the Social Sciences, 226. 118 produce a summary gamma score between anarchy and each c o n f l i c t and cooperation variable. A second drawback of using gamma i s that the value of gamma exaggerates the l e v e l of association when no t i e d ranks e x i s t s . 1 7 4 But since the data used contains a large number of t i e s , gamma's i n f l a t i o n a r y tendency i s less of an issue. Calculations of gamma were performed with SPSS-pc. Significance l e v e l s of gamma for analysis of BCOW data required consultation of a c r i t i c a l gamma values table f o r samples of ns40. 1 7 5 P-values calculations of p=.01 and p=.05 are from Champion's Table A12.176, and others are l i n e a r interpolations of the p-value-gamma re l a t i o n s h i p of these two p-values. 1 7 7 1 7 4 Agresti and Finley, Statistical Methods of the Social Sciences, 226. 1 7 5 Such a table can be found Champion, Basic Statistics for Social Research, 419. 1 7 6 Champion, Statistics for Social Sciences, 419. 1 7 7 Interpolated values are calculated at .01 p-values change per .0138 gamma change (.0138= { [.292-.223]/5} ). Based on t h i s interpolation gamma and i t s associated p-values are l i s t e d below: gamma .154 .168 .181 .195 .209 .223 .251 .264 .278 .292 two t a i l e d , .10 .09 .08 .07 .06 .05 .04 .03 .02 .01 p-value . 119 Kruskal-Wallis H-test The Kruskal-Wallis H-test i s a significance test of ordinal data for the difference of averages of three or more independent samples or v a r i a b l e s . 1 7 8 Champion179 and Gibbons each f i n d the Kruskal-Wallis test to be an appropriate test to accept or reject a l t e r n a t i v e hypotheses. 1 8 0 The test i s the nonparametric counter-part of the parametric F-test, and when applied to a two-sample case produces the same result as the d i f f e r e n t l y calculated Mann-Whitney U-test. 1 8 1 The advantage of the H-test over the U-test i s that the H-test can accommodate ns2 samples or v a r i a b l e s . 1 8 2 H-test calculations were performed with SPSS-pc. 1 7 8 Champion, Statistics for Social Sciences 2nd ed., 283-287; and Gibbons, Nonparametric Statistics: An Introduction, 43-54. 1 7 9 Jean Dickinson Gibbons, Nonparametric Measure of Association (Newberry Park, CA:Sage, 1993); and idem, Introduction to Nonparametric Statistics. 1 8 0 Champion, Statistics for Social Sciences-, Gibbons, Nonparametric Statistics: An Introduction; and idem, Nonparametric Measure of Association. 1 8 1 Gibbons, Nonparametric Measure of Association, 49. 1 8 2 Champion, Basic Statistics for Social Research 2nd ed, 275-76, 286. 120 Kendall Coefficient of Concordance Kendall's C o e f f i c i e n t of Concordance K i s a measure of association of na3 variables. The test assesses the concordance of variables' ordinal sequences by examining the consistency among variables' ordinal rankings. The Kendall S t a t i s t i c K ranges from 0 to 1 where a K=0 indicates no concurrence among the variables and a K=l indicates perfect concordance. Unlike other measurers of association, K does not range between -1 and 0. Regarding i t s a p p l i c a t i o n i n t h i s research, Kendall's test assesses the l e v e l and s t a t i s t i c a l significance of association among the anarchy and c o n f l i c t variables and among anarchy and cooperation variables. Kendall's test i s only applied when the H-test y i e l d s s i g n i f i c a n t difference of averages. Calculations of Kendall's Coefficient of Concordance follows the process outlined by Gibbons, 1 8 3 and are presented i n Appendix B. Wallflowers Kendall's tau and Sommer's d, two alte r n a t i v e ordinal measures of association are rejected because of t h e i r r e s t r i c t i o n s on t h e i r a b i l i t y to deal with t i e s and on row-1 8 3 Gibbons, Nonparametric Measure of Association. 121 column symmetry. While there i s a correction procedure f o r t i e s f or Kendall's tau, Champion reports that gamma i s a better measure of association when t i e d rankings are present. 1 8 4 He states that the correction f o r t i e d ranks for Kendall's tau produces a "small change i n the r e s u l t i n g tau values [and] provides l i t t l e reward f o r applying the co r r e c t i o n . " 1 8 5 This view i s a departure from his 19 7 0 1 8 6 recommendation that an advantage of Kendall's tau i s i t s correction for t i e d rankings. Sommer's d i s rejected because i t i s only able to correct for t i e d ranks on the dependent v a r i a b l e . 1 8 7 Therefore, when a large number of t i e d ranks exist, gamma i s a more appropriate measure than Kendall's tau or Sommer's d. Other tests, such as the t-te s t and the chi-squared test are rejected because they are inappropriate to use with ordinal l e v e l data. The Chi-Squared test, which assumes nominal l e v e l data, i s rejected, even though i t might be applied to ordinal data, because i t ignores information 1 8 4 Champion, Basic Statistics for Social Research, 324. 1 8 5 Champion, Basic Statistics for Social Research, 324. 1 8 6 Champion, Basic Statistics for Social Research, 218 . 1 8 7 Frank Kohout, Statistics for Social Science: A Coordinated Learning System (New York:Wiley, 1974), 228. 122 about the rankings. And the t- t e s t i s rejected because i t i s not appropriate since i t requires at least i n t e r v a l l e v e l data. 1 8 8 There i s a controversy about the appropriateness of using the t- t e s t with ordinal l e v e l data. For example, Bryman and Creamer 1 8 9 recognized that the s t r i c t conditions fo r using the t-test assume: (1) that the data i s i n t e r v a l l e v e l or higher; (2) that the d i s t r i b u t i o n i s normal; (3) and that variance i s homogeneous.190 Yet they argue that the t - t e s t i s robust even when these assumptions are v i o l a t e d , 1 9 1 and.report that the l e v e l of data assumption can be v i o l a t e d i f one has ordinal l e v e l data since "the test apply to numbers and not what those numbers s i g n i f y . " 1 9 2 Their p o s i t i o n i s upheld by Donald Zimmerman and Bruno Zumbo who report that many consider the t- t e s t to be "robust 1 8 8 Seigel, Nonparametric Statistics, 116; Champion, Basic Statistics for Social Research, 101. 1 8 9 Alan Bryman and Duncan Creamer, Quantitative Data Analysis for the Social Science (New York:Routledge, 1990) . 1 9 0 Bryman and Creamer, Quantitative Data Analysis for the Social Science, 116. 1 9 1 Bryman and Creamer, Quantitative Data Analysis for the Social Science, 116. 1 9 2 Bryman and Creamer, Quantitative Data Analysis for the Social Science, 116, 123 and do not recommend alternative [nonparametric tests] unless the l a t t e r two assumptions are v i o l a t e d . " 1 9 3 Zimmerman and Zumbo support t h i s claim by noting that the mathematical s t a t i s t i c a l l i t e r a t u r e considers "assumptions about normality [of the data] and homogeneity of variance are regarded as pertinent" where as "scales of measurement are generally ignored." 1 9 4 Further, Zimmerman and Zumbo claim that even when.the second and t h i r d assumptions are v i o l a t e d the t-test i s s t i l l robust. However, they caution use of the t-t e s t when the second and t h i r d assumptions are v i o l a t e d and the sample size i s small, since the robustness of the t-t e s t i s not known for small sample s i z e s . 1 9 5 Thus i t might seem that I have rejected what may be an appropriate test. However, i f such an error has been made, i t i s an error on the side of conservatism. 193 "The Relative Power of Parametric and Nonparametric S t a t i s t i c a l Methods" i n A Handbook for Data Analysis in the Behavioral Sciences: Methodological Issues ed. Gidion Keren and Charles Lewis (Lawrence Erlhawm:Hillsdale, N.J, 1993), 481. 1 9 4 Zimmerman and Zumbo, "Relative Power of Parametric and Nonparametric S t a t i s t i c a l , " 481. 1 9 5 Zimmerman and Zumbo, "Relative Power of Parametric and Nonparametric S t a t i s t i c a l , " 483. 124 Test Application and Projected Results Analyses of ICB and BCOW data sets w i l l assess the hypotheses i n both aggregated and matched-pair formats. Aggregated data i s analyzed i n the form of average c o n f l i c t and cooperation scores and samples of cooperation and c o n f l i c t variables associated with each l e v e l of s i m i l a r i t y of ends, s i m i l a r i t y of means, and anarchy. The Kruskal-Wallis H-test and the Kendall Coefficient of Concordance are applied to aggregate data. Matched-pair analysis i s a b i -variate case-wise analysis. Matched-pair data analyzes pairings between the s i m i l a r i t y of ends, the s i m i l a r i t y of means, and anarchy, on one hand, and each c o n f l i c t and cooperation variable, on the other hand. The Goodman-Kruskal gamma test i s applied to pair-wise data. The gamma test i s applied to the o r i g i n a l quad-chotomous coding form as well as t r i - and di-chotomous formats. These coding formats are used to off - s e t the ef f e c t of categorization on gamma values. In addition to gamma, matched-pair analysis includes c a l c u l a t i o n of the frequency and d i s t r i b u t i o n of cases where anarchy i s less than, equal to, or more than c o n f l i c t and cooperation variables. Aggregated and matched-p a i r data forms provide d i f f e r e n t perspectives to examine the hypotheses and should be considered as complimentary to 125 each other. Used i n tandem they add robustness to the analyses. Matched-pair analysis The Goodman-Kruskal gamma test i s applied to b i - v a r i a t e matched-pair data between anarchy, s i m i l a r i t y of ends, and s i m i l a r i t y of means, on one hand, and each of the c o n f l i c t and cooperation variables one the other hand. Gamma i s both an association and significance s t a t i s t i c . Calculations of gamma were performed with SPSS-pc. The Goodman-Kruskal test of BCOW data uses c o n f l i c t and cooperation variables which are collapsed into quad-chotomous, tri-chotomous, and dichotomous ordinal formats. These modifications are required so that c o n f l i c t and cooperation variable formats are consistent with the format of the anarchy variable. 1. 9 6 These modifications are v a l i d since ordinal data can be transformed so long as the 1 9 6 To collapse the continuous values between +3 and -3 into these ordinal categories, negative and p o s i t i v e values were placed into the top and bottom two categories, respectively. Then negative and po s i t i v e scores were divided into roughly equal portions with the highest negative and po s i t i v e scores placed i n the lowest and highest ordinal category. 126 transformation i s monotonic. 1 9 7 While the transformation may e f f e c t the magnitude and significance of results, i f the tested hypothesis i s v a l i d then re s u l t s from each format should also be v a l i d . Thus, the use of d i f f e r e n t formats add robustness to r e s u l t s . Matched-pair analysis also includes the c a l c u l a t i o n of the frequency and d i s t r i b u t i o n of cases where anarchy i s less than, more than, or equal to c o n f l i c t and cooperation variables. The cal c u l a t i o n i s simply the value of the anarchy variable minus the value of a c o n f l i c t or cooperation variable. The difference between the two variables y i e l d s the magnitude and frequency of an erroneously hypothesized relationship, when the difference i s not zero, and the frequency of accurately hypothesized relationships, when the difference i s zero. Then the frequencies are used to i l l u s t r a t e the d i s t r i b u t i o n of accurately and erroneously hypothesized relationships. Aggregated data analysis The Kruskal-Wallis H-test i s used to test the s t a t i s t i c a l significance of differences among c o n f l i c t and 1 9 7 Seigel and Castellan define a monotonic transformation as "one which preserves the ordering of objects." Nonparametric Statistics for the Behavioral Sciences . (McGraw-Hill, New York:1988), 26. 127 cooperation scores at each l e v e l of anarchy, and s i m i l a r i t i e s of ends and means. The H-test treats scores at each l e v e l as d i f f e r e n t independent samples and assesses whether or not a s t a t i s t i c a l l y s i g n i f i c a n t difference e x i s t s among them. If the difference of samples i s s t a t i s t i c a l l y s i g n i f i c a n t , the difference of the average scores of each sample can be treated as s t a t i s t i c a l l y s i g n i f i c a n t as well. The H-test w i l l i d e n t i f y c o n f l i c t and cooperation variables whose average scores are s t a t i s t i c a l l y s i g n i f i c a n t . Among these variables, the average scores w i l l be rank-ordered. The rank-ordering of average scores i s a v a l i d procedure since the difference among the raw scores i s s i g n i f i c a n t . Then, the rank-ordered average scores w i l l be tested for le v e l s of association with Kendall's c o e f f i c i e n t of concordance. Significance l e v e l The significance l e v e l f or s t a t i s t i c a l s i g nificance to accept the alternative hypotheses i s ps.20. While th i s i s a r e l a t i v e l y high p-value, Labovitz advises that at the exploration stage of research a .20 l e v e l i s appropriate. 1 9 8 Not only i s t h i s the i n i t i a l exploration 1 9 8 The decision to use the .20 l e v e l i s based on Labovitz's arguments that at the "exploration stage perhaps a .10 or .20 l e v e l would be s u f f i c i e n t " and that the 128 stage, but given the small n-size of 38 i n the BCOW data, a ps.20 seems to be reasonable. Therefore, i f the sign i f i c a n c e l e v e l i s equal to or less than .20 (ps20), the conclusion that the n u l l hypothesis can be rejected and the alt e r n a t i v e hypothesis accepted,will be made. Projected r e s u l t s Five sets of results must be present to accept d e f i n i t i v e l y the hypotheses. F i r s t , the ordinal ranking of average c o n f l i c t and cooperation scores w i l l be p o s i t i v e l y related (either concordantly or di r e c t l y ) to l e v e l s of anarchy, s i m i l a r i t y of ends, and s i m i l a r i t y of means. Regarding the s i m i l a r i t y of ends and s i m i l a r i t y of means, average c o n f l i c t and cooperation scores at high s i m i l a r i t y w i l l be higher than those at low s i m i l a r i t y . These re s u l t s are projected by HI and H2, respectively. Regarding anarchy, average c o n f l i c t and cooperation scores at each l e v e l of anarchy w i l l correspond to the l e v e l of anarchy from which the average i s taken. In other words, the ordinal l e v e l of average c o n f l i c t and cooperation scores w i l l correspond to the l e v e l of anarchy: the highest average sig n i f i c a n c e l e v e l can be higher for a smaller size n. Sandford Labovitz, " C r i t e r i a for selecting a si g n i f i c a n c e l e v e l : a note on the sacredness of .05" i n The Significance . Test Controversy--A Reader, ed. Denton E. Morrison and Ramon E. Henkel (Chicago:Aldine, 1970), 168, 170. 129 w i l l be at the highest l e v e l of anarchy, followed by the second highest averages at the second highest l e v e l of anarchy, the t h i r d highest averages at the t h i r d highest l e v e l of anarchy, and the lowest averages at the lowest l e v e l of anarchy. Second, not only w i l l anarchy, s i m i l a r i t y of means, and s i m i l a r i t y of ends scores be p o s i t i v e l y related to the average c o n f l i c t and cooperation scores, but the difference of average c o n f l i c t and cooperation samples at each l e v e l of anarchy and s i m i l a r i t y w i l l be s t a t i s t i c a l l y s i g n i f i c a n t . To assess t h e i r significance the Kruskal-Wallis i s employed. Third, i f the anarchy hypothesis i s v a l i d , Kendall's C o e f f i c i e n t of Concordance w i l l y i e l d s t a t i s t i c a l l y s i g n i f i c a n t association c o e f f i c i e n t s among the rankings of anarchy and the average cooperation variables' scores, as well as among the rankings of anarchy and c o n f l i c t variables' scores. -Fourth, regarding pair-wise results, the Goodman-Kruskal gamma results w i l l r e f l e c t p o s i t i v e and s t a t i s t i c a l l y s i g n i f i c a n t gamma c o e f f i c i e n t s between anarchy, s i m i l a r i t y of ends, and s i m i l a r i t y of means, on one hand, and cooperation and c o n f l i c t , on the other hand. F i f t h , the number of cases where the value of anarchy i s equal to the value of c o n f l i c t or cooperation w i l l be greater than the number of errors, cases where the value of anarchy i s less than or more than the value of c o n f l i c t of 130 cooperation. And, the frequency of error w i l l decrease as the magnitude or error increases. The most important hypothesis for the assessment of the anarchy tenet i s H3, since i t d i r e c t l y addresses the relationships between anarchy and cooperation and anarchy and c o n f l i c t . The hypotheses about means (Hi) and ends (H2) are important insofar as they shed l i g h t on the re l a t i o n s h i p between the s i m i l a r i t y of ends and means and c o n f l i c t and cooperation. However, they are not necessary to assess the v a l i d i t y of the t h i r d hypothesis. The anarchy hypothesis and set of expectations stand autonomously from the f i r s t two, given the interdependent ef f e c t of ends and means on rule making and rule enforcing. Thus the important f i n d i n g required to adhere with the general hypothesis i s the t h i r d set of expectations. Because s i m i l a r i t y of ends and means may not be related to level s of c o n f l i c t and/or cooperation independent from one another does not denote that i n t e r -a c t i v e l y they are not related to c o n f l i c t and/or cooperation. In other words, re j e c t i o n of hypotheses HI and H2 would not necessarily constitute grounds to rej e c t hypothesis H3\ However, support for HI or H2 would constitute support, a l b e i t preliminary, for H3. 131 CHAPTER 5 ANALYSIS OF BCOW DATA This chapter presents and analyzes test r e s u l t s of the relationships asserted i n the s i m i l a r i t y of ends, s i m i l a r i t y of means, and anarchy hypotheses. The test data i s the Behavioral Correlates of War (BCOW) data as reported by Russell Leng i n Interstate Crisis Behavior. The BCOW data i s a sample of fort y m i l i t a r i z e d i n t e r s t a t e c r i s e s (MIC). A MIC i s "when the threat, display, or use of force i s reciprocated by ... the other s i d e . " 1 9 9 Of the available MIC population (1816-1980, n=5 9 3 ) , 2 0 0 Leng focuses on MICs with less than f i f t y one interactions p r i o r to the termination of the c r i s i s or the onset of violence. 2 0 1 Then cases where randomly selected such that the d i s t r i b u t i o n of cases i n the periods 1816-1918, 1918-1945, and 1946-1980 r e f l e c t e d the same d i s t r i b u t i o n found i n the Gochman and Moaz population of Leng, Interstate Crisis Behavior, 26. Leng, Interstate Crisis Behavior, 26. Leng, Interstate Crisis Behavior, 26-27. 132 m i l i t a r i z e d international disputes (MID) . 2 0 2 Case se l e c t i o n was also designed to include a balance of war-ending and peace-ending MICs. 2 0 3 Sample size was l i m i t e d to f o r t y cases given the q u a l i t a t i v e treatment of each MIC. Of these cases two do not have complete data, leaving a sample of t h i r t y eight cases. Eight BCOW variables are used to test the three hypotheses. The anarchy variable i s composed of the BCOW inte r e s t and ca p a b i l i t y variables; cooperation variables are the r e c i p r o c i t y d i r e c t i o n and r e c i p r o c i t y distance variables, and th e i r sum, the cooperation index var i a b l e ; and, the c o n f l i c t variables are the h o s t i l i t y magnitude and h o s t i l i t y escalation rate variables, and t h e i r sum, the c o n f l i c t index variable. Summing of the two reciprocity- and h o s t i l i t y variables to produce a single variables for cooperation and c o n f l i c t , respectively i s a procedure used by Leng. . As noted i n chapter 4, three s t a t i s t i c a l tests are applied to test hypothesized relationships. Tests of data aggregated into averages include the Kruskal-Wallis H-test and the Kendall Coefficient of Concordance. The H-test i s used to test whether or not the difference of samples of c o n f l i c t and cooperation variables at each l e v e l of anarchy, 2 0 2 Leng, Interstate Crisis Behavior, 29. 2 0 3 Leng, Interstate Crisis Behavior, 29. 133 s i m i l a r i t y of ends, and s i m i l a r i t y of means are s t a t i s t i c a l l y s i g n i f i c a n t . If, for example, the samples of h o s t i l i t y magnitude scores at each l e v e l of anarchy are s t a t i s t i c a l l y d i f f e r e n t then averages calculated from these samples can also be accepted as s i g n i f i c a n t . The Kendall test i s then used to est a b l i s h the l e v e l association among variables whose difference of averages i s s i g n i f i c a n t per the Kruskal-Wallis H-test. For matched-pair data the Goodman-Kruskal test i s used to provide significance and association l e v e l s . Both average and matched-pair data are examined since each divulges d i f f e r e n t information about a r e l a t i o n s h i p . The Goodman-Kruskal test uses BCOW data collapsed into guad-chotomous, tri-chotomous, and dichotomous ordinal formats so that t h e i r format i s consistent with the format of the anarchy variable. These modifications are v a l i d since ordinal data can be transformed so long as the transformation i s monotonic. To collapse the values between +3 and -3 into ordinal formats, p o s i t i v e and negative values were placed into the top and bottom two categories, respectively. Then negative and p o s i t i v e scores were divided into roughly equal portions with the highest negative and p o s i t i v e scores placed i n the lowest and highest ordinal category. Results w i l l be presented i n the following sequence. F i r s t the average c o n f l i c t and cooperation scores w i l l be 134 discussed and compared to projected outcomes. Then the res u l t s of the Kruskal-Wallis H-test are discussed. Third and f i n a l l y , the Goodman-Kruskal and the difference i n actual and an expected random error d i s t r i b u t i o n s r e s u l t s are presented. This i s repeated for each hypothesis. Findings concerning the s i m i l a r i t y of means hypothesis If the hypothesis about the s i m i l a r i t y of means and cooperation and c o n f l i c t i s v a l i d , the average c o n f l i c t and cooperation scores should follow the patterns asserted i n chapter 4, and the difference of samples form which average scores are calculated should be s i g n i f i c a n t . In other words, the average of each c o n f l i c t and cooperation score at high s i m i l a r i t y of means w i l l be greater than those at low s i m i l a r i t y of means, and the difference of samples from which the averages are calculated w i l l be s t a t i s t i c a l l y s i g n i f i c a n t . Table 5.1 reports hypothesized results and average cooperation and c o n f l i c t scores at high and low s i m i l a r i t y of means, the ordinal magnitude of average scores, and the two-tailed significance of the Kruskal-Wallis H-test f o r the sign i f i c a n c e of the difference of samples from which are averages calculated. 135 As Table 5.1 indicates, the hypothesized average c o n f l i c t and cooperation scores are, with one exception, accurately represented i n the findings. The exception i s Table 5.1 Projected and Actual Findings: Similarity of Means and Cooperation and Conflict Cooperation Variables Conf l ic t Variables h o s t i l i t y s imi la r i ty reciproci ty cooperation escalation h o s t i l i t y oonf l ic t of means* distance direct ion index magnitude index LS hypo-t h e s i z e d r e s u l t s HS low h i g h low high low high low high low high low h i g h average scor e r e s u l t LS HS . 12 .12 . 03 . 08 - . 09 . 03 . 04 .18 .11 .26 .07 n=23 .44 n=15 average LS 1 o r d i n a l magnitude HS 2 Kr u s k a l -W a l l i s H-Test 2 0' . 91 .78 .66 .39 .52 .67 * LS=low s i m i l a r i t y HS=high S i m i l a r i t y the averages of the r e c i p r o c i t y d i r e c t i o n score. In t h i s case the magnitude of average scores i s the reverse of the Two-tailed. 136 hypothesized projection. The average scores i s higher at low s i m i l a r i t y than i t i s at low s i m i l a r i t y In the remaining f i v e relationships the actual and hypothesized results of the magnitude of average scores are the same. That i s , the ordinal magnitude of average cooperation and c o n f l i c t scores are greater at high s i m i l a r i t y than at low s i m i l a r i t y . In four of f i v e r elationships not only are the projected ordinal rankings correct, but average cooperation and c o n f l i c t scores associated with low s i m i l a r i t y of means are negative. This suggests that at low s i m i l a r i t y of means not only i s c o n f l i c t lower, but behavior i s cooperative and not just less c o n f l i c t i v e . These results support the acceptance of the hypothesis. Yet, the Kruskal-Wallis H-test of the sign i f i c a n c e of the difference of samples from which averages are calculated does not y i e l d a single s i g n i f i c a n t p-value. This f i n d i n g suggests that there i s no s t a t i s t i c a l l y s i g n i f i c a n t difference between samples of c o n f l i c t and cooperation variables at high and low level s of s i m i l a r i t y of means, and that the average scores and t h e i r magnitude cannot be accepted as s t a t i s t i c a l l y s i g n i f i c a n t . To f i n d results which support accepting the hypothesis, or more accurately part of the hypothesis, one must turn to tests of matched-pair data. Results of matched-pair analysis with the Goodman-Kruskal gamma c o e f f i c i e n t reveals 137 s t a t i s t i c a l l y s i g n i f i c a n t s i m i l a r i t y of means-conflict relat i o n s h i p s . Gamma c o e f f i c i e n t s for the s i m i l a r i t y of means-hostility magnitude and s i m i l a r i t y of means-conflict index relationships yielded p-values of ps.10. However, no s i g n i f i c a n t p-values resulted between the s i m i l a r i t y of means and cooperation variables. Table 5.2 reports these r e s u l t s . Table 5.2 Goodman-Kruskal gamma s i m i l a r i t y of means-cooperation & c o n f l i c t variables gamma p-value 2 Cooperation Variables Reciprocity Distance -.11 pa. 21 Reciprocity Direction -.02 pa.21 Cooperation Index .13 pa. 21 C o n f l i c t Variables H o s t i l i t y Rate .07 pa.21 H o s t i l i t y Magnitude .15 .10 Co n f l i c t Index .24 .05 The other analysis of matched-pair data i s an examination of the di s t r i b u t i o n s of accurately and erroneously hypothesized relationships. The hypothesized Two-tailed. 138 rel a t i o n s h i p projects that the difference between the s i m i l a r i t y of means and c o n f l i c t and cooperation variables i s zero ( s i m i l a r i t y of means minus c o n f l i c t variable=0). The hypothesis also projects that the frequency of cases where the difference i s zero w i l l be greater i n the findings than i n an expected random d i s t r i b u t i o n , and where the frequency of errors (when the difference i s not zero) w i l l be less than i n an expected random d i s t r i b u t i o n . If the actual d i s t r i b u t i o n i s sim i l a r to an expected random d i s t r i b u t i o n , the hypothesis cannot be accepted since the res u l t s indicate a random rather than an associative r e l a t i o n s h i p . Table 5.3 reports the d i s t r i b u t i o n of accurately and erroneously hypothesized relationships for actual and expected random d i s t r i b u t i o n s . Errors i n which the s i m i l a r i t y of means i s less than the c o n f l i c t or cooperation variable are reported in,the lover rows, and errors i n which the s i m i l a r i t y of means i s more than the c o n f l i c t or cooperation variables are reported i n the %under rows. Accurate projections are reported i n the %same rows. Table 5.3 reveals that the frequency of accurate projections does not d i f f e r from an expected random d i s t r i b u t i o n of accurate projections. The largest difference between actual and expected random d i s t r i b u t i o n s of accurate cases i s the 58% accurate case d i s t r i b u t i o n when 139 an expected random d i s t r i b u t i o n i s 50% (the s i m i l a r i t y of means-conflict index r e l a t i o n s h i p ) . Table 5.3 also reveals d i f f e r e n t trends f o r the er r o r d i s t r i b u t i o n s of cooperation-similarity of means and c o n f l i c t s i m i l a r i t y of means relationships. The cooperation-similarity of means d i s t r i b u t i o n of %under frequency i s more than twice of the %over frequency, and about 12 percentage points above the expected random d i s t r i b u t i o n . On the other hand, the c o n f l i c t - s i m i l a r i t y of means error d i s t r i b u t i o n s do not d i f f e r by more than a factor of about one half and are within 7 percentage points of the expected random d i s t r i b u t i o n . This finding suggest that when the hypothesis i s not accurate the tendency i s for cooperation scores to be smaller than the s i m i l a r i t y of means score. 140 Table 5.3 Similarity of Means Frequency of Errors in Actual and Random Distributions s i m i l a r i t y of m e a n s - r e c i p r o c i t y d i s t a n c e magnitude or e r r o r -1 0 1 frequency- 15 17 6 percent 39% 45% 16% a c t u a l random %under 39% 25% %same 45% 50% %over 16% 25% s i m i l a r i t y of m e a n s - r e c i p r o c i t y d i r e c t i o n magnitude or e r r o r -1 0 1 frequency 14 18 6 percent 37% 47% 16% a c t u a l random %under 37% 25% %same 47% 50% %over 16% 25% s i m i l a r i t y of means-cooperation index magnitude or e r r o r -1 . 0 1 frequency : 14 19 5 percent 37% 50% 13% a c t u a l random %under 37% 25% %same 50% 50% %over 13% 25% s i m i l a r i t y of m e a n s - h o s t i l i t y e s c a l a t i o n r a t e magnitude or e r r o r -1 0 1 frequency 7 21 10 percent 18% 55% 26% a c t u a l random %under 18% 25% %same 55% 50% %over 26% 25% 141 t a b l e 5.3 continued s i m i l a r i t y o f m e a n s - h o s t i l i t y magnitude magnitude o r e r r o r -1 0 1 frequency 9 21 8 percent 24% 55% 21% a c t u a l random %under 24% 25% %same 55% 50% %over 21% 25% s i m i l a r i t y o f m e a n s - c o n f l i c t index magnitude o r e r r o r -1 0 1 frequency 8 22 8 percent 21% 58% 21% a c t u a l random %under 21% 25% %same 58% 50% %over 21% 25% Similarity of means hypothesis: summary of findings Given the results of the Kruskal-Wallis and Goodman-Kruskal tests two conclusions can be made. F i r s t , evidence compels r e j e c t i o n of the portion of the hypothesis about the s i m i l a r i t y of ends and cooperation. While average c o n f l i c t and cooperation scores at high and low s i m i l a r i t y of means support the hypothesis, no s t a t i s t i c a l l y s i g n i f i c a n t p-value (ps.20) resulted from the H-test and no matched-pair gamma res u l t s yielded a s i g n i f i c a n t relationship (ps.20). Thus, 142 both tests concur that the cooperation-similarity of means relationships must be rejected. The s i m i l a r i t y of means-conflict portion of the hypothesis resulted i n mixed findings. The H-test does not y i e l d s i g n i f i c a n t results, while the matched-pair gamma test does. These results indicate the second conclusion, namely that the s i m i l a r i t y of means-conflict r e l a t i o n s h i p cannot be accepted as s t a t i s t i c a l l y s i g n i f i c a n t as a di r e c t r e l a t i o n s h i p . The H-test results preclude t h i s acceptance. However, s t a t i s t i c a l l y s i g n i f i c a n t gamma results suggest that the s i m i l a r i t y of means-conflict relationships can be accepted as a s t a t i s t i c a l l y s i g n i f i c a n t concordant re l a t i o n s h i p . Yet, modest gamma c o e f f i c i e n t s of .15 and .24 indicate a low l e v e l of association. The outright r e j e c t i o n of the cooperation portion of the hypothesis, the d i f f e r e n t H-test and gamma resu l t s of the c o n f l i c t portion of the hypothesis, and low gamma co e f f i c i e n t s suggest that the s i m i l a r i t y of means hypothesis i s not accurate. Results Concerning the Similarity of Ends Hypothesis The s i m i l a r i t y of ends hypothesis projects that at high s i m i l a r i t y of ends both c o n f l i c t and cooperation scores w i l l be higher than they are at low s i m i l a r i t y of ends. Table 5.4 reports average c o n f l i c t and cooperation scores 143 associated with high and low s i m i l a r i t y of ends, as well as t h e i r ordinal magnitude. Average cooperation and c o n f l i c t scores reveal the ordinal magnitudes projected by the hypothesis are accurately represented i n the findings. Each taverage c o n f l i c t and cooperation score at high s i m i l a r i t y of ends i s p o s i t i v e and higher than the those at low s i m i l a r i t y of means, where f i v e of six scores are negative. One important difference between the cooperation-s i m i l a r i t y of ends and c o n f l i c t - s i m i l a r i t y of ends relationships can be i d e n t i f i e d : the former i s s t a t i s t i c a l l y s i g n i f i c a n t while the l a t t e r i s not. The H-test of each cooperation variable i s s i g n i f i c a n t (ps.16), while no c o n f l i c t variable y i e l d s a s i g n i f i c a n t H-test. This suggests that while the c o n f l i c t - s i m i l a r i t y of ends portion of the hypothesis i s not s t a t i s t i c a l l y s i g n i f i c a n t , the cooperation-similarity of ends portion i s s i g n i f i c a n t . However, p r i o r to making a more f i n a l conclusion, consideration of matched-pair results i s warranted. Table 5.4 Hypothesized and Actual Findings: Similarity of Ends and Cooperation and Conflict 144 Cooperation Variables Conf l ic t Variables h o s t i l i t y s imi la r i ty reciproci ty cooperation escalation h o s t i l i t y con f l i c t of means* distance direct ion index rate magnitude index hypo-t h e s i z e d r e s u l t LS low HS h i g h low high low h i g h low high low h i g h low h i g h average scor e r e s u l t LS -.3 9 HS .12 .06 -.45 .003 .12 . 04 .11 .18 .12 .14 n = l l .24 n=27 average o r d i n a l magnitude LS HS K r u s k a l W a l l i s H-Test 2 0' . 13 .16 . 09 .25 .50 . 98 * LS=low s i m i l a r i t y HS=high S i m i l a r i t y •Two-tailed. 145 The Goodman-Kruskal results of matched-pair analysis are l i s t e d below i n Table 5 . 5 . With regard to c o n f l i c t , these r e s u l t s d i f f e r from those of the average scores just considered; with regard to cooperation, the re s u l t s concur with previous r e s u l t s . Diverging from H-test r e s u l t s of aggregated data, matched-pair results of the c o n f l i c t -s i m i l a r i t y of ends portion of the hypothesis y i e l d s t a t i s t i c a l l y s i g n i f i c a n t results ( p s . 0 9 ) . Concurring with aggregated data, matched-pair results support accepting the s i m i l a r i t y of end-cooperation portion of the hypothesis as s t a t i s t i c a l l y s i g n i f i c a n t ( p s . 0 1 ) . In each of the six matched-pair relationships, gamma c o e f f i c i e n t p-values. of p s . 0 9 i s achieved, with f i v e of s i x c o e f f i c i e n t s having p-values of p s . 0 1 . Each of the three Table 5.5 Goodman-Kruskal gamma s i m i l a r i t y of ends-cooperation & c o n f l i c t variables gamma p-value COOPERATION Reciprocity Distance Reciprocity Direction Cooperation Index .48 .41 .48 p<. 01 p< .01 p<.01 146 table 5.5 continued gamma p-value CONFLICT H o s t i l i t y Escalation Rate H o s t i l i t y Magnitude C o n f l i c t Index - .41 .17 .36 p<.01 p=. 09 p<.01 cooperation variables i s s i g n i f i c a n t at p<.01 while two of the three c o n f l i c t variables meets the same standard. Association c o e f f i c i e n t s among cooperation variables range from .41 to .48 while for c o n f l i c t variables they range from The d i s t r i b u t i o n of the frequency of errors of the hypothesis, reported i n Table 5.6, reveals that the hypothesis i s s l i g h t l y more accurate than an expected random d i s t r i b u t i o n , by about 13 percentage points, with regard to cooperation, and less than or about equal to an expected random d i s t r i b u t i o n with regard to c o n f l i c t . The error frequencies reveal that when the hypothesis i s not accurate the tendency i s for cooperation and c o n f l i c t variables to be less than the s i m i l a r i t y of ends, as indicated by the %over to be larger than the %under. This tendency i s more modest with regard to cooperation than i t i s with regard to c o n f l i c t . .17 to - .41. Table 5.6 Similarity of Means Frequency of Errors in Actual and Random Distributions s i m i l a r i t y o f e n d s - r e c i p r o c i t y d i s t a n c e magnitude o r e r r o r - 1 0 1 frequency 5 25 8 percent 13% 66% 21% a c t u a l random %under 13% 25% %same 66% 50% %over 21% 25% s i m i l a r i t y of e n d s - r e c i p r o c i t y d i r e c t i o n magnitude o r e r r o r -1 0 1 frequency 5 24 9 percent 13% 63% 24% a c t u a l random %under 13% 25% %same 63% 50% %over 24% 25% s i m i l a r i t y o f ends-cooperation index magnitude o r e r r o r -1 0 1 frequency 5 25 8 percent 13% 66% 21% a c t u a l random %under 13% 25% %same 66% 50% %over 21% 25% s i m i l a r i t y of e n d s - h o s t i l i t y e s c a l a t i o n r a t e magnitude o r e r r o r -1 0 1 frequency 5 13 20 percent 13% 34% 53% a c t u a l random %under 13% 25% %eame 34% 50% %over 53% 25% 148 Table 5.6 continued s i m i l a r i t y of e n d s - h o s t i l i t y magnitude magnitude o r e r r o r -1 0 1 frequency 4 19 15 percent 11% .50% 39% a c t u a l random %under 11% 25% %same 50% 50% %over 39% 25% s i m i l a r i t y o f e n d s - c o n f l i c t index magnitude o r e r r o r -1 0 1 frequency 3 20 15 percent 8% 53% 39% a c t u a l random %under 8% 25% %same 53% 50% %over 39% 25% S i m i l a r i t y of Ends Hypothesis: Summary of Results Both the Kruskal-Wallis and ,the Goodman-Kruskal matched-pair analysis results show a s i g n i f i c a n t r e l a t i o n s h i p between the s i m i l a r i t y of ends and cooperation (ps.16). However the two tests y i e l d d i f f e r e n t r e s u l t s about the c o n f l i c t - s i m i l a r i t y of ends re l a t i o n s h i p . The H-test analysis demands rejection, while the matched-pair analysis suggests acceptance. 149 With both average and matched-pair r e s u l t s concurring, the acceptance of the cooperation-similarity of ends portion of the hypothesis as s t a t i s t i c a l l y s i g n i f i c a n t i s convincing. Regarding the c o n f l i c t - s i m i l a r i t y of ends relationship, H-test results compel r e j e c t i o n of a d i r e c t r e l a t i o n s h i p . Gamma results support accepting a concordant r e l a t i o n s h i p between h o s t i l i t y magnitude and s i m i l a r i t y of ends and between c o n f l i c t index and s i m i l a r i t y of ends, and a discordant relationship between h o s t i l i t y e scalation rate and s i m i l a r i t y of ends. However, gamma c o e f f i c i e n t s are modest even at t h e i r highest magnitude (gammas.48). Results concerning the anarchy hypothesis The anarchy hypothesis asserts that (1) cooperation i s negatively associated with anarchy and (2) c o n f l i c t i s p o s i t i v e l y associated with anarchy. C o n f l i c t and cooperation scores are coded so that a higher score indicates more c o n f l i c t and less cooperation and a smaller score less c o n f l i c t and more cooperation. In terms of average c o n f l i c t and cooperation scores, the hypothesis projects: at the highest l e v e l of anarchy the scores w i l l be highest; that at the second highest l e v e l of anarchy the scores w i l l be second highest; that at the t h i r d highest l e v e l of anarchy the scores w i l l be t h i r d highest; and at the lowest l e v e l of anarchy the scores w i l l be lowest. 150 Thus, the ordinal magnitude of the average cooperation and c o n f l i c t scores at each l e v e l of anarchy should concur with the l e v e l of anarchy from which the average i s calculated, and an ordinal anarchy sequence of 1,2,3,4 should concur with average cooperation and c o n f l i c t sequences of 1,2,3,4. In addition to the tests used to test the s i m i l a r i t y of ends and means hypotheses, one test i s added f o r the t e s t i n g of the anarchy hypothesis. The test i s Kendall's c o e f f i c i e n t of Concordance. The test i s added since three or more ordinal categories for anarchy e x i s t . The analysis starts with results of the H-test, then addresses the frequency and magnitude of errors, and f i n a l l y r e s u l t s of the gamma test. The discussion of the c o n f l i c t -anarchy rela t i o n s h i p precedes the discussion of the cooperation-anarchy relationship. Anarchy-Conflict Relationship H-test r e s u l t s Neither the sequence of ordinal magnitudes of cooperation and c o n f l i c t average scores nor the H-test follow the pattern projected by the anarchy hypothesis. Results of the H-test do not y i e l d a single s i g n i f i c a n t 151 Table 5 .7 : ~ ' Hypothesized and A c t u a l F i n d i n g s : Anarchy and C o n f l i c t L e v e l of Anarchy C o n f l i c t V a r i a b l e s h o s t i l i t y e s c a l a t i o n r a t e h o s t i l i t y magnitude c o n f l i c t index h y p o t h e s i z e d r e s u 1 t s low med low med high h i g h low med low med high h i g h low med low med h i g h h i g h a v s e c r o a r g e e 0 . 01 0.37 0 . 06 0 .16 •0.26 0.64 0 .007 0 .24 •0.25 1. 01 0 . 06 0.40 n=10 n=01 n=13 n=14 a v e r a g e . o r d i n a 1 1 4 2 3 K r u s k a l -W a l l i s H-Test .42 .46 .71 152 a p-value (ps.20) for the difference of sample from which averages are calculated and, with the exception of anarchy l e v e l 1, the ordinal magnitude of average c o n f l i c t scores do not correspond to the l e v e l of anarchy from which they are derived. Where the sequence pf anarchy 1,2,3,4 i s projected to correspond to ordinal values of average c o n f l i c t scores of the same sequence, the actual sequence of the average c o n f l i c t scores i s 1,4,2,3. One explanation of the divergence of hypothesized and actual results i s that there i s only a single case of anarchy l e v e l 2. If t h i s case i s ignored, then the projected and actual ordinal correspondence i s accurate. However, even when anarchy l e v e l 2 i s ignored the H-test does not y i e l d a single s i g n i f i c a n t p-value. When the two middle lev e l s of anarchy are collapsed into a single category, the ordinal sequence of average c o n f l i c t and cooperation .scores correspond to the l e v e l of anarchy from which they are calculated. However, the H-test does not y i e l d a single s t a t i s t i c a l l y s i g n i f i c a n t p-value. Thus, the H-test compels re j e c t i o n of the anarchy c o n f l i c t hypothesis. Table 5.8 P r o j e c t e d and A c t u a l F i n d i n g s : Anarchy and C o n f l i c t , Tri-chotomous Form L e v e l o f Anarchy C o n f l i c t V a r i a b l e s h o s t i l i t y e s c a l a t i o n h o s t i l i t y r a t e magnitude c o n f l i c t index low low low medium medium medium hi g h high h i g h 0.01 0 . 08 0 .16 •0.26 0 ,05 0 .24 •0.25 n=10 0.13 n=14 0.40 n=14 K r u s k a l - W a l l i s .44 H-Test .67 154 Difference i n accurate and erroneous r e s u l t s i n actual and random d i s t r i b u t i o n s Table 5.9 reports the frequency and magnitude of errors i n the hypothesized anarchy-conflict relationship, as well as the frequency of accurately hypothesized r e l a t i o n s h i p s . The hypothesis projects that anarchy and c o n f l i c t are p o s i t i v e l y related. Therefore, the hypothesized difference between anarchy and each c o n f l i c t variable i s zero (anarchy-c o n f l i c t s , reported as %same). When the difference i s p o s i t i v e (reported as lover), the l e v e l of c o n f l i c t i s less than the l e v e l of anarchy; when the difference i s negative (reported as %under), the l e v e l of c o n f l i c t i s higher than the l e v e l of anarchy. Table 5.9 reports both expected random and actual d i s t r i b u t i o n s for accurately (%same) and erroneously (%under and lover) hypothesized rela t i o n s h i p s . Errors with s t r i c t and relaxed calculations are also reported. The expected random d i s t r i b u t i o n i s calculated i n the following manner. Since each variables' value ranges from 1 to 4, the permutations of possible b i - v a r i a t e value pairings i s 16. As such, an expected random d i s t r i b u t i o n projects that the frequency of each permutation i s 1/16, or 6.25%. If the d i s t r i b u t i o n i s random, the expectation i s that the (%under) and (lover) w i l l each be about 38%, since the expected frequency of value pairings i s 6 i n each category, 155 and 6/16 i s about 38%. On the other hand, the expected (%same) i n a random d i s t r i b u t i o n w i l l be about 25%, since the expected frequency of the value pairings i s 4 and 4/16, or 25%. The expectation i s that i f the hypothesis i s v a l i d random and actual results should d i f f e r . D i s t r i b u t i o n s of accurate and erroneous relationships are calculated with both relaxed and s t r i c t error c a l c u l a t i o n c r i t e r i a i n order to provide some leeway for the hypothesis. A relaxed hypothesis error c a l c u l a t i o n i s when the value of anarchy and c o n f l i c t d i f f e r s by more than 1. Table 5.9 Frequency of Errors in Actual and Random Distribution A n a r c h y - H o s t i l i t y E s c a l a t i o n magnitude of e r r o r - 3 - 2 - 1 0 1 2 3 frequency 2 2 8 3 9 9 5 p e r c e n t * 5% 5% 21% 8% 24% 24% 13% s t r i c t r e l a x e d a c t u a l random a c t u a l random % under 32% % same 8% % over 61% 38% 25% 38% % under 11% % same 53% % over 37% 19% 63% 19% 156 t a b l e 5.9 continued A n a r c h y - H o s t i l i t y Magnitude magnitude of e r r o r -3 -2 -1 0 1 2 3 frequency 1 3 6 10 7 7 4 pe r c e n t 3% 8% 16% 26% 18% 18% 11% " s t r i c t r e l a x e d a c t u a l random a c t u a l random % under 26% 38% % under 11% 19% % same 26% 25% % same 61% 63% % over 47% 38% % over 29% 19% A n a r c h y - C o n f l i c t Index magnitude of e r r o r -3 -2 -1 0 1 2 3 frequency 1 1 10 9 8 5 4 pe r c e n t 3% 3% 26% 24% 21% 13% 11% s t r i c t a c t u a l random % under 32% 38% % same 24% 25% % over 45% 38% r e l a x e d a c t u a l random % under 5% % same 71% % over 24% 19% 63% 19% percentages may not add up to 100% due to rounding e r r o r s The frequencies of actual accurate and erroneous r e s u l t s indicate that the hypothesized r e l a t i o n s h i p i s not accurate. The hypothesis projects a smaller frequency of errors i n actual results than the frequency of errors i n an expected random d i s t r i b u t i o n . That i s , i n terms of (lover), (%under), and (%same), the hypothesized projection i s that 157 the actual re s u l t of the d i s t r i b u t i o n of (%over) and (%under) w i l l be less than i n the expected random d i s t r i b u t i o n , while the actual (Isame) w i l l be greater than i n the expected random d i s t r i b u t i o n . Table 5.9 reveals that the only s i g n i f i c a n t deviation between actual and expected random d i s t r i b u t i o n s i s between (%same) and (lover) frequencies for the anarchy-hostility escalation rate. In t h i s case the actual frequency of (Isame) i s 8% where an expected random frequency i s 25%, and the actual frequency of (lover) i s 61% where an expected random frequency i s about 38%. Differences between actual and expected random d i s t r i b u t i o n i n the remaining relationships do not exceed 10%. While there i s a tendency for the actual frequency of (lover) to exceed an expected random frequency, the tendency i s not strong. When error the c a l c u l a t i o n i s relaxed, no s i g n i f i c a n t changes from s t r i c t l y calculated one occur. These results indicate that the actual d i s t r i b u t i o n of (Isame), (lunder), and (lover) i s not as hypothesized, and i s , with the one noted expectation, not s i g n i f i c a n t l y d i f f e r e n t from an expected random d i s t r i b u t i o n . This compels r e j e c t i o n the c o n f l i c t portion of the hypothesis. 158 Goodman-Kruskal gamma results Gamma resu l t s of matched-pair data, reported i n Table 5.10, supports acceptance of the n u l l hypothesis f o r the h o s t i l i t y magnitude and c o n f l i c t index. No coding format of the c o n f l i c t index-anarchy relationship y i e l d s a s i g n i f i c a n t p-value of ps.20, and the h o s t i l i t y magnitude-anarchy re l a t i o n s h i p i s only s i g n i f i c a n t i n di-chotomous coding format (p=.09). Table 5.10 Anarchy-Conflict: Goodman-Kruskal Gamma Results Anarchy quad-chotomoua Anarchy t r i -chotomous Anarchy d i -chotomous Average P P P P gamma value gamma value gamma value gamma va l u e H o s t i l i t y E s c a l a t i o n Rate -.27 .02 .36 .01 -.39 .01 -.34 ps.02 H o s t i l i t y Magnitude .NS .NS .17 .09 .17 p=.09 C o n f l i c t Index .NS .NS .NS NS 159 However, gamma results y i e l d s i g n i f i c a n t relationships between h o s t i l i t y rate and anarchy (ps.02) i n each of the three coding formats. These results suggest a s t a t i s t i c a l l y discordant re l a t i o n s h i p between h o s t i l i t y escalation rate and anarchy. However, these relationships are weak since the highest magnitude gamma c o e f f i c i e n t i s - . 3 9 . These gamma resu l t s also suggest that d i f f e r e n t aspects of c o n f l i c t have d i f f e r e n t relationships to anarchy, and that the hypothesized d i r e c t i o n i s not support i s the r e s u l t s . Therefore, these results concur with the other t e s t s : the hypothesized anarchy-conflict relationship i s rejected. Anarchy-conflict summary The results of the Goodman-Kruskal gamma, the Kruskal-Wallis H-test, and the s i m i l a r i t y of actual and expected random d i s t r i b u t i o n do not support accepting the anarchy-c o n f l i c t portion of the hypothesis. While the hypothesized average of c o n f l i c t and cooperation scores at each l e v e l of anarchy, and t h e i r correspondence to le v e l s of anarchy, are generally accurate (with the exception of anarchy l e v e l 2), the difference among the samples at each l e v e l of anarchy, from which the averages are calculated, are not s i g n i f i c a n t l y d i f f e r e n t . The only consistent s t a t i s t i c a l l y s i g n i f i c a n t relationship i s between h o s t i l i t y escalation rate and anarchy, where both H-test and gamma c o e f f i c i e n t s 160 i n each coding format y i e l d s i g n i f i c a n t p-values (gammas at ps.02 and H-test at p=.12). The only other c o n f l i c t v a r i a b le which has a s t a t i s t i c a l l y s i g n i f i c a n t r e l a t i o n s h i p to anarchy i s the h o s t i l i t y magnitude i n di-chotomous matched-pair form which y i e l d a gamma c o e f f i c i e n t of .17 (p=.09). The results of the frequency of error analysis suggests that the d i s t r i b u t i o n of accurate and inaccurate relationships i n the anarchy-conflict portion of the hypothesis does not, with the exception of the h o s t i l i t y escalation rate, d i f f e r s i g n i f i c a n t l y from an expected random d i s t r i b u t i o n . Based on these results the c o n f l i c t portion of the anarchy hypothesis cannot be accepted. Of notable interest i s the d i r e c t i o n of the anarchy-h o s t i l i t y escalation rate relationship, which i s the reverse of that projected. Instead of po s i t i v e r e l a t i o n s h i p between escalation rates and level s of anarchy, the rel a t i o n s h i p i s negative. Since c o n f l i c t scores are coded so that a higher score indicates more c o n f l i c t and a lower score indicates l e s s c o n f l i c t , the negative gamma c o e f f i c i e n t s of -.27, -.36, and -.39 (ps.02) indicate that as the l e v e l of anarchy increases the h o s t i l i t y escalation rate scores decrease. 161 Anarchy Cooperation Relationship H-test results The cooperation portion of the hypothesis projects that when anarchy i s higher cooperation w i l l be lower and that when anarchy i s lower cooperation w i l l be higher. As with the analysis of the conflict-anarchy relationship, the cooperation-anarchy relationship w i l l be examined with the same tests and data forms. Results of relationship between the ordinal values of average cooperation and anarchy scores are s i m i l a r to those between c o n f l i c t and anarchy. The projected correspondence and sequence of cooperation scores to le v e l s of anarchy do not follow the hypothesized pattern unless anarchy l e v e l 2 i s discounted. However, i n the case of the r e c i p r o c i t y direction-anarchy relationship ignoring the anarchy l e v e l 2 case does not eff e c t the discrepancy between projected and actual patterns. Of the three cooperation variables the only s i g n i f i c a n t H-test res u l t i s the r e c i p r o c i t y distance (p=.18). 162 Table 5.11 P r o j e c t e d and A c t u a l F i n d i n g s : Anarchy and Cooperation Scores L e v e l of Anarchy Cooperation V a r i a b l e s r e c i p r o c i t y d i s t a n c e r e c i p r o c i t y d i r e c t i o n c o o p e r a t i o n index h y p o t h e s i z e d r e s u 1 t s low med low med high h i g h low med low med h i g h high low med low med h i g h h i g h a v e r a g e s c o r -0 .32 -1.14 0 . 02 0 .21 •0.007 •0.60 0. 06 -0 . 52 -0 .32 •1 .74 0 . 09 0.15 n=10 n= 1 n=13 n=14 a v e r a g e o r d i n a 1 2 1 3 4 2 1 3 4 K r u s k a l -W a l l i s H-Test .18 .67 .34 163 When variables are collapsed into tri-chotomous form with l e v e l s 2 and 3 anarchy placed i n the same value category, the correspondence and sequence of average Table 5.12 P r o j e c t e d and A c t u a l F i n d i n g s : Anarchy and Coo p e r a t i o n tri-chotomous form L e v e l of Anarchy Cooperation V a r i a b l e s r e c i p r o c i t y d i s t a n c e r e c i p r o c i t y d i r e c t i o n c o o p e r a t i o n index h y P o t h e s i z e d a v e r a g a v e r a g e r e s u 1 t s s c o r e o r d i n a 1 low medium-high •0 .32 -0 . 06 0.21 1 2 3 low medium high •0 . 007 •0 . 02 •0 . 52 1 2 3 low medium hi g h •0.32 n=10 •0.04 n=14 0.15 n=14 1 2 3 K r u s k a l - W a l l i s H-Test .35 .66 .65 cooperation variable scores to anarchy are as hypothesized. However, no s i g n i f i c a n t H-test re s u l t i s present. These 164 r e s u l t s compel r e j e c t i o n of the anarchy cooperation portion of the hypothesis. Difference in accurate and erroneous results in actual and random distributions Table 5.13 reports the d i s t r i b u t i o n s of actual and expected random frequencies of accurately and erroneously hypothesized relationships. The frequency of erroneous and accurate relationships with both s t r i c t and relaxed error calculations are also reported. Table 5.13 Frequency of Errors in Actual and Random Distributions A n a r c h y - R e c i p r o c i t y D i s t a n c e magnitude of e r r o r -3 -2 -1 0 1 2 3 frequency 1 4 7 12 7 6 1' pe r c e n t 3% 11% 18% 32% 18% 16% 3% s t r i c t r e l a x e d a c t u a l random a c t u a l random % under 32% % same 32% % over 37% 38% 25% 38% % under 13% % same 68% % over 18% 19% 63% 19% 165 Table 5.13 cont i n u e d ' A n a r c h y - R e c i p r o c i t y D i r e c t i o n magnitude of e r r o r -3 -2 -1 0 1 2 3 frequency 1 4 8 13 5 3 4 p e r c e n t 3% 11% 21% 34% 13% 8% 11% s t r i c t r e l a x e d a c t u a l random a c t u a l random % under 34% 38% % under 13% 19% % same 34% 25% % same 68% 63% % over 32% 38% % over 18% 19% Anarchy-Cooperation Index -3 -2 -1 0 1 2 3 frequency 2 3 9 11 8 2 3 pe r c e n t 5% 8% 24% 29 s i 21% 5% 8% s t r i c t a c t u a l random % under 37% 38% % same 29% 25% % over 34% 38% r e l a x e d a c t u a l random % under 13% 19% % same 74% 63% % over 13% 19% Table 5.13 reveals that the difference between actual and random d i s t r i b u t i o n s are not s i g n i f i c a n t l y d i f f e r e n t . In fact, the largest difference between actual and expected random res u l t s with s t r i c t conditions i s 9% and with relaxed conditions i s 11%. The small difference between actual and expected random frequency d i s t r i b u t i o n s indicates that the hypothesized r e s u l t s are not achieved and that, based on t h i s evidence, 166 the anarchy cooperation portion of the hypothesis must be rejected. Goodman-Kruskal gamma results Matched-pair gamma significance r e s u l t s y i e l d are generally consistent with hypothesized r e s u l t s . Each of the three cooperation variables has a s i g n i f i c a n t r e l a t i o n s h i p Table 5.14 Anarchy-Cooperation Relationship: Goodman-Kruskal Gamma Results Anarchy quad-chotomous Anarchy t r i -chotomous Anarchy d i -chotomous Average P P P P gamma va l u e gamma value gamma va l u e gamma va l u e R e c i p r o c i t y .24 Di s t a n c e . 05 .31 .01 .39 .01 .31 ps.05 R e c i p r o c i t y .18 D i r e c t i o n 08 NS .22 .05 .20 ps.08 Coo p e r a t i o n .30 .01 Index NS .27 .02 .29 ps.02 to anarchy at p-values of ps.08. The exceptions are the r e c i p r o c i t y direction-anarchy and cooperation-anarchy relationships i n the tri-chotomous coding format, where no 167 s i g n i f i c a n t p-value i s found. However since gamma co e f f i c i e n t s do not exceed .39, the relationships are r e l a t i v e l y weak. These results suggest that while the hypothesis i s s t a t i s t i c a l l y s i g n i f i c a n t as a concordant relationship, the accuracy of the hypothesis i s modest, as indicated by gamma c o e f f i c i e n t s of gammas.39. Anarchy-cooperation summary While the results of the tests of the anarchy-cooperation portion of the hypothesis yielded d i f f e r e n t r e s u l t s , the preponderance of evidence suggest that the hypothesis i s not accurate. The Kruskal-Wallis H-test and the d i s t r i b u t i o n of accurate and inaccurate pairings compel rejection, while the matched-pair Goodman-Kruskal suggests acceptance, a l b e i t at a modest l e v e l of association. The Kruskal-Wallis H-test yielded a single s i g n i f i c a n t relationships between r e c i p r o c i t y distance and anarchy (p=.18). No other s i g n i f i c a n t relationship was found. Based on H-test results, the anarchy-cooperation portion of the hypothesis cannot be accepted. The d i s t r i b u t i o n of accurate and inaccurate anarchy-cooperation re s u l t s does not d i f f e r s i g n i f i c a n t l y from a random d i s t r i b u t i o n . This also compels r e j e c t i o n of the anarchy-cooperation portion of the hypothesis. 168 On the other hand, the Goodman-Kruskal matched-pair analysis yielded results which supports accepting the anarchy-cooperation portion of the hypothesis as concordant. Each set of anarchy-cooperation variables i n both quad- and di-chotomous form y i e l d s a s i g n i f i c a n t p-values of ps.08. The r e c i p r o c i t y d i r e c t i o n and the cooperation index variables were not s i g n i f i c a n t l y related to anarchy i n t r i -chotomous form. The gamma c o e f f i c i e n t s are a l l positive, i n d i c a t i n g that the hypothesized d i r e c t i o n of the relati o n s h i p i s accurate. Among the cooperation variables, r e c i p r o c i t y distance and anarchy yielded the highest gamma c o e f f i c i e n t s , with a range of .24 to .39 (ps.05). . These results c l e a r l y indicate support for accepting the anarchy-cooperation part of the hypothesis as s t a t i s t i c a l l y s i g n i f i c a n t , but only associated at a low l e v e l . H-test significance .results d i f f e r from gamma sign i f i c a n c e results because t h e i r c a l c u l a t i o n of variance i s d i f f e r e n t . The difference i s that balanced variance has a d i f f e r e n t e f f e c t on the significance of the H-test and gamma c o e f f i c i e n t s . Balanced variance i s when the number of under-estimated and over-estimated errors are roughly equal. When t h i s occurs gamma c o e f f i c i e n t s are smaller and have larger p-values. An unbalance variance (either over or under) i n the estimate can produce high gamma c o e f f i c i e n t s with low p-values since the ca l c u l a t i o n of gamma involves 169 calculations of the frequency of when one value i s equal to and higher (or lower) than the other. On the other hand, balanced variance has less of an influence on the H-s t a t i s t i c since i t i s calculated with average rankings where the process of taking an average o f f - s e t each side of balanced variance. Summary of Results S i m i l a r i t y of means hypothesis Regarding the hypothesis about the s i m i l a r i t y of means and c o n f l i c t and cooperation, the results of both the H-test and the matched-pair gamma test compel r e j e c t i o n of the cooperation-similarity of means portion of the hypothesis. Neither the Kruskal-Wallis nor the Goodman-Kruskal tests yielded a single relationship with p-values of p s . 2 0 . Therefore the cooperation-similarity of means portion of the hypothesis i s rejected. The two tests yielded d i f f e r e n t r e s u l t s concerning the c o n f l i c t portion of the s i m i l a r i t y of means hypothesis. While the hypothesized sequence of average ordinal scores i s accurate, the H-test does not resul t i n a single s i g n i f i c a n t p-value. On the other hand, the matched-pair gamma test resulted i n s i g n i f i c a n t relationships ( p s . 1 0 ) between the h o s t i l i t y magnitude and the c o n f l i c t index variables. 170 The difference i n the two tests regarding c o n f l i c t indicates that while the s i m i l a r i t y of means and c o n f l i c t r e l a t i o n s h i p i s s t a t i s t i c a l l y s i g n i f i c a n t , i t i s not associated at a high l e v e l , and can only be accepted as a s t a t i s t i c a l l y s i g n i f i c a n t concordant r e l a t i o n s h i p . Yet, gamma c o e f f i c i e n t s of .15 and .24 indicate only a small l e v e l of association between the s i m i l a r i t y of means and c o n f l i c t . Based on these results the s i m i l a r i t y of means hypothesis i s rejected as a v a l i d hypothesis. Similarity of ends hypothesis Both the matched-pair and difference of average tests concur that the hypothesis about the s i m i l a r i t y of ends and cooperation can be accepted as s t a t i s t i c a l l y s i g n i f i c a n t . Both the Kruskal-Wallis H-test and Goodman-Kruskal gamma test yielded p-values of .ps.16 and pis.01,. respectively. . The average cooperation scores are as the hypothesis projects, and the gamma c o e f f i c i e n t s ranged from .41 to .48 in d i c a t i n g a moderate l e v e l of association. With regard to c o n f l i c t , the two tests yielded d i f f e r e n t r e s u l t s . The Kruskal-Wallis H-test did not y i e l d a s'ingle s i g n i f i c a n t p-value for any c o n f l i c t variable, although the average ordinal scores are as the hypothesis projects. On the other hand the matched-pair Goodman-Kruskal gamma test yielded s i g n i f i c a n t p-values f o r each 171 c o n f l i c t variable-the s i m i l a r i t y of ends r e l a t i o n s h i p . In these relationships gamma c o e f f i c i e n t s ranged from -.41 to .17 (ps.09) i n d i c a t i n g a moderate discordant and weak concordant relationships. These results suggest that the c o n f l i c t - s i m i l a r i t y of ends relationship can be accepted as concordant with regard to h o s t i l i t y magnitude and c o n f l i c t index and discordant with regard to h o s t i l i t y escalation rate. Based on these re s u l t s the s i m i l a r i t y of ends hypothesis cannot be accepted since the cooperation portion of the hypothesis i s rejected as s t a t i s t i c a l l y s i g n i f i c a n t , and the c o n f l i c t portion of the hypothesis y i e l d s a rel a t i o n s h i p i n opposition to that hypothesized. Anarchy hypothesis Two conclusions can be drawn from the re s u l t s regarding the anarchy hypothesis. .First, there are no resu l t s which support acceptance of the anarchy c o n f l i c t portion of the hypothesis. H-test, error d i s t r i b u t i o n s , and gamma resu l t s concur that the anarchy c o n f l i c t portion of the hypothesis must be rejected. While the anarchy-hostility escalation rate was found to be s t a t i s t i c a l l y s i g n i f i c a n t with matched-p a i r tests, the relationship i s discordant and negative and not, as the hypothesis projects, p o s i t i v e . Second, while the anarchy-cooperation portion of the hypothesis i s supported i n matched-pair analysis, H-test and 172 the difference between actual and random error d i s t r i b u t i o n s r e s u l t s do not support accepting the hypothesis. The Kruskal-Wallis H-test only y i e l d one s i g n i f i c a n t H-test r e s u l t f o r r e c i p r o c i t y distance (p=.18), and the anarchy-cooperation index and anarchy-reciprocity d i r e c t i o n relationships are not s t a t i s t i c a l l y s i g n i f i c a n t . These re s u l t s suggest that r e j e c t i o n of t h i s portion of the hypothesis would i s most prudent. Given these conclusion and r e s u l t s , the hypothesis i s rejected. 173 Chapter 6 Analysis of International Crisis Behavior Data Set This chapter tests the three hypotheses with data from the International Crisis Behavior data s e t . 2 0 7 With one exception the format of the chapter follows that of the previous chapter. The same tests of significance and association used i n the previous chapter are used here. Section l of the chapter, which does not have a counterpart i n the previous chapter, i s a discussion of how the o r i g i n a l ICB data has been prepared for t e s t i n g purposes. Then, i n section 2 results from the t e s t i n g of the three hypotheses are presented and analyzed. Data Preparation The data u t i l i z e d i n t h i s chapter i s the F a l l 1994 version of the ICB data set. The ICB data set i s divided 2 0 7 I acknowledge with gratitude Dr. Jonathan Wilkenfeld, p r i n c i p l e investigator, International C r i s i s Behavior Project, who provided me with the l a t e s t update of the ICB data set i n the F a l l of 1994 at the University of Maryland, College Park. 174 into two separate data bases. One data base contains data on i n d i v i d u a l actors (actor l e v e l data) and the other contains data for each c r i s i s (system l e v e l data). Not a l l cases were appropriate to include i n the analysis. The actor l e v e l data contained data on 826 states i n c r i s i s and the system l e v e l data contained data on 390 c r i s e s . Of these cri s e s , c r i s e s which had one actor, more than f i v e actors, those with coding i r r e g u l a r i t i e s , and those which are not common to both actor and system l e v e l data sets are removed from consideration. 2 0 8 Single actor c r i s e s were excluded because they lacked data on a second actor. The absence of a second actor precluded using the anarchy model to generate an anarchy score. Of 390 c r i s e s 158 are single actor c r i s e s . Also excluded, because of infreguency, are cris e s with more than 5 actors. Of 390 c r i s e s , 10 are c r i s e s with more than 5 actors. This leaves 222 cases from the system l e v e l data and 577 cases from the a c t o r - l e v e l data. Then a l l intra-war cris e s were removed. These c r i s e s were eliminated because the hypotheses being tested involved cooperation and c o n f l i c t . As such, the war s e t t i n g where "(1) the c r i s i s was an integral part of an ongoing war; (2) at least one of the p r i n c i p l e adversaries i s a continuing 208 T h e p a l l 1994 version of the data was an advanced version which had not yet undergone f i n a l "cleaning". 175 actor i n that war" 2 0 9 would bias results towards c o n f l i c t . This l e f t 142 c r i s i s cases for analysis. To obtain data on cooperation and c o n f l i c t , the systems l e v e l data i s used. The unit of analysis i s the c r i s i s . To obtain data for actors' ends and means the actor l e v e l data i s used. Then the anarchy, s i m i l a r i t y of ends, and s i m i l a r i t y of means variables are calculated from the actor l e v e l data, so that each c r i s i s has associated with i t s i m i l a r i t y of ends, s i m i l a r i t y of means, and anarchy scores. Hypothesis testing Findings Regarding the S i m i l a r i t y of Means Hypothesis The s i m i l a r i t y of means hypothesis projects that c o n f l i c t w i l l be greater at high lev e l s of s i m i l a r i t y of means than at low s i m i l a r i t y of means. A pattern consistent with the hypothesis would be re f l e c t e d by higher average c o n f l i c t and cooperation variable scores at high l e v e l s of s i m i l a r i t y , and lower average c o n f l i c t and cooperation scores at low of s i m i l a r i t y of means. Average c o n f l i c t and cooperation scores at high and low s i m i l a r i t y of means are reported i n Table 6.1 As Table 6.1 indicates, the average c o n f l i c t and cooperation scores Wilkenfeld, "System l e v e l data," 14. 176 follow the projected pattern. Each average cooperation and c o n f l i c t variable i s higher at higher s i m i l a r i t y , and lower at lower s i m i l a r i t y . Table 6.1 Projected and Actual Average Conflict and Cooperation Scores and Significance by Level of Similarity of Means Conflict Variables Cooperation Variables C r i s i s Time Management of Technique Violence similarity Centrality Intensity Overall Conflict Cooperation of means Violence Violence Violence Index of Index h y p o t h e s i z e d LS low low low low low low low r e s u l t s HS high high high high high high high average LS 2.23 1.99 2.11 2.10 3.04 2.53 2.78 n=101 s c o r e r e s u l t s HS 2.54 2.30 2.32 2.38 3.17 2.85 3.01 n= 41 average LS 1 1 1 1 1 1 1 o r d i n a l magnitude HS 2 2 2 2 2 2 2 K r u s k a l -W a l l i s H - t e s t .17 .09 .25 .15 .30 .21 .41 The Kruskal-Wallis H-test i s u t i l i z e d to test the s t a t i s t i c a l significance of the difference of samples at each l e v e l of s i m i l a r i t y of means. The Kruskal-Wallis H-test y i e l d s s i g n i f i c a n t p-values i n three of the four c o n f l i c t variables (p-values of ps.17) and no s i g n i f i c a n t 177 difference of averages among the three cooperation variables (pa.21). The average c o n f l i c t and cooperation scores at high and low s i m i l a r i t y of means and the r e s u l t i n g Kruskal-Wallis p-values for difference among average cooperation and c o n f l i c t scores suggests that the hypothesis should be accepted for the c o n f l i c t - s i m i l a r i t y of means relationship, but should be rejected for cooperation-similarity of means rel a t i o n s h i p . Turning to matched-pair analysis of association between anarchy and c o n f l i c t variables, the Goodman-Kruskal gamma i s u t i l i z e d . The Goodman-Kruskal gamma results show that among c o n f l i c t variables only the c e n t r a l i t y of violence-s i m i l a r i t y of means relationship i s s i g n i f i c a n t (p=.14), while among cooperation variables both the timing of violence and the cooperation index variable are s i g n i f i c a n t (p=.04 and p=.06, respectively). The l e v e l of association indicated by the gamma co e f f i c i e n t s suggests a weak associations between the s i m i l a r i t y of means and the c e n t r a l i t y of violence v 178 Table 6.2 Gamma Results Similarity of Means and Conflict and Cooperation gamma p-value CONFLICT C e n t r a l i t y Violence .27 .14 Intensity Violence NS Overall Violence NS C o n f l i c t Index NS COOPERATION C r i s i s Management Technique NS Timing of Violence .37 .04 Cooperation Index .34 .06 (gamma=.27), between the s i m i l a r i t y of means and cooperation index (gamma=.34), and between the s i m i l a r i t y of means and c r i s i s management technique (gamma=.37). The next a n a l y t i c a l i s to examine actual and expected random error frequency d i s t r i b u t i o n s for the s i m i l a r i t y of means-conflict and s i m i l a r i t y of means-cooperation relationships are reported i n Table 6.3. Table 6.3 reports the frequency of errors where the s i m i l a r i t y of means i s larger than (lover), less than (%under), and equal to (%same) c o n f l i c t or cooperation variables. Recall than an 179 error i s when the s i m i l a r i t y of means does not equal the c o n f l i c t or cooperation variables. Table 6.3 reveals a s l i g h t tendency f o r the actual frequency of accurate results to be larger than that expected from a random sample. This tendency i s present i n six of seven c o n f l i c t and cooperative rela t i o n s h i p s . Among the c o n f l i c t - s i m i l a r i t y of means relationships the d i s t r i b u t i o n of I under errors tends to be larger the lover errors. Yet, the lunder frequency d i s t r i b u t i o n i s not generally d i f f e r e n t from an expected random d i s t r i b u t i o n , while the lunder frequency i s generally smaller than that i n an expected random frequency. Among the cooperation-similarity of means relationships the d i s t r i b u t i o n accurately hypothesized r e s u l t s does not d i f f e r greatly from an expected random d i s t r i b u t i o n . However, the d i s t r i b u t i o n of errors of i s d i f f e r e n t from an expected random d i s t r i b u t i o n of errors. Most notably the d i s t r i b u t i o n of errors i n which the value of cooperation i s greater than the s i m i l a r i t y of means (lover) i s about one-t h i r d of the expected random d i s t r i b u t i o n of errors, while the (lunder) i s greater than an expected random d i s t r i b u t i o n . 180 Table 6.3 Frequency of Errors in Actual and Random Distributions Similarity of Means s i m i l a r i t y o f m e a n s - c e n t r a l i t y o f v i o l e n c e magnitude frequency percent* or error -1 48 34% 0 78 551 a c t u a l 1 16 11% random lunder %same %over 341 551 111 25% 50% 25% s i m i l a r i t y o f m e a n s - i n t e n s i t y o f v i o l e n c e magnitude frequency percent or error -1 . 36 25% 0 84 59% a c t u a l 1 22 15% random lunder Isame %over 25% 59% 15% 25% 50% 25% s i m i l a r i t y o f m e a n s - o v e r a l l v i o l e n c e magnitude frequency percent or error -1 38 27% 0 80 56% a c t u a l 1 24 17% random %under Isame lover 27% 56% 17% 25% 501 251 s i m i l a r i t y o f m e a n s - c o n f l i c t index magnitude frequency percent or error -1 39 27% 0 82 58% a c t u a l , 1 21 151 random -%under Isame lover 27% 58% 15% 251 501 251 181 Table 6.3 continued s i m i l a r i t y o f means-t ime o f v i o l e n c e magnitude or error -1 0 1 frequency 12 77 53 percent 8% 54% 37% a c t u a l random %under 37% 25% %same 54% 50% %over 8% 25% s i m i l a r i t y o f m e a n s - c r i s i s management t e c h n i q u e magnitude or error -1 0 1 frequency 11 57 74 percent 8% 40% 52% a c t u a l random %under 52% 25% %same 40% 50% %over 8% 25% s i m i l a r i t y o f m e a n s - c o o p e r a t i o n index magnitude or error -1 0 1 frequency 58 73 11 percent 41% 51% 8% a c t u a l random %under 41% 25% %same 51% 50% %6ver 8% 25% •percentage may not add up to 100% due to rounding errors. S i m i l a r i t y of Means: Summary of Findings The re s u l t s of the two tests concerning the hypothesis about the s i m i l a r i t y of means yielded findings which suggest d i f f e r e n t relationships between the s i m i l a r i t y of means and di f f e r e n t aspects of cooperation and c o n f l i c t . Kruskal-182 Wallis H-test results indicate that d i r e c t relationships between the s i m i l a r i t y of means and c o n f l i c t are s i g n i f i c a n t i n three of four c o n f l i c t variables, while cooperation-s i m i l a r i t y of means relationship i s not s i g n i f i c a n t with any of the three cooperation variables. On the other hand, the matched-pair Goodman-Kruskal test r e s u l t s indicate that the cooperation-similarity of means relat i o n s h i p i s s i g n i f i c a n t i n two of three cases, while the c o n f l i c t - s i m i l a r i t y of means relat i o n s h i p i s s i g n i f i c a n t i n only one of four cases. The Goodman-Kruskal test also indicates that the association between c o n f l i c t and s i m i l a r i t y of means i s not as strong as between cooperation and s i m i l a r i t y of means (.34 and .37 vs .24). These res u l t s reveal that the c o n f l i c t - s i m i l a r i t y of means rel a t i o n s h i p i s s t a t i s t i c a l l y s i g n i f i c a n t as a d i r e c t relationship, but not a concordant or discordant one, while the cooperation-similarity of means relat i o n s h i p i s s t a t i s t i c a l l y s i g n i f i c a n t as a concordant rela t i o n s h i p but not a d i r e c t one. The results of the d i s t r i b u t i o n of accurately and erroneously hypothesized results also reveals the concordant c e n t r a l i t y of v i o l e n c e - s i m i l a r i t y of means and cooperation-s i m i l a r i t y of means relationships. These findings show that the relationship between s i m i l a r i t y of means and cooperation and c o n f l i c t i s not the same. While matched-pair results indicate one s i m i l a r i t y of 183 means-conflict and two s i m i l a r i t y means-cooperation relationships are s t a t i s t i c a l l y s i g n i f i c a n t , analysis of aggregated data indicates that three c o n f l i c t and no cooperation variables have s t a t i s t i c a l l y s i g n i f i c a n t differences at each l e v e l of s i m i l a r i t y . Of notable interest, the two tests concur i n only three instances: (1) that the c e n t r a l i t y of violence and s i m i l a r i t y of means relationship i s s t a t i s t i c a l l y s i g n i f i c a n t , (2) that the c r i s i s management technique-s i m i l a r i t y of means and (3) ov e r a l l v i o l e n c e - s i m i l a r i t y of means relationships are not s t a t i s t i c a l l y s i g n i f i c a n t . Given these results the hypothesis cannot be accepted. Findings regarding the s i m i l a r i t y of ends hypothesis The s i m i l a r i t y of ends hypothesis projects a pattern of higher c o n f l i c t and cooperation scores at high l e v e l s of s i m i l a r i t y , and lower c o n f l i c t and cooperation scores at lower l e v e l s of s i m i l a r i t y . Hypothesized and actual r e s u l t s are reported i n Table 6.4. Average c o n f l i c t and cooperation scores at high and low s i m i l a r i t y of ends follow the pattern projected i n the hypothesis. Each of the four c o n f l i c t variables y i e l d s s i g n i f i c a n t H-test r e s u l t s (at ps.02). Among the cooperation variables only the c r i s i s management technique i s s i g n i f i c a n t at a p-value of p=.05. 184 Table 6.4 Hypothesized and Actual Average Conflict and Cooperation Scores and Significance by Level of Similarity of Ends Conflict Variables Cooperation Variables •imilarity Centrality Intensity Overall Confliot C r i s i s Time Cooperation of ends Violence Violence Violence Index Management of Index Technique Violence h y p o t h e s i z e d LS low low low low low low low r e s u l t s HS high high high high high high high average LS 2.05 1.79 1.95 1.93 2.92 2.49 2.70 n=61 s c o r e r e s u l t HS 2.52 2.30 2.33 2.38 3.20 2.73 2.96 n=81 average 1 1 1 1 1 1 1 o r d i n a l magnitude 2 2 2 2 2 2 2 K r u s k a l W a l l i s H - t e s t .02 .002 .02 .006 .05 .31 .22 When the hypothesis i s tested with matched-pair data the r e s u l t s confirm acceptance of the s i m i l a r i t y of ends-c o n f l i c t r e l a t i o n s h i p . In these relationships the hypothesis can be accepted with p-values of ps.03 f o r each s i m i l a r i t y of end-conflict relationship. 185 Matched-pair results, reported i n Table 6.5, concerning cooperation suggest a s t a t i s t i c a l l y s i g n i f i c a n t concordant r e l a t i o n s h i p between timing of violence and s i m i l a r i t y of ends and cooperation index and s i m i l a r i t y of ends. Table 6.5 Similarity of Ends and Conflict and Cooperation Goodman-Kruskal gamma results gamma p-value CONFLICT Ce n t r a l i t y Violence .40 .01 Intensity Violence ; .45 .006 Overall Violence .34 .03 C o n f l i c t Index .42 .01 COOPERATION C r i s i s Management Technique NS Timing of Violence .24 .14 Cooperation Index .34 .04 Table 6.6 reports the d i s t r i b u t i o n of erroneous and accurate results for both actual and expected random d i s t r i b u t i o n s . Results of the error d i s t r i b u t i o n s exhibit a s l i g h t tendency for the accurately hypothesized r e s u l t s to 186 Table 6.7 Frequency of Errors in Actual and Random Distributions Similarity of Ends s i m i l a r i t y o f e n d s - c e n t r a l i t y o f v i o l e n c e magnitude or error -1 0 1 frequency 24 86 32 percent 17% 611 231 a c t u a l random lunder 171 251 Isame 611 501 lover 231 251 s i m i l a r i t y o f e n d s - i n t e n s i t y o f v i o l e n c e magnitude or error -1 0 1 frequency 16 84 42 percent 111 591 301 a c t u a l random lunder 111 251 Isame 591 501 lover 301 251 s i m i l a r i t y o f e n d s - o v e r a l l v i o l e n c e magnitude or error -1 0 1. frequency 18 80 44 percent 131 561 311 a c t u a l random lunder 131 251 Isame 561 501 lover 311 251 s i m i l a r i t y o f endB - c o n f l i c t index magnitude of error -1 0 1 frequency 18 84 40 percent 13% 591 281 a c t u a l random lunder 131 251 Isame 591 50% lover 281 251 187 Table 6.6 continued s i m i l a r i t y o f ends - t ime o f v i o l e n c e magnitude of error frequency percent -1 30 22% 0 81 57% 1 31 21% a c t u a l random %under %same %over 22% 57% 21% 25% 50% 25% s i m i l a r i t y o f e n d s - c r i s i s management t e c h n i q u e magnitude of error frequency percent -1 45 32% 0 75 53% - 1 22 15% a c t u a l random %under 32% 25% %same 53% 50% %over 15% 25% s i m i l a r i t y o f e n d s - c o o p e r a t i o n index magnitude of error -1 0 1 frequency 32 8 5 25 percent 23% 60% 18% a c t u a l random %under 23% 25% %same 60% 50% %over 18% 25% exceed the expected random d i s t r i b u t i o n . Yet t h i s tendency-i s not dramatic, ranging from three to eleven percentage points more than the 50% of the expected random d i s t r i b u t i o n . The error d i s t r i b u t i o n of hypothesized c o n f l i c t -s i m i l a r i t y of ends relationship exhibits a tendency where 188 c o n f l i c t i s less than the s i m i l a r i t y of means. The frequency of these cases exceeds cases where c o n f l i c t i s more than the s i m i l a r i t y of ends by more than a factor of two i n three c o n f l i c t relationships. This tendency i s the reverse of that found with cooperation relationships, where the tendency i s for s i m i l a r i t y of ends to be larger than cooperation. S i m i l a r i t y of Ends Hypothesis: Summary of Findings These re s u l t s indicate that the hypothesis about the rel a t i o n s h i p between s i m i l a r i t y of ends and c o n f l i c t i s accepted by standards of both Goodman-Kruskal gamma and the Kruskal-Wallis H-test. The results of the two tests suggest that the s t a t i s t i c a l significance of the c o n f l i c t - s i m i l a r i t y of ends rela t i o n s h i p hypothesis i s robust, and should be accepted. However, the two tests y i e l d d i f f e r e n t r e s u l t s regarding the cooperation part of the hypothesis. Neither test concurred with the other. This suggests that the cooperation-similarity of ends relationship i s not straight forward. As was the case with the s i m i l a r i t y of means, the di r e c t r e l a t i o n s h i p between cooperation and ends cannot be generally accepted since only the c r i s i s management technique y i e l d s a s i g n i f i c a n t H-test p-value (p=.05). On the other hand, the gamma c o e f f i c i e n t s for the timing of 189 v i o l e n c e - s i m i l a r i t y of ends and the cooperation index-s i m i l a r i t y of ends relationships y i e l d s i g n i f i c a n t p-values (ps.14). Given the disagreement of the tests the hypothesized cooperation-similarity of ends r e l a t i o n s h i p must be rejected. Findings regarding the anarchy hypothesis The discussion of results w i l l f i r s t address the c o n f l i c t portion of the hypothesis, and then turn to the cooperation part of the hypothesis. Anarchy-Conflict Relationship H-test and Kendall results The projected, actual average, and average ordinal magnitude of c o n f l i c t variables for each l e v e l of anarchy are reported below i n tables 6.7. At f i r s t glance the average c o n f l i c t scores i n Table 6.7 indicate mixed support 190 Table 6.7 Projected, Actual Average, and Actual Ordinal Conflict Scores by Level of Anarchy L e v e l o f A n a r c h y C o n f l i c t V a r i a b l e s C e n t r a l i t y I n t e n s i t y O v e r a l l C o n f l i c t V i o l e n c e V i o l e n c e V i o l e n c e Index h y p o t h e s i z e d r 1 low low low low e 8 2 med low med low med low med low U 1. 3 med high med high med high med high t B 4 high high high high a 1 2 06 1 81 1 98 1 95 n=52 v e s c 2 2 00 1 67 1 78 1 82 n= 9 r a o r 3 2 41 2 18 2 24 2 28 n=4 9 g e e 4 2 69 2 47 2 47 2 54 n=32 a v e r a g e o r d i n a 1 2 1 3 4 Kruskal-Wallis H-test .10 .01 .08 .04 191 f o r accepting the hypothesis. Among the c o n f l i c t variables the average ordinal magnitude of each c o n f l i c t v a r i a ble i s 2,1,3,4 when the projection i s 1,2,3,4. The anomaly between projected and actual i s that each c o n f l i c t variables' actual average score for anarchy l e v e l s 1 and 2 i s reversed. That i s , instead of the projected 1,2 ordinal sequence of c o n f l i c t variables for anarchy l e v e l s l and 2, respectively, the actual ordinal sequence i s 2,1. This denotes that when anarchy i s equal to 2 the l e v e l of c o n f l i c t i s lower than when anarchy i s equal to 1, a pattern which i s contrary to that hypothesized. 2 1 0 The actual ordinal sequence f o r c o n f l i c t variables at level s 3 and 4 of anarchy i s the same as the hypothesized pattern, with a higher average c o n f l i c t scores at anarchy l e v e l 4 than at anarchy l e v e l 3. The difference of samples from which average scores are calculated i s s i g n i f i c a n t at p-values of ps.10, as measured by the Kruskal-Wallis H-test. This indicates that the hypothesis about anarchy and c o n f l i c t i s v a l i d , and that with the noted caveat, i t i s also substantively accurate. The agreement of the ordinal magnitude of average scores can be measures with Kendall's C o e f f i c i e n t of 2 1 0 This reversal of actual and projected rankings also occurred i n the BCOW data. And, as with BCOW, the ICB anarchy l e v e l 2 compromises a dis-proportionally small part of the sample, at only 6% of the cases (in BCOW anarchy l e v e l 2 compromised 3% of the cases). 192 Concordance. Taking the average c o n f l i c t scores i n ordinal sequence i s v a l i d since the difference among the samples from which averages are calculated i s s i g n i f i c a n t at p-values of ps.10. As such, using ordinal values of the averages, the Kendall C o e f f i c i e n t of Concordance can be applied, and measures of association among c o n f l i c t and anarchy variables can be calculated. These findings, which are reported i n Table 6.8, indicates than the ranking of anarchy and c o n f l i c t variables are associated at the .93 l e v e l and that the .93 l e v e l i s s i g n i f i c a n t at p=.002. However the above noted proj ected ranking anomaly of c o n f l i c t scores at the lowest two lev e l s of anarchy indicates that the hypothesis i s not represented i n the data at the two lowest lev e l s of anarchy. This suggests that the hypotheses i s not accurate at lower l e v e l s of anarchy or c o n f l i c t . Yet, i t must be noted that the number of cases i n anarchy at l e v e l 2 i s n=9, and represents a disproportional small number of cases (about 6% of 142). When these nine cases are discounted actual and projected ordinal rankings are i d e n t i c a l . The Kendall c o e f f i c i e n t with the nine anarchy=2 cases removed i s a perfect 1.00 with a p-value of p=.05. 193 Table 6.8 Kendall Coefficient of Concordance Anarchy Centrality Violence Intensity Violence Overall Violence a l l anarchy levels 1 2 3 4 2 1 3 4 2 1 3 4 2 1 3 4 W= . 93 ps . 002 anarchy l e v e l 2 removed 1. 2 2 2 W=1.00 3 4 3 4 3 4 3 4 p= . 05 When anarchy i s collapsed into tri-chotomous form, actual and hypothesized results are the same for every c o n f l i c t variable. The lowest c o n f l i c t scores occur at the lowest l e v e l of anarchy, the middle range c o n f l i c t scores occur at the middle l e v e l of c o n f l i c t , and the highest c o n f l i c t scores occur at .the highest l e v e l of anarchy. There i s no exception. Collapsing the smaller-n (n=9) anarchy l e v e l 2 with anarchy l e v e l 3 into a mid-range value i n tri-chotomdus form removes both the in-congruencies of the o r i g i n a l quad-chotomous form, as Table 6.9 indicates. The difference of averages are also s i g n i f i c a n t at the ps.08 l e v e l . 194 Table 6.9 Projected, Actual Average, and Actual Ordinal Conflict and Cooperation Variables Scores by Level of Anarchy in in Tri-chotomous Form L e v e l o f A n a r c h y C o n f l i c t V a r i a b l e s C e n t r a l i t y I n t e n s i t y O v e r a l l C o n f l i c t V i o l e n c e V i o l e n c e V i o l e n c e Index h y low low low P r 1 low o e t s h u 2 med med med med e 1 s t i s 3 high high high high z e d a v s 1 2 . 06 1.81 1. 98 1.95 n=52 e c r o 2 2.34 2 .10 2 .17 2.21 n=58 a r a e 3 2.69 2.47 2.47 2.54 n=32 e a o • v r 1 1 . 1 1 1 e d "\ r i 2 2 2 2 2 a n g a 3 3 3 3 3 e 1 Kruskal-Wallis H-test . 07 .01 . 08 .03 195 The rankings of the average ordinal scores i n t r i -chotomous format correspond p e r f e c t l y to each other with a Kendall C o e f f i c i e n t W value of 1.00 at p=.05, as indicated i n Table 6.10. Table 6.10 Kendall C o e f f i c i e n t of Concordance Among Anarchy and C o n f l i c t Variables, Tri-chotomous Anarchy 1 2 3 Cen t r a l i t y Violence 1 2 3 Intensity Violence 1 2 3 Overall Violence 1 2 3 W=1.00 p=.05 Difference i n accurate and erroneous r e s u l t s i n actual and random d i s t r i b u t i o n s Table 6.11 reports the frequency of accurate and erroneous re s u l t s for actual and expected random d i s t r i b u t i o n s for each of the c o n f l i c t variables with both s t r i c t and relaxed error calculations. Recall that a s t r i c t hypothesis i s the d i r e c t r e l a t i o n where any deviation from the hypothesis i s treated as an error, and the relaxed 196 hypothesis i s when an error i s when the anarchy-conflict variable difference i s greater than 1. Table 6.11 reveals that the difference between actual and expected random d i s t r i b u t i o n s i s n e g l i g i b l e . While the percentage of actual accurate results (%same) tends to be higher that an expected random d i s t r i b u t i o n , the difference between the two does not exceed 14%. Also revealed i s a tendency for lover errors to exceed lunder errors i n the anarchy-conflict relationship. This suggest that when anarchy and c o n f l i c t are not equal, the tendency i s f o r the l e v e l of c o n f l i c t to be less than the l e v e l of anarchy. When error calculations are relaxed the hypothesized r e s u l t s f a i r less well. Under relaxed conditions, the difference between actual and an expected random d i s t r i b u t i o n are less that under s t r i c t conditions. These results suggest that the c o n f l i c t portion of hypothesis i s not accurate and cannot be accepted: the s i m i l a r i t y of actual and random d i s t r i b u t i o n s are too close to accept the actual patter as an associative r e l a t i o n s h i p . Table 6.11 Difference of Actual and Random Distributions A n a r c h y - C o n t r a i l t y o f V i o l e n c e magnitude o f e r r o r -3 -2 -1 0 1 2 3 f r e q u e n c y 11 11 13 56 25 19 7 p e r c e n t 8%. 8% 9% 39% 18% .13% 5% s t r i c t r e l a x e d a c t u a l random a c t u a l randc % under 25% 38% % under 15% 19% % same 39% 25% % same 66% 63% % over 36% 38% % over 18% 19% A n a r c h y - I n t e n s i t y o f V i o l e n c e magnitude o f e r r o r -3 -2 -1 0 1 2 3 f r e q u e n c y 4 11 13 54 30 23 7 p e r c e n t 3% 8% 9% 38% 21% 16% 5% s t r i c t r e l a x e d a c t u a l random a c t u a l randc % under 20% 38% , % under 11% 19% % same 38% 25% • % same 68% 63% % over 42% 38% % over 21% 19% A n a r c h y - O v e r a l l V i o l e n c e magnitude o f e r r o r -3 -2 -1 0 • 1 2 3 frequency 4 13 18 4 9 29 23 6 percent 3% 9% 13% 35% 20% 16% 4% s t r i c t a c t u a l random % under 25% 38% % same 35% 25% % over 41% 38%. r e l a x e d a c t u a l random % under % same % over 12% 68% 20% 19% 63% 19% 198 table 6.11 continued A n a r c h y - C o n f l i c t Index magnitude o f e r r o r -3 -2 -1 0 1 2 3 frequency 11 5 19 54 23 23 7 percent 8% 4% s t r i c t 13% 38% 16% 16% r e l a x e d 5% a c t u a l random a c t u a l random % under 25% 38% % under 11% 19% % same 39% 25% % same 68% 63% % over 37% 38% % over 21% 19% Goodman-Kruskal gamma resu l t s Turning to matched-pair analysis of association between anarchy and c o n f l i c t variables, the Goodman-Kruskal gamma i s u t i l i z e d as i t was for the hypotheses about the s i m i l a r i t y of ends and means. Table 6.12 reports matched-pair gamma res u l t s of each c o n f l i c t and anarchy variable. Results from the matched-pair gamma test y i e l d s s i g n i f i c a n t relationships (ps.ll) between each of the c o n f l i c t variables and anarchy i n each coding format. With one exception each p-value i s ps.01. This denotes that the hypothesis about c o n f l i c t and anarchy can be accepted as a s t a t i s t i c a l l y s i g n i f i c a n t relationship. However association l e v e l s between anarchy and c o n f l i c t are moderate at best. Knowledge of the l e v e l of anarchy only increases the accuracy the c o n f l i c t estimate by 25% to 53%. 199 Table 6.12 Summary of gamma coefficients Anarchy quad-chotomous Anarchy t r i -chotomous Anarchy d i -chotomous Average gamma p - v a l u e gamma p - v a l u e gamma p - v a l u e gamma p - v a l u e Centrality Violence Intensity Violence Overall Violence Co n f l i c t Index .25 .01 .33 .001 .25 .01 .29 .004 .27 .001 .37 .002 .31 .001 .29 .01 .53 .001 .53 .006 .36 .11 .49 .02 .35 ps.01 .41 ps.006 .31 p s . l l .33 ps.02 Conflict-Anarchy Summary of Results The Kruskal-Wallis H-test, Goodman-Kruskal gamma, and Kendall's Coefficient of Concordance each y i e l d support to accept the alternative hypothesis as s t a t i s t i c a l l y s i g n i f i c a n t . S t a t i s t i c a l l y s i g n i f i c a n t relationships exists between and among anarchy and each of c o n f l i c t variables. Association l e v e l s generated with the Kendall and Goodman-Kruskal tests, as well as the Kruskal-Wallis H-test also indicate that substantive differences e x i s t between anarchy and each of the c o n f l i c t variables. The ranges of s t a t i s t i c a l l y s i g n i f i c a n t association, which varies from highs of .93 and 1.00 (Kendall W coe f f i c i e n t s ) to lows of 200 .25 to .53 (gamma c o e f f i c i e n t s ) , indicate that the substantive accuracy of the c o n f l i c t portion of the hypothesis i s varied. In relationships analyzed with aggregated data (averages), the association i s quite high (1.00 and .93) while i n matched-pair analyses i t i s quite a b i t lower (.25 to .53). This suggests that i n general, as indicated by aggregated data, the c o n f l i c t and anarchy portion of the hypothesis i s s i g n i f i c a n t and accurate. When matched-pair data i s considered, the accuracy of the c o n f l i c t portion of the hypothesis i s not as accurate as with more general average data. However, the s i m i l a r i t y of actual and random d i s t r i b u t i o n s of errors does not support accepting the anarchy-conflict hypothesis. Taken together these results suggest that the anarchy-c o n f l i c t relationship i s not acceptably described by the hypothesis. The fact that the matched-pair re s u l t s do not y i e l d consistently higher association c o e f f i c i e n t s (gammas.53) imply that while the conflict-anarchy r e l a t i o n s h i p i s s t a t i s t i c a l l y s i g n i f i c a n t , the accuracy of the hypothesis i s not high. The pattern of higher gamma values when the coding form i s more general (dichotomous and tri-chotomous coding formats), as well as the high Kendall W c o e f f i c i e n t s , indicate that the accuracy of the hypothesis with regard to c o n f l i c t i s accurate i n more general terms, but not s p e c i f i c terms. 201 Despite r e s u l t s which compel accepting the c o n f l i c t -anarchy rela t i o n s h i p as s t a t i s t i c a l l y s i g n i f i c a n t , and accurate i n general terms, the f a i l u r e of the tests of the rel a t i o n s h i p to y i e l d higher l e v e l s of association i n matched-pair analyses, as well as the modest difference between actual and random results, compel r e j e c t i o n of t h i s portion of the hypothesis as accurate. Anarchy-Cooperation Relationship The following examination of the rel a t i o n s h i p between anarchy and cooperation follows the same form as the examination between anarchy and c o n f l i c t . The same procedures and tests are applied. H-test r e s u l t s Among the cooperation variables, with the exception of the cooperation index variable, there i s at f i r s t glance l i t t l e s i m i l a r i t y between the actual and hypothesized r e s u l t s . While the magnitude of the cooperation index's actual and hypothesized average scores are the same, the hypothesized magnitude of the c r i s i s management technique and timing of violence averages do not correspond to actual r e s u l t s . Where the hypothesis projects the sequence of ordinal magnitudes of average scores to be 1,2,3,4, they are 202 3,2,1,4 for c r i s i s management technique and 1,4,2,3 and f o r the timing of violence. In the case of the timing of violence the hypothesis i s accurate only at the lowest l e v e l of anarchy. In the case of the c r i s i s management technique the hypothesis i s accurate at anarchy l e v e l s 2 and 4. Regarding the anarchy-timing of violence relationship, the hypothesized results are accurate when the anarchy l e v e l 2 cases (n=9, or 6% of cases) are discounted. When these cases are removed from analysis, actual and hypothesized average ordinal magnitudes are the same. However, discounting anarchy l e v e l 2 cases does not change the d i s -congruence of the c r i s i s management technique variable's actual and hypothesized r e s u l t s . Thus i n the case of c r i s i s management technique there i s questionable representation of the hypothesized r e s u l t s . The difference of samples from which averages scores are calculated for c r i s i s management technique i s s t a t i s t i c a l l y s i g n i f i c a n t at p=.12. On the other hand, the difference of the samples from which the time of violence and cooperation index variables are not s i g n i f i c a n t at ps .20. The cooperation variables' relationship to anarchy suggest that the projections are not accurate i n at least two important ways. F i r s t , hypothesized projections f a i l to 203 Table 6.13 Projected, Actual Average, and Actual Ordinal Cooperation Variables Scores by Level of Anarchy, Quad-chotomous Form L e v e l o f Anarchy C o o p e r a t i o n V a r i a b l e s C r i s i s Time Management o f Techn ique V i o l e n c e C o o p e r a t i o n Index h y p o t h e s i z e d a v e r a g e r e s u 1 t s a v s e c r o a r g e e o r d i n a 1 1 2 3 4 low med low med high high 2 . 98 2 .56 2 .10 3 .34 3 2 1 4 low med low med high high 2.38 3 .11 2-. 6 9 2 . 78 1 4 2 3 low med low med high high 2 .68 2 .83 2 .90 3 . 06 1 2 .3 4 n=52 n=09 n=49 n=32 Kruskal-Wallis H-test ,12 .28 .59 204 be accurate with respect to cooperation at mid-range l e v e l s of anarchy. Hypothesized projections are not accurate with respect to the anarchy-crisis management technique and the anarchy-timing of violence relationships, while i t i s correct f o r the anarchy-cooperation index r e l a t i o n s h i p . Second, the only variables with s t a t i s t i c a l l y s i g n i f i c a n t d i f f e r e n t sample at each l e v e l of anarchy i s the c r i s i s management technique. However, the averages of the samples do not correspond to hypothesized sequence of magnitudes. The next analysis i s to collapse the two middle rankings of anarchy lev e l s 2 and 3 together. Anarchy i n t h i s tri-chotomous form, and the average cooperation scores associated with each l e v e l of anarchy are l i s t e d i n table 6.14. The average cooperation scores at each l e v e l of anarchy are the same as the hypothesized r e s u l t . However, the difference of samples from which averages are calculated i s only s i g n i f i c a n t (p=.13) i n the case of the c r i s i s management technique. These results compel r e j e c t i o n of the cooperation part of the hypothesis. 205 Table 6.14 Projected, Actual Average, and Actual Ordinal Cooperation Variables Scores by Level of Anarchy, Tri-chotomous Form L e v e l o f A n a r c h y C o o p e r a t i o n V a r i a b l e s C r i s i s Management Technique Time o f V i o l e n c e C o o p e r a t i o n Index h y p o t • h e s i z e d r e S U 1 t s a v s e c r o, a . r g e e low med high 2 . 98 3 . 02 3 .34 low low med med high high 2 .38 2 .76 2 .78 2.68 n=52 2.89 n=58 3.06 n=32 a v e r a 9 e o r d i n a 1 Kruskal-Wallis H-test .13 .24 .43 206 Difference in accurate and erroneous results in actual and random distributions Table 6.15 reports actual and expected random d i s t r i b u t i o n s of accurate and erroneous results with both s t r i c t and relaxed error calculations. The table indicates that the difference between actual and expected random d i s t r i b u t i o n s of accurate and erroneous re s u l t s i s quite small, and that the hypothesized relationship i s not represented i n the r e s u l t s . Relaxing the error c a l c u l a t i o n does not improve the difference between an expected random d i s t r i b u t i o n and actual r e s u l t s . This indicates that the hypothesized anarchy-conflict relationships i s not accurate. One trend of interest i s the difference between %under and lover error frequencies i n the c r i s i s management technique-anarchy relationship. In t h i s case the lunder error frequency exceed lover error frequencies by a factor of three. ^ Table 6.15 Frequency and Magnitude of Errors A n a r c h y - C r i s i s Management Technique magnitude o f e r r o r s -3 -2 -1 0 1 2 3 frequency percent 20 22 14% 15% 32 23% 41 29% 14 10% 9 6% 4 3 s t r i c t r e l a x e d a c t u a l random a c t u a l random % under 52% 38% % under 3 0% 19% % same 29% 25% % same 61% 63% % over 19% 38% % over 9% 19% A n a r c h y - T i m e o f V i o l e n c e magnitude o f e r r o r s -3 -2 -1 0 1 2 3 frequency percent 16 12 11% 8% 27 19% 41 29% 25 18% 17 12% . 4 3% s t r i c t r e l a x e d a c t u a l random a c t u a l random % under 39% 38% % under 2 0% 19% % same 29% 25% % same 65% 63% % over 32% 38% % over 15% 19% A n a r c h y - c o o p e r a t i o n index magnitude o f e r r o r s -3 -2 -1 0 1 2 3 frequency percentage 10 18 7% 13% 32 23% 36 25% 30 21% 12 8% 4 3 s t r i c t a c t u a l random r e l a x e d a c t u a l random % under 42% % same 25% % over 32% 38% 25% 38% % under 2 0% % same 6 9% % over 11% 19% 63% 19% 208 Goodman-Kruskal gamma resu l t s The re s u l t s of the matched-pair gamma test y i e l d s s i g n i f i c a n t relationships between anarchy and cooperation index i n each coding format. However, the coding format does e f f e c t the magnitude and significance of the c r i s i s management technique-anarchy and timing of violence-anarchy rel a t i o n s h i p s . The anarchy-crisis management technique r e l a t i o n s h i p i s s i g n i f i c a n t i n quad- and tri-chotomous formats (ps.04), and the anarchy-cooperation index re l a t i o n s h i p i s only s i g n i f i c a n t i n the dichotomous format(p=.16). However, the association l e v e l s which are less than .54 only indicates a moderate relat i o n s h i p at best. The Goodman-Kruskal gamma test r e s u l t s indicate that the anarchy-cooperation relationship i s neither as important nor as strong as the hypothesis suggests. While gamma res u l t s of the cooperation index-anarchy and c r i s i s management technique y i e l d s i g n i f i c a n t relationships (ps.16), the association of .14 to .54 i s moderate even at the highest l e v e l . 209 Table 6.16 Summary of anarchy-cooperation gamma c o e f f i c i e n t s Anarchy Anarchy Anarchy quad- t r i - d i -chotomous chotomous chotomous Average gamma p - v a l u e gamma p - v a l u e gamma p - v a l u e gamma p - v a l u e C r i s i s Management Technique .21 .04 .27 .02 NS .24 ps . 04 Timing of Violence NS NS .44 .04 .44 ps.04 Cooperation Index .14 .16 .33 .01 .54 .01 .34 ps.16 Cooperation-anarchy results summary The general conclusion about the cooperation-anarchy re l a t i o n s h i p i s that some; aspects of the rel a t i o n s h i p are s t a t i s t i c a l l y s i g n i f i c a n t . Most notable i s the sig n i f i c a n c e of the c r i s i s management technique-anarchy relationship, which y i e l d s s i g n i f i c a n t H-test (p=.12) and s i g n i f i c a n t gamma c o e f f i c i e n t s i n two of the three coding forms, as well as an error d i s t r i b u t i o n d i f f e r e n t from an expected random d i s t r i b u t i o n . Yet, the sequence of average c r i s i s management technique scores at each l e v e l of anarchy i s only as hypothesized i n tri-chotomous format, and gamma association c o e f f i c i e n t s are moderate at best (gammas.54). 210 While the cooperation index-anarchy rel a t i o n s h i p y i e l d s s i g n i f i c a n t gamma c o e f f i c i e n t s i n each coding format, and the time of violence-anarchy relationship i s s i g n i f i c a n t i n matched-pair dichotomous format, no s i g n i f i c a n t H-test r e s u l t , and an error d i s t r i b u t i o n s i m i l a r to an expected random d i s t r i b u t i o n undermine general acceptance of t h i s portion of the hypothesis. These results compel r e j e c t i o n of the cooperation part of the hypothesis. Summary of Results S i m i l a r i t y of Means Hypothesis The s i m i l a r i t y of means hypothesis y i e l d s d i f f e r e n t r e s u l t s about cooperation and c o n f l i c t . S i g n i f i c a n t H-test res u l t s of three of four c o n f l i c t variables, and the gamma res u l t s of only a single -significant gamma c o e f f i c i e n t indicate that the s i m i l a r i t y of means-conflict r e l a t i o n s h i p i s generally s t a t i s t i c a l l y s i g n i f i c a n t as a d i r e c t but not a concordant relationship. On the other hand, the s t a t i s t i c a l l y s i g n i f i c a n t gamma results i n two of three cooperation variables and i n s i g n i f i c a n t H-test r e s u l t s indicate that the cooperation-anarchy relationship i s not di r e c t but concordant. Yet, the gamma c o e f f i c i e n t s of .37 and .34 are quite low. These results indicate that the 211 hypothesized cooperation and c o n f l i c t relationships are not accurate and that the hypothesis must be rejected. S i m i l a r i t y of Ends Hypothesis The s i m i l a r i t y of ends hypothesis also produced d i f f e r e n t results regarding cooperation and c o n f l i c t . Both s i g n i f i c a n t H-test and gamma results indicate that the s i m i l a r i t y of ends-conflict relationship i s s t a t i s t i c a l l y s i g n i f i c a n t and moderately accurate (gammas from .34 to • 45) . Regarding cooperation, the hypothesis yielded d i f f e r e n t r e s u l t s with respect to d i f f e r e n t aspects of cooperation. The c r i s i s management technique i s s i g n i f i c a n t as a d i r e c t but not concordant or discordant relationship as the s i g n i f i c a n t H-test and i n s i g n i f i c a n t gamma resu l t s indicates. On the other, hand the timing of violence and the cooperation index variables do not y i e l d a s i g n i f i c a n t difference of sample from which averages are calculated, while t h e i r gamma c o e f f i c i e n t s are s i g n i f i c a n t . This indicates that the time of violence and the cooperation index relationships are concordant but not d i r e c t . The contending H-test and gamma results compel r e j e c t i o n of the cooperation-similarity relationship, and as a re s u l t the hypothesis must be rejected. 212 Anarchy hypothesis Test of the anarchy hypothesis do not y i e l d support to accept the hypothesis. The cooperation portion of the hypothesis does not generally y i e l d results which support i t s acceptance. While the c r i s i s management technique does possess a s t a t i s t i c a l l y s i g n i f i c a n t H-test re s u l t s , the other two cooperation variables do not. And, the gamma res u l t s only y i e l d consistent results i n the case of the cooperation index variable. But, th i s support i s mixed, as indicated by the wide range of gamma c o e f f i c i e n t s (.14 to .54) depending on the coding format. The errors of hypothesized results are too simi l a r to an expected random occurrence of errors to accept i t . While the error frequency of the c r i s i s management technique i s d i f f e r e n t from an expected random d i s t r i b u t i o n , the other two cooperation variables' error frequencies are s i m i l a r to an expected random d i s t r i b u t i o n . S t a t i s t i c a l l y s i g n i f i c a n t H-test and gamma res u l t s , as well as Kendall results, support the acceptance of the c o n f l i c t portion of the hypothesis. Clearly supportive aggregated H-test and Kendall results are present. Yet, the f a i l u r e of gamma c o e f f i c i e n t s to exceed .53 weakens the support of the more general H-test and Kendall r e s u l t s . Thus, i f the hypothesis only s p e c i f i e d c o n f l i c t i t would be more accurate. But even with t h i s s p e c i f i c a t i o n the 213 r e l a t i v e l y weak association of matched-pair r e s u l t s would compel serious question about the accuracy of such a hypothesis. 214 Chapter 7 Conclusion: ICB and BCOW Results in Comparison This chapter analyses test results from both the BCOW and ICB data sets regarding the v a l i d i t y and accuracy of the of the s i m i l a r i t y of means, the s i m i l a r i t y of ends, the anarchy tenet hypotheses. Recall that accepting or rej e c t i n g the hypotheses i s based on the f i v e c r i t e r i a put for t h i n chapter 4. They are: 1. the ordinal ranking of average c o n f l i c t and cooperation scores w i l l be p o s i t i v e l y related to le v e l s of anarchy, s i m i l a r i t y of ends, and s i m i l a r i t y of means,-2. the difference of average c o n f l i c t and cooperation samples at each l e v e l of anarchy and s i m i l a r i t y w i l l be s t a t i s t i c a l l y s i g n i f i c a n t based on res u l t from the Kruskal-Wallis test; 3. Kendall's Coefficient of Concordance res u l t s w i l l y i e l d s t a t i s t i c a l l y s i g n i f i c a n t and high association c o e f f i c i e n t s among the rankings of average scores,- 2 1 1 4. regarding pair-wise results, the Goodman-Kruskal gamma results w i l l r e f l e c t p o s i t i v e and s t a t i s t i c a l l y s i g n i f i c a n t gamma c o e f f i c i e n t s between anarchy, s i m i l a r i t y of ends, and 2 1 1 This set of results i s not necessary for the s i m i l a r i t y of means and s i m i l a r i t y of ends hypotheses since the test cannot be performed with k categories where k<3. 215 s i m i l a r i t y of means, on one hand, and cooperation and c o n f l i c t , on the other hand; 5. the number of cases where the value of anarchy i s equal to the value of c o n f l i c t or cooperation w i l l be greater than the number of errors (cases where the value of anarchy i s less than or more than the value of c o n f l i c t of cooperation). The following discussion w i l l f i r s t address test r e s u l t s of the s i m i l a r i t y of means hypothesis, then the s i m i l a r i t y of ends hypothesis, and f i n a l l y the anarchy tenet hypothesis. In each of these sections we w i l l examine the outcomes from the two data sets. We w i l l f i r s t examine the f i v e c r i t e r i a , which w i l l be compared for concurrence with hypothesized outcomes. Then the results of the H-tests and Goodman-Kruskal w i l l be addressed. S i m i l a r i t y of Means, C o n f l i c t , and Cooperation: V a l i d i t y and Accuracy of HI BCOW and ICB data test results of the s i m i l a r i t y of means hypothesis are reported below i n Table 7.1. The re s u l t s are mixed. While some c r i t e r i a support of accepting the hypothesis, others demand rej e c t i n g i t . On balance, the re s u l t s as a whole suggest that the hypothesis should not be accepted. Generally concurring with the hypothesis are the re s u l t s of the ordinal magnitudes of the average c o n f l i c t 216 Table 7.1 BCOW and ICB Results Adherence to the Five Criteria of the Similarity of Means Hypothesis CONFLICT HYPOTHESIS ICB BCOW data data COOPERATION HYPOTHESIS ICB BCOW data data correspondence YES of actual and hypothesized ordinal sequences YES YES MIXED 2. significant difference of averages YES NO NO NO 3. significant N/A Kendall high Kendall N/A N/A N/A N/A N/A N/A N/A 4. significant NO gamma high gamma NO YES NO YES NO NO NA 5. error distribution accurate NO NO NO NO 217 and cooperation scores at each l e v e l of s i m i l a r i t y of means: at low s i m i l a r i t y of means the c o n f l i c t and cooperation scores are lower than t h e i r averages at high s i m i l a r i t y of means. However, the significance of these r e s u l t s must be questioned since only the ICB c o n f l i c t variables yielded s i g n i f i c a n t p-values for the differences between the magnitude of the ordinal scores. The H-test r e s u l t s from the two data sets concur that the s i m i l a r i t y of means-cooperation relationships are not s i g n i f i c a n t . And, the res u l t s from the s i m i l a r i t y of means-conflict relationships Of the two data sets contest sig n i f i c a n c e : the ICB s i m i l a r i t y of means-conflict relationships y i e l d three of four relationships as s i g n i f i c a n t while the BCOW s i m i l a r i t y of means-conflict does not y i e l d a single s i g n i f i c a n t H-test r e s u l t . Results from the remaining two c r i t e r i a also make i t d i f f i c u l t to accept the s i m i l a r i t y of means hypothesis. Gamma c o e f f i c i e n t s generated from each data sets contends re s u l t s from the other. While BCOW c o n f l i c t relationships are generally s i g n i f i c a n t , ICB c o n f l i c t relationships are not. And while ICB cooperation relationships are si g n i f i c a n t , BCOW ones are not. The contention of sign i f i c a n c e between the two data sets make i t d i f f i c u l t to accept the hypothesis. The d i s t r i b u t i o n or errors c r i t e r i o n , that the actual frequency of errors w i l l be less 218 that errors i n an expected random d i s t r i b u t i o n s , also suggest that the hypothesis should be rejected. Individual gamma and H-test results are reported below i n Table 7.2. Results from Table 7.2 highlight why the hypothesis should be rejected. F i r s t , the d i s t r i b u t i o n of s i g n i f i c a n t and i n - s i g n i f i c a n t p-values i s such that only one variable ( c e n t r a l i t y of violence) yielded s i g n i f i c a n t p-values with both the H-test and Goodman-Kruskal t e s t . Among the other twelve variables, f i v e yielded p-values above p=.20 for both H-test and Goodman-Kruskal gamma t e s t , 2 1 2 and the remaining seven variables yielded contentious r e s u l t s between the H-test and Goodman-Kruskal t e s t : where one test indicates significance, the other test does not. Second, the difference between actual and expected random d i s t r i b u t i o n s does not r e f l e c t hypothesized r e s u l t s . While actual results incur less errors than an expected random d i s t r i b u t i o n , as hypothesized, the difference i s s l i g h t and ranges from one to nine percentage points more less than the expected random d i s t r i b u t i o n . 2 1 2 Both tests concur that each of the three BCOW cooperation variables r e c i p r o c i t y distance, r e c i p r o c i t y d i r e c t i o n BCOW cooperation index, the ICB c r i s i s management technique and ove r a l l violence, and BCOW h o s t i l i t y e scalation relationships to s i m i l a r i t y of means are not s t a t i s t i c a l l y s i g n i f i c a n t . Table 7.2 Similarity of Means Hypothesis H-test and gamma Significance Results Hostility-Escalation Rate H - t e s t p - v a l u e p - v a l u e gamma v a l u e when p s . Centrality Violence .17 .14 .27 Intensity Violence . 09 Overall Violence . ns . ns ICB Conflict Index .15 . ns H o s t i l i t y Magnitude .10 . 15 BCOW Conflict Index . 05 .24 C r i s i s Management Technique . ns-Timing of Violence , 04 .37 ICB Cooperation Index . ns . 06 .34 Reciprocity Distance Reciprocity Direction . ns BCOW Cooperation Index 220 These findings indicate a relationship which i s not captured by the hypothesis. Clearly the hypothesis cannot be accepted, and must be rejected. The fact that the magnitude of average c o n f l i c t and cooperation scores at each l e v e l of s i m i l a r i t y of means correspond to the hypothesized pattern, and that s t a t i s t i c a l tests y i e l d some s i g n i f i c a n t r e s u l t s , indicates that increasing the s p e c i f i c i t y of the hypothesis may y i e l d acceptable r e s u l t s . 2 1 3 This conclusion i s not surprising given the contention between the power preponderance and power p a r i t y arguments. 2 1 4 Similarity of Ends, Conflict, and Cooperation: Validity and Accuracy of H2 As with the s i m i l a r i t y of means hypothesis, the ordinal magnitude of the c o n f l i c t and cooperation variables at each l e v e l of s i m i l a r i t y concur with hypothesized r e s u l t s . Cooperation and c o n f l i c t scores are higher at high l e v e l s of s i m i l a r i t y of ends than they are at lower at the low s i m i l a r i t y of ends. These results as, well as the those based on the other c r i t e r i a , are reported below i n Table 7.3. 2 1 3 Types of spe c i f i c a t i o n s are discussed further below a f t e r r e s u l t s from each of the three hypotheses are discussed. 2 1 4 See Levy, "Causes of war," for a review. Table 7.3 BCOW and ICB Results Adherence to the Five Criteria of the Similarity of Ends Hypothesis CONFLICT HYPOTHESIS COOPERATION HYPOTHESIS ICB BCOW ICB BCOW data data data data CORRESPONDENCE YES YES YES YES OF ACTUAL AND HYPOTHESIZED ORDINAL SEQUENCES SIGNIFICANT YES NO NO YES DIFFERENCE OF AVERAGES significant N/A N/A N/A N/A Kendall high Kendall N/A N/A N/A N/A SIGNIFICANT YES YES YES . YES GAMMA HIGH GAMMA NO NO NO NO ERROR DISTRIBUTION YES YES NO NO ACCURATE 222 Regarding the second c r i t e r i o n , the sig n i f i c a n c e of the difference of ordinal magnitudes, ICB data y i e l d s s i g n i f i c a n t H-test results for the c o n f l i c t - s i m i l a r i t y of ends relationships (ps.02), while the BCOW does not y i e l d a single p-value (ps.20) for the c o n f l i c t - s i m i l a r i t y of means rela t i o n s h i p s . Inter-data set contention i s also present i n res u l t s of the cooperation portion of the hypothesis. The BCOW cooperation-similarity of means re l a t i o n s h i p yielded s i g n i f i c a n t H-test p-values, while the ICB cooperation-similarity of means relationships only yielded a single s i g n i f i c a n t p-value. Thus, with regard to the H-test, where one data set suggests that the hypothesis i s v a l i d , the other suggest i t i s not. Goodman-Kruskal gamma results, with one exception, concur with hypothesized r e s u l t s . Both BCOW and ICB c o n f l i c t - s i m i l a r i t y of ends and cooperation-similarity of ends relationships yielded gamma c o e f f i c i e n t s with s i g n i f i c a n t p-values (ps.16). Yet gamma c o e f f i c i e n t s are moderate and do not r i s e above 0.48. The f i f t h c r i t e r i o n , the d i s t r i b u t i o n of errors, r e f l e c t s the same contention between data sets as with the H-test r e s u l t s : the re s u l t s from one data set contradict the results of the other. Individual gamma and H-test results are reported below i n Table 7.4. Two trends are nearly immediately evident i n Table 7.4. The f i r s t trend i s the nearly perfect acceptance of the hypothesis according to gamma re s u l t s . Each 223 s i m i l a r i t y of ends-conflict, and with a single exception each s i m i l a r i t y of ends-cooperation r e l a t i o n s h i p i s found to be a s t a t i s t i c a l l y s i g n i f i c a n t . The exception i s the c r i s i s management technique-similarity of ends re l a t i o n s h i p . The second trend i s more complex and involves discrepancy between tests and between data sets. Here we e s s e n t i a l l y f i n d a discrepancy between ICB and BCOW sets, as well as a discrepancy between H-test and gamma sig n i f i c a n c e r e s u l t s . This pattern i s manifest i n two ways. F i r s t , when results from one data set y i e l d s r e s u l t s which support accepting the hypothesis, results from the other data set do not. When ICB results suggest the hypothesis i s v a l i d (ICB c o n f l i c t relationships) BCOW resu l t s contest i t s acceptance, and when BCOW results suggest the hypothesis i s v a l i d (BCOW.cooperation relationships) ICB resu l t s contest i t s acceptance. Second, when one s t a t i s t i c suggest v a l i d i t y the other does not. . The discrepancy between the two data sets i s c l e a r l y v i s i b l e i s the ICB results of the ICB c o n f l i c t - s i m i l a r i t y and the BCOW cooperation-similarity of ends r e s u l t s . The former ICB test results of s i m i l a r i t y of ends-conflict yielded s i g n i f i c a n t p-values for both H-test and Goodman-Kruskal test r e s u l t s . The l a t t e r , the BCOW s i m i l a r i t y of ends-cooperation results also yielded s i g n i f i c a n t p-values Table 7.4 Similarity of Ends Hypothesis H-test and gamma Significance Results H - t e s t p - v a l u e cj annua p - v a l u e gamma v a l u e when p s . 2 0 Centrality Violence .02 .01 .40 Intensity Violence . 01 . 01 .45 Overall Violence , 02 . 01 .34 ICB Conflict Index . 01 , 01 .42 H o s t i l i t y Escalation Rate , 01 .41 H o s t i l i t y Magnitude . 09 . 17 BCOW Conflict Index . 01 .36 C r i s i s Management Technique . 05 Timing of Violence ,14 .24 ICB Cooperation Index . 04 .34 Reciprocity Distance .13 , 01 .48 Reciprocity Direction .16 . 01 .41 BCOW Cooperation Index . 09 . 01 .48 •^;>>^ >„. 225 for both H-test and Goodman-Kruskal test. However, where one data set yielded s i g n i f i c a n t p-values with both tests, the other data set y i e l d s mixed r e s u l t s . The second pattern i s found i n the discrepancy of res u l t s between tests. Discrepancy between the two data sets i s found i n the Goodman-Kruskal gamma res u l t s which c l e a r l y support accepting the s i m i l a r i t y of ends-conflict rela t i o n s h i p , and the H-test results which are much les s supportive. This pattern i s manifest i n the ICB cooperation and BCOW c o n f l i c t relationships. In these relationships when one test y i e l d s s t a t i s t i c a l s i g n i f i c a n t r e s u l t s , the other test does not: when BCOW c o n f l i c t and ICB cooperation relationships y i e l d s i g n i f i c a n t H-test re s u l t s gamma sign i f i c a n c e results are not s i g n i f i c a n t . These results reveal that the s i m i l a r i t y of ends-cooperation relationship i s not captured by the hypothesis. When both tests of one data set are supportive of accepting the hypothesized relationships, results from the other data set mandate re j e c t i o n . For example the clear cut supportive r e s u l t s of the ICB c o n f l i c t relationships are undermined by contending BCOW c o n f l i c t r e s u l t s . Given these re s u l t s the hypothesis cannot be accepted. While re s u l t s supportive of accepting of the hypothesis are present, clear-cut contending results mandate i t s r e j e c t i o n . 226 Anarchy, C o n f l i c t , and Cooperation: V a l i d i t y and Accuracy of H3 Two general l i n e s of conclusions can be made regarding the v a l i d i t y and accuracy Of the anarchy hypothesis. F i r s t , while some results are support accepting the hypothesis, such support i s not adequate to accept the hypothesis as both v a l i d and accurate. The general hypotheses must be rejected because i t does not meet the f i v e c r i t e r i a necessary to d e f i n i t i v e l y accept the hypotheses. Second, d i f f e r e n t aspects of cooperation and c o n f l i c t have d i f f e r e n t r e l ationships to anarchy. Discussion of these r e s u l t s w i l l f i r s t consider the cooperation portion of the hypothesis, and then proceed to the c o n f l i c t portion of the hypothesis. In each of these sections we w i l l f i r s t consider the outcome of the f i v e c r i t e r i a , and then turn to an examination of the r e s u l t s i n more d e t a i l . Anarchy-Cooperation Results Table 7.5 i s a summary report of how well the cooperation portion of the anarchy tenet hypothesis fared with the f i v e c r i t e r i a necessary to d e f i n i t i v e l y aspect the hypothesis. Table 7.5 reveals that the hypothesized correspondence between the magnitude of cooperation scores 227 and anarchy scores is accurate in both the ICB results and BCOW results. Results from both.data sets also concur that the difference of the magnitudes is not significant. Table 7.5 BCOW and ICB Results Adherence to the Five Criteria of the Cooperation-Anarchy Hypothesis ANARCHY COOPERATION ICB Anarchy BCOW Anarchy correspondence of actual and hypothesized ordinal sequences Yes YES 2. significant difference of averages NO NO 3. significant Kendall high Kendall N/A N/A N/A N/A 4. significant gamma high gamma NO NO NO NO 5. error distribution accurate YES NO 228 Further undermining accepting the hypothesis i s the non-significance of the gamma c o e f f i c i e n t s , which when s i g n i f i c a n t are below .54. Results of the two data sets diverge with regard to the difference between actual and expected random d i s t r i b u t i o n s . On th i s c r i t e r i o n , the ICB res u l t s are as hypothesize and the BCOW resu l t s are not. Results from the two data sets' actual and expected random d i s t r i b u t i o n s diverge. The ICB actual d i s t r i b u t i o n i s as hypothesized while the BCOW i s not. Taken as a whole, the r e s u l t s suggests the hypothesis must be rejected. Table 7.6 reports s p e c i f i c c o e f f i c i e n t s and p-values from tests of the anarchy-cooperation hypothesis. The most important i n d i c a t i o n that the hypothesis must be rejected i s that neither the H-test nor the Goodman-Kruskal test yielded consistent p-values needed to accept the hypothesis. Among the six cooperation variables only the c r i s i s management technique y i e l d s s i g n i f i c a n t H-test p-values i n both coding formats. The only other s i g n i f i c a n t H-test p-value i n either coding format i s fo r the BCOW r e c i p r o c i t y direction-anarchy relationship i n the quad-chotomous format. The Goodman-Kruskal gamma significance r e s u l t s provides some ground exist to accept the cooperation portion of the hypothesis. S i g n i f i c a n t gamma p-values are found f o r two variables i n each of the three coding formats (ICB cooperation index and BCOW r e c i p r o c i t y distance). Among the 229 Table 7.6 Cooperation-Anarchy Relationships H-test and gamma Significance Results in each coding format Gamma quad-gamma p t r i -gamma p di-gamma p H -test quad- t r i -P P C r i s i s Management Technique .21 .04 .27 .02 ns 12 13 Timing of Violence ns ns .44 .04 ns ns ICB Cooperation Index 14 .16 .33 .01 .54 .01 ns ns Reciprocity Distance 24 .05 .31 .01 .39 .01 18 ns Reciprocity D i r e c t i o n 18 .08 ns .22 .05 ns ns BCOW Cooperation Index ,30 .01 ns .27 .02 ns ns 230 remaining four variables, one (timing of violence) i s s i g n i f i c a n t i n one coding format, and the other three are s i g n i f i c a n t i n two of the three coding formats (BCOW cooperation index, r e c i p r o c i t y d i r e c t i o n , and c r i s i s management technique). It i s important to note that generally both the gamma values increase as the number of categories decreases. This suggests that these increases may be a function of the gamma calc u l a t i o n , which i s sensitive to coding formats, rather than an ac t u a l l y stronger relationship. Despite the significance of the gamma scores i t i s d i f f i c u l t to accept the hypothesis s o l e l y on the basis of these t h e i r r e s u l t s . More support from the H-test, error d i s t r i b u t i o n s , and more consistency across coding formats are required. However, even i f one where to accept the anarchy cooperation relationship as v a l i d , since the gamma c o e f f i c i e n t does not y i e l d a value greater than .54 the association between cooperation and anarchy i s low at best. In the end diverging gamma c o e f f i c i e n t and sign i f i c a n c e r e s u l t s based on coding format, low gamma c o e f f i c i e n t s , and very l i t t l e support from H-test results, as well as diverging r e s u l t s from the d i s t r i b u t i o n s of errors demands that the anarchy-cooperation portion of the hypothesis be rejected. 231 Anarchy-Conflict Results Table 7.7 reports results of the conflict portion of the anarchy tenet hypothesis according to the five c r i t e r i a . As with previous results, both data sets find that actual and hypothesized average conflict scores correspond to the Table 7.7 Summary of Results ANARCHY CONFLICT ICB Anarchy BCOW Anarchy correspondence of actual and hypothesized ordinal sequences significant difference of averages significant Kendall high Kendall significant gamma high gamma error distribution accurate YES YES YES YES YES YES NO YES NO N/A N/A NO NO NO 232 ordinal magnitude of the anarchy variables. And, also following an e a r l i e r pattern, the two data sets diverge of the s i g n i f i c a n c e of the concurrence: ICB resu l t s y i e lded s i g n i f i c a n t p-values while BCOW resu l t s did not. Among the s i g n i f i c a n t ICB scores, Kendall c o e f f i c i e n t score were both high and s i g n i f i c a n t . Gamma scores followed the same pattern as ICB results, with ICB relationships s i g n i f i c a n t and BCOW one not s i g n i f i c a n t . And, the error d i s t r i b u t i o n was as hypothesized i n the case of ICB data, but not so i n the case of BCOW. Thus while the ICB data does meet the c r i t e r i a , the BCOW results do not. Table 7.8 reports s p e c i f i c c o e f f i c i e n t s and p-values for each anarchy-conflict relationship from each data set. An examination of Table 7.8 reveals that the divergence of the two data sets noted above i s quite dramatic. While ICB res u l t suggest a s i g n i f i c a n t and moderate relationship, BCOW resul t s do not. With one exception ICB p-values are well below the l e v e l required to reject the hypothesis. And, both the H-test and the Goodman-Kruskal gamma test yielded s i g n i f i c a n t r e s u l t s . The ICB anarchy-conflict r e s u l t s y i e l d the strongest among any previous findings to accept the hypothesis. Yet despite ICB results, which support accepting the hypothesis, BCOW results which y i e l d a single anarchy-c o n f l i c t relationship that i s s i g n i f i c a n t i n each coding 233 format, and not a single H-test p-value, make i t d i f f i c u l t to accept the hypothesis. Table 7.8 Conflict-Anarchy Relationships H-test and Gamma Significance Results Gamma H -test quad-gamma p t r i -gamma p d i -gamma p quad-P t r i -P C e n t r a l i t y Violence 25 .01 .27 .001 ' .53 .001 10 07 Intensity Violence 33 .001 .37 .002 .53 .006 01 .01 Overall Violence 25 .01 .31 .001 .36 .11 08 .08 ICB C o n f l i c t Index 29 .004 .29 .01 .49 .02 04 03 H o s t i l i t y Escalation Rate 27 .02 -.36 .01 -.39 .01 ns ns H o s t i l i t y Magnitude ns ns .17 .09 ns ns BCOW Co n f l i c t Index ns ns ns ns ns 234 Discussion Two res u l t s are of p a r t i c u l a r importance. F i r s t , the anarchy hypothesis must be rejected. The preponderance of evidence does not support i t s acceptance. The assertion that higher l e v e l s of c o n f l i c t and lower l e v e l s of cooperation are related to anarchy does not have enough empirical support. Second, while the anarchy hypothesis must be rejected as i t stands, evidence does support the hypothesis that d i f f e r e n t aspects of c o n f l i c t and cooperation have d i f f e r e n t relationships to anarchy. Results also suggest that some aspects of c o n f l i c t and cooperation relationships cannot be dismissed and taken to be unrelated to anarchy. Just because the hypothesis i s rejected as a general hypothesis does not preclude that p a r t i c u l a r aspects of c o n f l i c t or cooperation are not related to anarchy i n important ways. For example, the finding that ICB c o n f l i c t variables are p o s i t i v e l y associated with anarchy and the BCOW h o s t i l i t y e scalation rate i s negatively associated indicates that l e v e l and rate of c o n f l i c t do not share the same rel a t i o n s h i p to anarchy. This c l e a r l y indicates that d i f f e r e n t aspects of c o n f l i c t - - l e v e l s of c o n f l i c t and rate of change i n conflict--have d i f f e r e n t relationships to anarchy. •235 These two findings indicate that i f relationships between anarchy and c o n f l i c t , and between anarchy and cooperation are specified, some types of c o n f l i c t i v e and cooperative behaviors may have important relationships with anarchy. For example the finding that h o s t i l i t y escalation has a negative relationship to anarchy, while violence has a po s i t i v e relationship, and that c r i s i s management technique has a p o s i t i v e relationship to anarchy, while the timing of violence-anarchy relationship i s not s t a t i s t i c a l l y s i g n i f i c a n t , suggest that increased s p e c i f i c a t i o n may lead to better r e s u l t s . From a broader perspective, the results suggests that th e o r i s t s of inte r - s t a t e cooperation and c o n f l i c t who use the anarchy hypothesis as a cornerstone must re-evaluate t h e i r t h e o r e t i c a l foundation. Evidence does not support upholding anarchy as a d i r e c t cause of c o n f l i c t , nor as a di r e c t obstruction to cooperation. And, res u l t s do not indicate that c o n f l i c t and cooperation have opposite relationships to anarchy. The hypothesis that the anarchy-cooperation and anarchy-conflict relationships are similar, with exception to t h e i r direction, cannot be accepted. Aside from r e j e c t i n g the anarchy tenet, the research serio u s l y challenges neorealist theory, insofar as the assumptions made and the circumstances studied are assumptions and circumstances which favor neorealist claims. The assumptions made at the onset about the character of 236 in t e r n a t i o n a l actors and t h e i r environment are neorealist assumptions, and the evidence selected, conditions of states i n c r i s i s , are conditions when neorealism should be accurate. Yet, the r e s u l t s are not supportive of the core ne o r e a l i s t proposition, that anarchy i s the master variable f o r explanations of the recurrence of c o n f l i c t and the absence of cooperation among states. This seriously undermines neorealist assertions about anarchy and i t s consequences. Likewise, theorists who point to the erosion of anarchy as an important factor for increases i n i n t e r -state cooperation must also re-evaluate t h e i r assumptions. The findings also shows that the v a l i d i t y of the anarchy hypothesis increases as the generality of the d e f i n i t i o n s of anarchy, c o n f l i c t , and cooperation increases. The very high Kendall c o e f f i c i e n t s (.93 and 1.00), as well as the increase of gamma c o e f f i c i e n t s when coding formats are more general (change .between quad-chotomous to t r i -chotomous formats), and correctly^ hypothesized relationships between average c o n f l i c t and cooperation variables and anarchy, are clear indications of t h i s dynamic. This suggest that when relationships are generalized and abstracted, the hypothesis i s more accurate than i t s more focused from. 2 1 5 2 1 5 Note that the increased focus i s increasing the number of categories, and not increasing conditional s p e c i f i c a t i o n s of when the hypothesis i s more or less 237 These findings also suggest that hypotheses about anarchy, cooperation, and c o n f l i c t should not be dismissed and ignored i n future research. Clearly some cooperative and c o n f l i c t i v e behaviors are related to anarchy. The fact that the strength of these associations varies from weak to strong suggests that i t would be imprudent to dismiss the importance of anarchy i n s p e c i f i c conditions. This may seems obvious. Yet neither the tenet nor many of i t s adherents assert c o n d i t i o n a l i t y . For example, J e r v i s does not q u a l i f y his claim [T]hat the lack of an international sovereign not only permits wars to occur, but also makes i t d i f f i c u l t f or states to ... arr i v e at a goal that they recognize as being i n t h e i r common i n t e r e s t . 2 1 6 Empirical r e s u l t s found here suggest that s p e c i f i c i t y and c o n d i t i o n a l i t y need to be incorporated into the anarchy tenet hypotheses. The blanket statement that anarchy causes c o n f l i c t and impedes cooperation cannot be accepted with out such s p e c i f i c i t y . Such s p e c i f i c i t y needs to include the d i f f e r e n t i a t i o n between the anarchy-conflict and anarchy-cooperation relationships, as well as among d i f f e r e n t types of cooperation and c o n f l i c t . accurate. 2 1 6 J e r v i s , "Cooperation Under the Security Dilemma", 167. 238 While t h i s conclusion i s consistent with others, 2 1 7 one contribution of the research to t h i s issue i s that i t provides an empirical basis to reconsider the importance of anarchy f o r states' c o n f l i c t i v e and cooperative behaviors. S p e c i f i c a t i o n s of conditions f o r when anarchy i s more or less important can include both longitudinal and actor based variab l e s . In fact, one prominent and p r i n c i p l e supporter of the anarchy tenet, Kenneth Waltz, concedes that the only s t r u c t u r a l change occurred i n 1945, when the structure of the system changed form multi-polar to bi-pol a r systems. While the s p e c i f i c a t i o n i s i n terms of p o l a r i t y , the longitudinal segment i s as well defined. Examples of actor based s p e c i f i c a t i o n s are the currently vogue and often asserted democratic peace theories and t h e i r companion empirical r e s u l t s . 2 1 8 These analysts s p e c i f i c a t i o n and condition i s that the democratic status of a state as an important variable which mitigates the impact of anarchy on both c o n f l i c t and cooperation. In the same vein, others have argued that the defensive or offensive posture of m i l i t a r y c a p a b i l i t i e s are important factors f o r 2 1 7 Morgan, "Multilateralism and security: prospects i n Europe." 2 1 8 For a concise review see David L. Rousseau, Christopher Gelpi, Dan Reiter, and Paul K. Huth, "Assessing the Dyadic Nature of the Democratic Peace, 1918-1988," American Political Science Review 90, no.3 (1995). 239 the l i k e l i h o o d of c o n f l i c t . 2 1 9 These actor based c h a r a c t e r i s t i c s , and undoubtedly others as well, are areas where s p e c i f i c i t y can increase the l e v e l of confidence one has i n t h e i r r e j e c t i o n or acceptance of the anarchy tenet. 2 2 0 Expanding the research to non c r i s i s conditions may also f a c i l i t a t e increasing the s p e c i f i c a t i o n of when the anarchy tenet i s more or less v a l i d and accurate. While the use of c r i s i s data as test data of the anarchy hypotheses i s appropriate because the theoreti c a l underpinnings and assumptions of the test hypotheses are directed p r i m a r i l y at these types of interactions. As such, i f the hypotheses are not v a l i d or accurate i n the conditions which t h e i r t h e o r e t i c a l base are primarily directed, then they must be a l l the more suspect i n other types of interactions. Yet, since the conclusions are based on analysis of int e r n a t i o n a l c r i s e s , a few caveats and grounds where future research can be directed should be noted. E a r l i e r I argued that one advantage of using c r i s i s data i s that cooperative and c o n f l i c t i v e behaviors are more c l e a r l y evident i n c r i s i s conditions than i n other types of conditions. In non-crisis conditions mixed-motives can and 2 1 9 Thomas J. Christensen and Jack Snyder, "Chain gangs and passed bucks: predicting a l l i a n c e patterns i n multi-p o l a r i t y , " International Organization 44, no.2 (1990). 2 2 0 Theise variables were not included i n t h i s analysis since the test hypothesis does not st i p u l a t e intervening or conditional variables. 240 often do r e s u l t i n cross-cutting i n t e r e s t s that can obscure manifestation of c o n f l i c t i v e and cooperative behaviors. What one actor holds up as cooperative behavior can be, and often i s , the r e s u l t of coercive arm-twisting. While non-vi o l e n t , such occurrences are hardly examples of cooperative behavior on the part of the actor with the sore elbow. 2 2 1 Thus an advantage of c r i s i s based research i s that cooperative and c o n f l i c t behaviors are r e l a t i v e l y more detectable and more c l e a r l y manifested than i n n o n - c r i s i s conditions. This advantage i s b e n e f i c i a l to test the anarchy tenet, which makes l i t t l e or no s p e c i f i c a t i o n on the scope of the a p p l i c a b i l i t y of the hypotheses. Yet, the use of c r i s i s - d a t a and the findings also serves to l i m i t the a b i l i t y to generalize r e s u l t s to non-c r i s i s conditions. Thus while results demand that the hypothesis be rejected as i t stands with l i t t l e or no s p e c i f i c i t y , t h i s does not indicate that i f s p e c i f i c a t i o n i s introduced, or i f d i f f e r e n t i n t e r a c t i o n data i s used, r e s u l t s might be d i f f e r e n t . 2 2 1 Examples include many less developed countries i n WTO negotiations, who were worse off i n the WTO than they were under GATT, yet f e l t compelled to j o i n WTO since they would be better off as WTO signatories than non-signatories (Stephen D. Krasner, "Failed Reeds: Ins t i t u t i o n s i n the International System," (Paper prepared for the Conference on What i s Institutionalism?, University of Maryland October 14-15, 1994). A more recent example i s the attempt by the Moi regime i n Kenya to redress international p o l i t i c a l matter i n the face the IMF's funding withdrawal. 241 If data based on other types of interactions were used r e s u l t s may indeed be d i f f e r e n t . If such a difference were to ex i s t , part of the difference may be a r e s u l t s of the i n a b i l i t y to accurately code data f o r l e v e l s of cooperation and c o n f l i c t . As noted e a r l i e r , i n other types of interactions where mixed-motives are more prevalent, and coding cooperative and c o n f l i c t i v e behaviors i s l i k e l y to incur more errors, since a behavior w i l l l i k e l y be a mixture of both cooperative and c o n f l i c t i v e impulses. In many ways the e a r l i e r c a l l s for analysts to d i f f e r e n t i a t e between the anarchy-conflict and anarchy-cooperation relationships, and for theorists to increase s p e c i f i c a t i o n of when anarchy i s more and less important, (both are based on empirical findings), i s t a c i t recommendation that interactions other that c r i s i s conditions should be investigated i n future work. Indeed, results of such examinations may y i e l d d i f f e r e n t conclusions than those reached here. Yet, t h i s i s consistent with the conclusion reached above, that more s p e c i f i c i t y needs to be introduced to the anarchy tenet hypotheses. Aside from te s t i n g a foundational assumption of int e r n a t i o n a l relations theory, and r e j e c t i n g i t s empirical accuracy and v a l i d i t y , t h i s research has also contributed an method to generate a measure of anarchy. The method s p e c i f i e s d i f f e r e n t l e v e l s of anarchy i n terms which can be 242 measured, and provides th e o r e t i c a l grounding about the rel a t i o n s h i p between d i f f e r e n t l e v e l s of anarchy and c o n f l i c t and cooperation. The conception of anarchy as non-ubiquitous i s c e r t a i n l y d i f f e r e n t from that advanced i n the mainstream. Conceiving of anarchy i n terms of aggregated actor c h a r a c t e r i s t i c s i s a d i r e c t challenge to the common thinking that anarchy i s uni v e r s a l l y s i m i l a r : i f the actor c h a r a c t e r i s t i c s change, t h e i r aggregation can change, and anarchy can also change. Not only have I argued t h i s , but I also argue that d i f f e r e n t l e v e l s of anarchy can exis t concurrently among d i f f e r e n t sets of actors. This i s , i t seems, an even more r a d i c a l departure from the mainstream concept of anarchy. The t h e o r e t i c a l and empirical e f f o r t of the research point to the need for more the o r e t i c a l and empirical work regarding the anarchy tenet: theory needs to specify when anarchy i s more and less important and the s p e c i f i c a t i o n needs to be done i n a manner that hypotheses can be generated and tested. 243 Bibliography Ashley, Richard K. 1988. Untying the sovereign state: a double reading of the anarchy problematique. Millennium 17, no.2:227-262. Axelrod, Robert. 1984. The Evolution of Cooperation. New York:Basic Books. Axelrod, Robert and Robert Keohane. 1985. Achieving cooperation under anarchy: strategies and i n s t i t u t i o n s World Politics 38, no.1:228-254. Agresti, Alan and Barbara Finlay. 1986. Statistical Methods of the Social Sciences, 2nd ed. San Francisco:Dellen. von Bertalanffy, Ludwig. 1956. General systems theory. General Systems 1:1-16. Black, Antony. 1993. Political Thought In Europe 1250-1450. New York:Cambridge. Bodin, Jean. Republique, [1583]. In On Sovereignty trans. and ed. by J u l i a n H. Franklin. New York:Cambridge, 1992. . Bossuet, Jacques Benigne. [1870]. Politique tiree des propre paroles de l'Ecriture Sainte, trans L. Pearce Williams. In The Absolutism of Louis XIV, ed. Brian Tierney, Donald Kagan, and L. Pearce Williams. New York:Random House, 1967. Bryman, Alan and Duncan Creamer. 1990. Quantitative Data Analysis for the Social Science. New York:Routledge. B u l l , Headley. 1977. The Anarchical Society: A Study of Order in World Politics. New York:Columbia. Burns, J.H. 1990. The idea of absolutism. In Absolutism in . Seventeenth-Century Europe, ed John M i l l e r , 16-37. London:Macmillam. 244 Buzan, Barry. 1983. People, States, and Fear:The National Security Problem in International Relations. Chapel H i l l , N.C.:University of North Carolina Press. Buzan, Barry, Charles Jones, and Richard L i t t l e . 1993. The Logic of Anarchy: Neorealism to Structural Realism. New York:Columbia. C a m i l l e r i , Joseph A. 1990. Rethinking sovereignty i n a shrinking, fragmented world. In Contending Sovereignties: Redefining Political Community, ed. R.B.J. Walker and Saul H. Mendlowitz, 56-72. Boulder, C0:Lynn Reinner. Cerny, P h i l l i p G. 1993. P l u r i l a t e r a l i s m , s t r u c t u r a l d i f f e r e n t i a t i o n and functional c o n f l i c t i n the post-cold war world order. Millennium 22, no.1:27-51. Champion, Dean J. 1970. Basic Statistics for Social Research. Scranton:Chandler. . 1981. Basic Statistics for Social Research 2nd. ed. Scranton:Chandler. Christensen, Thomas J. and Jack Snyder. 1990. Chain gangs and passed bucks: predicting a l l i a n c e patterns i n m u l t i - p o l a r i t y International Organization 44, no.2:137-68. Deutsch, Karl W. 1957. Political Community in the North Atlantic Area. Princeton NJ:Princeton University Press. Dudney, Daniel. Binding powers, bound states: the l o g i c of g e o p o l i t i c s and negarchy. Paper presented and the 34th International Studies Association Meeting, Washington D.C., March, 1994. Fisher, Markus. 1992. Feudal Europe, 800-1300: communal discourse and c o n f l i c t u a l practices International Organization 46, no.2:427-465. Franklin, J u l i a n H., ed. 1992. On Sovereignty. New York:Cambridge. Gibbons, Jean Dickinson. 1993. Nonparametric Statistics: An Introduction Newberry Park, CA:Sage. 245 . 1993. Nonparametric Measure of Association. Newberry Park, CA:Sage. G i l p i n , Robert. 1981. War and Change in World Politics. New York:Cambridge. Gottemoeller, Rose E. 1987/88. Land attack cruise missiles Adelphi Papers No. 226 (Winter). Greico, Joseph M. 1988. Anarchy and the l i m i t s of cooperation. International Organization 42, no.3:485-507. - . 1988. Realist theory and the problem of international cooperation. Journal of Politics 50, no.3:600-624. . 1990. Cooperation Among Nation. I t h i c a : C o r n e l l . . 1993. Understanding the problem of international cooperation. In Neorealism and Neoliberals, ed. David Baldwin, 116-143. New York^Columbia University Press. • _. 1995. Anarchy and the l i m i t s of cooperation: a r e a l i s t c r i t i q u e of the newest l i b e r a l i n s t i t u t i o n a l i s m . In Controversies in International Relations Theory: Realism and the Neoliberal Challenge, ed. Charles W. Kegley, 151-171. New York:St. Martin's Press. H a l l , John A. and G. John Ikenbeery. 1989. The State. Minneapolis:University of Minnesota Press. Hass, Peter M. 1992. Epistemic communities and international environmental protection. International Organization 46, no.1:1-36. Held, David. 1989. Political Theory and the Modern State: Essays on State, Power, and Democracy. Stanford, CA:Stanford University Press. Hinsley, F. H. 1986. Sovereignty, 2nd ed. New York:Cambridge. H o l s t i , K. J . 1985. The Dividing Discipline. Boston:Allen & Unwin. 246 . 1990. The comparative analysis of foreign p o l i c y : some notes on the p i t f a l l s and paths to theory. In The Political Economy of Foreign Policy in South East Asia, ed. David Warfel and Bruce Burton, 9-20. New York:St. Martin's Press. _ . 1992. Government without governance: polyarchy i n the nineteenth century european int e r n a t i o n a l p o l i t i c s . In Governance Without Government: Order and Change in World Politics, ed. James N. Rosenau and Otto-Ernst Czempiel, 30-57. New York: Cambridge. Hopf, Ted. 1995. Domesticizing anarchy: state i d e n t i t y ( s ) and i n t e r s t a t e security. Paper Presented at the 3 6th International Studies Association Conference, Chicago, February 1995 James, Patrick. 1993. Neorealism as a research enterprise. Jnternational Political Science Review 14, no.2:123-148. Je r v i s , Robert. 1978. Cooperation under the security dilemma. World Politics 30, no.2:167-214. . 1988 Realism, game theory, and cooperation. World Politics 40, no.1:317-349. de Jouvenel, Bertrand. 1957. Sovereignty and Inquiry into Political Good, trans. J.F. Huntington. Chicago:University of Chicago Press. Kaplan, Morton A. 1963. System and Process in International Politics. New York:John Wiley and Sons. Kegley, Charles W. J r . , ed. 1995. Controversies in International Relations Theory: Realism and the Neoliberal Challenge. New York:St. Martin's. Kegely, Charles W. J r . and Gregory A. Raymond. 1990. When Trust Breaks Down: Alliance Norms and World Politics. Columbia:University of South Carolina. Keren, Gidion and Charles Lewis, ed. 1993. A Handbook for Data Analysis in the Behavioral Sciences: Methodological Issues. H i l l s d a l e , N.J:Lawrence Erlhawm. Kiernan, V.G. 1980. State and Society in Europe 1550-1650. New York:St. Martin's Press. 247 Keohane, Robert 0. 1984. A f t e r Hegemony: Cooperation and Discord in the World Political Economy. Princeton N.J.:Princeton. ' . 1986. Reciprocity i n international r e l a t i o n s . International Organization 40, no.1:1-27. _. 1989. International Institutions and State Power: Essays in International Relations Theories. Boulder CO:Westview. Keohane, Robert 0. and Joseph S. Nye. 1977. Power and Interdependence: World Politics in Transition. Boston:Little Brown. Knorr, Klaus and James N. Rosenau, ed. 1969. Contending Approaches to International Politics. Princeton, N.J.:Princeton University Press. Knutsen, Torbjorn L. 1992. A History of International Relations Theory. New York Manchester. Kohout, Frank. 1974. Statistics for Social Science: A Coordinated Learning System. New York:Wiley. Krasner, Stephen. 1983. Structural causes and regime consequences. In International Regimes, 1-23. I t h i c a New York:Cornell. . 19-76. State power and the structure of international trade. World Politics 28 no.3:317-347. . Fa i l e d Reeds:' Institutions on the International System. Paper presented at the Conference on What i s I n s t i t u t i o n a l i s m Now? University of Maryland, College Park, Maryland, October 06, 1994. Lake, David A. 1996. Anarchy, hierarchy and the v a r i e t y of international r e l a t i o n s . International Organization 50 no.1:1-33 Labovitz, Sandford. 1970. C r i t e r i a f o r s e l e c t i n g a si g n i f i c a n c e l e v e l : a note on the sacredness of .05. In The Significance Test Controversy--A Reader, ed. Denton E. Morrison and Ramon E. Henkel, 166-171. Chicago:Aldine. 248 Layne, Christopher. 1995. The unipolar i l l u s s i o n : why new great powers w i l l r i s e . " In Perils of Anarchy: Contemporary Realism and International Security, ed. Michael E. Brown, Sean M . Lynn-Jones, and Steve E. M i l l e r , 131-154. Cambridge:MIT Press Leng, Russell J. 1993. Interstate Crisis Behavior 1816-1980: Realism Versus Reciprocity. New York:Cambridge. Levy, Jack S. 1989. The causes*b"f war: a Review of theories and evidence. In Behavior, Society, and Nuclear War, Volume One, ed. P h i l i p Tetlock et. al.', 209-313. New York:Oxford. Lijph a r t , Arend. 1974. The structure of the t h e o r e t i c a l revolution i n international r e l a t i o n s . International Studies Quarterly 18, no.1:41-74. Lipson, Charles. 1984. International cooperation i n economic and security a f f a i r s . World Politics 37, no.1:1-23 Mansbach, Richard W. and John A. Vasquez. 1981. In Search of Theory: A New Paradigm for Global Politics. New York:Columbia. Mercer, Jonathan. 1995. Anarchy, s e l f help, r e l a t i v e gains. International Organization 49, no.2:229-252. Merritt, Richard L. and Bruce M. Russett, ed. 1981. From National Development to Global Community: Essays in Honor of Karl W. Deutsh. London:George, Allen , and Unwin. M i l l e r , Lynn H. 1985. Global Order: Values and Power in International Politics. Boulder CO:Westview 1985. Milner, Helen. 1991. The assumption of anarchy i n international relations theory: a c r i t i q u e . Review of International Studies 17:67-85. Moravcsik, Andrew. 1992. Liberalism and International Relations Theory. CFIA Working Paper Cambridge:CFIA Harvard University. Morgan, Patrick M. 1993. M u l t i l a t e r a l i s m and security: prospects i n Europe. In Multilateralism Matter: The Theory and Praxis of an Institutional Form, ed. John Gerald Ruggie, 327-364. New York:Columbia. 249 Morgenthau, Hans. 1954. Politics Among Nations.: The Struggle for Power and Peace, 2nd ed. New York:Knopf. Morrison, Denton E. and Ramon E. Henkel, ed. 1970. The Significance Test Controversy--A Reader. Chicago: Aldine. Morse, Edward L. 1976. Modernization and the Transformation of International Relations. New York .-The Free Press. Nordlinger, E r i c . 1981. The Autonomy of the Democratic State Cambridge:Harvard. North, Robert. Discussion at the 1994 Annual Meeting of the American P o l i t i c a l Science Association, Washington D.C., 1994. Onuf, Nicholas Greenwood. 1989. World of Our Making: Rules and Rule in Social Theory and International Relations. Columbia:South Carolina. Onuf, Nicholas Greenwood and Frank F. Klink. 1989. Anarchy, authority, and rule. International Studies Quarterly 33, no.2:149-173. Oye, Kenneth. 1985. Explaining cooperation under anarchy: hypotheses and strategies. World Politics 38, no.1:1-24. Palmer, R.R. and Joel Colton. 1967. The France of Louis XIV-a b r i e f perspective.; In The Absolutism of Louis XIV-- The End of Anarchy or the Beginning of Tyranny?, ed. Brian Tierney, Donald Kagan, and L. Pearce Williams, 12-27. New York:Random House. Philpott, Daniel. 1995. Sovereignty: an introduction and b r i e f history. Journal of International Affairs 48, no.2:353-368. Puchala, Donald J . 1981. Integration theory and the study of international r e l a t i o n s . In From National Development to Global Community: Essays in Honor of Karl W. Deutsh, ed. Richard L. Merritt and Bruce M. Russett, 145-165. London:George, Allen , and Unwin. Rapoport, Anatol. 1969. The Strategy of Conscience. New York:Schocken Books. 250 Rosenau, James N. and Otto-Ernst Czempiel, ed. 1992. Governance Without Government: Order and Change in World Politics. New York:Cambridge. Rousseau, David L., Christopher Gelpi, Dan Reiter, and Paul K. Huth. 1996. Assessing the Dyadic Nature of the Democratic Peace, 1918-1988. American Political Science Review 90, no.3:512-537. Rittenberger, Volker and Michael Ziirn, ed. 1990. International Regimes in East-West Policies. London:Printer Publishers. Ruggie, John Gerald. 1986. Continuity and transformation i n the world p o l i t y : toward a neorealist synthesis. In Neorealism and it Critics, ed. Robert 0. Keohane, 131-157. New York:Columbia University Press. . 1982. International regimes, transactions and change: embedded l i b e r a l i s m i n the post-war economic order. International Organization 36, no.2:379-415. ' ' ed. 1994. Multilateralism Matters The Theory and Praxis of an Institutional Form. New York:Columbia. Schroder, Paul. 1995. Anarchy i n the 19th Century. In Perils of Anarchy: Contemporary Realism and International Security, ed. Michael E. Brown, Sean M. Lynn-Jones, and Steve E. M i l l e r , 437-453. Cambridge:MIT Press 1995. . Seigel, Sidney. 1956. Nonparametric Statistics for the Behavioral Sciences. New York:McGraw-Hill. Seigel, Sidney and N. John Castellan, J r . 1988. Nonparametric Statistics for the Behavioral Sciences. New York:McGraw-Hill. Singer, John David. 1961. The leve l s of analysis problem. In The International. System: Theoretical Essays, ed. Klaus Knorr and James N. Rosenau, 63-83. New York:Free Press, 1961. _. 1969. The Incomplete Theorist: Insight Without Evidence. In Contending Approaches to International Politics, ed. Klaus Knorr and James N. Rosenau, 62-86. Princeton, N.J.:Princeton University Press. 251 Snidal, Duncan. 1985. Coordination versus prisoners' dilemma: implications f o r international cooperation and regimes. American Political Science Review 79, no.4:923-942. . 1985. Snidal, Duncan. The Game theory of international p o l i t i c s . World Politics 38, no.1:25-57. 1991. Relative gains and the pattern of international cooperation. American Political Science Review 85, no.3:701-726. Stein, Arthur A. 1990. Why Nations Cooperated: Circumstance and Choice in International Relations. Ithica:Cornell University Press. . 1983. Coordination and collaboration: regimes i n an anarchic world. In International Regimes, ed. Stephen Krasner, 115-140. Ithica, New . York:Cornell. Thompson, Janice E. 1995. State sovereignty i n int e r n a t i o n a l r e l a t i o n s : bridging the gap between theory and empirical research. International Studies Quarterly 39, no.2:213-233. Tierney, Brian, Donald Kagan, and L. Pearce Williams. 1967. The Absolutism of Louis XIV-- The End of Anarchy or the Beginning of Tyranny?. New York:Random House. Wagner, Harrison. 1983. The theory of games and the problem of international cooperation. American Political Science Review 77, no.2:330-347. Walker, R.B.J, and Saul H. Mendlowitz, ed., 1990. Contending Sovereignties: Redefining Political Community. Boulder, CO:Lynn Reinner. Waltz, Kenneth N. 1979. Theory of International Politics. Reading MA:Addison-Wesley. . 1956. Man, the State, and War: A Theoretical Analysis. New York:Columbia. . 1995. Realist thought and neorealist theory. In Controversies in International Relations Theory: Realism and the Neoliberal Challenge, ed. Charles W. Kegley, 67-82. New York:St. Martin. 252 Watson, Adam. 1992. The Evolution of International Society: A Comparative Historical Analysis. New York:Routledge. Wendt, Alexander. 1992. Anarchy i s what states make of i t : the s o c i a l construction of power p o l i t i c s . International Organization 46, no.2:391-426. Weber, Steve. 1991. Cooperation and Discord in U.S.-Soviet Arms Control. Princeton, N.J.:Princeton. Wight, Martin. 1986. Power Politics. New York:Penguin. Wilkenfeld, Jonathan et. a l . 1989. Crisis in the Twentieth Century, Volume II. New York:Permagon. Wilkenfeld, Jonathan and Michael Brecher. 1989. Code Book for ICB1, System Level Data, photocopied. College Park, MD:University of Maryland. Wilkenfeld, Jonathan and Michael Brecher. 1989. Code Book for ICB2, Actor Level Data, photocopied. College Park, MD:University of Maryland. Young, Oran. 1978. Anarchy and s o c i a l choice: r e f l e c t i o n s on the international p o l i t y . World Politics 30, no.2:241-263. Zacher, Mark. 1992. The decaying p i l l a r s of the Westphalian temple: implications for international order and governance. In Governance Without Government: Order and Change in World Politics, ed. James N. Rosenau and Otto-Ernst Czempiel, 58-101. New York:Cambridge, 1992. Zacher, Mark W. and Richard A. Matthew. L i b e r a l international theory: common threads, divergent strands. Paper Presented at 88th annual American P o l i t i c a l Science Association Conference, 1992. Zimmerman, Donald and Bruno Zumbo. 1993. The r e l a t i v e power of parametric and nonparametric s t a t i s t i c a l methods. In A Handbook for Data Analysis in the Behavioral Sciences: Methodological Issues, ed. Gidion Keren and Charles Lewis. H i l l s d a l e , 429-443. N.J:Lawrence Erlhawm. 253 APPENDIX A Calculation Method of Proportional Similarity of Attributes Score for n=3,4,5 Actors This Appendix outlines the procedures to generate and standardize s i m i l a r i t y of ends and means values and anarchy scores for n>2 actors and y attributes where yan. In order to assess the l e v e l of ( d i s ) s i m i l a r i t y a method which can be applied to cases with n>2 actors which provides at least ordinal l e v e l information about the l e v e l of ( d i s ) s i m i l a r i t y for a l l actors needs to be devised. Such a measure i s contained i n c a l c u l a t i n g the proportional s i m i l a r i t y of attributes (PSA) , 2 2 2 The PSA i s the c a l c u l a t i o n of the proportion of s i m i l a r and d i s - s i m i l a r a t t r i b u t e s among actors. . 2 2 3 The PSA score ranges from 0 2 2 2 The dyadic evaluation used i n the BCOW analysis i s consistent with the PSA. While a p p l i c a t i o n of PSA to a two actor cases i s possible, the simpler "look-and-see" method used i s much easier and equally v a l i d . 2 2 3 While i t may be desirous to include some type of sc a l i n g which increase p r e c i s i o n of measure, such sca l i n g i s not necessary to test the hypotheses under examination. Any sc a l i n g would greatly complicate an analysis when s i m i l a r i t y l e v e l s are generated with the PSA method (which i s discussed further below) and standardized for the number of actors. Increasing the pre c i s i o n of measures may also introduce spurious accuracy. 254 to 1 where the higher the score the moire s i m i l a r the attr i b u t e s among actors, and the lower the score the l e s s s i m i l a r the att r i b u t e s . The c a l c u l a t i o n of the PSA score proceeds through four steps. F i r s t , the maximum number of possible a t t r i b u t e d i s t r i b u t i o n s d i f f e r e n t i a t e d by s i m i l a r i t y of at t r i b u t e s i s calculated for n=3, n=4, and n=5 actors, where the number of attr i b u t e s (k) i s greater than or equal to n. 2 2 4 These d i s t r i b u t i o n s are referred to as N. To calculate N one l i s t s each possible N d i s t r i b u t i o n given n-actors and k-attributes. This i s accomplished by l i s t i n g the most si m i l a r N d i s t r i b u t i o n , where each actor has the same attribute, to the least s i m i l a r N d i s t r i b u t i o n , where no two actors have the same at t r i b u t e . For kan, when n=3 there i s maximum of 3 N d i s t r i b u t i o n s ; when n=4 there i s a maximum of 5 N distributions,- and when n=5 there i s a maximum of 7 N d i s t r i b u t i o n s . These l i s t i n g s are located below i n Figure 4.10. The numbers i n Figure A l . l r e f e r to a single a t t r i b u t e which can be any of the attributes of the varia b l e . For example i f n=4 and the variable i s colour of ink used and the range colours i s red, blue, black, and green, 1 can 2 2 4 The number of attributes k must be greater than or equal to n (kan) for the maximum number of a t t r i b u t e d i s t r i b u t i o n s (N) to occur. If the number of attributes k i s less than n (k<n), then the maximum number of at t r i b u t e d i s t r i b u t i o n s (N) i s less than those l i s t e d i n Figure 4.10. 255 r e f e r to any of these colours so long as the reference i s held constant i n the analysis. In permutation (1) { l l l l } the series of four l ' s indicates that each ink user uses the same colour of ink; i t t e l l s use nothing about which colour i s being used. In permutation (2) {1112} three ink user use the same colour ink while the fourth uses a d i f f e r e n t colour. In permutation (3) {1122} two ink user use the same colour ink (noted with l's) which i s d i f f e r e n t i n colour from the shared colour the other two use (noted with 2's). In permutation (4) {1123} two ink users use the same colour (noted with l's) while the remaining two ink users use two colours of ink which are d i f f e r e n t from the f i r s t two users and d i f f e r e n t from each other (noted with 2 and 3). And i n permutation (5) each ink user uses a colour of ink which i s d i f f e r e n t from every other ink user (noted with 1 2 3 4). Figure A l . l N d i s t r i b u t i o n s for n Actors when n=3,4,5 3 Actors A B G (Si) 1 • 1 1 (N2) 1 1 2 (N3) 1 2 3 4 Actors A B C D (ND 1 1 1 1 (N2) 1 1 1 2 (N3) 1 1 2 2 (N4) 1 1 2 3 (N5) 1 2 3 4 5 Actors A B C D E (ND .1 1. 1 1 1 (N2) T 1 1 1- 2 (N3) 1 1 1 2 2 (N4) 1 1 1 2 3 (N5) 1 1 2 2 3 (N6) 1 1 2 3 4 (N7) 1 . 2 3 4 5 257 Second, the maximum possible number of unique p a i r s of actors given the number of c r i s i s actors are l i s t e d . 2 2 5 This step i s accomplished by creating an n by n matrix where n=the number of actors. Those c e l l s i n the matrix which l i e on the northwest to southeast diagonal are ignored since they represents pairings of the same variables. Those c e l l s which l i e above the northwest to southeast diagonal are ignored since they are redundant to those below the diagonal. The remaining c e l l s , below the diagonal, are the v a l i d c e l l s , 2 2 6 The t h i r d step i n the generation of PSA scores i s a count of the number of equality and non-equality pairs i s taken i n the n-by-n matrix created i n step two. This i s accomplished by plugging i n the a t t r i b u t e values from each of the N d i s t r i b u t i o n l i s t e d above and comparing the value of the row and column attributes i n the v a l i d c e l l s . The s p e c i f i c ordering of actors along row and column i s only important i n so f a r as the same actor occupies the same row and column po s i t i o n . If the two attributes are the same, then an equal sign i s placed n the c e l l ; i f the two att r i b u t e s are d i f f e r e n t , then a not equal sign i s placed i n 2 2 5 In a 5 actor c r i s i s there i s a maximum of ten unique p a i r s of states; i n a four actor c r i s i s there i s a maximum of 6 unique pairings; and i n a three actor cases there i s a maximum of 3 unique parings. 2 2 6 The number of c e l l s i s the same to the number generated by the formula: n!/2!-(n!). 258 the c e l l . This process i s conducted f o r each N d i s t r i b u t i o n f o r n= 3,4, and 5 actors i n Figures 4.11 to 4.13. The PSA scores are l i s t e d below each matrix. F i n a l l y the proportion of the number of equal p a i r to the maximum number of equal pairs i s taken. This l a s t step r e s u l t s i n the PSA score which i s a number that ranges from 0 to l . A PSA score of 1 represent the case when the number of pairs-wise e q u a l i t i e s i s equal to the maximum number of pair-wise e q u a l i t i e s . The denotes that a l l actors a t t r i b u t e s are the same and thus a very high l e v e l of s i m i l a r i t y between of among actors. A PSA score of 0 represents the case.when the number of pair-wise e q u a l i t i e s i s at the minimum l e v e l of 0. Thus, a PSA score of 0 denotes a case where each actor has a d i f f e r e n t a t t r i b u t e and thus a very low l e v e l of s i m i l a r i t y i s present. Thus, the higher the PSA score, the more si m i l a r are the att r i b u t e s among the actors. The lower the PSA score, the lower the s i m i l a r i t y of attributes among actors. Table A1.2 PSA score c a l c u l a t i o n s f o r n=3, n=4 and n=5 actors 3 actor d i s t r i b u t i o n (Nl) A B c 1 1 1 A 1 X X X B 1 = X X C 1 = X number of signs: 3= 0* maximum number of = signs: 3 PSA= 3/3 - .1.0 d i s t r i b u t i o n (N2) A B C 1 1 2 A 1 X X X B 1 = X X C 2 * X number of signs: 1= maximum .number Of = signs PSA= 1/3 S 0.33 d i s t r i b u t i o n (N3) A B C 1 2 3 A 1 X X X B 2 X X C 3 * X number of signs: 0= 3* maximum number of = signs: 3 PSA= 0/3 =0.0 Table A1.21 four actor d i s t r i b u t i o n (Nl) A B C D 1 1 1 1 A 1 X X X X B 1 X X X C 1 = = X X D 1 = = X number of signs: 6= 0* maximum number of = signs: 6 PSAs 6/6 « 1.0 d i s t r i b u t i o n (N2) d i s t r i b u t i o n (N4) A B c D A B c D 1 1 1 2 1 1 2 3 A 1 X X X X A 1 . X X X X B 1 X X X B 1 = X X X C 1 = X X C 2 X X D 2 * X D 3 X number of signs: 3= 3* number of signs: 1= 5 maximum number of = signs: 6 maximum number of = signs: PSA= 3/6 = 0.5 PSA= 1/6 = .165 d i s t r i b u t i o n (N3) d i s t r i b u t i o n (N5) A B c D A B C D 1 1 2 2 1 2 3 4 A 1 X X X X A 1 X X X X B 1 X X X B 2 X X X C 2 * X X C 3 X X D 2 * = X D 4 * X number of signs: 2= 4* maximum number of = signs: 6 PSA= 2/6 = 0.33 number of signs: 0= 6 maximum number of = signs: PSA= 0/6 =0.0 Table A1.23 f i v e actor d i s t r i b u t i o n Nl A B c D E 1 1 1 1 1 A 1 X X X X X B 1 = X X X X C 1 = . = X X X D 1 = = = X X E 1 = = X number of signs: 10= 0* maximum number of = signs: 10 PSA= 10/10 =1.0 d i s t r i b u t i o n N2 d i s t r i b u t i o n N5 A B c D E A B c D E 1 1 1 1 2 1 1 2 2 3 A 1 X X X X X A 1 X X X X X B 1 X X X X B 1 = X X X X C 1 = X X X C 2 X X X D 1 = = X X D 2 it = X X E 2 * it it X E 3 it ?t X number of signs: 6 = 4* number of signs: 2 = 8* maximum number of = signs : 10 maximum number of = signs :10 PSA= 6/10 = 0.6 PSA= 2/10 B 0.2 d i s t r i b u t i o n N3 d i s t r i b u t i o n N6 A B C D E A B C D E 1 1 1 2 2 1' 1 2 3 4 A 1 X X X X X A 1 X X X X X B 1 X X X X B 1 ' = X X X X C 1 = = X X X C 2 * X X X D 2 * X X D 3 X X E 2 * y • • = X •' E 4 it it it X number of signs: 4 = 6* number of signs: 1 = Sit maximum number of = signs :10 maximum number of = signs :10 PSA= 4/10 = .4 PSA= 1/10 B 0.1 d i s t r i b u t i o n N4 d i s t r i b u t i o n N7 A B C D E A B C D E 1 1 1 2 3 1 2 3 4 5 A 1 X X X X X A 1 X X X X X B 1 = X X X X B 2 it X X X X C 1 .= X X X C 3 * X X X D 2 •* * X X D 4 * * It X X E 3 * it X E 5 it .it It it X number of signs: 3= 7* maximum number of = signs: 10 PSA= 3/10 =0.3 number of signs: 0= 10* maximum number of = signs:10 PSA= 0/10 =0.0 262 Once the PSA scores have been obtained an anarchy score can be generated. The anarchy score i s calculated using the anarchy model matrix with the PSA scores f o r the s i m i l a r i t y of ends i n the column placement and the PSA s i m i l a r i t y of means i n the row placement. The most s i m i l a r PSA score i s placed i n the most northwest c e l l , and i n descending order i n the PSA scores f o r ends follow along the west to east axis and the PSA scores for means on the north to south axis. This procedure generates nine anarchy scores for n=3 actors, twenty f i v e anarchy scores for n=4 actors, and f o r t y nine anarchy scores for n=5 actors, and i s i l l u s t r a t e d below i n Figure 4.14. Note that t h i s procedure produces the universe of possible anarchy scores for 3, 4, and 5 actors cases for k categories when kan. If k<n, then the range of possible anarchy and PSA scores diminishes. For example, i f k=2 and n=5 then the lowest possible PSA score i s PSA=0.4, or the PSA score obtained i n d i s t r i b u t i o n N3 i n Table 4.13. Figure Al.3 Calculation of Anarchy for no3, n=4, and n=5 actors using Ends and Means PSA scores n=3 PSA ends PSA MEANS 1 .33 0 1 9 6 3 .33 8 5 2 0 7 4 1 n=4 P S A e n d s PSA MEANS 1 .5 .33 .165 0 1 25 20 15 10 5 .5 24 19 14 9 4 .33 23 18 13 8 3 .165 22 17 12 7 2 0 21 16 11 6 1 n=5 e n d s 1 .6 .4 .3 .2 .1 0 1 49 42 : 35 28 21 14 7 :6 48 41 34 27 20 13 6 .4 47 40 33 26 19 12 5 .3 46 39 32 25 18 11 4 .2 45 38 . 31 24 17 10 3 . 1 44 37 30 23 16 9 2 0 43 36 29 22 15 8 1 264 The PSA scores re sensitive to the number of actors. To standardize PSA scores so that they are actor number in s e n s i t i v e one modification i s required. The PSA scores are categorized into high s i m i l a r i t y and low s i m i l a r i t y values. PSA scores greater than or equal to 0.5 are coded as high s i m i l a r i t y and a PSA score l e s s than 0.5 are coded as low s i m i l a r i t y . This process i s i l l u s t r a t e d below i n Figure 4.15. Figure A1.4 ANARCHY, SIMILARITY OF ENDS (ESIM) AND MEANS (MSIM) PSA Recodings ESIM & MSIM ANARCHY 0 1 1 2 3 4 PSA t t t t t t code t t ' • t t t t for: t t t t t t n=2 0 1 1 2 3 4 n=3 0 to .3 1 2,3 4,5 6,7 8,9 n=4 0 to. 4 .6. to 1 2-7 8-13 14-19 20-25 n=5 0 to .4 •6 to 1 1-12 13-24 25-37 38-4 9 • The decision to use the 0.5 l e v e l as the d i v i d i n g between high and low l e v e l s of s i m i l a r i t y i s not an a r b i t r a r y one. The basis for the decision i s that below the 265 0.5 PSA l e v e l attributes d i f f e r more than those at or above 0.5. From these actor number i n s e n s i t i v e PSA scores am anarchy score i s calculated for the c r i s i s as a whole. 266 Appendix B Calculations of Kendall Coefficient of Concordance W, S, Maximum S, and significance level Note: a l l significance levels where generated using a c r i t i c a l S table according to the method outlined by Gibbons in Nonparametric Measure of Association. Anarchy and Conflict Variables, Quad-chotomous Form Anarchy 1. 00 2 00 3.00 4.00 1 0 2 0 3 0 4 0 C e n t . V i o l 2 . 06 2 00 2.41 2.69 2 0 1 0 3 0 4 0 I n t n . V i o l 1.81 1 67 2.18 2.47 2 0 1 0 3 0 4 0 Over.Vio 1 . 98 1 78 2.24 2.47 2 0 1 0 3 0 4 0 1 sum of column 7 0 5 0 12 0 16 0 2 mean of 1 10 0 3 (mean-sum) A2 9 0 25 0 4 0 36 0 S = 74 . 0 4 sum Of 3 S = 74 0 Max S = 80 . 0 1 Max S = 4 0 8 0 12 0 16 0 W = 0 . 9 2 mean of 1 10 0 3 (mean-sum) "'2 36 0 4 . 0 4 . 0 36 0 n=4, k=4 4 sum Of 3 Max S = 80 0 sig=.002 W = S/Max s 0 . 9 A n a r c h y and C o n f l i c t V a r i a b l e s , Quad-chotomous Form, A n a r c h y L e v e l 2 removed Anarchy 1.00 3.00 4.00 1 6 2 0 3 0 Cent.Viol 2 . 06 2.41 2.69 1 0 2 0 3 0 Intn.Viol 1.81 2.18 2.47 1 0 2 0 3 0 Over.Vio 1. 98 2.24 2.47 1 0 2 0 3 0 1 sum of column 4 0 8 0 12 0 2 mean of 1 8 0 3 (mean-sum)A2 16 0 0 0 16 0 S = 32.0 4 sum Of 3 S = 32 0 Max S = 32 . 0 1 Max S « 4 0 8 0 12 0 W = 1.0 2 mean of 1 8 0 3 (mean-sum)A2 16 0 0 .0 16 0 n=3, k=4 4 sum Of 3 Max S = 32 0 sig=.028 W = S/Max s = 1 0 A n a r c h y and C o n f l i c t V a r i a b l e s ; T r i - c h o t o m o u s Form Anarchy 1. 00 2 00 3.00 1 0 2 0 3 0 Cent.Viol 2 . 06 2 34 2.69 1 0 2 0 3 0 Int.Viol 1.81 2 10 2.47 1 0 2 0 3 0 Over.Viol 1.98 2 17 2 .47 1 0 2 0 3 0 1 sum of column 4 0 8 0 12 0 2 mean of 1 8 0 3 (mean-sum)A2 16 0 0 0 16 0 S = 32 . 0 4 sum Of 3 S = 32 0 Max S = 32 . 0 1 Max S = 4 0 8 0 12 0 W = 1. 0 2 mean of 1 8 0 3 (mean-sum)A2 16 0 0 0 16 . 0 n=3, k=4 4 sum Of 3 Max S = 32 0 sig=.028 W = S/Max s = 1 0 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0088081/manifest

Comment

Related Items